]>
Commit | Line | Data |
---|---|---|
b0bbbd85 RS |
1 | /* C Compatible Compiler Preprocessor (CCCP) |
2 | Copyright (C) 1986, 1987, 1989, 1992 Free Software Foundation, Inc. | |
3 | Written by Paul Rubin, June 1986 | |
4 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
5 | ||
6 | This program is free software; you can redistribute it and/or modify it | |
7 | under the terms of the GNU General Public License as published by the | |
8 | Free Software Foundation; either version 2, or (at your option) any | |
9 | later version. | |
10 | ||
11 | This program is distributed in the hope that it will be useful, | |
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | GNU General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU General Public License | |
17 | along with this program; if not, write to the Free Software | |
18 | Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
19 | ||
20 | In other words, you are welcome to use, share and improve this program. | |
21 | You are forbidden to forbid anyone else to use, share and improve | |
22 | what you give them. Help stamp out software-hoarding! */ | |
23 | \f | |
24 | typedef unsigned char U_CHAR; | |
25 | ||
26 | #ifdef EMACS | |
27 | #define NO_SHORTNAMES | |
28 | #include "../src/config.h" | |
29 | #ifdef open | |
30 | #undef open | |
31 | #undef read | |
32 | #undef write | |
33 | #endif /* open */ | |
34 | #endif /* EMACS */ | |
35 | ||
36 | /* The macro EMACS is defined when cpp is distributed as part of Emacs, | |
37 | for the sake of machines with limited C compilers. */ | |
38 | #ifndef EMACS | |
39 | #include "config.h" | |
40 | #endif /* not EMACS */ | |
41 | ||
42 | #ifndef STANDARD_INCLUDE_DIR | |
43 | #define STANDARD_INCLUDE_DIR "/usr/include" | |
44 | #endif | |
45 | ||
46 | #ifndef LOCAL_INCLUDE_DIR | |
47 | #define LOCAL_INCLUDE_DIR "/usr/local/include" | |
48 | #endif | |
49 | ||
50 | #include "pcp.h" | |
51 | ||
52 | #ifndef STDC_VALUE | |
53 | #define STDC_VALUE 1 | |
54 | #endif | |
55 | ||
56 | /* In case config.h defines these. */ | |
57 | #undef bcopy | |
58 | #undef bzero | |
59 | #undef bcmp | |
60 | ||
61 | #include <sys/types.h> | |
62 | #include <sys/stat.h> | |
63 | #include <ctype.h> | |
64 | #include <stdio.h> | |
65 | #include <signal.h> | |
66 | ||
67 | #ifndef VMS | |
68 | #include <sys/file.h> | |
69 | #ifndef USG | |
70 | #include <sys/time.h> /* for __DATE__ and __TIME__ */ | |
71 | #include <sys/resource.h> | |
72 | #else | |
73 | #define index strchr | |
74 | #define rindex strrchr | |
75 | #include <time.h> | |
76 | #include <fcntl.h> | |
77 | #endif /* USG */ | |
78 | #endif /* not VMS */ | |
79 | ||
80 | extern char *index (); | |
81 | extern char *rindex (); | |
82 | ||
83 | /* VMS-specific definitions */ | |
84 | #ifdef VMS | |
85 | #include <time.h> | |
86 | #include <errno.h> /* This defines "errno" properly */ | |
87 | #include <perror.h> /* This defines sys_errlist/sys_nerr properly */ | |
88 | #include <descrip.h> | |
89 | #define O_RDONLY 0 /* Open arg for Read/Only */ | |
90 | #define O_WRONLY 1 /* Open arg for Write/Only */ | |
91 | #define read(fd,buf,size) VMS_read(fd,buf,size) | |
92 | #define write(fd,buf,size) VMS_write(fd,buf,size) | |
93 | #define open(fname,mode,prot) VMS_open(fname,mode,prot) | |
94 | #define fopen(fname,mode) VMS_fopen(fname,mode) | |
95 | #define freopen(fname,mode,ofile) VMS_freopen(fname,mode,ofile) | |
96 | static int VMS_read (); | |
97 | static int VMS_write (); | |
98 | static int VMS_open (); | |
99 | static FILE * VMS_fopen (); | |
100 | static FILE * VMS_freopen (); | |
101 | static void hack_vms_include_specification (); | |
102 | typedef struct { unsigned :16, :16, :16; } vms_ino_t; | |
103 | #define ino_t vms_ino_t | |
104 | #ifdef __GNUC__ | |
105 | #define BSTRING /* VMS/GCC supplies the bstring routines */ | |
106 | #endif /* __GNUC__ */ | |
107 | #endif /* VMS */ | |
108 | ||
109 | #ifndef O_RDONLY | |
110 | #define O_RDONLY 0 | |
111 | #endif | |
112 | ||
113 | #undef MIN | |
114 | #undef MAX | |
115 | #define MIN(X,Y) ((X) < (Y) ? (X) : (Y)) | |
116 | #define MAX(X,Y) ((X) > (Y) ? (X) : (Y)) | |
117 | ||
118 | #ifndef S_ISREG | |
119 | #define S_ISREG(m) (((m) & S_IFMT) == S_IFREG) | |
120 | #endif | |
121 | ||
122 | /* Exported declarations. */ | |
123 | ||
124 | char *xmalloc (); | |
125 | void error (); | |
126 | void warning (); | |
127 | ||
128 | /* External declarations. */ | |
129 | ||
130 | extern char *getenv (); | |
131 | extern FILE *fdopen (); | |
132 | extern char *version_string; | |
133 | extern struct tm *localtime (); | |
134 | extern int sys_nerr; | |
135 | extern char *sys_errlist[]; | |
136 | ||
137 | #ifndef errno | |
138 | extern int errno; | |
139 | #endif | |
140 | ||
141 | /* Forward declarations. */ | |
142 | ||
143 | struct directive; | |
144 | struct file_buf; | |
145 | struct arglist; | |
146 | struct argdata; | |
147 | ||
148 | #if defined(USG) || defined(VMS) | |
149 | #ifndef BSTRING | |
150 | void bcopy (); | |
151 | void bzero (); | |
152 | int bcmp (); | |
153 | #endif | |
154 | #endif | |
155 | ||
156 | /* These functions are declared to return int instead of void since they | |
157 | are going to be placed in a table and some old compilers have trouble with | |
158 | pointers to functions returning void. */ | |
159 | ||
160 | static int do_define (); | |
161 | static int do_line (); | |
162 | static int do_include (); | |
163 | static int do_undef (); | |
164 | static int do_error (); | |
165 | static int do_pragma (); | |
166 | static int do_ident (); | |
167 | static int do_if (); | |
168 | static int do_xifdef (); | |
169 | static int do_else (); | |
170 | static int do_elif (); | |
171 | static int do_endif (); | |
172 | static int do_sccs (); | |
173 | static int do_once (); | |
174 | static int do_assert (); | |
175 | static int do_unassert (); | |
176 | static int do_warning (); | |
177 | ||
178 | static void add_import (); | |
179 | static void deps_output (); | |
180 | static void make_undef (); | |
181 | static void make_definition (); | |
182 | static void make_assertion (); | |
183 | static void path_include (); | |
184 | static void initialize_builtins (); | |
185 | static void initialize_char_syntax (); | |
186 | static void dump_arg_n (); | |
187 | static void dump_defn_1 (); | |
188 | static void delete_macro (); | |
189 | static void trigraph_pcp (); | |
190 | static void rescan (); | |
191 | static void finclude (); | |
192 | static void validate_else (); | |
193 | static int comp_def_part (); | |
194 | static void error_from_errno (); | |
195 | static void error_with_line (); | |
196 | void pedwarn (); | |
197 | static void pedwarn_with_file_and_line (); | |
198 | static void fatal (); | |
199 | void fancy_abort (); | |
200 | static void pfatal_with_name (); | |
201 | static void perror_with_name (); | |
202 | static void print_containing_files (); | |
203 | static int lookup_import (); | |
204 | static int check_preconditions (); | |
205 | static void pcfinclude (); | |
206 | static void pcstring_used (); | |
207 | static void write_output (); | |
208 | static int check_macro_name (); | |
209 | static int compare_defs (); | |
210 | static int compare_token_lists (); | |
211 | static int eval_if_expression (); | |
212 | static int discard_comments (); | |
213 | static int delete_newlines (); | |
214 | static int line_for_error (); | |
215 | static int hashf (); | |
216 | static int file_size_and_mode (); | |
217 | ||
218 | static struct arglist *read_token_list (); | |
219 | static void free_token_list (); | |
220 | ||
221 | static struct hashnode *install (); | |
222 | struct hashnode *lookup (); | |
223 | ||
224 | static struct assertion_hashnode *assertion_install (); | |
225 | static struct assertion_hashnode *assertion_lookup (); | |
226 | ||
227 | static char *xrealloc (); | |
228 | static char *xcalloc (); | |
229 | static char *savestring (); | |
230 | ||
231 | static void delete_assertion (); | |
232 | static void macroexpand (); | |
233 | static void dump_all_macros (); | |
234 | static void conditional_skip (); | |
235 | static void skip_if_group (); | |
236 | static void output_line_command (); | |
237 | ||
238 | /* Last arg to output_line_command. */ | |
239 | enum file_change_code {same_file, enter_file, leave_file}; | |
240 | ||
241 | static int grow_outbuf (); | |
242 | static int handle_directive (); | |
243 | static void memory_full (); | |
244 | ||
245 | static U_CHAR *macarg1 (); | |
246 | static char *macarg (); | |
247 | ||
248 | static U_CHAR *skip_to_end_of_comment (); | |
249 | static U_CHAR *skip_quoted_string (); | |
250 | static U_CHAR *skip_paren_group (); | |
251 | ||
252 | static char *check_precompiled (); | |
253 | static struct macrodef create_definition (); | |
254 | static void dump_single_macro (); | |
255 | \f | |
256 | #ifndef FAILURE_EXIT_CODE | |
257 | #define FAILURE_EXIT_CODE 33 /* gnu cc command understands this */ | |
258 | #endif | |
259 | ||
260 | #ifndef SUCCESS_EXIT_CODE | |
261 | #define SUCCESS_EXIT_CODE 0 /* 0 means success on Unix. */ | |
262 | #endif | |
263 | ||
264 | /* Name under which this program was invoked. */ | |
265 | ||
266 | static char *progname; | |
267 | ||
268 | /* Nonzero means handle C++ comment syntax and use | |
269 | extra default include directories for C++. */ | |
270 | ||
271 | static int cplusplus; | |
272 | ||
273 | /* Nonzero means handle #import, for objective C. */ | |
274 | ||
275 | static int objc; | |
276 | ||
277 | /* Nonzero means this is an assembly file, and allow | |
278 | unknown directives, which could be comments. */ | |
279 | ||
280 | static int lang_asm; | |
281 | ||
282 | /* Current maximum length of directory names in the search path | |
283 | for include files. (Altered as we get more of them.) */ | |
284 | ||
285 | static int max_include_len; | |
286 | ||
287 | /* Nonzero means turn NOTREACHED into #pragma NOTREACHED etc */ | |
288 | ||
289 | static int lint = 0; | |
290 | ||
291 | /* Nonzero means copy comments into the output file. */ | |
292 | ||
293 | static int put_out_comments = 0; | |
294 | ||
295 | /* Nonzero means don't process the ANSI trigraph sequences. */ | |
296 | ||
297 | static int no_trigraphs = 0; | |
298 | ||
299 | /* Nonzero means print the names of included files rather than | |
300 | the preprocessed output. 1 means just the #include "...", | |
301 | 2 means #include <...> as well. */ | |
302 | ||
303 | static int print_deps = 0; | |
304 | ||
305 | /* Nonzero means print names of header files (-H). */ | |
306 | ||
307 | static int print_include_names = 0; | |
308 | ||
309 | /* Nonzero means don't output line number information. */ | |
310 | ||
311 | static int no_line_commands; | |
312 | ||
313 | /* dump_only means inhibit output of the preprocessed text | |
314 | and instead output the definitions of all user-defined | |
315 | macros in a form suitable for use as input to cccp. | |
316 | dump_names means pass #define and the macro name through to output. | |
317 | dump_definitions means pass the whole definition (plus #define) through | |
318 | */ | |
319 | ||
320 | static enum {dump_none, dump_only, dump_names, dump_definitions} | |
321 | dump_macros = dump_none; | |
322 | ||
323 | /* Nonzero means pass all #define and #undef directives which we actually | |
324 | process through to the output stream. This feature is used primarily | |
325 | to allow cc1 to record the #defines and #undefs for the sake of | |
326 | debuggers which understand about preprocessor macros, but it may | |
327 | also be useful with -E to figure out how symbols are defined, and | |
328 | where they are defined. */ | |
329 | static int debug_output = 0; | |
330 | ||
331 | /* Holds local startup time. */ | |
332 | static struct tm *timebuf = NULL; | |
333 | ||
334 | /* Nonzero indicates special processing used by the pcp program. The | |
335 | special effects of this mode are: | |
336 | ||
337 | Inhibit all macro expansion, except those inside #if directives. | |
338 | ||
339 | Process #define directives normally, and output their contents | |
340 | to the output file. | |
341 | ||
342 | Output preconditions to pcp_outfile indicating all the relevant | |
343 | preconditions for use of this file in a later cpp run. | |
344 | */ | |
345 | static FILE *pcp_outfile; | |
346 | ||
347 | /* Nonzero means we are inside an IF during a -pcp run. In this mode | |
348 | macro expansion is done, and preconditions are output for all macro | |
349 | uses requiring them. */ | |
350 | static int pcp_inside_if; | |
351 | ||
352 | /* Nonzero means never to include precompiled files. */ | |
353 | static int no_precomp; | |
354 | ||
355 | /* Nonzero means give all the error messages the ANSI standard requires. */ | |
356 | ||
357 | int pedantic; | |
358 | ||
359 | /* Nonzero means try to make failure to fit ANSI C an error. */ | |
360 | ||
361 | static int pedantic_errors; | |
362 | ||
363 | /* Nonzero means don't print warning messages. -w. */ | |
364 | ||
365 | static int inhibit_warnings = 0; | |
366 | ||
367 | /* Nonzero means warn if slash-star appears in a comment. */ | |
368 | ||
369 | static int warn_comments; | |
370 | ||
371 | /* Nonzero means warn if a macro argument is (or would be) | |
372 | stringified with -traditional. */ | |
373 | ||
374 | static int warn_stringify; | |
375 | ||
376 | /* Nonzero means warn if there are any trigraphs. */ | |
377 | ||
378 | static int warn_trigraphs; | |
379 | ||
380 | /* Nonzero means turn warnings into errors. */ | |
381 | ||
382 | static int warnings_are_errors; | |
383 | ||
384 | /* Nonzero means try to imitate old fashioned non-ANSI preprocessor. */ | |
385 | ||
386 | int traditional; | |
387 | ||
388 | /* Nonzero causes output not to be done, | |
389 | but directives such as #define that have side effects | |
390 | are still obeyed. */ | |
391 | ||
392 | static int no_output; | |
393 | ||
394 | /* Nonzero means that we have finished processing the command line options. | |
395 | This flag is used to decide whether or not to issue certain errors | |
396 | and/or warnings. */ | |
397 | ||
398 | static int done_initializing = 0; | |
399 | \f | |
400 | /* I/O buffer structure. | |
401 | The `fname' field is nonzero for source files and #include files | |
402 | and for the dummy text used for -D and -U. | |
403 | It is zero for rescanning results of macro expansion | |
404 | and for expanding macro arguments. */ | |
405 | #define INPUT_STACK_MAX 200 | |
406 | static struct file_buf { | |
407 | char *fname; | |
408 | /* Filename specified with #line command. */ | |
409 | char *nominal_fname; | |
410 | /* Record where in the search path this file was found. | |
411 | For #include_next. */ | |
412 | struct file_name_list *dir; | |
413 | int lineno; | |
414 | int length; | |
415 | U_CHAR *buf; | |
416 | U_CHAR *bufp; | |
417 | /* Macro that this level is the expansion of. | |
418 | Included so that we can reenable the macro | |
419 | at the end of this level. */ | |
420 | struct hashnode *macro; | |
421 | /* Value of if_stack at start of this file. | |
422 | Used to prohibit unmatched #endif (etc) in an include file. */ | |
423 | struct if_stack *if_stack; | |
424 | /* Object to be freed at end of input at this level. */ | |
425 | U_CHAR *free_ptr; | |
426 | /* True if this is a header file included using <FILENAME>. */ | |
427 | char system_header_p; | |
428 | } instack[INPUT_STACK_MAX]; | |
429 | ||
430 | static int last_error_tick; /* Incremented each time we print it. */ | |
431 | static int input_file_stack_tick; /* Incremented when the status changes. */ | |
432 | ||
433 | /* Current nesting level of input sources. | |
434 | `instack[indepth]' is the level currently being read. */ | |
435 | static int indepth = -1; | |
436 | #define CHECK_DEPTH(code) \ | |
437 | if (indepth >= (INPUT_STACK_MAX - 1)) \ | |
438 | { \ | |
439 | error_with_line (line_for_error (instack[indepth].lineno), \ | |
440 | "macro or `#include' recursion too deep"); \ | |
441 | code; \ | |
442 | } | |
443 | ||
444 | /* Current depth in #include directives that use <...>. */ | |
445 | static int system_include_depth = 0; | |
446 | ||
447 | typedef struct file_buf FILE_BUF; | |
448 | ||
449 | /* The output buffer. Its LENGTH field is the amount of room allocated | |
450 | for the buffer, not the number of chars actually present. To get | |
451 | that, subtract outbuf.buf from outbuf.bufp. */ | |
452 | ||
453 | #define OUTBUF_SIZE 10 /* initial size of output buffer */ | |
454 | static FILE_BUF outbuf; | |
455 | ||
456 | /* Grow output buffer OBUF points at | |
457 | so it can hold at least NEEDED more chars. */ | |
458 | ||
459 | #define check_expand(OBUF, NEEDED) \ | |
460 | (((OBUF)->length - ((OBUF)->bufp - (OBUF)->buf) <= (NEEDED)) \ | |
461 | ? grow_outbuf ((OBUF), (NEEDED)) : 0) | |
462 | ||
463 | struct file_name_list | |
464 | { | |
465 | struct file_name_list *next; | |
466 | char *fname; | |
467 | /* If the following is nonzero, it is a macro name. | |
468 | Don't include the file again if that macro is defined. */ | |
469 | U_CHAR *control_macro; | |
470 | }; | |
471 | ||
472 | /* #include "file" looks in source file dir, then stack. */ | |
473 | /* #include <file> just looks in the stack. */ | |
474 | /* -I directories are added to the end, then the defaults are added. */ | |
475 | static struct default_include { char *fname; int cplusplus; } include_defaults_array[] | |
476 | #ifdef INCLUDE_DEFAULTS | |
477 | = INCLUDE_DEFAULTS; | |
478 | #else | |
479 | = { | |
480 | /* Pick up GNU C++ specific include files. */ | |
481 | { GPLUSPLUS_INCLUDE_DIR, 1}, | |
482 | { GCC_INCLUDE_DIR, 0}, | |
483 | #ifdef CROSS_COMPILE | |
484 | /* For cross-compilation, this dir name is generated | |
485 | automatically in Makefile.in. */ | |
486 | { CROSS_INCLUDE_DIR, 0 }, | |
487 | #else /* not CROSS_COMPILE */ | |
488 | { LOCAL_INCLUDE_DIR, 0}, | |
489 | /* Some systems have an extra dir of include files. */ | |
490 | #ifdef SYSTEM_INCLUDE_DIR | |
491 | { SYSTEM_INCLUDE_DIR, 0}, | |
492 | #endif | |
493 | { STANDARD_INCLUDE_DIR, 0}, | |
494 | #endif /* not CROSS_COMPILE */ | |
495 | { 0, 0} | |
496 | }; | |
497 | #endif /* no INCLUDE_DEFAULTS */ | |
498 | ||
499 | /* The code looks at the defaults through this pointer, rather than through | |
500 | the constant structure above. This pointer gets changed if an environment | |
501 | variable specifies other defaults. */ | |
502 | static struct default_include *include_defaults = include_defaults_array; | |
503 | ||
504 | static struct file_name_list *include = 0; /* First dir to search */ | |
505 | /* First dir to search for <file> */ | |
506 | static struct file_name_list *first_bracket_include = 0; | |
507 | static struct file_name_list *last_include = 0; /* Last in chain */ | |
508 | ||
509 | /* Chain of include directories to put at the end of the other chain. */ | |
510 | static struct file_name_list *after_include = 0; | |
511 | static struct file_name_list *last_after_include = 0; /* Last in chain */ | |
512 | ||
513 | /* List of included files that contained #pragma once. */ | |
514 | static struct file_name_list *dont_repeat_files = 0; | |
515 | ||
516 | /* List of other included files. | |
517 | If ->control_macro if nonzero, the file had a #ifndef | |
518 | around the entire contents, and ->control_macro gives the macro name. */ | |
519 | static struct file_name_list *all_include_files = 0; | |
520 | ||
521 | /* Global list of strings read in from precompiled files. This list | |
522 | is kept in the order the strings are read in, with new strings being | |
523 | added at the end through stringlist_tailp. We use this list to output | |
524 | the strings at the end of the run. | |
525 | */ | |
526 | static STRINGDEF *stringlist; | |
527 | static STRINGDEF **stringlist_tailp = &stringlist; | |
528 | ||
529 | ||
530 | /* Structure returned by create_definition */ | |
531 | typedef struct macrodef MACRODEF; | |
532 | struct macrodef | |
533 | { | |
534 | struct definition *defn; | |
535 | U_CHAR *symnam; | |
536 | int symlen; | |
537 | }; | |
538 | ||
539 | \f | |
540 | /* Structure allocated for every #define. For a simple replacement | |
541 | such as | |
542 | #define foo bar , | |
543 | nargs = -1, the `pattern' list is null, and the expansion is just | |
544 | the replacement text. Nargs = 0 means a functionlike macro with no args, | |
545 | e.g., | |
546 | #define getchar() getc (stdin) . | |
547 | When there are args, the expansion is the replacement text with the | |
548 | args squashed out, and the reflist is a list describing how to | |
549 | build the output from the input: e.g., "3 chars, then the 1st arg, | |
550 | then 9 chars, then the 3rd arg, then 0 chars, then the 2nd arg". | |
551 | The chars here come from the expansion. Whatever is left of the | |
552 | expansion after the last arg-occurrence is copied after that arg. | |
553 | Note that the reflist can be arbitrarily long--- | |
554 | its length depends on the number of times the arguments appear in | |
555 | the replacement text, not how many args there are. Example: | |
556 | #define f(x) x+x+x+x+x+x+x would have replacement text "++++++" and | |
557 | pattern list | |
558 | { (0, 1), (1, 1), (1, 1), ..., (1, 1), NULL } | |
559 | where (x, y) means (nchars, argno). */ | |
560 | ||
561 | typedef struct definition DEFINITION; | |
562 | struct definition { | |
563 | int nargs; | |
564 | int length; /* length of expansion string */ | |
565 | int predefined; /* True if the macro was builtin or */ | |
566 | /* came from the command line */ | |
567 | U_CHAR *expansion; | |
568 | int line; /* Line number of definition */ | |
569 | char *file; /* File of definition */ | |
570 | struct reflist { | |
571 | struct reflist *next; | |
572 | char stringify; /* nonzero if this arg was preceded by a | |
573 | # operator. */ | |
574 | char raw_before; /* Nonzero if a ## operator before arg. */ | |
575 | char raw_after; /* Nonzero if a ## operator after arg. */ | |
576 | int nchars; /* Number of literal chars to copy before | |
577 | this arg occurrence. */ | |
578 | int argno; /* Number of arg to substitute (origin-0) */ | |
579 | } *pattern; | |
580 | union { | |
581 | /* Names of macro args, concatenated in reverse order | |
582 | with comma-space between them. | |
583 | The only use of this is that we warn on redefinition | |
584 | if this differs between the old and new definitions. */ | |
585 | U_CHAR *argnames; | |
586 | } args; | |
587 | }; | |
588 | ||
589 | /* different kinds of things that can appear in the value field | |
590 | of a hash node. Actually, this may be useless now. */ | |
591 | union hashval { | |
592 | int ival; | |
593 | char *cpval; | |
594 | DEFINITION *defn; | |
595 | KEYDEF *keydef; | |
596 | }; | |
597 | ||
598 | ||
599 | /* The structure of a node in the hash table. The hash table | |
600 | has entries for all tokens defined by #define commands (type T_MACRO), | |
601 | plus some special tokens like __LINE__ (these each have their own | |
602 | type, and the appropriate code is run when that type of node is seen. | |
603 | It does not contain control words like "#define", which are recognized | |
604 | by a separate piece of code. */ | |
605 | ||
606 | /* different flavors of hash nodes --- also used in keyword table */ | |
607 | enum node_type { | |
608 | T_DEFINE = 1, /* the `#define' keyword */ | |
609 | T_INCLUDE, /* the `#include' keyword */ | |
610 | T_INCLUDE_NEXT, /* the `#include_next' keyword */ | |
611 | T_IMPORT, /* the `#import' keyword */ | |
612 | T_IFDEF, /* the `#ifdef' keyword */ | |
613 | T_IFNDEF, /* the `#ifndef' keyword */ | |
614 | T_IF, /* the `#if' keyword */ | |
615 | T_ELSE, /* `#else' */ | |
616 | T_PRAGMA, /* `#pragma' */ | |
617 | T_ELIF, /* `#elif' */ | |
618 | T_UNDEF, /* `#undef' */ | |
619 | T_LINE, /* `#line' */ | |
620 | T_ERROR, /* `#error' */ | |
621 | T_WARNING, /* `#warning' */ | |
622 | T_ENDIF, /* `#endif' */ | |
623 | T_SCCS, /* `#sccs', used on system V. */ | |
624 | T_IDENT, /* `#ident', used on system V. */ | |
625 | T_ASSERT, /* `#assert', taken from system V. */ | |
626 | T_UNASSERT, /* `#unassert', taken from system V. */ | |
627 | T_SPECLINE, /* special symbol `__LINE__' */ | |
628 | T_DATE, /* `__DATE__' */ | |
629 | T_FILE, /* `__FILE__' */ | |
630 | T_BASE_FILE, /* `__BASE_FILE__' */ | |
631 | T_INCLUDE_LEVEL, /* `__INCLUDE_LEVEL__' */ | |
632 | T_VERSION, /* `__VERSION__' */ | |
633 | T_SIZE_TYPE, /* `__SIZE_TYPE__' */ | |
634 | T_PTRDIFF_TYPE, /* `__PTRDIFF_TYPE__' */ | |
635 | T_WCHAR_TYPE, /* `__WCHAR_TYPE__' */ | |
636 | T_TIME, /* `__TIME__' */ | |
637 | T_CONST, /* Constant value, used by `__STDC__' */ | |
638 | T_MACRO, /* macro defined by `#define' */ | |
639 | T_DISABLED, /* macro temporarily turned off for rescan */ | |
640 | T_SPEC_DEFINED, /* special `defined' macro for use in #if statements */ | |
641 | T_PCSTRING, /* precompiled string (hashval is KEYDEF *) */ | |
642 | T_UNUSED /* Used for something not defined. */ | |
643 | }; | |
644 | ||
645 | struct hashnode { | |
646 | struct hashnode *next; /* double links for easy deletion */ | |
647 | struct hashnode *prev; | |
648 | struct hashnode **bucket_hdr; /* also, a back pointer to this node's hash | |
649 | chain is kept, in case the node is the head | |
650 | of the chain and gets deleted. */ | |
651 | enum node_type type; /* type of special token */ | |
652 | int length; /* length of token, for quick comparison */ | |
653 | U_CHAR *name; /* the actual name */ | |
654 | union hashval value; /* pointer to expansion, or whatever */ | |
655 | }; | |
656 | ||
657 | typedef struct hashnode HASHNODE; | |
658 | ||
659 | /* Some definitions for the hash table. The hash function MUST be | |
660 | computed as shown in hashf () below. That is because the rescan | |
661 | loop computes the hash value `on the fly' for most tokens, | |
662 | in order to avoid the overhead of a lot of procedure calls to | |
663 | the hashf () function. Hashf () only exists for the sake of | |
664 | politeness, for use when speed isn't so important. */ | |
665 | ||
666 | #define HASHSIZE 1403 | |
667 | static HASHNODE *hashtab[HASHSIZE]; | |
668 | #define HASHSTEP(old, c) ((old << 2) + c) | |
669 | #define MAKE_POS(v) (v & 0x7fffffff) /* make number positive */ | |
670 | ||
671 | /* Symbols to predefine. */ | |
672 | ||
673 | #ifdef CPP_PREDEFINES | |
674 | static char *predefs = CPP_PREDEFINES; | |
675 | #else | |
676 | static char *predefs = ""; | |
677 | #endif | |
678 | \f | |
679 | /* We let tm.h override the types used here, to handle trivial differences | |
680 | such as the choice of unsigned int or long unsigned int for size_t. | |
681 | When machines start needing nontrivial differences in the size type, | |
682 | it would be best to do something here to figure out automatically | |
683 | from other information what type to use. */ | |
684 | ||
685 | /* The string value for __size_type__. */ | |
686 | ||
687 | #ifndef SIZE_TYPE | |
688 | #define SIZE_TYPE "long unsigned int" | |
689 | #endif | |
690 | ||
691 | /* The string value for __ptrdiff_type__. */ | |
692 | ||
693 | #ifndef PTRDIFF_TYPE | |
694 | #define PTRDIFF_TYPE "long int" | |
695 | #endif | |
696 | ||
697 | /* The string value for __wchar_type__. */ | |
698 | ||
699 | #ifndef WCHAR_TYPE | |
700 | #define WCHAR_TYPE "int" | |
701 | #endif | |
702 | \f | |
703 | /* In the definition of a #assert name, this structure forms | |
704 | a list of the individual values asserted. | |
705 | Each value is itself a list of "tokens". | |
706 | These are strings that are compared by name. */ | |
707 | ||
708 | struct tokenlist_list { | |
709 | struct tokenlist_list *next; | |
710 | struct arglist *tokens; | |
711 | }; | |
712 | ||
713 | struct assertion_hashnode { | |
714 | struct assertion_hashnode *next; /* double links for easy deletion */ | |
715 | struct assertion_hashnode *prev; | |
716 | /* also, a back pointer to this node's hash | |
717 | chain is kept, in case the node is the head | |
718 | of the chain and gets deleted. */ | |
719 | struct assertion_hashnode **bucket_hdr; | |
720 | int length; /* length of token, for quick comparison */ | |
721 | U_CHAR *name; /* the actual name */ | |
722 | /* List of token-sequences. */ | |
723 | struct tokenlist_list *value; | |
724 | }; | |
725 | ||
726 | typedef struct assertion_hashnode ASSERTION_HASHNODE; | |
727 | ||
728 | /* Some definitions for the hash table. The hash function MUST be | |
729 | computed as shown in hashf () below. That is because the rescan | |
730 | loop computes the hash value `on the fly' for most tokens, | |
731 | in order to avoid the overhead of a lot of procedure calls to | |
732 | the hashf () function. Hashf () only exists for the sake of | |
733 | politeness, for use when speed isn't so important. */ | |
734 | ||
735 | #define ASSERTION_HASHSIZE 37 | |
736 | static ASSERTION_HASHNODE *assertion_hashtab[ASSERTION_HASHSIZE]; | |
737 | ||
738 | /* Nonzero means inhibit macroexpansion of what seem to be | |
739 | assertion tests, in rescan. For #if. */ | |
740 | static int assertions_flag; | |
741 | \f | |
742 | /* `struct directive' defines one #-directive, including how to handle it. */ | |
743 | ||
744 | struct directive { | |
745 | int length; /* Length of name */ | |
746 | int (*func)(); /* Function to handle directive */ | |
747 | char *name; /* Name of directive */ | |
748 | enum node_type type; /* Code which describes which directive. */ | |
749 | char angle_brackets; /* Nonzero => <...> is special. */ | |
750 | char traditional_comments; /* Nonzero: keep comments if -traditional. */ | |
751 | char pass_thru; /* Copy preprocessed directive to output file. */ | |
752 | }; | |
753 | ||
754 | /* Here is the actual list of #-directives, most-often-used first. */ | |
755 | ||
756 | static struct directive directive_table[] = { | |
757 | { 6, do_define, "define", T_DEFINE, 0, 1}, | |
758 | { 2, do_if, "if", T_IF}, | |
759 | { 5, do_xifdef, "ifdef", T_IFDEF}, | |
760 | { 6, do_xifdef, "ifndef", T_IFNDEF}, | |
761 | { 5, do_endif, "endif", T_ENDIF}, | |
762 | { 4, do_else, "else", T_ELSE}, | |
763 | { 4, do_elif, "elif", T_ELIF}, | |
764 | { 4, do_line, "line", T_LINE}, | |
765 | { 7, do_include, "include", T_INCLUDE, 1}, | |
766 | { 12, do_include, "include_next", T_INCLUDE_NEXT, 1}, | |
767 | { 6, do_include, "import", T_IMPORT, 1}, | |
768 | { 5, do_undef, "undef", T_UNDEF}, | |
769 | { 5, do_error, "error", T_ERROR}, | |
770 | { 7, do_warning, "warning", T_WARNING}, | |
771 | #ifdef SCCS_DIRECTIVE | |
772 | { 4, do_sccs, "sccs", T_SCCS}, | |
773 | #endif | |
774 | { 6, do_pragma, "pragma", T_PRAGMA, 0, 0, 1}, | |
775 | { 5, do_ident, "ident", T_IDENT, 0, 0, 1}, | |
776 | { 6, do_assert, "assert", T_ASSERT}, | |
777 | { 8, do_unassert, "unassert", T_UNASSERT}, | |
778 | { -1, 0, "", T_UNUSED}, | |
779 | }; | |
780 | ||
781 | /* When a directive handler is called, | |
782 | this points to the # that started the directive. */ | |
783 | U_CHAR *directive_start; | |
784 | ||
785 | /* table to tell if char can be part of a C identifier. */ | |
786 | U_CHAR is_idchar[256]; | |
787 | /* table to tell if char can be first char of a c identifier. */ | |
788 | U_CHAR is_idstart[256]; | |
789 | /* table to tell if c is horizontal space. */ | |
790 | U_CHAR is_hor_space[256]; | |
791 | /* table to tell if c is horizontal or vertical space. */ | |
792 | static U_CHAR is_space[256]; | |
793 | ||
794 | #define SKIP_WHITE_SPACE(p) do { while (is_hor_space[*p]) p++; } while (0) | |
795 | #define SKIP_ALL_WHITE_SPACE(p) do { while (is_space[*p]) p++; } while (0) | |
796 | ||
797 | static int errors = 0; /* Error counter for exit code */ | |
798 | ||
799 | /* Zero means dollar signs are punctuation. | |
800 | -$ stores 0; -traditional may store 1. Default is 1 for VMS, 0 otherwise. | |
801 | This must be 0 for correct processing of this ANSI C program: | |
802 | #define foo(a) #a | |
803 | #define lose(b) foo(b) | |
804 | #define test$ | |
805 | lose(test) */ | |
806 | static int dollars_in_ident; | |
807 | #ifndef DOLLARS_IN_IDENTIFIERS | |
808 | #define DOLLARS_IN_IDENTIFIERS 1 | |
809 | #endif | |
810 | ||
811 | static FILE_BUF expand_to_temp_buffer (); | |
812 | ||
813 | static DEFINITION *collect_expansion (); | |
814 | ||
815 | /* Stack of conditionals currently in progress | |
816 | (including both successful and failing conditionals). */ | |
817 | ||
818 | struct if_stack { | |
819 | struct if_stack *next; /* for chaining to the next stack frame */ | |
820 | char *fname; /* copied from input when frame is made */ | |
821 | int lineno; /* similarly */ | |
822 | int if_succeeded; /* true if a leg of this if-group | |
823 | has been passed through rescan */ | |
824 | U_CHAR *control_macro; /* For #ifndef at start of file, | |
825 | this is the macro name tested. */ | |
826 | enum node_type type; /* type of last directive seen in this group */ | |
827 | }; | |
828 | typedef struct if_stack IF_STACK_FRAME; | |
829 | static IF_STACK_FRAME *if_stack = NULL; | |
830 | ||
831 | /* Buffer of -M output. */ | |
832 | static char *deps_buffer; | |
833 | ||
834 | /* Number of bytes allocated in above. */ | |
835 | static int deps_allocated_size; | |
836 | ||
837 | /* Number of bytes used. */ | |
838 | static int deps_size; | |
839 | ||
840 | /* Number of bytes since the last newline. */ | |
841 | static int deps_column; | |
842 | ||
843 | /* File name which deps are being written to. | |
844 | This is 0 if deps are being written to stdout. */ | |
845 | static char *deps_file = 0; | |
846 | ||
847 | /* Nonzero means -I- has been seen, | |
848 | so don't look for #include "foo" the source-file directory. */ | |
849 | static int ignore_srcdir; | |
850 | \f | |
851 | /* Handler for SIGPIPE. */ | |
852 | ||
853 | static void | |
854 | pipe_closed (signo) | |
855 | /* If this is missing, some compilers complain. */ | |
856 | int signo; | |
857 | { | |
858 | fatal ("output pipe has been closed"); | |
859 | } | |
860 | ||
861 | int | |
862 | main (argc, argv) | |
863 | int argc; | |
864 | char **argv; | |
865 | { | |
866 | int st_mode; | |
867 | long st_size; | |
868 | char *in_fname, *out_fname; | |
869 | char *p; | |
870 | int f, i; | |
871 | FILE_BUF *fp; | |
872 | char **pend_files = (char **) xmalloc (argc * sizeof (char *)); | |
873 | char **pend_defs = (char **) xmalloc (argc * sizeof (char *)); | |
874 | char **pend_undefs = (char **) xmalloc (argc * sizeof (char *)); | |
875 | char **pend_assertions = (char **) xmalloc (argc * sizeof (char *)); | |
876 | char **pend_includes = (char **) xmalloc (argc * sizeof (char *)); | |
877 | ||
878 | /* Record the option used with each element of pend_assertions. | |
879 | This is preparation for supporting more than one option for making | |
880 | an assertion. */ | |
881 | char **pend_assertion_options = (char **) xmalloc (argc * sizeof (char *)); | |
882 | int inhibit_predefs = 0; | |
883 | int no_standard_includes = 0; | |
884 | int missing_newline = 0; | |
885 | ||
886 | /* Non-0 means don't output the preprocessed program. */ | |
887 | int inhibit_output = 0; | |
888 | ||
889 | /* Stream on which to print the dependency information. */ | |
890 | FILE *deps_stream = 0; | |
891 | /* Target-name to write with the dependency information. */ | |
892 | char *deps_target = 0; | |
893 | ||
894 | #ifdef RLIMIT_STACK | |
895 | /* Get rid of any avoidable limit on stack size. */ | |
896 | { | |
897 | struct rlimit rlim; | |
898 | ||
899 | /* Set the stack limit huge so that alloca (particularly stringtab | |
900 | * in dbxread.c) does not fail. */ | |
901 | getrlimit (RLIMIT_STACK, &rlim); | |
902 | rlim.rlim_cur = rlim.rlim_max; | |
903 | setrlimit (RLIMIT_STACK, &rlim); | |
904 | } | |
905 | #endif /* RLIMIT_STACK defined */ | |
906 | ||
907 | progname = argv[0]; | |
908 | #ifdef VMS | |
909 | { | |
910 | /* Remove directories from PROGNAME. */ | |
911 | char *s; | |
912 | ||
913 | progname = savestring (argv[0]); | |
914 | ||
915 | if (!(s = rindex (progname, ']'))) | |
916 | s = rindex (progname, ':'); | |
917 | if (s) | |
918 | strcpy (progname, s+1); | |
919 | if (s = rindex (progname, '.')) | |
920 | *s = '\0'; | |
921 | } | |
922 | #endif | |
923 | ||
924 | in_fname = NULL; | |
925 | out_fname = NULL; | |
926 | ||
927 | /* Initialize is_idchar to allow $. */ | |
928 | dollars_in_ident = 1; | |
929 | initialize_char_syntax (); | |
930 | dollars_in_ident = DOLLARS_IN_IDENTIFIERS > 0; | |
931 | ||
932 | no_line_commands = 0; | |
933 | no_trigraphs = 1; | |
934 | dump_macros = dump_none; | |
935 | no_output = 0; | |
936 | cplusplus = 0; | |
937 | ||
938 | #ifdef SIGPIPE | |
939 | signal (SIGPIPE, pipe_closed); | |
940 | #endif | |
941 | ||
942 | for (i = 0; include_defaults[i].fname; i++) | |
943 | max_include_len = MAX (max_include_len, | |
944 | strlen (include_defaults[i].fname)); | |
945 | /* Leave room for making file name longer when converting to VMS syntax. */ | |
946 | #ifdef VMS | |
947 | max_include_len += 10; | |
948 | #endif | |
949 | ||
950 | bzero (pend_files, argc * sizeof (char *)); | |
951 | bzero (pend_defs, argc * sizeof (char *)); | |
952 | bzero (pend_undefs, argc * sizeof (char *)); | |
953 | bzero (pend_assertions, argc * sizeof (char *)); | |
954 | bzero (pend_includes, argc * sizeof (char *)); | |
955 | ||
956 | /* Process switches and find input file name. */ | |
957 | ||
958 | for (i = 1; i < argc; i++) { | |
959 | if (argv[i][0] != '-') { | |
960 | if (out_fname != NULL) | |
961 | fatal ("Usage: %s [switches] input output", argv[0]); | |
962 | else if (in_fname != NULL) | |
963 | out_fname = argv[i]; | |
964 | else | |
965 | in_fname = argv[i]; | |
966 | } else { | |
967 | switch (argv[i][1]) { | |
968 | ||
969 | case 'i': | |
970 | if (!strcmp (argv[i], "-include")) { | |
971 | if (i + 1 == argc) | |
972 | fatal ("Filename missing after -include option"); | |
973 | else | |
974 | pend_includes[i] = argv[i+1], i++; | |
975 | } | |
976 | if (!strcmp (argv[i], "-imacros")) { | |
977 | if (i + 1 == argc) | |
978 | fatal ("Filename missing after -imacros option"); | |
979 | else | |
980 | pend_files[i] = argv[i+1], i++; | |
981 | } | |
982 | /* Add directory to end of path for includes. */ | |
983 | if (!strcmp (argv[i], "-idirafter")) { | |
984 | struct file_name_list *dirtmp; | |
985 | ||
986 | dirtmp = (struct file_name_list *) | |
987 | xmalloc (sizeof (struct file_name_list)); | |
988 | dirtmp->next = 0; /* New one goes on the end */ | |
989 | dirtmp->control_macro = 0; | |
990 | if (after_include == 0) | |
991 | after_include = dirtmp; | |
992 | else | |
993 | last_after_include->next = dirtmp; | |
994 | last_after_include = dirtmp; /* Tail follows the last one */ | |
995 | ||
996 | if (i + 1 == argc) | |
997 | fatal ("Directory name missing after -idirafter option"); | |
998 | else | |
999 | dirtmp->fname = argv[++i]; | |
1000 | ||
1001 | if (strlen (dirtmp->fname) > max_include_len) | |
1002 | max_include_len = strlen (dirtmp->fname); | |
1003 | } | |
1004 | break; | |
1005 | ||
1006 | case 'o': | |
1007 | if (out_fname != NULL) | |
1008 | fatal ("Output filename specified twice"); | |
1009 | if (i + 1 == argc) | |
1010 | fatal ("Filename missing after -o option"); | |
1011 | out_fname = argv[++i]; | |
1012 | if (!strcmp (out_fname, "-")) | |
1013 | out_fname = ""; | |
1014 | break; | |
1015 | ||
1016 | case 'p': | |
1017 | if (!strcmp (argv[i], "-pedantic")) | |
1018 | pedantic = 1; | |
1019 | else if (!strcmp (argv[i], "-pedantic-errors")) { | |
1020 | pedantic = 1; | |
1021 | pedantic_errors = 1; | |
1022 | } else if (!strcmp (argv[i], "-pcp")) { | |
1023 | char *pcp_fname = argv[++i]; | |
1024 | pcp_outfile = | |
1025 | ((pcp_fname[0] != '-' || pcp_fname[1] != '\0') | |
1026 | ? fopen (pcp_fname, "w") | |
1027 | : fdopen (dup (fileno (stdout)), "w")); | |
1028 | if (pcp_outfile == 0) | |
1029 | pfatal_with_name (pcp_fname); | |
1030 | no_precomp = 1; | |
1031 | } | |
1032 | break; | |
1033 | ||
1034 | case 't': | |
1035 | if (!strcmp (argv[i], "-traditional")) { | |
1036 | traditional = 1; | |
1037 | if (dollars_in_ident > 0) | |
1038 | dollars_in_ident = 1; | |
1039 | } else if (!strcmp (argv[i], "-trigraphs")) { | |
1040 | no_trigraphs = 0; | |
1041 | } | |
1042 | break; | |
1043 | ||
1044 | case 'l': | |
1045 | if (! strcmp (argv[i], "-lang-c")) | |
1046 | cplusplus = 0, objc = 0; | |
1047 | if (! strcmp (argv[i], "-lang-c++")) | |
1048 | cplusplus = 1, objc = 0; | |
1049 | if (! strcmp (argv[i], "-lang-objc")) | |
1050 | objc = 1, cplusplus = 0; | |
1051 | if (! strcmp (argv[i], "-lang-objc++")) | |
1052 | objc = 1, cplusplus = 1; | |
1053 | if (! strcmp (argv[i], "-lang-asm")) | |
1054 | lang_asm = 1; | |
1055 | if (! strcmp (argv[i], "-lint")) | |
1056 | lint = 1; | |
1057 | break; | |
1058 | ||
1059 | case '+': | |
1060 | cplusplus = 1; | |
1061 | break; | |
1062 | ||
1063 | case 'w': | |
1064 | inhibit_warnings = 1; | |
1065 | break; | |
1066 | ||
1067 | case 'W': | |
1068 | if (!strcmp (argv[i], "-Wtrigraphs")) | |
1069 | warn_trigraphs = 1; | |
1070 | else if (!strcmp (argv[i], "-Wno-trigraphs")) | |
1071 | warn_trigraphs = 0; | |
1072 | else if (!strcmp (argv[i], "-Wcomment")) | |
1073 | warn_comments = 1; | |
1074 | else if (!strcmp (argv[i], "-Wno-comment")) | |
1075 | warn_comments = 0; | |
1076 | else if (!strcmp (argv[i], "-Wcomments")) | |
1077 | warn_comments = 1; | |
1078 | else if (!strcmp (argv[i], "-Wno-comments")) | |
1079 | warn_comments = 0; | |
1080 | else if (!strcmp (argv[i], "-Wtraditional")) | |
1081 | warn_stringify = 1; | |
1082 | else if (!strcmp (argv[i], "-Wno-traditional")) | |
1083 | warn_stringify = 0; | |
1084 | else if (!strcmp (argv[i], "-Werror")) | |
1085 | warnings_are_errors = 1; | |
1086 | else if (!strcmp (argv[i], "-Wno-error")) | |
1087 | warnings_are_errors = 0; | |
1088 | else if (!strcmp (argv[i], "-Wall")) | |
1089 | { | |
1090 | warn_trigraphs = 1; | |
1091 | warn_comments = 1; | |
1092 | } | |
1093 | break; | |
1094 | ||
1095 | case 'M': | |
1096 | if (!strcmp (argv[i], "-M")) | |
1097 | print_deps = 2; | |
1098 | else if (!strcmp (argv[i], "-MM")) | |
1099 | print_deps = 1; | |
1100 | else if (!strcmp (argv[i], "-MD")) | |
1101 | print_deps = 2; | |
1102 | else if (!strcmp (argv[i], "-MMD")) | |
1103 | print_deps = 1; | |
1104 | /* For -MD and -MMD options, write deps on file named by next arg. */ | |
1105 | if (!strcmp (argv[i], "-MD") | |
1106 | || !strcmp (argv[i], "-MMD")) { | |
1107 | i++; | |
1108 | deps_file = argv[i]; | |
1109 | deps_stream = fopen (argv[i], "a"); | |
1110 | if (deps_stream == 0) | |
1111 | pfatal_with_name (argv[i]); | |
1112 | } else { | |
1113 | /* For -M and -MM, write deps on standard output | |
1114 | and suppress the usual output. */ | |
1115 | deps_stream = stdout; | |
1116 | inhibit_output = 1; | |
1117 | } | |
1118 | break; | |
1119 | ||
1120 | case 'd': | |
1121 | { | |
1122 | char *p = argv[i] + 2; | |
1123 | char c; | |
1124 | while (c = *p++) { | |
1125 | /* Arg to -d specifies what parts of macros to dump */ | |
1126 | switch (c) { | |
1127 | case 'M': | |
1128 | dump_macros = dump_only; | |
1129 | no_output = 1; | |
1130 | break; | |
1131 | case 'N': | |
1132 | dump_macros = dump_names; | |
1133 | break; | |
1134 | case 'D': | |
1135 | dump_macros = dump_definitions; | |
1136 | break; | |
1137 | } | |
1138 | } | |
1139 | } | |
1140 | break; | |
1141 | ||
1142 | case 'g': | |
1143 | if (argv[i][2] == '3') | |
1144 | debug_output = 1; | |
1145 | break; | |
1146 | ||
1147 | case 'v': | |
1148 | fprintf (stderr, "GNU CPP version %s", version_string); | |
1149 | #ifdef TARGET_VERSION | |
1150 | TARGET_VERSION; | |
1151 | #endif | |
1152 | fprintf (stderr, "\n"); | |
1153 | break; | |
1154 | ||
1155 | case 'H': | |
1156 | print_include_names = 1; | |
1157 | break; | |
1158 | ||
1159 | case 'D': | |
1160 | { | |
1161 | char *p, *p1; | |
1162 | ||
1163 | if (argv[i][2] != 0) | |
1164 | p = argv[i] + 2; | |
1165 | else if (i + 1 == argc) | |
1166 | fatal ("Macro name missing after -D option"); | |
1167 | else | |
1168 | p = argv[++i]; | |
1169 | ||
1170 | pend_defs[i] = p; | |
1171 | } | |
1172 | break; | |
1173 | ||
1174 | case 'A': | |
1175 | { | |
1176 | char *p, *p1; | |
1177 | ||
1178 | if (argv[i][2] != 0) | |
1179 | p = argv[i] + 2; | |
1180 | else if (i + 1 == argc) | |
1181 | fatal ("Assertion missing after -A option"); | |
1182 | else | |
1183 | p = argv[++i]; | |
1184 | ||
1185 | if (!strcmp (p, "-")) { | |
1186 | /* -A- eliminates all predefined macros and assertions. | |
1187 | Let's include also any that were specified earlier | |
1188 | on the command line. That way we can get rid of any | |
1189 | that were passed automatically in from GCC. */ | |
1190 | int j; | |
1191 | inhibit_predefs = 1; | |
1192 | for (j = 0; j < i; j++) | |
1193 | pend_defs[j] = pend_assertions[j] = 0; | |
1194 | } else { | |
1195 | pend_assertions[i] = p; | |
1196 | pend_assertion_options[i] = "-A"; | |
1197 | } | |
1198 | } | |
1199 | break; | |
1200 | ||
1201 | case 'U': /* JF #undef something */ | |
1202 | if (argv[i][2] != 0) | |
1203 | pend_undefs[i] = argv[i] + 2; | |
1204 | else if (i + 1 == argc) | |
1205 | fatal ("Macro name missing after -U option"); | |
1206 | else | |
1207 | pend_undefs[i] = argv[i+1], i++; | |
1208 | break; | |
1209 | ||
1210 | case 'C': | |
1211 | put_out_comments = 1; | |
1212 | break; | |
1213 | ||
1214 | case 'E': /* -E comes from cc -E; ignore it. */ | |
1215 | break; | |
1216 | ||
1217 | case 'P': | |
1218 | no_line_commands = 1; | |
1219 | break; | |
1220 | ||
1221 | case '$': /* Don't include $ in identifiers. */ | |
1222 | dollars_in_ident = 0; | |
1223 | break; | |
1224 | ||
1225 | case 'I': /* Add directory to path for includes. */ | |
1226 | { | |
1227 | struct file_name_list *dirtmp; | |
1228 | ||
1229 | if (! ignore_srcdir && !strcmp (argv[i] + 2, "-")) | |
1230 | ignore_srcdir = 1; | |
1231 | else { | |
1232 | dirtmp = (struct file_name_list *) | |
1233 | xmalloc (sizeof (struct file_name_list)); | |
1234 | dirtmp->next = 0; /* New one goes on the end */ | |
1235 | dirtmp->control_macro = 0; | |
1236 | if (include == 0) | |
1237 | include = dirtmp; | |
1238 | else | |
1239 | last_include->next = dirtmp; | |
1240 | last_include = dirtmp; /* Tail follows the last one */ | |
1241 | if (argv[i][2] != 0) | |
1242 | dirtmp->fname = argv[i] + 2; | |
1243 | else if (i + 1 == argc) | |
1244 | fatal ("Directory name missing after -I option"); | |
1245 | else | |
1246 | dirtmp->fname = argv[++i]; | |
1247 | if (strlen (dirtmp->fname) > max_include_len) | |
1248 | max_include_len = strlen (dirtmp->fname); | |
1249 | if (ignore_srcdir && first_bracket_include == 0) | |
1250 | first_bracket_include = dirtmp; | |
1251 | } | |
1252 | } | |
1253 | break; | |
1254 | ||
1255 | case 'n': | |
1256 | if (!strcmp (argv[i], "-nostdinc")) | |
1257 | /* -nostdinc causes no default include directories. | |
1258 | You must specify all include-file directories with -I. */ | |
1259 | no_standard_includes = 1; | |
1260 | else if (!strcmp (argv[i], "-noprecomp")) | |
1261 | no_precomp = 1; | |
1262 | break; | |
1263 | ||
1264 | case 'u': | |
1265 | /* Sun compiler passes undocumented switch "-undef". | |
1266 | Let's assume it means to inhibit the predefined symbols. */ | |
1267 | inhibit_predefs = 1; | |
1268 | break; | |
1269 | ||
1270 | case '\0': /* JF handle '-' as file name meaning stdin or stdout */ | |
1271 | if (in_fname == NULL) { | |
1272 | in_fname = ""; | |
1273 | break; | |
1274 | } else if (out_fname == NULL) { | |
1275 | out_fname = ""; | |
1276 | break; | |
1277 | } /* else fall through into error */ | |
1278 | ||
1279 | default: | |
1280 | fatal ("Invalid option `%s'", argv[i]); | |
1281 | } | |
1282 | } | |
1283 | } | |
1284 | ||
1285 | /* Add dirs from CPATH after dirs from -I. */ | |
1286 | /* There seems to be confusion about what CPATH should do, | |
1287 | so for the moment it is not documented. */ | |
1288 | /* Some people say that CPATH should replace the standard include dirs, | |
1289 | but that seems pointless: it comes before them, so it overrides them | |
1290 | anyway. */ | |
1291 | p = (char *) getenv ("CPATH"); | |
1292 | if (p != 0 && ! no_standard_includes) | |
1293 | path_include (p); | |
1294 | ||
1295 | /* Now that dollars_in_ident is known, initialize is_idchar. */ | |
1296 | initialize_char_syntax (); | |
1297 | ||
1298 | /* Initialize output buffer */ | |
1299 | ||
1300 | outbuf.buf = (U_CHAR *) xmalloc (OUTBUF_SIZE); | |
1301 | outbuf.bufp = outbuf.buf; | |
1302 | outbuf.length = OUTBUF_SIZE; | |
1303 | ||
1304 | /* Do partial setup of input buffer for the sake of generating | |
1305 | early #line directives (when -g is in effect). */ | |
1306 | ||
1307 | fp = &instack[++indepth]; | |
1308 | if (in_fname == NULL) | |
1309 | in_fname = ""; | |
1310 | fp->nominal_fname = fp->fname = in_fname; | |
1311 | fp->lineno = 0; | |
1312 | ||
1313 | /* Install __LINE__, etc. Must follow initialize_char_syntax | |
1314 | and option processing. */ | |
1315 | initialize_builtins (fp, &outbuf); | |
1316 | ||
1317 | /* Do standard #defines and assertions | |
1318 | that identify system and machine type. */ | |
1319 | ||
1320 | if (!inhibit_predefs) { | |
1321 | char *p = (char *) alloca (strlen (predefs) + 1); | |
1322 | strcpy (p, predefs); | |
1323 | while (*p) { | |
1324 | char *q; | |
1325 | while (*p == ' ' || *p == '\t') | |
1326 | p++; | |
1327 | /* Handle -D options. */ | |
1328 | if (p[0] == '-' && p[1] == 'D') { | |
1329 | q = &p[2]; | |
1330 | while (*p && *p != ' ' && *p != '\t') | |
1331 | p++; | |
1332 | if (*p != 0) | |
1333 | *p++= 0; | |
1334 | if (debug_output) | |
1335 | output_line_command (fp, &outbuf, 0, same_file); | |
1336 | make_definition (q, &outbuf); | |
1337 | while (*p == ' ' || *p == '\t') | |
1338 | p++; | |
1339 | } else if (p[0] == '-' && p[1] == 'A') { | |
1340 | /* Handle -A options (assertions). */ | |
1341 | char *assertion; | |
1342 | char *past_name; | |
1343 | char *value; | |
1344 | char *past_value; | |
1345 | char *termination; | |
1346 | int save_char; | |
1347 | ||
1348 | assertion = &p[2]; | |
1349 | past_name = assertion; | |
1350 | /* Locate end of name. */ | |
1351 | while (*past_name && *past_name != ' ' | |
1352 | && *past_name != '\t' && *past_name != '(') | |
1353 | past_name++; | |
1354 | /* Locate `(' at start of value. */ | |
1355 | value = past_name; | |
1356 | while (*value && (*value == ' ' || *value == '\t')) | |
1357 | value++; | |
1358 | if (*value++ != '(') | |
1359 | abort (); | |
1360 | while (*value && (*value == ' ' || *value == '\t')) | |
1361 | value++; | |
1362 | past_value = value; | |
1363 | /* Locate end of value. */ | |
1364 | while (*past_value && *past_value != ' ' | |
1365 | && *past_value != '\t' && *past_value != ')') | |
1366 | past_value++; | |
1367 | termination = past_value; | |
1368 | while (*termination && (*termination == ' ' || *termination == '\t')) | |
1369 | termination++; | |
1370 | if (*termination++ != ')') | |
1371 | abort (); | |
1372 | if (*termination && *termination != ' ' && *termination != '\t') | |
1373 | abort (); | |
1374 | /* Temporarily null-terminate the value. */ | |
1375 | save_char = *termination; | |
1376 | *termination = '\0'; | |
1377 | /* Install the assertion. */ | |
1378 | make_assertion ("-A", assertion); | |
1379 | *termination = (char) save_char; | |
1380 | p = termination; | |
1381 | while (*p == ' ' || *p == '\t') | |
1382 | p++; | |
1383 | } else { | |
1384 | abort (); | |
1385 | } | |
1386 | } | |
1387 | } | |
1388 | ||
1389 | /* Now handle the command line options. */ | |
1390 | ||
1391 | /* Do assertions specified with -A. */ | |
1392 | for (i = 1; i < argc; i++) | |
1393 | if (pend_assertions[i]) | |
1394 | make_assertion (pend_assertion_options[i], pend_assertions[i]); | |
1395 | ||
1396 | /* Do defines specified with -D. */ | |
1397 | for (i = 1; i < argc; i++) | |
1398 | if (pend_defs[i]) { | |
1399 | if (debug_output) | |
1400 | output_line_command (fp, &outbuf, 0, same_file); | |
1401 | make_definition (pend_defs[i], &outbuf); | |
1402 | } | |
1403 | ||
1404 | /* Do undefines specified with -U. */ | |
1405 | for (i = 1; i < argc; i++) | |
1406 | if (pend_undefs[i]) { | |
1407 | if (debug_output) | |
1408 | output_line_command (fp, &outbuf, 0, same_file); | |
1409 | make_undef (pend_undefs[i], &outbuf); | |
1410 | } | |
1411 | ||
1412 | done_initializing = 1; | |
1413 | ||
1414 | { /* read the appropriate environment variable and if it exists | |
1415 | replace include_defaults with the listed path. */ | |
1416 | char *epath = 0; | |
1417 | switch ((objc << 1) + cplusplus) | |
1418 | { | |
1419 | case 0: | |
1420 | epath = getenv ("C_INCLUDE_PATH"); | |
1421 | break; | |
1422 | case 1: | |
1423 | epath = getenv ("CPLUS_INCLUDE_PATH"); | |
1424 | break; | |
1425 | case 2: | |
1426 | epath = getenv ("OBJC_INCLUDE_PATH"); | |
1427 | break; | |
1428 | case 3: | |
1429 | epath = getenv ("OBJCPLUS_INCLUDE_PATH"); | |
1430 | break; | |
1431 | } | |
1432 | /* If the environment var for this language is set, | |
1433 | add to the default list of include directories. */ | |
1434 | if (epath) { | |
1435 | char *nstore = (char *) alloca (strlen (epath) + 2); | |
1436 | int num_dirs; | |
1437 | char *startp, *endp; | |
1438 | ||
1439 | for (num_dirs = 1, startp = epath; *startp; startp++) | |
1440 | if (*startp == ':') | |
1441 | num_dirs++; | |
1442 | include_defaults | |
1443 | = (struct default_include *) xmalloc ((num_dirs | |
1444 | * sizeof (struct default_include)) | |
1445 | + sizeof (include_defaults_array)); | |
1446 | startp = endp = epath; | |
1447 | num_dirs = 0; | |
1448 | while (1) { | |
1449 | if ((*endp == ':') || (*endp == '\0')) { | |
1450 | strncpy (nstore, startp, endp-startp); | |
1451 | if (endp == startp) | |
1452 | strcpy (nstore, "."); | |
1453 | else | |
1454 | nstore[endp-startp] = '\0'; | |
1455 | ||
1456 | max_include_len = MAX (max_include_len, endp-startp+2); | |
1457 | include_defaults[num_dirs].fname = savestring (nstore); | |
1458 | include_defaults[num_dirs].cplusplus = cplusplus; | |
1459 | num_dirs++; | |
1460 | if (*endp == '\0') | |
1461 | break; | |
1462 | endp = startp = endp + 1; | |
1463 | } else | |
1464 | endp++; | |
1465 | } | |
1466 | /* Put the usual defaults back in at the end. */ | |
1467 | bcopy (include_defaults_array, &include_defaults[num_dirs], | |
1468 | sizeof (include_defaults_array)); | |
1469 | } | |
1470 | } | |
1471 | ||
1472 | /* Unless -fnostdinc, | |
1473 | tack on the standard include file dirs to the specified list */ | |
1474 | if (!no_standard_includes) { | |
1475 | struct default_include *p = include_defaults; | |
1476 | char *specd_prefix = getenv ("GCC_EXEC_PREFIX"); | |
1477 | char *default_prefix = savestring (GCC_INCLUDE_DIR); | |
1478 | int default_len = 0; | |
1479 | /* Remove the `include' from /usr/local/lib/gcc.../include. */ | |
1480 | if (!strcmp (default_prefix + strlen (default_prefix) - 8, "/include")) { | |
1481 | default_len = strlen (default_prefix) - 7; | |
1482 | default_prefix[default_len] = 0; | |
1483 | } | |
1484 | /* Search "translated" versions of GNU directories. | |
1485 | These have /usr/local/lib/gcc... replaced by specd_prefix. */ | |
1486 | if (specd_prefix != 0 && default_len != 0) | |
1487 | for (p = include_defaults; p->fname; p++) { | |
1488 | /* Some standard dirs are only for C++. */ | |
1489 | if (!p->cplusplus || cplusplus) { | |
1490 | /* Does this dir start with the prefix? */ | |
1491 | if (!strncmp (p->fname, default_prefix, default_len)) { | |
1492 | /* Yes; change prefix and add to search list. */ | |
1493 | struct file_name_list *new | |
1494 | = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
1495 | int this_len = strlen (specd_prefix) + strlen (p->fname) - default_len; | |
1496 | char *str = (char *) xmalloc (this_len + 1); | |
1497 | strcpy (str, specd_prefix); | |
1498 | strcat (str, p->fname + default_len); | |
1499 | new->fname = str; | |
1500 | new->control_macro = 0; | |
1501 | ||
1502 | /* Add elt to tail of list. */ | |
1503 | if (include == 0) | |
1504 | include = new; | |
1505 | else | |
1506 | last_include->next = new; | |
1507 | /* Make sure list for #include <...> also has the standard dirs. */ | |
1508 | if (ignore_srcdir && first_bracket_include == 0) | |
1509 | first_bracket_include = new; | |
1510 | /* Record new tail. */ | |
1511 | last_include = new; | |
1512 | /* Update max_include_len if necessary. */ | |
1513 | if (this_len > max_include_len) | |
1514 | max_include_len = this_len; | |
1515 | } | |
1516 | } | |
1517 | } | |
1518 | /* Search ordinary names for GNU include directories. */ | |
1519 | for (p = include_defaults; p->fname; p++) { | |
1520 | /* Some standard dirs are only for C++. */ | |
1521 | if (!p->cplusplus || cplusplus) { | |
1522 | struct file_name_list *new | |
1523 | = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
1524 | new->control_macro = 0; | |
1525 | /* Add elt to tail of list. */ | |
1526 | if (include == 0) | |
1527 | include = new; | |
1528 | else | |
1529 | last_include->next = new; | |
1530 | /* Make sure list for #include <...> also has the standard dirs. */ | |
1531 | if (ignore_srcdir && first_bracket_include == 0) | |
1532 | first_bracket_include = new; | |
1533 | /* Record new tail. */ | |
1534 | last_include = new; | |
1535 | new->fname = p->fname; | |
1536 | } | |
1537 | } | |
1538 | } | |
1539 | ||
1540 | /* Tack the after_include chain at the end of the include chain. */ | |
1541 | if (last_include) | |
1542 | last_include->next = after_include; | |
1543 | else | |
1544 | include = after_include; | |
1545 | if (ignore_srcdir && first_bracket_include == 0) | |
1546 | first_bracket_include = after_include; | |
1547 | ||
1548 | /* Terminate the after_include chain. */ | |
1549 | if (last_after_include) | |
1550 | last_after_include->next = 0; | |
1551 | ||
1552 | /* Scan the -imacros files before the main input. | |
1553 | Much like #including them, but with no_output set | |
1554 | so that only their macro definitions matter. */ | |
1555 | ||
1556 | no_output++; | |
1557 | for (i = 1; i < argc; i++) | |
1558 | if (pend_files[i]) { | |
1559 | int fd = open (pend_files[i], O_RDONLY, 0666); | |
1560 | if (fd < 0) { | |
1561 | perror_with_name (pend_files[i]); | |
1562 | return FAILURE_EXIT_CODE; | |
1563 | } | |
1564 | finclude (fd, pend_files[i], &outbuf, 0, 0); | |
1565 | } | |
1566 | no_output--; | |
1567 | ||
1568 | /* Copy the entire contents of the main input file into | |
1569 | the stacked input buffer previously allocated for it. */ | |
1570 | ||
1571 | /* JF check for stdin */ | |
1572 | if (in_fname == NULL || *in_fname == 0) { | |
1573 | in_fname = ""; | |
1574 | f = 0; | |
1575 | } else if ((f = open (in_fname, O_RDONLY, 0666)) < 0) | |
1576 | goto perror; | |
1577 | ||
1578 | /* Either of two environment variables can specify output of deps. | |
1579 | Its value is either "OUTPUT_FILE" or "OUTPUT_FILE DEPS_TARGET", | |
1580 | where OUTPUT_FILE is the file to write deps info to | |
1581 | and DEPS_TARGET is the target to mention in the deps. */ | |
1582 | ||
1583 | if (print_deps == 0 | |
1584 | && (getenv ("SUNPRO_DEPENDENCIES") != 0 | |
1585 | || getenv ("DEPENDENCIES_OUTPUT") != 0)) { | |
1586 | char *spec = getenv ("DEPENDENCIES_OUTPUT"); | |
1587 | char *s; | |
1588 | char *output_file; | |
1589 | ||
1590 | if (spec == 0) { | |
1591 | spec = getenv ("SUNPRO_DEPENDENCIES"); | |
1592 | print_deps = 2; | |
1593 | } | |
1594 | else | |
1595 | print_deps = 1; | |
1596 | ||
1597 | s = spec; | |
1598 | /* Find the space before the DEPS_TARGET, if there is one. */ | |
1599 | /* Don't use `index'; that causes trouble on USG. */ | |
1600 | while (*s != 0 && *s != ' ') s++; | |
1601 | if (*s != 0) { | |
1602 | deps_target = s + 1; | |
1603 | output_file = (char *) xmalloc (s - spec + 1); | |
1604 | bcopy (spec, output_file, s - spec); | |
1605 | output_file[s - spec] = 0; | |
1606 | } | |
1607 | else { | |
1608 | deps_target = 0; | |
1609 | output_file = spec; | |
1610 | } | |
1611 | ||
1612 | deps_file = output_file; | |
1613 | deps_stream = fopen (output_file, "a"); | |
1614 | if (deps_stream == 0) | |
1615 | pfatal_with_name (output_file); | |
1616 | } | |
1617 | ||
1618 | /* For -M, print the expected object file name | |
1619 | as the target of this Make-rule. */ | |
1620 | if (print_deps) { | |
1621 | deps_allocated_size = 200; | |
1622 | deps_buffer = (char *) xmalloc (deps_allocated_size); | |
1623 | deps_buffer[0] = 0; | |
1624 | deps_size = 0; | |
1625 | deps_column = 0; | |
1626 | ||
1627 | if (deps_target) { | |
1628 | deps_output (deps_target, 0); | |
1629 | deps_output (":", 0); | |
1630 | } else if (*in_fname == 0) | |
1631 | deps_output ("-: ", 0); | |
1632 | else { | |
1633 | int len; | |
1634 | char *p = in_fname; | |
1635 | char *p1 = p; | |
1636 | /* Discard all directory prefixes from P. */ | |
1637 | while (*p1) { | |
1638 | if (*p1 == '/') | |
1639 | p = p1 + 1; | |
1640 | p1++; | |
1641 | } | |
1642 | /* Output P, but remove known suffixes. */ | |
1643 | len = strlen (p); | |
1644 | if (p[len - 2] == '.' && p[len - 1] == 'c') | |
1645 | deps_output (p, len - 2); | |
1646 | else if (p[len - 2] == '.' && p[len - 1] == 'C') | |
1647 | deps_output (p, len - 2); | |
1648 | else if (p[len - 3] == '.' | |
1649 | && p[len - 2] == 'c' | |
1650 | && p[len - 1] == 'c') | |
1651 | deps_output (p, len - 3); | |
1652 | else if (p[len - 2] == '.' && p[len - 1] == 's') | |
1653 | deps_output (p, len - 2); | |
1654 | else if (p[len - 2] == '.' && p[len - 1] == 'S') | |
1655 | deps_output (p, len - 2); | |
1656 | else if (p[len - 2] == '.' && p[len - 1] == 'm') | |
1657 | deps_output (p, len - 2); | |
1658 | else | |
1659 | deps_output (p, 0); | |
1660 | /* Supply our own suffix. */ | |
1661 | deps_output (".o : ", 0); | |
1662 | deps_output (in_fname, 0); | |
1663 | deps_output (" ", 0); | |
1664 | } | |
1665 | } | |
1666 | ||
1667 | file_size_and_mode (f, &st_mode, &st_size); | |
1668 | fp->nominal_fname = fp->fname = in_fname; | |
1669 | fp->lineno = 1; | |
1670 | fp->system_header_p = 0; | |
1671 | /* JF all this is mine about reading pipes and ttys */ | |
1672 | if (! S_ISREG (st_mode)) { | |
1673 | /* Read input from a file that is not a normal disk file. | |
1674 | We cannot preallocate a buffer with the correct size, | |
1675 | so we must read in the file a piece at the time and make it bigger. */ | |
1676 | int size; | |
1677 | int bsize; | |
1678 | int cnt; | |
1679 | U_CHAR *bufp; | |
1680 | ||
1681 | bsize = 2000; | |
1682 | size = 0; | |
1683 | fp->buf = (U_CHAR *) xmalloc (bsize + 2); | |
1684 | bufp = fp->buf; | |
1685 | for (;;) { | |
1686 | cnt = read (f, bufp, bsize - size); | |
1687 | if (cnt < 0) goto perror; /* error! */ | |
1688 | if (cnt == 0) break; /* End of file */ | |
1689 | size += cnt; | |
1690 | bufp += cnt; | |
1691 | if (bsize == size) { /* Buffer is full! */ | |
1692 | bsize *= 2; | |
1693 | fp->buf = (U_CHAR *) xrealloc (fp->buf, bsize + 2); | |
1694 | bufp = fp->buf + size; /* May have moved */ | |
1695 | } | |
1696 | } | |
1697 | fp->length = size; | |
1698 | } else { | |
1699 | /* Read a file whose size we can determine in advance. | |
1700 | For the sake of VMS, st_size is just an upper bound. */ | |
1701 | long i; | |
1702 | fp->length = 0; | |
1703 | fp->buf = (U_CHAR *) xmalloc (st_size + 2); | |
1704 | ||
1705 | while (st_size > 0) { | |
1706 | i = read (f, fp->buf + fp->length, st_size); | |
1707 | if (i <= 0) { | |
1708 | if (i == 0) break; | |
1709 | goto perror; | |
1710 | } | |
1711 | fp->length += i; | |
1712 | st_size -= i; | |
1713 | } | |
1714 | } | |
1715 | fp->bufp = fp->buf; | |
1716 | fp->if_stack = if_stack; | |
1717 | ||
1718 | /* Make sure data ends with a newline. And put a null after it. */ | |
1719 | ||
1720 | if ((fp->length > 0 && fp->buf[fp->length - 1] != '\n') | |
1721 | /* Backslash-newline at end is not good enough. */ | |
1722 | || (fp->length > 1 && fp->buf[fp->length - 2] == '\\')) { | |
1723 | fp->buf[fp->length++] = '\n'; | |
1724 | missing_newline = 1; | |
1725 | } | |
1726 | fp->buf[fp->length] = '\0'; | |
1727 | ||
1728 | /* Unless inhibited, convert trigraphs in the input. */ | |
1729 | ||
1730 | if (!no_trigraphs) | |
1731 | trigraph_pcp (fp); | |
1732 | ||
1733 | /* Now that we know the input file is valid, open the output. */ | |
1734 | ||
1735 | if (!out_fname || !strcmp (out_fname, "")) | |
1736 | out_fname = "stdout"; | |
1737 | else if (! freopen (out_fname, "w", stdout)) | |
1738 | pfatal_with_name (out_fname); | |
1739 | ||
1740 | output_line_command (fp, &outbuf, 0, same_file); | |
1741 | ||
1742 | /* Scan the -include files before the main input. */ | |
1743 | ||
1744 | for (i = 1; i < argc; i++) | |
1745 | if (pend_includes[i]) { | |
1746 | int fd = open (pend_includes[i], O_RDONLY, 0666); | |
1747 | if (fd < 0) { | |
1748 | perror_with_name (pend_includes[i]); | |
1749 | return FAILURE_EXIT_CODE; | |
1750 | } | |
1751 | finclude (fd, pend_includes[i], &outbuf, 0, 0); | |
1752 | } | |
1753 | ||
1754 | /* Scan the input, processing macros and directives. */ | |
1755 | ||
1756 | rescan (&outbuf, 0); | |
1757 | ||
1758 | if (pedantic && missing_newline) | |
1759 | pedwarn ("file does not end in newline"); | |
1760 | ||
1761 | /* Now we have processed the entire input | |
1762 | Write whichever kind of output has been requested. */ | |
1763 | ||
1764 | if (dump_macros == dump_only) | |
1765 | dump_all_macros (); | |
1766 | else if (! inhibit_output) { | |
1767 | write_output (); | |
1768 | } | |
1769 | ||
1770 | if (print_deps) { | |
1771 | /* Don't actually write the deps file if compilation has failed. | |
1772 | Delete it instead. */ | |
1773 | if (errors > 0 && deps_file != 0) | |
1774 | unlink (deps_file); | |
1775 | else { | |
1776 | fputs (deps_buffer, deps_stream); | |
1777 | putc ('\n', deps_stream); | |
1778 | if (deps_stream != stdout) { | |
1779 | fclose (deps_stream); | |
1780 | if (ferror (deps_stream)) | |
1781 | fatal ("I/O error on output"); | |
1782 | } | |
1783 | } | |
1784 | } | |
1785 | ||
1786 | if (ferror (stdout)) | |
1787 | fatal ("I/O error on output"); | |
1788 | ||
1789 | if (errors) | |
1790 | exit (FAILURE_EXIT_CODE); | |
1791 | exit (SUCCESS_EXIT_CODE); | |
1792 | ||
1793 | perror: | |
1794 | pfatal_with_name (in_fname); | |
1795 | return 0; | |
1796 | } | |
1797 | \f | |
1798 | /* Given a colon-separated list of file names PATH, | |
1799 | add all the names to the search path for include files. */ | |
1800 | ||
1801 | static void | |
1802 | path_include (path) | |
1803 | char *path; | |
1804 | { | |
1805 | char *p; | |
1806 | ||
1807 | p = path; | |
1808 | ||
1809 | if (*p) | |
1810 | while (1) { | |
1811 | char *q = p; | |
1812 | char *name; | |
1813 | struct file_name_list *dirtmp; | |
1814 | ||
1815 | /* Find the end of this name. */ | |
1816 | while (*q != 0 && *q != ':') q++; | |
1817 | if (p == q) { | |
1818 | /* An empty name in the path stands for the current directory. */ | |
1819 | name = (char *) xmalloc (2); | |
1820 | name[0] = '.'; | |
1821 | name[1] = 0; | |
1822 | } else { | |
1823 | /* Otherwise use the directory that is named. */ | |
1824 | name = (char *) xmalloc (q - p + 1); | |
1825 | bcopy (p, name, q - p); | |
1826 | name[q - p] = 0; | |
1827 | } | |
1828 | ||
1829 | dirtmp = (struct file_name_list *) | |
1830 | xmalloc (sizeof (struct file_name_list)); | |
1831 | dirtmp->next = 0; /* New one goes on the end */ | |
1832 | dirtmp->control_macro = 0; | |
1833 | if (include == 0) | |
1834 | include = dirtmp; | |
1835 | else | |
1836 | last_include->next = dirtmp; | |
1837 | last_include = dirtmp; /* Tail follows the last one */ | |
1838 | dirtmp->fname = name; | |
1839 | if (strlen (dirtmp->fname) > max_include_len) | |
1840 | max_include_len = strlen (dirtmp->fname); | |
1841 | if (ignore_srcdir && first_bracket_include == 0) | |
1842 | first_bracket_include = dirtmp; | |
1843 | ||
1844 | /* Advance past this name. */ | |
1845 | p = q; | |
1846 | if (*p == 0) | |
1847 | break; | |
1848 | /* Skip the colon. */ | |
1849 | p++; | |
1850 | } | |
1851 | } | |
1852 | \f | |
1853 | /* Pre-C-Preprocessor to translate ANSI trigraph idiocy in BUF | |
1854 | before main CCCP processing. Name `pcp' is also in honor of the | |
1855 | drugs the trigraph designers must have been on. | |
1856 | ||
1857 | Using an extra pass through the buffer takes a little extra time, | |
1858 | but is infinitely less hairy than trying to handle trigraphs inside | |
1859 | strings, etc. everywhere, and also makes sure that trigraphs are | |
1860 | only translated in the top level of processing. */ | |
1861 | ||
1862 | static void | |
1863 | trigraph_pcp (buf) | |
1864 | FILE_BUF *buf; | |
1865 | { | |
1866 | register U_CHAR c, *fptr, *bptr, *sptr; | |
1867 | int len; | |
1868 | ||
1869 | fptr = bptr = sptr = buf->buf; | |
1870 | while ((sptr = (U_CHAR *) index (sptr, '?')) != NULL) { | |
1871 | if (*++sptr != '?') | |
1872 | continue; | |
1873 | switch (*++sptr) { | |
1874 | case '=': | |
1875 | c = '#'; | |
1876 | break; | |
1877 | case '(': | |
1878 | c = '['; | |
1879 | break; | |
1880 | case '/': | |
1881 | c = '\\'; | |
1882 | break; | |
1883 | case ')': | |
1884 | c = ']'; | |
1885 | break; | |
1886 | case '\'': | |
1887 | c = '^'; | |
1888 | break; | |
1889 | case '<': | |
1890 | c = '{'; | |
1891 | break; | |
1892 | case '!': | |
1893 | c = '|'; | |
1894 | break; | |
1895 | case '>': | |
1896 | c = '}'; | |
1897 | break; | |
1898 | case '-': | |
1899 | c = '~'; | |
1900 | break; | |
1901 | case '?': | |
1902 | sptr--; | |
1903 | continue; | |
1904 | default: | |
1905 | continue; | |
1906 | } | |
1907 | len = sptr - fptr - 2; | |
1908 | if (bptr != fptr && len > 0) | |
1909 | bcopy (fptr, bptr, len); /* BSD doc says bcopy () works right | |
1910 | for overlapping strings. In ANSI | |
1911 | C, this will be memmove (). */ | |
1912 | bptr += len; | |
1913 | *bptr++ = c; | |
1914 | fptr = ++sptr; | |
1915 | } | |
1916 | len = buf->length - (fptr - buf->buf); | |
1917 | if (bptr != fptr && len > 0) | |
1918 | bcopy (fptr, bptr, len); | |
1919 | buf->length -= fptr - bptr; | |
1920 | buf->buf[buf->length] = '\0'; | |
1921 | if (warn_trigraphs && fptr != bptr) | |
1922 | warning ("%d trigraph(s) encountered", (fptr - bptr) / 2); | |
1923 | } | |
1924 | \f | |
1925 | /* Move all backslash-newline pairs out of embarrassing places. | |
1926 | Exchange all such pairs following BP | |
1927 | with any potentially-embarrasing characters that follow them. | |
1928 | Potentially-embarrassing characters are / and * | |
1929 | (because a backslash-newline inside a comment delimiter | |
1930 | would cause it not to be recognized). */ | |
1931 | ||
1932 | static void | |
1933 | newline_fix (bp) | |
1934 | U_CHAR *bp; | |
1935 | { | |
1936 | register U_CHAR *p = bp; | |
1937 | register int count = 0; | |
1938 | ||
1939 | /* First count the backslash-newline pairs here. */ | |
1940 | ||
1941 | while (1) { | |
1942 | if (p[0] == '\\') { | |
1943 | if (p[1] == '\n') | |
1944 | p += 2, count++; | |
1945 | else if (p[1] == '\r' && p[2] == '\n') | |
1946 | p += 3, count++; | |
1947 | else | |
1948 | break; | |
1949 | } else | |
1950 | break; | |
1951 | } | |
1952 | ||
1953 | /* What follows the backslash-newlines is not embarrassing. */ | |
1954 | ||
1955 | if (count == 0 || (*p != '/' && *p != '*')) | |
1956 | return; | |
1957 | ||
1958 | /* Copy all potentially embarrassing characters | |
1959 | that follow the backslash-newline pairs | |
1960 | down to where the pairs originally started. */ | |
1961 | ||
1962 | while (*p == '*' || *p == '/') | |
1963 | *bp++ = *p++; | |
1964 | ||
1965 | /* Now write the same number of pairs after the embarrassing chars. */ | |
1966 | while (count-- > 0) { | |
1967 | *bp++ = '\\'; | |
1968 | *bp++ = '\n'; | |
1969 | } | |
1970 | } | |
1971 | ||
1972 | /* Like newline_fix but for use within a directive-name. | |
1973 | Move any backslash-newlines up past any following symbol constituents. */ | |
1974 | ||
1975 | static void | |
1976 | name_newline_fix (bp) | |
1977 | U_CHAR *bp; | |
1978 | { | |
1979 | register U_CHAR *p = bp; | |
1980 | register int count = 0; | |
1981 | ||
1982 | /* First count the backslash-newline pairs here. */ | |
1983 | while (1) { | |
1984 | if (p[0] == '\\') { | |
1985 | if (p[1] == '\n') | |
1986 | p += 2, count++; | |
1987 | else if (p[1] == '\r' && p[2] == '\n') | |
1988 | p += 3, count++; | |
1989 | else | |
1990 | break; | |
1991 | } else | |
1992 | break; | |
1993 | } | |
1994 | ||
1995 | /* What follows the backslash-newlines is not embarrassing. */ | |
1996 | ||
1997 | if (count == 0 || !is_idchar[*p]) | |
1998 | return; | |
1999 | ||
2000 | /* Copy all potentially embarrassing characters | |
2001 | that follow the backslash-newline pairs | |
2002 | down to where the pairs originally started. */ | |
2003 | ||
2004 | while (is_idchar[*p]) | |
2005 | *bp++ = *p++; | |
2006 | ||
2007 | /* Now write the same number of pairs after the embarrassing chars. */ | |
2008 | while (count-- > 0) { | |
2009 | *bp++ = '\\'; | |
2010 | *bp++ = '\n'; | |
2011 | } | |
2012 | } | |
2013 | \f | |
2014 | /* Look for lint commands in comments. | |
2015 | ||
2016 | When we come in here, ibp points into a comment. Limit is as one expects. | |
2017 | scan within the comment -- it should start, after lwsp, with a lint command. | |
2018 | If so that command is returned as a (constant) string. | |
2019 | ||
2020 | Upon return, any arg will be pointed to with argstart and will be | |
2021 | arglen long. Note that we don't parse that arg since it will just | |
2022 | be printed out again. | |
2023 | */ | |
2024 | ||
2025 | static char * | |
2026 | get_lintcmd (ibp, limit, argstart, arglen, cmdlen) | |
2027 | register U_CHAR *ibp; | |
2028 | register U_CHAR *limit; | |
2029 | U_CHAR **argstart; /* point to command arg */ | |
2030 | int *arglen, *cmdlen; /* how long they are */ | |
2031 | { | |
2032 | long linsize; | |
2033 | register U_CHAR *numptr; /* temp for arg parsing */ | |
2034 | ||
2035 | *arglen = 0; | |
2036 | ||
2037 | SKIP_WHITE_SPACE (ibp); | |
2038 | ||
2039 | if (ibp >= limit) return NULL; | |
2040 | ||
2041 | linsize = limit - ibp; | |
2042 | ||
2043 | /* Oh, I wish C had lexical functions... hell, I'll just open-code the set */ | |
2044 | if ((linsize >= 10) && !strncmp (ibp, "NOTREACHED", 10)) { | |
2045 | *cmdlen = 10; | |
2046 | return "NOTREACHED"; | |
2047 | } | |
2048 | if ((linsize >= 8) && !strncmp (ibp, "ARGSUSED", 8)) { | |
2049 | *cmdlen = 8; | |
2050 | return "ARGSUSED"; | |
2051 | } | |
2052 | if ((linsize >= 11) && !strncmp (ibp, "LINTLIBRARY", 8)) { | |
2053 | *cmdlen = 8; | |
2054 | return "LINTLIBRARY"; | |
2055 | } | |
2056 | if ((linsize >= 7) && !strncmp (ibp, "VARARGS", 7)) { | |
2057 | *cmdlen = 7; | |
2058 | ibp += 7; linsize -= 7; | |
2059 | if ((linsize == 0) || ! isdigit (*ibp)) return "VARARGS"; | |
2060 | ||
2061 | /* OK, read a number */ | |
2062 | for (numptr = *argstart = ibp; (numptr < limit) && isdigit (*numptr); | |
2063 | numptr++); | |
2064 | *arglen = numptr - *argstart; | |
2065 | return "VARARGS"; | |
2066 | } | |
2067 | return NULL; | |
2068 | } | |
2069 | \f | |
2070 | /* | |
2071 | * The main loop of the program. | |
2072 | * | |
2073 | * Read characters from the input stack, transferring them to the | |
2074 | * output buffer OP. | |
2075 | * | |
2076 | * Macros are expanded and push levels on the input stack. | |
2077 | * At the end of such a level it is popped off and we keep reading. | |
2078 | * At the end of any other kind of level, we return. | |
2079 | * #-directives are handled, except within macros. | |
2080 | * | |
2081 | * If OUTPUT_MARKS is nonzero, keep Newline markers found in the input | |
2082 | * and insert them when appropriate. This is set while scanning macro | |
2083 | * arguments before substitution. It is zero when scanning for final output. | |
2084 | * There are three types of Newline markers: | |
2085 | * * Newline - follows a macro name that was not expanded | |
2086 | * because it appeared inside an expansion of the same macro. | |
2087 | * This marker prevents future expansion of that identifier. | |
2088 | * When the input is rescanned into the final output, these are deleted. | |
2089 | * These are also deleted by ## concatenation. | |
2090 | * * Newline Space (or Newline and any other whitespace character) | |
2091 | * stands for a place that tokens must be separated or whitespace | |
2092 | * is otherwise desirable, but where the ANSI standard specifies there | |
2093 | * is no whitespace. This marker turns into a Space (or whichever other | |
2094 | * whitespace char appears in the marker) in the final output, | |
2095 | * but it turns into nothing in an argument that is stringified with #. | |
2096 | * Such stringified arguments are the only place where the ANSI standard | |
2097 | * specifies with precision that whitespace may not appear. | |
2098 | * | |
2099 | * During this function, IP->bufp is kept cached in IBP for speed of access. | |
2100 | * Likewise, OP->bufp is kept in OBP. Before calling a subroutine | |
2101 | * IBP, IP and OBP must be copied back to memory. IP and IBP are | |
2102 | * copied back with the RECACHE macro. OBP must be copied back from OP->bufp | |
2103 | * explicitly, and before RECACHE, since RECACHE uses OBP. | |
2104 | */ | |
2105 | ||
2106 | static void | |
2107 | rescan (op, output_marks) | |
2108 | FILE_BUF *op; | |
2109 | int output_marks; | |
2110 | { | |
2111 | /* Character being scanned in main loop. */ | |
2112 | register U_CHAR c; | |
2113 | ||
2114 | /* Length of pending accumulated identifier. */ | |
2115 | register int ident_length = 0; | |
2116 | ||
2117 | /* Hash code of pending accumulated identifier. */ | |
2118 | register int hash = 0; | |
2119 | ||
2120 | /* Current input level (&instack[indepth]). */ | |
2121 | FILE_BUF *ip; | |
2122 | ||
2123 | /* Pointer for scanning input. */ | |
2124 | register U_CHAR *ibp; | |
2125 | ||
2126 | /* Pointer to end of input. End of scan is controlled by LIMIT. */ | |
2127 | register U_CHAR *limit; | |
2128 | ||
2129 | /* Pointer for storing output. */ | |
2130 | register U_CHAR *obp; | |
2131 | ||
2132 | /* REDO_CHAR is nonzero if we are processing an identifier | |
2133 | after backing up over the terminating character. | |
2134 | Sometimes we process an identifier without backing up over | |
2135 | the terminating character, if the terminating character | |
2136 | is not special. Backing up is done so that the terminating character | |
2137 | will be dispatched on again once the identifier is dealt with. */ | |
2138 | int redo_char = 0; | |
2139 | ||
2140 | /* 1 if within an identifier inside of which a concatenation | |
2141 | marker (Newline -) has been seen. */ | |
2142 | int concatenated = 0; | |
2143 | ||
2144 | /* While scanning a comment or a string constant, | |
2145 | this records the line it started on, for error messages. */ | |
2146 | int start_line; | |
2147 | ||
2148 | /* Line where a newline was first seen in a string constant. */ | |
2149 | int multiline_string_line = 0; | |
2150 | ||
2151 | /* Record position of last `real' newline. */ | |
2152 | U_CHAR *beg_of_line; | |
2153 | ||
2154 | /* Pop the innermost input stack level, assuming it is a macro expansion. */ | |
2155 | ||
2156 | #define POPMACRO \ | |
2157 | do { ip->macro->type = T_MACRO; \ | |
2158 | if (ip->free_ptr) free (ip->free_ptr); \ | |
2159 | --indepth; } while (0) | |
2160 | ||
2161 | /* Reload `rescan's local variables that describe the current | |
2162 | level of the input stack. */ | |
2163 | ||
2164 | #define RECACHE \ | |
2165 | do { ip = &instack[indepth]; \ | |
2166 | ibp = ip->bufp; \ | |
2167 | limit = ip->buf + ip->length; \ | |
2168 | op->bufp = obp; \ | |
2169 | check_expand (op, limit - ibp); \ | |
2170 | beg_of_line = 0; \ | |
2171 | obp = op->bufp; } while (0) | |
2172 | ||
2173 | if (no_output && instack[indepth].fname != 0) | |
2174 | skip_if_group (&instack[indepth], 1); | |
2175 | ||
2176 | obp = op->bufp; | |
2177 | RECACHE; | |
2178 | beg_of_line = ibp; | |
2179 | ||
2180 | /* Our caller must always put a null after the end of | |
2181 | the input at each input stack level. */ | |
2182 | if (*limit != 0) | |
2183 | abort (); | |
2184 | ||
2185 | while (1) { | |
2186 | c = *ibp++; | |
2187 | *obp++ = c; | |
2188 | ||
2189 | switch (c) { | |
2190 | case '\\': | |
2191 | if (ibp >= limit) | |
2192 | break; | |
2193 | if (*ibp == '\n') { | |
2194 | /* Always merge lines ending with backslash-newline, | |
2195 | even in middle of identifier. */ | |
2196 | ++ibp; | |
2197 | ++ip->lineno; | |
2198 | --obp; /* remove backslash from obuf */ | |
2199 | break; | |
2200 | } | |
2201 | /* Otherwise, backslash suppresses specialness of following char, | |
2202 | so copy it here to prevent the switch from seeing it. | |
2203 | But first get any pending identifier processed. */ | |
2204 | if (ident_length > 0) | |
2205 | goto specialchar; | |
2206 | *obp++ = *ibp++; | |
2207 | break; | |
2208 | ||
2209 | case '#': | |
2210 | if (assertions_flag) { | |
2211 | /* Copy #foo (bar lose) without macro expansion. */ | |
2212 | SKIP_WHITE_SPACE (ibp); | |
2213 | while (is_idchar[*ibp]) | |
2214 | *obp++ = *ibp++; | |
2215 | SKIP_WHITE_SPACE (ibp); | |
2216 | if (*ibp == '(') { | |
2217 | ip->bufp = ibp; | |
2218 | skip_paren_group (ip); | |
2219 | bcopy (ibp, obp, ip->bufp - ibp); | |
2220 | obp += ip->bufp - ibp; | |
2221 | ibp = ip->bufp; | |
2222 | } | |
2223 | } | |
2224 | ||
2225 | /* If this is expanding a macro definition, don't recognize | |
2226 | preprocessor directives. */ | |
2227 | if (ip->macro != 0) | |
2228 | goto randomchar; | |
2229 | if (ident_length) | |
2230 | goto specialchar; | |
2231 | ||
2232 | /* # keyword: a # must be first nonblank char on the line */ | |
2233 | if (beg_of_line == 0) | |
2234 | goto randomchar; | |
2235 | { | |
2236 | U_CHAR *bp; | |
2237 | ||
2238 | /* Scan from start of line, skipping whitespace, comments | |
2239 | and backslash-newlines, and see if we reach this #. | |
2240 | If not, this # is not special. */ | |
2241 | bp = beg_of_line; | |
2242 | while (1) { | |
2243 | if (is_hor_space[*bp]) | |
2244 | bp++; | |
2245 | else if (*bp == '\\' && bp[1] == '\n') | |
2246 | bp += 2; | |
2247 | else if (*bp == '/' && bp[1] == '*') { | |
2248 | bp += 2; | |
2249 | while (!(*bp == '*' && bp[1] == '/')) | |
2250 | bp++; | |
2251 | bp += 2; | |
2252 | } | |
2253 | else if ((cplusplus || objc) && *bp == '/' && bp[1] == '/') { | |
2254 | bp += 2; | |
2255 | while (*bp++ != '\n') ; | |
2256 | } | |
2257 | else break; | |
2258 | } | |
2259 | if (bp + 1 != ibp) | |
2260 | goto randomchar; | |
2261 | } | |
2262 | ||
2263 | /* This # can start a directive. */ | |
2264 | ||
2265 | --obp; /* Don't copy the '#' */ | |
2266 | ||
2267 | ip->bufp = ibp; | |
2268 | op->bufp = obp; | |
2269 | if (! handle_directive (ip, op)) { | |
2270 | #ifdef USE_C_ALLOCA | |
2271 | alloca (0); | |
2272 | #endif | |
2273 | /* Not a known directive: treat it as ordinary text. | |
2274 | IP, OP, IBP, etc. have not been changed. */ | |
2275 | if (no_output && instack[indepth].fname) { | |
2276 | /* If not generating expanded output, | |
2277 | what we do with ordinary text is skip it. | |
2278 | Discard everything until next # directive. */ | |
2279 | skip_if_group (&instack[indepth], 1); | |
2280 | RECACHE; | |
2281 | beg_of_line = ibp; | |
2282 | break; | |
2283 | } | |
2284 | ++obp; /* Copy the '#' after all */ | |
2285 | goto randomchar; | |
2286 | } | |
2287 | #ifdef USE_C_ALLOCA | |
2288 | alloca (0); | |
2289 | #endif | |
2290 | /* A # directive has been successfully processed. */ | |
2291 | /* If not generating expanded output, ignore everything until | |
2292 | next # directive. */ | |
2293 | if (no_output && instack[indepth].fname) | |
2294 | skip_if_group (&instack[indepth], 1); | |
2295 | obp = op->bufp; | |
2296 | RECACHE; | |
2297 | beg_of_line = ibp; | |
2298 | break; | |
2299 | ||
2300 | case '\"': /* skip quoted string */ | |
2301 | case '\'': | |
2302 | /* A single quoted string is treated like a double -- some | |
2303 | programs (e.g., troff) are perverse this way */ | |
2304 | ||
2305 | if (ident_length) | |
2306 | goto specialchar; | |
2307 | ||
2308 | start_line = ip->lineno; | |
2309 | ||
2310 | /* Skip ahead to a matching quote. */ | |
2311 | ||
2312 | while (1) { | |
2313 | if (ibp >= limit) { | |
2314 | if (traditional) { | |
2315 | if (ip->macro != 0) { | |
2316 | /* try harder: this string crosses a macro expansion boundary */ | |
2317 | POPMACRO; | |
2318 | RECACHE; | |
2319 | continue; | |
2320 | } | |
2321 | } else { | |
2322 | error_with_line (line_for_error (start_line), | |
2323 | "unterminated string or character constant"); | |
2324 | error_with_line (multiline_string_line, | |
2325 | "possible real start of unterminated constant"); | |
2326 | multiline_string_line = 0; | |
2327 | } | |
2328 | break; | |
2329 | } | |
2330 | *obp++ = *ibp; | |
2331 | switch (*ibp++) { | |
2332 | case '\n': | |
2333 | ++ip->lineno; | |
2334 | ++op->lineno; | |
2335 | /* Traditionally, end of line ends a string constant with no error. | |
2336 | So exit the loop and record the new line. */ | |
2337 | if (traditional) { | |
2338 | beg_of_line = ibp; | |
2339 | goto while2end; | |
2340 | } | |
2341 | if (pedantic || c == '\'') { | |
2342 | error_with_line (line_for_error (start_line), | |
2343 | "unterminated string or character constant"); | |
2344 | goto while2end; | |
2345 | } | |
2346 | if (multiline_string_line == 0) | |
2347 | multiline_string_line = ip->lineno - 1; | |
2348 | break; | |
2349 | ||
2350 | case '\\': | |
2351 | if (ibp >= limit) | |
2352 | break; | |
2353 | if (*ibp == '\n') { | |
2354 | /* Backslash newline is replaced by nothing at all, | |
2355 | but keep the line counts correct. */ | |
2356 | --obp; | |
2357 | ++ibp; | |
2358 | ++ip->lineno; | |
2359 | } else { | |
2360 | /* ANSI stupidly requires that in \\ the second \ | |
2361 | is *not* prevented from combining with a newline. */ | |
2362 | while (*ibp == '\\' && ibp[1] == '\n') { | |
2363 | ibp += 2; | |
2364 | ++ip->lineno; | |
2365 | } | |
2366 | *obp++ = *ibp++; | |
2367 | } | |
2368 | break; | |
2369 | ||
2370 | case '\"': | |
2371 | case '\'': | |
2372 | if (ibp[-1] == c) | |
2373 | goto while2end; | |
2374 | break; | |
2375 | } | |
2376 | } | |
2377 | while2end: | |
2378 | break; | |
2379 | ||
2380 | case '/': | |
2381 | if (*ibp == '\\' && ibp[1] == '\n') | |
2382 | newline_fix (ibp); | |
2383 | ||
2384 | if (*ibp != '*' | |
2385 | && !((cplusplus || objc) && *ibp == '/')) | |
2386 | goto randomchar; | |
2387 | if (ip->macro != 0) | |
2388 | goto randomchar; | |
2389 | if (ident_length) | |
2390 | goto specialchar; | |
2391 | ||
2392 | if (*ibp == '/') { | |
2393 | /* C++ style comment... */ | |
2394 | start_line = ip->lineno; | |
2395 | ||
2396 | --ibp; /* Back over the slash */ | |
2397 | --obp; | |
2398 | ||
2399 | /* Comments are equivalent to spaces. */ | |
2400 | if (! put_out_comments) | |
2401 | *obp++ = ' '; | |
2402 | else { | |
2403 | /* must fake up a comment here */ | |
2404 | *obp++ = '/'; | |
2405 | *obp++ = '/'; | |
2406 | } | |
2407 | { | |
2408 | U_CHAR *before_bp = ibp+2; | |
2409 | ||
2410 | while (ibp < limit) { | |
2411 | if (*ibp++ == '\n') { | |
2412 | ibp--; | |
2413 | if (put_out_comments) { | |
2414 | bcopy (before_bp, obp, ibp - before_bp); | |
2415 | obp += ibp - before_bp; | |
2416 | } | |
2417 | break; | |
2418 | } | |
2419 | } | |
2420 | break; | |
2421 | } | |
2422 | } | |
2423 | ||
2424 | /* Ordinary C comment. Skip it, optionally copying it to output. */ | |
2425 | ||
2426 | start_line = ip->lineno; | |
2427 | ||
2428 | ++ibp; /* Skip the star. */ | |
2429 | ||
2430 | /* If this cpp is for lint, we peek inside the comments: */ | |
2431 | if (lint) { | |
2432 | U_CHAR *argbp; | |
2433 | int cmdlen, arglen; | |
2434 | char *lintcmd = get_lintcmd (ibp, limit, &argbp, &arglen, &cmdlen); | |
2435 | ||
2436 | if (lintcmd != NULL) { | |
2437 | /* I believe it is always safe to emit this newline: */ | |
2438 | obp[-1] = '\n'; | |
2439 | bcopy ("#pragma lint ", obp, 13); | |
2440 | obp += 13; | |
2441 | bcopy (lintcmd, obp, cmdlen); | |
2442 | obp += cmdlen; | |
2443 | ||
2444 | if (arglen != 0) { | |
2445 | *(obp++) = ' '; | |
2446 | bcopy (argbp, obp, arglen); | |
2447 | obp += arglen; | |
2448 | } | |
2449 | ||
2450 | /* OK, now bring us back to the state we were in before we entered | |
2451 | this branch. We need #line b/c the newline for the pragma | |
2452 | could fuck things up. */ | |
2453 | output_line_command (ip, op, 0, same_file); | |
2454 | *(obp++) = ' '; /* just in case, if comments are copied thru */ | |
2455 | *(obp++) = '/'; | |
2456 | } | |
2457 | } | |
2458 | ||
2459 | /* Comments are equivalent to spaces. | |
2460 | Note that we already output the slash; we might not want it. | |
2461 | For -traditional, a comment is equivalent to nothing. */ | |
2462 | if (! put_out_comments) { | |
2463 | if (traditional) | |
2464 | obp--; | |
2465 | else | |
2466 | obp[-1] = ' '; | |
2467 | } | |
2468 | else | |
2469 | *obp++ = '*'; | |
2470 | ||
2471 | { | |
2472 | U_CHAR *before_bp = ibp; | |
2473 | ||
2474 | while (ibp < limit) { | |
2475 | switch (*ibp++) { | |
2476 | case '/': | |
2477 | if (warn_comments && ibp < limit && *ibp == '*') | |
2478 | warning("`/*' within comment"); | |
2479 | break; | |
2480 | case '*': | |
2481 | if (*ibp == '\\' && ibp[1] == '\n') | |
2482 | newline_fix (ibp); | |
2483 | if (ibp >= limit || *ibp == '/') | |
2484 | goto comment_end; | |
2485 | break; | |
2486 | case '\n': | |
2487 | ++ip->lineno; | |
2488 | /* Copy the newline into the output buffer, in order to | |
2489 | avoid the pain of a #line every time a multiline comment | |
2490 | is seen. */ | |
2491 | if (!put_out_comments) | |
2492 | *obp++ = '\n'; | |
2493 | ++op->lineno; | |
2494 | } | |
2495 | } | |
2496 | comment_end: | |
2497 | ||
2498 | if (ibp >= limit) | |
2499 | error_with_line (line_for_error (start_line), | |
2500 | "unterminated comment"); | |
2501 | else { | |
2502 | ibp++; | |
2503 | if (put_out_comments) { | |
2504 | bcopy (before_bp, obp, ibp - before_bp); | |
2505 | obp += ibp - before_bp; | |
2506 | } | |
2507 | } | |
2508 | } | |
2509 | break; | |
2510 | ||
2511 | case '$': | |
2512 | if (!dollars_in_ident) | |
2513 | goto randomchar; | |
2514 | goto letter; | |
2515 | ||
2516 | case '0': case '1': case '2': case '3': case '4': | |
2517 | case '5': case '6': case '7': case '8': case '9': | |
2518 | /* If digit is not part of identifier, it starts a number, | |
2519 | which means that following letters are not an identifier. | |
2520 | "0x5" does not refer to an identifier "x5". | |
2521 | So copy all alphanumerics that follow without accumulating | |
2522 | as an identifier. Periods also, for sake of "3.e7". */ | |
2523 | ||
2524 | if (ident_length == 0) { | |
2525 | while (ibp < limit) { | |
2526 | while (ibp < limit && ibp[0] == '\\' && ibp[1] == '\n') { | |
2527 | ++ip->lineno; | |
2528 | ibp += 2; | |
2529 | } | |
2530 | c = *ibp++; | |
2531 | /* ".." terminates a preprocessing number. This is useless for C | |
2532 | code but useful for preprocessing other things. */ | |
2533 | if (!isalnum (c) && (c != '.' || *ibp == '.') && c != '_') { | |
2534 | --ibp; | |
2535 | break; | |
2536 | } | |
2537 | *obp++ = c; | |
2538 | /* A sign can be part of a preprocessing number | |
2539 | if it follows an e. */ | |
2540 | if (c == 'e' || c == 'E') { | |
2541 | while (ibp < limit && ibp[0] == '\\' && ibp[1] == '\n') { | |
2542 | ++ip->lineno; | |
2543 | ibp += 2; | |
2544 | } | |
2545 | if (ibp < limit && (*ibp == '+' || *ibp == '-')) { | |
2546 | *obp++ = *ibp++; | |
2547 | /* But traditional C does not let the token go past the sign. */ | |
2548 | if (traditional) | |
2549 | break; | |
2550 | } | |
2551 | } | |
2552 | } | |
2553 | break; | |
2554 | } | |
2555 | /* fall through */ | |
2556 | ||
2557 | case '_': | |
2558 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2559 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2560 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
2561 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
2562 | case 'y': case 'z': | |
2563 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
2564 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
2565 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
2566 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
2567 | case 'Y': case 'Z': | |
2568 | letter: | |
2569 | ident_length++; | |
2570 | /* Compute step of hash function, to avoid a proc call on every token */ | |
2571 | hash = HASHSTEP (hash, c); | |
2572 | break; | |
2573 | ||
2574 | case '\n': | |
2575 | /* If reprocessing a macro expansion, newline is a special marker. */ | |
2576 | if (ip->macro != 0) { | |
2577 | /* Newline White is a "funny space" to separate tokens that are | |
2578 | supposed to be separate but without space between. | |
2579 | Here White means any whitespace character. | |
2580 | Newline - marks a recursive macro use that is not | |
2581 | supposed to be expandable. */ | |
2582 | ||
2583 | if (*ibp == '-') { | |
2584 | /* Newline - inhibits expansion of preceding token. | |
2585 | If expanding a macro arg, we keep the newline -. | |
2586 | In final output, it is deleted. */ | |
2587 | if (! concatenated) { | |
2588 | ident_length = 0; | |
2589 | hash = 0; | |
2590 | } | |
2591 | ibp++; | |
2592 | if (!output_marks) { | |
2593 | obp--; | |
2594 | } else { | |
2595 | /* If expanding a macro arg, keep the newline -. */ | |
2596 | *obp++ = '-'; | |
2597 | } | |
2598 | } else if (is_space[*ibp]) { | |
2599 | /* Newline Space does not prevent expansion of preceding token | |
2600 | so expand the preceding token and then come back. */ | |
2601 | if (ident_length > 0) | |
2602 | goto specialchar; | |
2603 | ||
2604 | /* If generating final output, newline space makes a space. */ | |
2605 | if (!output_marks) { | |
2606 | obp[-1] = *ibp++; | |
2607 | /* And Newline Newline makes a newline, so count it. */ | |
2608 | if (obp[-1] == '\n') | |
2609 | op->lineno++; | |
2610 | } else { | |
2611 | /* If expanding a macro arg, keep the newline space. | |
2612 | If the arg gets stringified, newline space makes nothing. */ | |
2613 | *obp++ = *ibp++; | |
2614 | } | |
2615 | } else abort (); /* Newline followed by something random? */ | |
2616 | break; | |
2617 | } | |
2618 | ||
2619 | /* If there is a pending identifier, handle it and come back here. */ | |
2620 | if (ident_length > 0) | |
2621 | goto specialchar; | |
2622 | ||
2623 | beg_of_line = ibp; | |
2624 | ||
2625 | /* Update the line counts and output a #line if necessary. */ | |
2626 | ++ip->lineno; | |
2627 | ++op->lineno; | |
2628 | if (ip->lineno != op->lineno) { | |
2629 | op->bufp = obp; | |
2630 | output_line_command (ip, op, 1, same_file); | |
2631 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
2632 | obp = op->bufp; | |
2633 | } | |
2634 | break; | |
2635 | ||
2636 | /* Come here either after (1) a null character that is part of the input | |
2637 | or (2) at the end of the input, because there is a null there. */ | |
2638 | case 0: | |
2639 | if (ibp <= limit) | |
2640 | /* Our input really contains a null character. */ | |
2641 | goto randomchar; | |
2642 | ||
2643 | /* At end of a macro-expansion level, pop it and read next level. */ | |
2644 | if (ip->macro != 0) { | |
2645 | obp--; | |
2646 | ibp--; | |
2647 | /* If traditional, and we have an identifier that ends here, | |
2648 | process it now, so we get the right error for recursion. */ | |
2649 | if (traditional && ident_length | |
2650 | && ! is_idchar[*instack[indepth - 1].bufp]) { | |
2651 | redo_char = 1; | |
2652 | goto randomchar; | |
2653 | } | |
2654 | POPMACRO; | |
2655 | RECACHE; | |
2656 | break; | |
2657 | } | |
2658 | ||
2659 | /* If we don't have a pending identifier, | |
2660 | return at end of input. */ | |
2661 | if (ident_length == 0) { | |
2662 | obp--; | |
2663 | ibp--; | |
2664 | op->bufp = obp; | |
2665 | ip->bufp = ibp; | |
2666 | goto ending; | |
2667 | } | |
2668 | ||
2669 | /* If we do have a pending identifier, just consider this null | |
2670 | a special character and arrange to dispatch on it again. | |
2671 | The second time, IDENT_LENGTH will be zero so we will return. */ | |
2672 | ||
2673 | /* Fall through */ | |
2674 | ||
2675 | specialchar: | |
2676 | ||
2677 | /* Handle the case of a character such as /, ', " or null | |
2678 | seen following an identifier. Back over it so that | |
2679 | after the identifier is processed the special char | |
2680 | will be dispatched on again. */ | |
2681 | ||
2682 | ibp--; | |
2683 | obp--; | |
2684 | redo_char = 1; | |
2685 | ||
2686 | default: | |
2687 | ||
2688 | randomchar: | |
2689 | ||
2690 | if (ident_length > 0) { | |
2691 | register HASHNODE *hp; | |
2692 | ||
2693 | /* We have just seen an identifier end. If it's a macro, expand it. | |
2694 | ||
2695 | IDENT_LENGTH is the length of the identifier | |
2696 | and HASH is its hash code. | |
2697 | ||
2698 | The identifier has already been copied to the output, | |
2699 | so if it is a macro we must remove it. | |
2700 | ||
2701 | If REDO_CHAR is 0, the char that terminated the identifier | |
2702 | has been skipped in the output and the input. | |
2703 | OBP-IDENT_LENGTH-1 points to the identifier. | |
2704 | If the identifier is a macro, we must back over the terminator. | |
2705 | ||
2706 | If REDO_CHAR is 1, the terminating char has already been | |
2707 | backed over. OBP-IDENT_LENGTH points to the identifier. */ | |
2708 | ||
2709 | if (!pcp_outfile || pcp_inside_if) { | |
2710 | startagain: | |
2711 | for (hp = hashtab[MAKE_POS (hash) % HASHSIZE]; hp != NULL; | |
2712 | hp = hp->next) { | |
2713 | ||
2714 | if (hp->length == ident_length) { | |
2715 | int obufp_before_macroname; | |
2716 | int op_lineno_before_macroname; | |
2717 | register int i = ident_length; | |
2718 | register U_CHAR *p = hp->name; | |
2719 | register U_CHAR *q = obp - i; | |
2720 | int disabled; | |
2721 | ||
2722 | if (! redo_char) | |
2723 | q--; | |
2724 | ||
2725 | do { /* All this to avoid a strncmp () */ | |
2726 | if (*p++ != *q++) | |
2727 | goto hashcollision; | |
2728 | } while (--i); | |
2729 | ||
2730 | /* We found a use of a macro name. | |
2731 | see if the context shows it is a macro call. */ | |
2732 | ||
2733 | /* Back up over terminating character if not already done. */ | |
2734 | if (! redo_char) { | |
2735 | ibp--; | |
2736 | obp--; | |
2737 | } | |
2738 | ||
2739 | /* Save this as a displacement from the beginning of the output | |
2740 | buffer. We can not save this as a position in the output | |
2741 | buffer, because it may get realloc'ed by RECACHE. */ | |
2742 | obufp_before_macroname = (obp - op->buf) - ident_length; | |
2743 | op_lineno_before_macroname = op->lineno; | |
2744 | ||
2745 | if (hp->type == T_PCSTRING) { | |
2746 | pcstring_used (hp); /* Mark the definition of this key | |
2747 | as needed, ensuring that it | |
2748 | will be output. */ | |
2749 | break; /* Exit loop, since the key cannot have a | |
2750 | definition any longer. */ | |
2751 | } | |
2752 | ||
2753 | /* Record whether the macro is disabled. */ | |
2754 | disabled = hp->type == T_DISABLED; | |
2755 | ||
2756 | /* This looks like a macro ref, but if the macro was disabled, | |
2757 | just copy its name and put in a marker if requested. */ | |
2758 | ||
2759 | if (disabled) { | |
2760 | #if 0 | |
2761 | /* This error check caught useful cases such as | |
2762 | #define foo(x,y) bar(x(y,0), y) | |
2763 | foo(foo, baz) */ | |
2764 | if (traditional) | |
2765 | error ("recursive use of macro `%s'", hp->name); | |
2766 | #endif | |
2767 | ||
2768 | if (output_marks) { | |
2769 | check_expand (op, limit - ibp + 2); | |
2770 | *obp++ = '\n'; | |
2771 | *obp++ = '-'; | |
2772 | } | |
2773 | break; | |
2774 | } | |
2775 | ||
2776 | /* If macro wants an arglist, verify that a '(' follows. | |
2777 | first skip all whitespace, copying it to the output | |
2778 | after the macro name. Then, if there is no '(', | |
2779 | decide this is not a macro call and leave things that way. */ | |
2780 | if ((hp->type == T_MACRO || hp->type == T_DISABLED) | |
2781 | && hp->value.defn->nargs >= 0) | |
2782 | { | |
2783 | U_CHAR *old_ibp = ibp; | |
2784 | U_CHAR *old_obp = obp; | |
2785 | int old_iln = ip->lineno; | |
2786 | int old_oln = op->lineno; | |
2787 | ||
2788 | while (1) { | |
2789 | /* Scan forward over whitespace, copying it to the output. */ | |
2790 | if (ibp == limit && ip->macro != 0) { | |
2791 | POPMACRO; | |
2792 | RECACHE; | |
2793 | old_ibp = ibp; | |
2794 | old_obp = obp; | |
2795 | old_iln = ip->lineno; | |
2796 | old_oln = op->lineno; | |
2797 | } | |
2798 | /* A comment: copy it unchanged or discard it. */ | |
2799 | else if (*ibp == '/' && ibp+1 != limit && ibp[1] == '*') { | |
2800 | if (put_out_comments) { | |
2801 | *obp++ = '/'; | |
2802 | *obp++ = '*'; | |
2803 | } else if (! traditional) { | |
2804 | *obp++ = ' '; | |
2805 | } | |
2806 | ibp += 2; | |
2807 | while (ibp + 1 != limit | |
2808 | && !(ibp[0] == '*' && ibp[1] == '/')) { | |
2809 | /* We need not worry about newline-marks, | |
2810 | since they are never found in comments. */ | |
2811 | if (*ibp == '\n') { | |
2812 | /* Newline in a file. Count it. */ | |
2813 | ++ip->lineno; | |
2814 | ++op->lineno; | |
2815 | } | |
2816 | if (put_out_comments) | |
2817 | *obp++ = *ibp++; | |
2818 | else | |
2819 | ibp++; | |
2820 | } | |
2821 | ibp += 2; | |
2822 | if (put_out_comments) { | |
2823 | *obp++ = '*'; | |
2824 | *obp++ = '/'; | |
2825 | } | |
2826 | } | |
2827 | else if (is_space[*ibp]) { | |
2828 | *obp++ = *ibp++; | |
2829 | if (ibp[-1] == '\n') { | |
2830 | if (ip->macro == 0) { | |
2831 | /* Newline in a file. Count it. */ | |
2832 | ++ip->lineno; | |
2833 | ++op->lineno; | |
2834 | } else if (!output_marks) { | |
2835 | /* A newline mark, and we don't want marks | |
2836 | in the output. If it is newline-hyphen, | |
2837 | discard it entirely. Otherwise, it is | |
2838 | newline-whitechar, so keep the whitechar. */ | |
2839 | obp--; | |
2840 | if (*ibp == '-') | |
2841 | ibp++; | |
2842 | else { | |
2843 | if (*ibp == '\n') | |
2844 | ++op->lineno; | |
2845 | *obp++ = *ibp++; | |
2846 | } | |
2847 | } else { | |
2848 | /* A newline mark; copy both chars to the output. */ | |
2849 | *obp++ = *ibp++; | |
2850 | } | |
2851 | } | |
2852 | } | |
2853 | else break; | |
2854 | } | |
2855 | if (*ibp != '(') { | |
2856 | /* It isn't a macro call. | |
2857 | Put back the space that we just skipped. */ | |
2858 | ibp = old_ibp; | |
2859 | obp = old_obp; | |
2860 | ip->lineno = old_iln; | |
2861 | op->lineno = old_oln; | |
2862 | /* Exit the for loop. */ | |
2863 | break; | |
2864 | } | |
2865 | } | |
2866 | ||
2867 | /* This is now known to be a macro call. | |
2868 | Discard the macro name from the output, | |
2869 | along with any following whitespace just copied. */ | |
2870 | obp = op->buf + obufp_before_macroname; | |
2871 | op->lineno = op_lineno_before_macroname; | |
2872 | ||
2873 | /* Expand the macro, reading arguments as needed, | |
2874 | and push the expansion on the input stack. */ | |
2875 | ip->bufp = ibp; | |
2876 | op->bufp = obp; | |
2877 | macroexpand (hp, op); | |
2878 | ||
2879 | /* Reexamine input stack, since macroexpand has pushed | |
2880 | a new level on it. */ | |
2881 | obp = op->bufp; | |
2882 | RECACHE; | |
2883 | break; | |
2884 | } | |
2885 | hashcollision: | |
2886 | ; | |
2887 | } /* End hash-table-search loop */ | |
2888 | } | |
2889 | ident_length = hash = 0; /* Stop collecting identifier */ | |
2890 | redo_char = 0; | |
2891 | concatenated = 0; | |
2892 | } /* End if (ident_length > 0) */ | |
2893 | } /* End switch */ | |
2894 | } /* End per-char loop */ | |
2895 | ||
2896 | /* Come here to return -- but first give an error message | |
2897 | if there was an unterminated successful conditional. */ | |
2898 | ending: | |
2899 | if (if_stack != ip->if_stack) { | |
2900 | char *str; | |
2901 | switch (if_stack->type) { | |
2902 | case T_IF: | |
2903 | str = "if"; | |
2904 | break; | |
2905 | case T_IFDEF: | |
2906 | str = "ifdef"; | |
2907 | break; | |
2908 | case T_IFNDEF: | |
2909 | str = "ifndef"; | |
2910 | break; | |
2911 | case T_ELSE: | |
2912 | str = "else"; | |
2913 | break; | |
2914 | case T_ELIF: | |
2915 | str = "elif"; | |
2916 | break; | |
2917 | } | |
2918 | error_with_line (line_for_error (if_stack->lineno), | |
2919 | "unterminated `#%s' conditional", str); | |
2920 | } | |
2921 | if_stack = ip->if_stack; | |
2922 | } | |
2923 | \f | |
2924 | /* | |
2925 | * Rescan a string into a temporary buffer and return the result | |
2926 | * as a FILE_BUF. Note this function returns a struct, not a pointer. | |
2927 | * | |
2928 | * OUTPUT_MARKS nonzero means keep Newline markers found in the input | |
2929 | * and insert such markers when appropriate. See `rescan' for details. | |
2930 | * OUTPUT_MARKS is 1 for macroexpanding a macro argument separately | |
2931 | * before substitution; it is 0 for other uses. | |
2932 | */ | |
2933 | static FILE_BUF | |
2934 | expand_to_temp_buffer (buf, limit, output_marks, assertions) | |
2935 | U_CHAR *buf, *limit; | |
2936 | int output_marks, assertions; | |
2937 | { | |
2938 | register FILE_BUF *ip; | |
2939 | FILE_BUF obuf; | |
2940 | int length = limit - buf; | |
2941 | U_CHAR *buf1; | |
2942 | int odepth = indepth; | |
2943 | int save_assertions_flag = assertions_flag; | |
2944 | ||
2945 | assertions_flag = assertions; | |
2946 | ||
2947 | if (length < 0) | |
2948 | abort (); | |
2949 | ||
2950 | /* Set up the input on the input stack. */ | |
2951 | ||
2952 | buf1 = (U_CHAR *) alloca (length + 1); | |
2953 | { | |
2954 | register U_CHAR *p1 = buf; | |
2955 | register U_CHAR *p2 = buf1; | |
2956 | ||
2957 | while (p1 != limit) | |
2958 | *p2++ = *p1++; | |
2959 | } | |
2960 | buf1[length] = 0; | |
2961 | ||
2962 | /* Set up to receive the output. */ | |
2963 | ||
2964 | obuf.length = length * 2 + 100; /* Usually enough. Why be stingy? */ | |
2965 | obuf.bufp = obuf.buf = (U_CHAR *) xmalloc (obuf.length); | |
2966 | obuf.fname = 0; | |
2967 | obuf.macro = 0; | |
2968 | obuf.free_ptr = 0; | |
2969 | ||
2970 | CHECK_DEPTH ({return obuf;}); | |
2971 | ||
2972 | ++indepth; | |
2973 | ||
2974 | ip = &instack[indepth]; | |
2975 | ip->fname = 0; | |
2976 | ip->nominal_fname = 0; | |
2977 | ip->system_header_p = 0; | |
2978 | ip->macro = 0; | |
2979 | ip->free_ptr = 0; | |
2980 | ip->length = length; | |
2981 | ip->buf = ip->bufp = buf1; | |
2982 | ip->if_stack = if_stack; | |
2983 | ||
2984 | ip->lineno = obuf.lineno = 1; | |
2985 | ||
2986 | /* Scan the input, create the output. */ | |
2987 | rescan (&obuf, output_marks); | |
2988 | ||
2989 | /* Pop input stack to original state. */ | |
2990 | --indepth; | |
2991 | ||
2992 | if (indepth != odepth) | |
2993 | abort (); | |
2994 | ||
2995 | /* Record the output. */ | |
2996 | obuf.length = obuf.bufp - obuf.buf; | |
2997 | ||
2998 | assertions_flag = save_assertions_flag; | |
2999 | return obuf; | |
3000 | } | |
3001 | \f | |
3002 | /* | |
3003 | * Process a # directive. Expects IP->bufp to point after the '#', as in | |
3004 | * `#define foo bar'. Passes to the command handler | |
3005 | * (do_define, do_include, etc.): the addresses of the 1st and | |
3006 | * last chars of the command (starting immediately after the # | |
3007 | * keyword), plus op and the keyword table pointer. If the command | |
3008 | * contains comments it is copied into a temporary buffer sans comments | |
3009 | * and the temporary buffer is passed to the command handler instead. | |
3010 | * Likewise for backslash-newlines. | |
3011 | * | |
3012 | * Returns nonzero if this was a known # directive. | |
3013 | * Otherwise, returns zero, without advancing the input pointer. | |
3014 | */ | |
3015 | ||
3016 | static int | |
3017 | handle_directive (ip, op) | |
3018 | FILE_BUF *ip, *op; | |
3019 | { | |
3020 | register U_CHAR *bp, *cp; | |
3021 | register struct directive *kt; | |
3022 | register int ident_length; | |
3023 | U_CHAR *resume_p; | |
3024 | ||
3025 | /* Nonzero means we must copy the entire command | |
3026 | to get rid of comments or backslash-newlines. */ | |
3027 | int copy_command = 0; | |
3028 | ||
3029 | U_CHAR *ident, *after_ident; | |
3030 | ||
3031 | bp = ip->bufp; | |
3032 | ||
3033 | /* Record where the directive started. do_xifdef needs this. */ | |
3034 | directive_start = bp - 1; | |
3035 | ||
3036 | /* Skip whitespace and \-newline. */ | |
3037 | while (1) { | |
3038 | if (is_hor_space[*bp]) { | |
3039 | if ((*bp == '\f' || *bp == '\v') && pedantic) | |
3040 | pedwarn ("%s in preprocessing directive", | |
3041 | *bp == '\f' ? "formfeed" : "vertical tab"); | |
3042 | bp++; | |
3043 | } else if (*bp == '/' && bp[1] == '*') { | |
3044 | ip->bufp = bp; | |
3045 | skip_to_end_of_comment (ip, &ip->lineno); | |
3046 | bp = ip->bufp; | |
3047 | } else if (*bp == '\\' && bp[1] == '\n') { | |
3048 | bp += 2; ip->lineno++; | |
3049 | } else break; | |
3050 | } | |
3051 | ||
3052 | /* Now find end of directive name. | |
3053 | If we encounter a backslash-newline, exchange it with any following | |
3054 | symbol-constituents so that we end up with a contiguous name. */ | |
3055 | ||
3056 | cp = bp; | |
3057 | while (1) { | |
3058 | if (is_idchar[*cp]) | |
3059 | cp++; | |
3060 | else { | |
3061 | if (*cp == '\\' && cp[1] == '\n') | |
3062 | name_newline_fix (cp); | |
3063 | if (is_idchar[*cp]) | |
3064 | cp++; | |
3065 | else break; | |
3066 | } | |
3067 | } | |
3068 | ident_length = cp - bp; | |
3069 | ident = bp; | |
3070 | after_ident = cp; | |
3071 | ||
3072 | /* A line of just `#' becomes blank. */ | |
3073 | ||
3074 | if (ident_length == 0 && *after_ident == '\n') { | |
3075 | ip->bufp = after_ident; | |
3076 | return 1; | |
3077 | } | |
3078 | ||
3079 | if (ident_length == 0 || !is_idstart[*ident]) { | |
3080 | U_CHAR *p = ident; | |
3081 | while (is_idchar[*p]) { | |
3082 | if (*p < '0' || *p > '9') | |
3083 | break; | |
3084 | p++; | |
3085 | } | |
3086 | /* Handle # followed by a line number. */ | |
3087 | if (p != ident && !is_idchar[*p]) { | |
3088 | static struct directive line_directive_table[] = { | |
3089 | { 4, do_line, "line", T_LINE}, | |
3090 | }; | |
3091 | if (pedantic) | |
3092 | pedwarn ("`#' followed by integer"); | |
3093 | after_ident = ident; | |
3094 | kt = line_directive_table; | |
3095 | goto old_linenum; | |
3096 | } | |
3097 | ||
3098 | /* Avoid error for `###' and similar cases unless -pedantic. */ | |
3099 | if (p == ident) { | |
3100 | while (*p == '#' || is_hor_space[*p]) p++; | |
3101 | if (*p == '\n') { | |
3102 | if (pedantic && !lang_asm) | |
3103 | warning ("invalid preprocessor directive"); | |
3104 | return 0; | |
3105 | } | |
3106 | } | |
3107 | ||
3108 | if (!lang_asm) | |
3109 | error ("invalid preprocessor directive name"); | |
3110 | ||
3111 | return 0; | |
3112 | } | |
3113 | ||
3114 | /* | |
3115 | * Decode the keyword and call the appropriate expansion | |
3116 | * routine, after moving the input pointer up to the next line. | |
3117 | */ | |
3118 | for (kt = directive_table; kt->length > 0; kt++) { | |
3119 | if (kt->length == ident_length && !strncmp (kt->name, ident, ident_length)) { | |
3120 | register U_CHAR *buf; | |
3121 | register U_CHAR *limit; | |
3122 | int unterminated; | |
3123 | int junk; | |
3124 | int *already_output = 0; | |
3125 | ||
3126 | /* Nonzero means do not delete comments within the directive. | |
3127 | #define needs this when -traditional. */ | |
3128 | int keep_comments; | |
3129 | ||
3130 | old_linenum: | |
3131 | ||
3132 | limit = ip->buf + ip->length; | |
3133 | unterminated = 0; | |
3134 | keep_comments = traditional && kt->traditional_comments; | |
3135 | /* #import is defined only in Objective C, or when on the NeXT. */ | |
3136 | if (kt->type == T_IMPORT && !(objc || lookup ("__NeXT__", -1, -1))) | |
3137 | break; | |
3138 | ||
3139 | /* Find the end of this command (first newline not backslashed | |
3140 | and not in a string or comment). | |
3141 | Set COPY_COMMAND if the command must be copied | |
3142 | (it contains a backslash-newline or a comment). */ | |
3143 | ||
3144 | buf = bp = after_ident; | |
3145 | while (bp < limit) { | |
3146 | register U_CHAR c = *bp++; | |
3147 | switch (c) { | |
3148 | case '\\': | |
3149 | if (bp < limit) { | |
3150 | if (*bp == '\n') { | |
3151 | ip->lineno++; | |
3152 | copy_command = 1; | |
3153 | } | |
3154 | bp++; | |
3155 | } | |
3156 | break; | |
3157 | ||
3158 | case '\'': | |
3159 | case '\"': | |
3160 | bp = skip_quoted_string (bp - 1, limit, ip->lineno, &ip->lineno, ©_command, &unterminated); | |
3161 | /* Don't bother calling the directive if we already got an error | |
3162 | message due to unterminated string. Skip everything and pretend | |
3163 | we called the directive. */ | |
3164 | if (unterminated) { | |
3165 | if (traditional) { | |
3166 | /* Traditional preprocessing permits unterminated strings. */ | |
3167 | ip->bufp = bp; | |
3168 | goto endloop1; | |
3169 | } | |
3170 | ip->bufp = bp; | |
3171 | return 1; | |
3172 | } | |
3173 | break; | |
3174 | ||
3175 | /* <...> is special for #include. */ | |
3176 | case '<': | |
3177 | if (!kt->angle_brackets) | |
3178 | break; | |
3179 | while (*bp && *bp != '>') bp++; | |
3180 | break; | |
3181 | ||
3182 | case '/': | |
3183 | if (*bp == '\\' && bp[1] == '\n') | |
3184 | newline_fix (bp); | |
3185 | if (*bp == '*' | |
3186 | || ((cplusplus || objc) && *bp == '/')) { | |
3187 | U_CHAR *obp = bp - 1; | |
3188 | ip->bufp = bp + 1; | |
3189 | skip_to_end_of_comment (ip, &ip->lineno); | |
3190 | bp = ip->bufp; | |
3191 | /* No need to copy the command because of a comment at the end; | |
3192 | just don't include the comment in the directive. */ | |
3193 | if (bp == limit || *bp == '\n') { | |
3194 | bp = obp; | |
3195 | goto endloop1; | |
3196 | } | |
3197 | /* Don't remove the comments if -traditional. */ | |
3198 | if (! keep_comments) | |
3199 | copy_command++; | |
3200 | } | |
3201 | break; | |
3202 | ||
3203 | case '\f': | |
3204 | case '\v': | |
3205 | if (pedantic) | |
3206 | pedwarn ("%s in preprocessing directive", | |
3207 | c == '\f' ? "formfeed" : "vertical tab"); | |
3208 | break; | |
3209 | ||
3210 | case '\n': | |
3211 | --bp; /* Point to the newline */ | |
3212 | ip->bufp = bp; | |
3213 | goto endloop1; | |
3214 | } | |
3215 | } | |
3216 | ip->bufp = bp; | |
3217 | ||
3218 | endloop1: | |
3219 | resume_p = ip->bufp; | |
3220 | /* BP is the end of the directive. | |
3221 | RESUME_P is the next interesting data after the directive. | |
3222 | A comment may come between. */ | |
3223 | ||
3224 | /* If a directive should be copied through, and -E was given, | |
3225 | pass it through before removing comments. */ | |
3226 | if (!no_output && kt->pass_thru && put_out_comments) { | |
3227 | int len; | |
3228 | ||
3229 | /* Output directive name. */ | |
3230 | check_expand (op, kt->length + 2); | |
3231 | /* Make sure # is at the start of a line */ | |
3232 | if (op->bufp > op->buf && op->bufp[-1] != '\n') { | |
3233 | op->lineno++; | |
3234 | *op->bufp++ = '\n'; | |
3235 | } | |
3236 | *op->bufp++ = '#'; | |
3237 | bcopy (kt->name, op->bufp, kt->length); | |
3238 | op->bufp += kt->length; | |
3239 | ||
3240 | /* Output arguments. */ | |
3241 | len = (bp - buf); | |
3242 | check_expand (op, len); | |
3243 | bcopy (buf, op->bufp, len); | |
3244 | op->bufp += len; | |
3245 | /* Take account of any (escaped) newlines just output. */ | |
3246 | while (--len >= 0) | |
3247 | if (buf[len] == '\n') | |
3248 | op->lineno++; | |
3249 | ||
3250 | already_output = &junk; | |
3251 | } /* Don't we need a newline or #line? */ | |
3252 | ||
3253 | if (copy_command) { | |
3254 | register U_CHAR *xp = buf; | |
3255 | /* Need to copy entire command into temp buffer before dispatching */ | |
3256 | ||
3257 | cp = (U_CHAR *) alloca (bp - buf + 5); /* room for cmd plus | |
3258 | some slop */ | |
3259 | buf = cp; | |
3260 | ||
3261 | /* Copy to the new buffer, deleting comments | |
3262 | and backslash-newlines (and whitespace surrounding the latter). */ | |
3263 | ||
3264 | while (xp < bp) { | |
3265 | register U_CHAR c = *xp++; | |
3266 | *cp++ = c; | |
3267 | ||
3268 | switch (c) { | |
3269 | case '\n': | |
3270 | abort (); /* A bare newline should never part of the line. */ | |
3271 | break; | |
3272 | ||
3273 | /* <...> is special for #include. */ | |
3274 | case '<': | |
3275 | if (!kt->angle_brackets) | |
3276 | break; | |
3277 | while (xp < bp && c != '>') { | |
3278 | c = *xp++; | |
3279 | if (c == '\\' && xp < bp && *xp == '\n') | |
3280 | xp++; | |
3281 | else | |
3282 | *cp++ = c; | |
3283 | } | |
3284 | break; | |
3285 | ||
3286 | case '\\': | |
3287 | if (*xp == '\n') { | |
3288 | xp++; | |
3289 | cp--; | |
3290 | if (cp != buf && is_space[cp[-1]]) { | |
3291 | while (cp != buf && is_space[cp[-1]]) cp--; | |
3292 | cp++; | |
3293 | SKIP_WHITE_SPACE (xp); | |
3294 | } else if (is_space[*xp]) { | |
3295 | *cp++ = *xp++; | |
3296 | SKIP_WHITE_SPACE (xp); | |
3297 | } | |
3298 | } | |
3299 | break; | |
3300 | ||
3301 | case '\'': | |
3302 | case '\"': | |
3303 | { | |
3304 | register U_CHAR *bp1 | |
3305 | = skip_quoted_string (xp - 1, limit, ip->lineno, 0, 0, 0); | |
3306 | while (xp != bp1) | |
3307 | if (*xp == '\\') { | |
3308 | if (*++xp != '\n') | |
3309 | *cp++ = '\\'; | |
3310 | else | |
3311 | xp++; | |
3312 | } else | |
3313 | *cp++ = *xp++; | |
3314 | } | |
3315 | break; | |
3316 | ||
3317 | case '/': | |
3318 | if (*xp == '*' | |
3319 | || ((cplusplus || objc) && *xp == '/')) { | |
3320 | ip->bufp = xp + 1; | |
3321 | /* If we already copied the command through, | |
3322 | already_output != 0 prevents outputting comment now. */ | |
3323 | skip_to_end_of_comment (ip, already_output); | |
3324 | if (keep_comments) | |
3325 | while (xp != ip->bufp) | |
3326 | *cp++ = *xp++; | |
3327 | /* Delete or replace the slash. */ | |
3328 | else if (traditional) | |
3329 | cp--; | |
3330 | else | |
3331 | cp[-1] = ' '; | |
3332 | xp = ip->bufp; | |
3333 | } | |
3334 | } | |
3335 | } | |
3336 | ||
3337 | /* Null-terminate the copy. */ | |
3338 | ||
3339 | *cp = 0; | |
3340 | } else | |
3341 | cp = bp; | |
3342 | ||
3343 | ip->bufp = resume_p; | |
3344 | ||
3345 | /* Some directives should be written out for cc1 to process, | |
3346 | just as if they were not defined. And sometimes we're copying | |
3347 | definitions through. */ | |
3348 | ||
3349 | if (!no_output && already_output == 0 | |
3350 | && (kt->pass_thru | |
3351 | || (kt->type == T_DEFINE | |
3352 | && (dump_macros == dump_names | |
3353 | || dump_macros == dump_definitions)))) { | |
3354 | int len; | |
3355 | ||
3356 | /* Output directive name. */ | |
3357 | check_expand (op, kt->length + 1); | |
3358 | *op->bufp++ = '#'; | |
3359 | bcopy (kt->name, op->bufp, kt->length); | |
3360 | op->bufp += kt->length; | |
3361 | ||
3362 | if (kt->pass_thru || dump_macros == dump_definitions) { | |
3363 | /* Output arguments. */ | |
3364 | len = (cp - buf); | |
3365 | check_expand (op, len); | |
3366 | bcopy (buf, op->bufp, len); | |
3367 | op->bufp += len; | |
3368 | } | |
3369 | } /* Don't we need a newline or #line? */ | |
3370 | ||
3371 | /* Call the appropriate command handler. buf now points to | |
3372 | either the appropriate place in the input buffer, or to | |
3373 | the temp buffer if it was necessary to make one. cp | |
3374 | points to the first char after the contents of the (possibly | |
3375 | copied) command, in either case. */ | |
3376 | (*kt->func) (buf, cp, op, kt); | |
3377 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
3378 | ||
3379 | return 1; | |
3380 | } | |
3381 | } | |
3382 | ||
3383 | /* It is deliberate that we don't warn about undefined directives. | |
3384 | That is the responsibility of cc1. */ | |
3385 | return 0; | |
3386 | } | |
3387 | \f | |
3388 | static char *monthnames[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
3389 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
3390 | }; | |
3391 | ||
3392 | /* | |
3393 | * expand things like __FILE__. Place the expansion into the output | |
3394 | * buffer *without* rescanning. | |
3395 | */ | |
3396 | ||
3397 | static void | |
3398 | special_symbol (hp, op) | |
3399 | HASHNODE *hp; | |
3400 | FILE_BUF *op; | |
3401 | { | |
3402 | char *buf; | |
3403 | int i, len; | |
3404 | int true_indepth; | |
3405 | FILE_BUF *ip = NULL; | |
3406 | ||
3407 | int paren = 0; /* For special `defined' keyword */ | |
3408 | ||
3409 | if (pcp_outfile && pcp_inside_if | |
3410 | && hp->type != T_SPEC_DEFINED && hp->type != T_CONST) | |
3411 | error ("Predefined macro `%s' used inside `#if' during precompilation", | |
3412 | hp->name); | |
3413 | ||
3414 | for (i = indepth; i >= 0; i--) | |
3415 | if (instack[i].fname != NULL) { | |
3416 | ip = &instack[i]; | |
3417 | break; | |
3418 | } | |
3419 | if (ip == NULL) { | |
3420 | error ("cccp error: not in any file?!"); | |
3421 | return; /* the show must go on */ | |
3422 | } | |
3423 | ||
3424 | switch (hp->type) { | |
3425 | case T_FILE: | |
3426 | case T_BASE_FILE: | |
3427 | { | |
3428 | char *string; | |
3429 | if (hp->type == T_FILE) | |
3430 | string = ip->nominal_fname; | |
3431 | else | |
3432 | string = instack[0].nominal_fname; | |
3433 | ||
3434 | if (string) | |
3435 | { | |
3436 | buf = (char *) alloca (3 + strlen (string)); | |
3437 | sprintf (buf, "\"%s\"", string); | |
3438 | } | |
3439 | else | |
3440 | buf = "\"\""; | |
3441 | ||
3442 | break; | |
3443 | } | |
3444 | ||
3445 | case T_INCLUDE_LEVEL: | |
3446 | true_indepth = 0; | |
3447 | for (i = indepth; i >= 0; i--) | |
3448 | if (instack[i].fname != NULL) | |
3449 | true_indepth++; | |
3450 | ||
3451 | buf = (char *) alloca (8); /* Eight bytes ought to be more than enough */ | |
3452 | sprintf (buf, "%d", true_indepth - 1); | |
3453 | break; | |
3454 | ||
3455 | case T_VERSION: | |
3456 | buf = (char *) alloca (3 + strlen (version_string)); | |
3457 | sprintf (buf, "\"%s\"", version_string); | |
3458 | break; | |
3459 | ||
3460 | case T_SIZE_TYPE: | |
3461 | buf = (char *) alloca (3 + strlen (SIZE_TYPE)); | |
3462 | sprintf (buf, "%s", SIZE_TYPE); | |
3463 | break; | |
3464 | ||
3465 | case T_PTRDIFF_TYPE: | |
3466 | buf = (char *) alloca (3 + strlen (PTRDIFF_TYPE)); | |
3467 | sprintf (buf, "%s", PTRDIFF_TYPE); | |
3468 | break; | |
3469 | ||
3470 | case T_WCHAR_TYPE: | |
3471 | buf = (char *) alloca (3 + strlen (WCHAR_TYPE)); | |
3472 | sprintf (buf, "%s", WCHAR_TYPE); | |
3473 | break; | |
3474 | ||
3475 | case T_CONST: | |
3476 | buf = (char *) alloca (4 * sizeof (int)); | |
3477 | sprintf (buf, "%d", hp->value.ival); | |
3478 | if (pcp_inside_if && pcp_outfile) | |
3479 | /* Output a precondition for this macro use */ | |
3480 | fprintf (pcp_outfile, "#define %s %d\n", hp->name, hp->value.ival); | |
3481 | break; | |
3482 | ||
3483 | case T_SPECLINE: | |
3484 | buf = (char *) alloca (10); | |
3485 | sprintf (buf, "%d", ip->lineno); | |
3486 | break; | |
3487 | ||
3488 | case T_DATE: | |
3489 | case T_TIME: | |
3490 | buf = (char *) alloca (20); | |
3491 | if (hp->type == T_DATE) | |
3492 | sprintf (buf, "\"%s %2d %4d\"", monthnames[timebuf->tm_mon], | |
3493 | timebuf->tm_mday, timebuf->tm_year + 1900); | |
3494 | else | |
3495 | sprintf (buf, "\"%02d:%02d:%02d\"", timebuf->tm_hour, timebuf->tm_min, | |
3496 | timebuf->tm_sec); | |
3497 | break; | |
3498 | ||
3499 | case T_SPEC_DEFINED: | |
3500 | buf = " 0 "; /* Assume symbol is not defined */ | |
3501 | ip = &instack[indepth]; | |
3502 | SKIP_WHITE_SPACE (ip->bufp); | |
3503 | if (*ip->bufp == '(') { | |
3504 | paren++; | |
3505 | ip->bufp++; /* Skip over the paren */ | |
3506 | SKIP_WHITE_SPACE (ip->bufp); | |
3507 | } | |
3508 | ||
3509 | if (!is_idstart[*ip->bufp]) | |
3510 | goto oops; | |
3511 | if (hp = lookup (ip->bufp, -1, -1)) { | |
3512 | if (pcp_outfile && pcp_inside_if | |
3513 | && hp->value.defn->predefined) | |
3514 | /* Output a precondition for this macro use. */ | |
3515 | fprintf (pcp_outfile, "#define %s\n", hp->name); | |
3516 | buf = " 1 "; | |
3517 | } | |
3518 | else | |
3519 | if (pcp_outfile && pcp_inside_if) { | |
3520 | /* Output a precondition for this macro use */ | |
3521 | U_CHAR *cp = ip->bufp; | |
3522 | fprintf (pcp_outfile, "#undef "); | |
3523 | while (is_idchar[*cp]) /* Ick! */ | |
3524 | fputc (*cp++, pcp_outfile); | |
3525 | putc ('\n', pcp_outfile); | |
3526 | } | |
3527 | while (is_idchar[*ip->bufp]) | |
3528 | ++ip->bufp; | |
3529 | SKIP_WHITE_SPACE (ip->bufp); | |
3530 | if (paren) { | |
3531 | if (*ip->bufp != ')') | |
3532 | goto oops; | |
3533 | ++ip->bufp; | |
3534 | } | |
3535 | break; | |
3536 | ||
3537 | oops: | |
3538 | ||
3539 | error ("`defined' without an identifier"); | |
3540 | break; | |
3541 | ||
3542 | default: | |
3543 | error ("cccp error: invalid special hash type"); /* time for gdb */ | |
3544 | abort (); | |
3545 | } | |
3546 | len = strlen (buf); | |
3547 | check_expand (op, len); | |
3548 | bcopy (buf, op->bufp, len); | |
3549 | op->bufp += len; | |
3550 | ||
3551 | return; | |
3552 | } | |
3553 | ||
3554 | \f | |
3555 | /* Routines to handle #directives */ | |
3556 | ||
3557 | /* Handle #include and #import. | |
3558 | This function expects to see "fname" or <fname> on the input. */ | |
3559 | ||
3560 | static int | |
3561 | do_include (buf, limit, op, keyword) | |
3562 | U_CHAR *buf, *limit; | |
3563 | FILE_BUF *op; | |
3564 | struct directive *keyword; | |
3565 | { | |
3566 | int importing = (keyword->type == T_IMPORT); | |
3567 | int skip_dirs = (keyword->type == T_INCLUDE_NEXT); | |
3568 | static int import_warning = 0; | |
3569 | char *fname; /* Dynamically allocated fname buffer */ | |
3570 | char *pcftry; | |
3571 | char *pcfname; | |
3572 | U_CHAR *fbeg, *fend; /* Beginning and end of fname */ | |
3573 | ||
3574 | struct file_name_list *search_start = include; /* Chain of dirs to search */ | |
3575 | struct file_name_list dsp[1]; /* First in chain, if #include "..." */ | |
3576 | struct file_name_list *searchptr; | |
3577 | int flen; | |
3578 | ||
3579 | int f; /* file number */ | |
3580 | ||
3581 | int retried = 0; /* Have already tried macro | |
3582 | expanding the include line*/ | |
3583 | FILE_BUF trybuf; /* It got expanded into here */ | |
3584 | int system_header_p = 0; /* 0 for "...", 1 for <...> */ | |
3585 | int pcf = -1; | |
3586 | char *pcfbuf; | |
3587 | int pcfbuflimit; | |
3588 | int pcfnum; | |
3589 | f= -1; /* JF we iz paranoid! */ | |
3590 | ||
3591 | if (importing && !instack[indepth].system_header_p && !import_warning) { | |
3592 | import_warning = 1; | |
3593 | warning ("using `#import' is not recommended"); | |
3594 | fprintf (stderr, "The fact that a certain header file need not be processed more than once\n"); | |
3595 | fprintf (stderr, "should be indicated in the header file, not where it is used.\n"); | |
3596 | fprintf (stderr, "The best way to do this is with a conditional of this form:\n\n"); | |
3597 | fprintf (stderr, " #ifndef _FOO_H_INCLUDED\n"); | |
3598 | fprintf (stderr, " #define _FOO_H_INCLUDED\n"); | |
3599 | fprintf (stderr, " ... <real contents of file> ...\n"); | |
3600 | fprintf (stderr, " #endif /* Not _FOO_H_INCLUDED */\n\n"); | |
3601 | fprintf (stderr, "Then users can use `#include' any number of times.\n"); | |
3602 | fprintf (stderr, "GNU C automatically avoids processing the file more than once\n"); | |
3603 | fprintf (stderr, "when it is equipped with such a conditional.\n"); | |
3604 | } | |
3605 | ||
3606 | get_filename: | |
3607 | ||
3608 | fbeg = buf; | |
3609 | SKIP_WHITE_SPACE (fbeg); | |
3610 | /* Discard trailing whitespace so we can easily see | |
3611 | if we have parsed all the significant chars we were given. */ | |
3612 | while (limit != fbeg && is_hor_space[limit[-1]]) limit--; | |
3613 | ||
3614 | switch (*fbeg++) { | |
3615 | case '\"': | |
3616 | fend = fbeg; | |
3617 | while (fend != limit && *fend != '\"') | |
3618 | fend++; | |
3619 | if (*fend == '\"' && fend + 1 == limit) { | |
3620 | FILE_BUF *fp; | |
3621 | ||
3622 | /* We have "filename". Figure out directory this source | |
3623 | file is coming from and put it on the front of the list. */ | |
3624 | ||
3625 | /* If -I- was specified, don't search current dir, only spec'd ones. */ | |
3626 | if (ignore_srcdir) break; | |
3627 | ||
3628 | for (fp = &instack[indepth]; fp >= instack; fp--) | |
3629 | { | |
3630 | int n; | |
3631 | char *ep,*nam; | |
3632 | ||
3633 | if ((nam = fp->nominal_fname) != NULL) { | |
3634 | /* Found a named file. Figure out dir of the file, | |
3635 | and put it in front of the search list. */ | |
3636 | dsp[0].next = search_start; | |
3637 | search_start = dsp; | |
3638 | #ifndef VMS | |
3639 | ep = rindex (nam, '/'); | |
3640 | #else /* VMS */ | |
3641 | ep = rindex (nam, ']'); | |
3642 | if (ep == NULL) ep = rindex (nam, '>'); | |
3643 | if (ep == NULL) ep = rindex (nam, ':'); | |
3644 | if (ep != NULL) ep++; | |
3645 | #endif /* VMS */ | |
3646 | if (ep != NULL) { | |
3647 | n = ep - nam; | |
3648 | dsp[0].fname = (char *) alloca (n + 1); | |
3649 | strncpy (dsp[0].fname, nam, n); | |
3650 | dsp[0].fname[n] = '\0'; | |
3651 | if (n > max_include_len) max_include_len = n; | |
3652 | } else { | |
3653 | dsp[0].fname = 0; /* Current directory */ | |
3654 | } | |
3655 | break; | |
3656 | } | |
3657 | } | |
3658 | break; | |
3659 | } | |
3660 | goto fail; | |
3661 | ||
3662 | case '<': | |
3663 | fend = fbeg; | |
3664 | while (fend != limit && *fend != '>') fend++; | |
3665 | if (*fend == '>' && fend + 1 == limit) { | |
3666 | system_header_p = 1; | |
3667 | /* If -I-, start with the first -I dir after the -I-. */ | |
3668 | if (first_bracket_include) | |
3669 | search_start = first_bracket_include; | |
3670 | break; | |
3671 | } | |
3672 | goto fail; | |
3673 | ||
3674 | default: | |
3675 | fail: | |
3676 | if (retried) { | |
3677 | if (importing) | |
3678 | error ("`#import' expects \"fname\" or <fname>"); | |
3679 | else | |
3680 | error ("`#include' expects \"fname\" or <fname>"); | |
3681 | return 0; | |
3682 | } else { | |
3683 | trybuf = expand_to_temp_buffer (buf, limit, 0, 0); | |
3684 | buf = (U_CHAR *) alloca (trybuf.bufp - trybuf.buf + 1); | |
3685 | bcopy (trybuf.buf, buf, trybuf.bufp - trybuf.buf); | |
3686 | limit = buf + (trybuf.bufp - trybuf.buf); | |
3687 | free (trybuf.buf); | |
3688 | retried++; | |
3689 | goto get_filename; | |
3690 | } | |
3691 | } | |
3692 | ||
3693 | /* For #include_next, skip in the search path | |
3694 | past the dir in which the containing file was found. */ | |
3695 | if (skip_dirs) { | |
3696 | FILE_BUF *fp; | |
3697 | for (fp = &instack[indepth]; fp >= instack; fp--) | |
3698 | if (fp->fname != NULL) { | |
3699 | /* fp->dir is null if the containing file was specified | |
3700 | with an absolute file name. In that case, don't skip anything. */ | |
3701 | if (fp->dir) | |
3702 | search_start = fp->dir->next; | |
3703 | break; | |
3704 | } | |
3705 | } | |
3706 | ||
3707 | flen = fend - fbeg; | |
3708 | /* Allocate this permanently, because it gets stored in the definitions | |
3709 | of macros. */ | |
3710 | fname = (char *) xmalloc (max_include_len + flen + 2); | |
3711 | /* + 2 above for slash and terminating null. */ | |
3712 | ||
3713 | /* See if we already included this file and we can tell in advance | |
3714 | (from a #ifndef around its contents last time) | |
3715 | that there is no need to include it again. */ | |
3716 | { | |
3717 | struct file_name_list *l = all_include_files; | |
3718 | strncpy (fname, fbeg, flen); | |
3719 | fname[flen] = 0; | |
3720 | for (; l; l = l->next) | |
3721 | if (! strcmp (fname, l->fname) | |
3722 | && l->control_macro | |
3723 | && lookup (l->control_macro, -1, -1)) | |
3724 | return 0; | |
3725 | } | |
3726 | ||
3727 | /* If specified file name is absolute, just open it. */ | |
3728 | ||
3729 | if (*fbeg == '/') { | |
3730 | strncpy (fname, fbeg, flen); | |
3731 | fname[flen] = 0; | |
3732 | if (importing) | |
3733 | f = lookup_import (fname); | |
3734 | else | |
3735 | f = open (fname, O_RDONLY, 0666); | |
3736 | if (f == -2) | |
3737 | return 0; /* Already included this file */ | |
3738 | } else { | |
3739 | /* Search directory path, trying to open the file. | |
3740 | Copy each filename tried into FNAME. */ | |
3741 | ||
3742 | for (searchptr = search_start; searchptr; searchptr = searchptr->next) { | |
3743 | if (searchptr->fname) { | |
3744 | /* The empty string in a search path is ignored. | |
3745 | This makes it possible to turn off entirely | |
3746 | a standard piece of the list. */ | |
3747 | if (searchptr->fname[0] == 0) | |
3748 | continue; | |
3749 | strcpy (fname, searchptr->fname); | |
3750 | strcat (fname, "/"); | |
3751 | fname[strlen (fname) + flen] = 0; | |
3752 | } else { | |
3753 | fname[0] = 0; | |
3754 | } | |
3755 | strncat (fname, fbeg, flen); | |
3756 | #ifdef VMS | |
3757 | /* Change this 1/2 Unix 1/2 VMS file specification into a | |
3758 | full VMS file specification */ | |
3759 | if (searchptr->fname && (searchptr->fname[0] != 0)) { | |
3760 | /* Fix up the filename */ | |
3761 | hack_vms_include_specification (fname); | |
3762 | } else { | |
3763 | /* This is a normal VMS filespec, so use it unchanged. */ | |
3764 | strncpy (fname, fbeg, flen); | |
3765 | fname[flen] = 0; | |
3766 | } | |
3767 | #endif /* VMS */ | |
3768 | if (importing) | |
3769 | f = lookup_import (fname); | |
3770 | else | |
3771 | f = open (fname, O_RDONLY, 0666); | |
3772 | if (f == -2) | |
3773 | return 0; /* Already included this file */ | |
3774 | if (f >= 0) | |
3775 | break; | |
3776 | } | |
3777 | } | |
3778 | ||
3779 | if (f < 0) { | |
3780 | /* A file that was not found. */ | |
3781 | ||
3782 | strncpy (fname, fbeg, flen); | |
3783 | fname[flen] = 0; | |
3784 | error_from_errno (fname); | |
3785 | ||
3786 | /* For -M, add this file to the dependencies. */ | |
3787 | if (print_deps > (system_header_p || (system_include_depth > 0))) { | |
3788 | /* Break the line before this. */ | |
3789 | deps_output ("", 0); | |
3790 | ||
3791 | /* If it was requested as a system header file, | |
3792 | then assume it belongs in the first place to look for such. */ | |
3793 | if (system_header_p) { | |
3794 | for (searchptr = search_start; searchptr; searchptr = searchptr->next) { | |
3795 | if (searchptr->fname) { | |
3796 | if (searchptr->fname[0] == 0) | |
3797 | continue; | |
3798 | deps_output (searchptr->fname, 0); | |
3799 | deps_output ("/", 0); | |
3800 | break; | |
3801 | } | |
3802 | } | |
3803 | } | |
3804 | /* Otherwise, omit the directory, as if the file existed | |
3805 | in the directory with the source. */ | |
3806 | deps_output (fbeg, flen); | |
3807 | deps_output (" ", 0); | |
3808 | } | |
3809 | } else { | |
3810 | struct stat stat_f; | |
3811 | ||
3812 | /* Check to see if this include file is a once-only include file. | |
3813 | If so, give up. */ | |
3814 | ||
3815 | struct file_name_list* ptr; | |
3816 | ||
3817 | for (ptr = dont_repeat_files; ptr; ptr = ptr->next) { | |
3818 | if (!strcmp (ptr->fname, fname)) { | |
3819 | close (f); | |
3820 | return 0; /* This file was once'd. */ | |
3821 | } | |
3822 | } | |
3823 | ||
3824 | for (ptr = all_include_files; ptr; ptr = ptr->next) { | |
3825 | if (!strcmp (ptr->fname, fname)) | |
3826 | break; /* This file was included before. */ | |
3827 | } | |
3828 | ||
3829 | if (ptr == 0) { | |
3830 | /* This is the first time for this file. */ | |
3831 | /* Add it to list of files included. */ | |
3832 | ||
3833 | ptr = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
3834 | ptr->control_macro = 0; | |
3835 | ptr->next = all_include_files; | |
3836 | all_include_files = ptr; | |
3837 | ptr->fname = savestring (fname); | |
3838 | ||
3839 | /* For -M, add this file to the dependencies. */ | |
3840 | if (print_deps > (system_header_p || (system_include_depth > 0))) { | |
3841 | deps_output ("", 0); | |
3842 | deps_output (fname, 0); | |
3843 | deps_output (" ", 0); | |
3844 | } | |
3845 | } | |
3846 | ||
3847 | /* Handle -H option. */ | |
3848 | if (print_include_names) | |
3849 | fprintf (stderr, "%s\n", fname); | |
3850 | ||
3851 | if (system_header_p) | |
3852 | system_include_depth++; | |
3853 | ||
3854 | /* Actually process the file. */ | |
3855 | add_import (f, fname); /* Record file on "seen" list for #import. */ | |
3856 | ||
3857 | pcftry = (char *) alloca (strlen (fname) + 30); | |
3858 | pcfbuf = 0; | |
3859 | pcfnum = 0; | |
3860 | ||
3861 | fstat (f, &stat_f); | |
3862 | ||
3863 | if (!no_precomp) | |
3864 | do { | |
3865 | sprintf (pcftry, "%s%d", fname, pcfnum++); | |
3866 | ||
3867 | pcf = open (pcftry, O_RDONLY, 0666); | |
3868 | if (pcf != -1) | |
3869 | { | |
3870 | struct stat s; | |
3871 | ||
3872 | fstat (pcf, &s); | |
3873 | if (bcmp (&stat_f.st_ino, &s.st_ino, sizeof (s.st_ino)) | |
3874 | || stat_f.st_dev != s.st_dev) | |
3875 | { | |
3876 | pcfbuf = check_precompiled (pcf, fname, &pcfbuflimit); | |
3877 | /* Don't need it any more. */ | |
3878 | close (pcf); | |
3879 | } | |
3880 | else | |
3881 | { | |
3882 | /* Don't need it at all. */ | |
3883 | close (pcf); | |
3884 | break; | |
3885 | } | |
3886 | } | |
3887 | } while (pcf != -1 && !pcfbuf); | |
3888 | ||
3889 | /* Actually process the file */ | |
3890 | if (pcfbuf) { | |
3891 | pcfname = xmalloc (strlen (pcftry) + 1); | |
3892 | strcpy (pcfname, pcftry); | |
3893 | pcfinclude (pcfbuf, pcfbuflimit, fname, op); | |
3894 | } | |
3895 | else | |
3896 | finclude (f, fname, op, system_header_p, searchptr); | |
3897 | ||
3898 | if (system_header_p) | |
3899 | system_include_depth--; | |
3900 | } | |
3901 | return 0; | |
3902 | } | |
3903 | ||
3904 | /* Process the contents of include file FNAME, already open on descriptor F, | |
3905 | with output to OP. | |
3906 | SYSTEM_HEADER_P is 1 if this file was specified using <...>. | |
3907 | DIRPTR is the link in the dir path through which this file was found, | |
3908 | or 0 if the file name was absolute. */ | |
3909 | ||
3910 | static void | |
3911 | finclude (f, fname, op, system_header_p, dirptr) | |
3912 | int f; | |
3913 | char *fname; | |
3914 | FILE_BUF *op; | |
3915 | int system_header_p; | |
3916 | struct file_name_list *dirptr; | |
3917 | { | |
3918 | int st_mode; | |
3919 | long st_size; | |
3920 | long i; | |
3921 | FILE_BUF *fp; /* For input stack frame */ | |
3922 | int missing_newline = 0; | |
3923 | ||
3924 | CHECK_DEPTH (return;); | |
3925 | ||
3926 | if (file_size_and_mode (f, &st_mode, &st_size) < 0) | |
3927 | goto nope; /* Impossible? */ | |
3928 | ||
3929 | fp = &instack[indepth + 1]; | |
3930 | bzero (fp, sizeof (FILE_BUF)); | |
3931 | fp->nominal_fname = fp->fname = fname; | |
3932 | fp->length = 0; | |
3933 | fp->lineno = 1; | |
3934 | fp->if_stack = if_stack; | |
3935 | fp->system_header_p = system_header_p; | |
3936 | fp->dir = dirptr; | |
3937 | ||
3938 | if (S_ISREG (st_mode)) { | |
3939 | fp->buf = (U_CHAR *) alloca (st_size + 2); | |
3940 | fp->bufp = fp->buf; | |
3941 | ||
3942 | /* Read the file contents, knowing that st_size is an upper bound | |
3943 | on the number of bytes we can read. */ | |
3944 | while (st_size > 0) { | |
3945 | i = read (f, fp->buf + fp->length, st_size); | |
3946 | if (i <= 0) { | |
3947 | if (i == 0) break; | |
3948 | goto nope; | |
3949 | } | |
3950 | fp->length += i; | |
3951 | st_size -= i; | |
3952 | } | |
3953 | } | |
3954 | else { | |
3955 | /* Cannot count its file size before reading. | |
3956 | First read the entire file into heap and | |
3957 | copy them into buffer on stack. */ | |
3958 | ||
3959 | U_CHAR *bufp; | |
3960 | U_CHAR *basep; | |
3961 | int bsize = 2000; | |
3962 | ||
3963 | st_size = 0; | |
3964 | basep = (U_CHAR *) xmalloc (bsize + 2); | |
3965 | bufp = basep; | |
3966 | ||
3967 | for (;;) { | |
3968 | i = read (f, bufp, bsize - st_size); | |
3969 | if (i < 0) | |
3970 | goto nope; /* error! */ | |
3971 | if (i == 0) | |
3972 | break; /* End of file */ | |
3973 | st_size += i; | |
3974 | bufp += i; | |
3975 | if (bsize == st_size) { /* Buffer is full! */ | |
3976 | bsize *= 2; | |
3977 | basep = (U_CHAR *) xrealloc (basep, bsize + 2); | |
3978 | bufp = basep + st_size; /* May have moved */ | |
3979 | } | |
3980 | } | |
3981 | fp->buf = (U_CHAR *) alloca (st_size + 2); | |
3982 | fp->bufp = fp->buf; | |
3983 | bcopy (basep, fp->buf, st_size); | |
3984 | fp->length = st_size; | |
3985 | free (basep); | |
3986 | } | |
3987 | ||
3988 | /* Close descriptor now, so nesting does not use lots of descriptors. */ | |
3989 | close (f); | |
3990 | ||
3991 | if (!no_trigraphs) | |
3992 | trigraph_pcp (fp); | |
3993 | ||
3994 | if ((fp->length > 0 && fp->buf[fp->length - 1] != '\n') | |
3995 | /* Backslash-newline at end is not good enough. */ | |
3996 | || (fp->length > 1 && fp->buf[fp->length - 2] == '\\')) { | |
3997 | fp->buf[fp->length++] = '\n'; | |
3998 | missing_newline = 1; | |
3999 | } | |
4000 | fp->buf[fp->length] = '\0'; | |
4001 | ||
4002 | indepth++; | |
4003 | input_file_stack_tick++; | |
4004 | ||
4005 | output_line_command (fp, op, 0, enter_file); | |
4006 | rescan (op, 0); | |
4007 | ||
4008 | if (pedantic && missing_newline) | |
4009 | pedwarn ("file does not end in newline"); | |
4010 | ||
4011 | indepth--; | |
4012 | input_file_stack_tick++; | |
4013 | output_line_command (&instack[indepth], op, 0, leave_file); | |
4014 | return; | |
4015 | ||
4016 | nope: | |
4017 | ||
4018 | perror_with_name (fname); | |
4019 | close (f); | |
4020 | } | |
4021 | ||
4022 | /* Record that inclusion of the file named FILE | |
4023 | should be controlled by the macro named MACRO_NAME. | |
4024 | This means that trying to include the file again | |
4025 | will do something if that macro is defined. */ | |
4026 | ||
4027 | static void | |
4028 | record_control_macro (file, macro_name) | |
4029 | char *file; | |
4030 | U_CHAR *macro_name; | |
4031 | { | |
4032 | struct file_name_list *new; | |
4033 | ||
4034 | for (new = all_include_files; new; new = new->next) { | |
4035 | if (!strcmp (new->fname, file)) { | |
4036 | new->control_macro = macro_name; | |
4037 | return; | |
4038 | } | |
4039 | } | |
4040 | ||
4041 | /* If the file is not in all_include_files, something's wrong. */ | |
4042 | abort (); | |
4043 | } | |
4044 | \f | |
4045 | /* Maintain and search list of included files, for #import. */ | |
4046 | ||
4047 | #define IMPORT_HASH_SIZE 31 | |
4048 | ||
4049 | struct import_file { | |
4050 | char *name; | |
4051 | ino_t inode; | |
4052 | dev_t dev; | |
4053 | struct import_file *next; | |
4054 | }; | |
4055 | ||
4056 | /* Hash table of files already included with #include or #import. */ | |
4057 | ||
4058 | static struct import_file *import_hash_table[IMPORT_HASH_SIZE]; | |
4059 | ||
4060 | /* Hash a file name for import_hash_table. */ | |
4061 | ||
4062 | static int | |
4063 | import_hash (f) | |
4064 | char *f; | |
4065 | { | |
4066 | int val = 0; | |
4067 | ||
4068 | while (*f) val += *f++; | |
4069 | return (val%IMPORT_HASH_SIZE); | |
4070 | } | |
4071 | ||
4072 | /* Search for file FILENAME in import_hash_table. | |
4073 | Return -2 if found, either a matching name or a matching inode. | |
4074 | Otherwise, open the file and return a file descriptor if successful | |
4075 | or -1 if unsuccessful. */ | |
4076 | ||
4077 | static int | |
4078 | lookup_import (filename) | |
4079 | char *filename; | |
4080 | { | |
4081 | struct import_file *i; | |
4082 | int h; | |
4083 | int hashval; | |
4084 | struct stat sb; | |
4085 | int fd; | |
4086 | ||
4087 | hashval = import_hash (filename); | |
4088 | ||
4089 | /* Attempt to find file in list of already included files */ | |
4090 | i = import_hash_table[hashval]; | |
4091 | ||
4092 | while (i) { | |
4093 | if (!strcmp (filename, i->name)) | |
4094 | return -2; /* return found */ | |
4095 | i = i->next; | |
4096 | } | |
4097 | /* Open it and try a match on inode/dev */ | |
4098 | fd = open (filename, O_RDONLY, 0666); | |
4099 | if (fd < 0) | |
4100 | return fd; | |
4101 | fstat (fd, &sb); | |
4102 | for (h = 0; h < IMPORT_HASH_SIZE; h++) { | |
4103 | i = import_hash_table[h]; | |
4104 | while (i) { | |
4105 | /* Compare the inode and the device. | |
4106 | Supposedly on some systems the inode is not a scalar. */ | |
4107 | if (!bcmp (&i->inode, &sb.st_ino, sizeof (sb.st_ino)) | |
4108 | && i->dev == sb.st_dev) { | |
4109 | close (fd); | |
4110 | return -2; /* return found */ | |
4111 | } | |
4112 | i = i->next; | |
4113 | } | |
4114 | } | |
4115 | return fd; /* Not found, return open file */ | |
4116 | } | |
4117 | ||
4118 | /* Add the file FNAME, open on descriptor FD, to import_hash_table. */ | |
4119 | ||
4120 | static void | |
4121 | add_import (fd, fname) | |
4122 | int fd; | |
4123 | char *fname; | |
4124 | { | |
4125 | struct import_file *i; | |
4126 | int hashval; | |
4127 | struct stat sb; | |
4128 | ||
4129 | hashval = import_hash (fname); | |
4130 | fstat (fd, &sb); | |
4131 | i = (struct import_file *)xmalloc (sizeof (struct import_file)); | |
4132 | i->name = (char *)xmalloc (strlen (fname)+1); | |
4133 | strcpy (i->name, fname); | |
4134 | bcopy (&sb.st_ino, &i->inode, sizeof (sb.st_ino)); | |
4135 | i->dev = sb.st_dev; | |
4136 | i->next = import_hash_table[hashval]; | |
4137 | import_hash_table[hashval] = i; | |
4138 | } | |
4139 | \f | |
4140 | /* Load the specified precompiled header into core, and verify its | |
4141 | preconditions. PCF indicates the file descriptor to read, which must | |
4142 | be a regular file. FNAME indicates the file name of the original | |
4143 | header. *LIMIT will be set to an address one past the end of the file. | |
4144 | If the preconditions of the file are not satisfied, the buffer is | |
4145 | freed and we return 0. If the preconditions are satisfied, return | |
4146 | the address of the buffer following the preconditions. The buffer, in | |
4147 | this case, should never be freed because various pieces of it will | |
4148 | be referred to until all precompiled strings are output at the end of | |
4149 | the run. | |
4150 | */ | |
4151 | static char * | |
4152 | check_precompiled (pcf, fname, limit) | |
4153 | int pcf; | |
4154 | char *fname; | |
4155 | char **limit; | |
4156 | { | |
4157 | int st_mode; | |
4158 | long st_size; | |
4159 | int length = 0; | |
4160 | char *buf; | |
4161 | char *dollar_loc; | |
4162 | int i; | |
4163 | char *cp; | |
4164 | ||
4165 | if (pcp_outfile) | |
4166 | return 0; | |
4167 | ||
4168 | if (file_size_and_mode (pcf, &st_mode, &st_size) < 0) | |
4169 | return 0; | |
4170 | ||
4171 | if (S_ISREG (st_mode)) | |
4172 | { | |
4173 | buf = xmalloc (st_size + 2); | |
4174 | while (st_size > 0) | |
4175 | { | |
4176 | i = read (pcf, buf + length, st_size); | |
4177 | if (i < 0) | |
4178 | goto nope; | |
4179 | if (i == 0) | |
4180 | break; | |
4181 | length += i; | |
4182 | st_size -= i; | |
4183 | } | |
4184 | } | |
4185 | else | |
4186 | abort (); | |
4187 | ||
4188 | if (length > 0 && buf[length-1] != '\n') | |
4189 | buf[length++] = '\n'; | |
4190 | buf[length] = '\0'; | |
4191 | ||
4192 | *limit = buf + length; | |
4193 | ||
4194 | /* File is in core. Check the preconditions. */ | |
4195 | if (!check_preconditions (buf)) | |
4196 | goto nope; | |
4197 | for (cp = buf; *cp; cp++) | |
4198 | ; | |
4199 | #ifdef DEBUG_PCP | |
4200 | fprintf (stderr, "Using preinclude %s\n", fname); | |
4201 | #endif | |
4202 | return cp + 1; | |
4203 | ||
4204 | nope: | |
4205 | #ifdef DEBUG_PCP | |
4206 | fprintf (stderr, "Cannot use preinclude %s\n", fname); | |
4207 | #endif | |
4208 | free (buf); | |
4209 | return 0; | |
4210 | } | |
4211 | ||
4212 | /* PREC (null terminated) points to the preconditions of a | |
4213 | precompiled header. These are a series of #define and #undef | |
4214 | lines which must match the current contents of the hash | |
4215 | table. */ | |
4216 | static int | |
4217 | check_preconditions (prec) | |
4218 | char *prec; | |
4219 | { | |
4220 | MACRODEF mdef; | |
4221 | char *lineend; | |
4222 | ||
4223 | while (*prec) { | |
4224 | lineend = (char *) index (prec, '\n'); | |
4225 | ||
4226 | if (*prec++ != '#') { | |
4227 | error ("Bad format encountered while reading precompiled file"); | |
4228 | return 0; | |
4229 | } | |
4230 | if (!strncmp (prec, "define", 6)) { | |
4231 | HASHNODE *hp; | |
4232 | ||
4233 | prec += 6; | |
4234 | mdef = create_definition (prec, lineend, 0); | |
4235 | ||
4236 | if (mdef.defn == 0) | |
4237 | abort(); | |
4238 | ||
4239 | if ((hp = lookup (mdef.symnam, mdef.symlen, -1)) == NULL | |
4240 | || (hp->type != T_MACRO && hp->type != T_CONST) | |
4241 | || (hp->type == T_MACRO | |
4242 | && !compare_defs (mdef.defn, hp->value.defn) | |
4243 | && (mdef.defn->length != 2 | |
4244 | || mdef.defn->expansion[0] != '\n' | |
4245 | || mdef.defn->expansion[1] != ' '))) | |
4246 | return 0; | |
4247 | } else if (!strncmp (prec, "undef", 5)) { | |
4248 | char *name; | |
4249 | int len; | |
4250 | ||
4251 | prec += 5; | |
4252 | while (is_hor_space[*prec]) | |
4253 | prec++; | |
4254 | name = prec; | |
4255 | while (is_idchar[*prec]) | |
4256 | prec++; | |
4257 | len = prec - name; | |
4258 | ||
4259 | if (lookup (name, len, -1)) | |
4260 | return 0; | |
4261 | } else { | |
4262 | error ("Bad format encountered while reading precompiled file"); | |
4263 | return 0; | |
4264 | } | |
4265 | prec = lineend + 1; | |
4266 | } | |
4267 | /* They all passed successfully */ | |
4268 | return 1; | |
4269 | } | |
4270 | ||
4271 | /* Process the main body of a precompiled file. BUF points to the | |
4272 | string section of the file, following the preconditions. LIMIT is one | |
4273 | character past the end. NAME is the name of the file being read | |
4274 | in. OP is the main output buffer */ | |
4275 | static void | |
4276 | pcfinclude (buf, limit, name, op) | |
4277 | U_CHAR *buf, *limit, *name; | |
4278 | FILE_BUF *op; | |
4279 | { | |
4280 | FILE_BUF tmpbuf; | |
4281 | int nstrings; | |
4282 | U_CHAR *cp = buf; | |
4283 | ||
4284 | /* First in the file comes 4 bytes indicating the number of strings, */ | |
4285 | /* in network byte order. (MSB first). */ | |
4286 | nstrings = *cp++; | |
4287 | nstrings = (nstrings << 8) | *cp++; | |
4288 | nstrings = (nstrings << 8) | *cp++; | |
4289 | nstrings = (nstrings << 8) | *cp++; | |
4290 | ||
4291 | /* Looping over each string... */ | |
4292 | while (nstrings--) { | |
4293 | U_CHAR *string_start; | |
4294 | U_CHAR *endofthiskey; | |
4295 | STRINGDEF *str; | |
4296 | int nkeys; | |
4297 | ||
4298 | /* Each string starts with a STRINGDEF structure (str), followed */ | |
4299 | /* by the text of the string (string_start) */ | |
4300 | ||
4301 | /* First skip to a longword boundary */ | |
4302 | if ((int)cp & 3) | |
4303 | cp += 4 - ((int)cp & 3); | |
4304 | ||
4305 | /* Now get the string. */ | |
4306 | str = (STRINGDEF *) cp; | |
4307 | string_start = cp += sizeof (STRINGDEF); | |
4308 | ||
4309 | for (; *cp; cp++) /* skip the string */ | |
4310 | ; | |
4311 | ||
4312 | /* We need to macro expand the string here to ensure that the | |
4313 | proper definition environment is in place. If it were only | |
4314 | expanded when we find out it is needed, macros necessary for | |
4315 | its proper expansion might have had their definitions changed. */ | |
4316 | tmpbuf = expand_to_temp_buffer (string_start, cp++, 0, 0); | |
4317 | /* Lineno is already set in the precompiled file */ | |
4318 | str->contents = tmpbuf.buf; | |
4319 | str->len = tmpbuf.length; | |
4320 | str->writeflag = 0; | |
4321 | str->filename = name; | |
4322 | str->output_mark = outbuf.bufp - outbuf.buf; | |
4323 | ||
4324 | str->chain = 0; | |
4325 | *stringlist_tailp = str; | |
4326 | stringlist_tailp = &str->chain; | |
4327 | ||
4328 | /* Next comes a fourbyte number indicating the number of keys */ | |
4329 | /* for this string. */ | |
4330 | nkeys = *cp++; | |
4331 | nkeys = (nkeys << 8) | *cp++; | |
4332 | nkeys = (nkeys << 8) | *cp++; | |
4333 | nkeys = (nkeys << 8) | *cp++; | |
4334 | ||
4335 | /* If this number is -1, then the string is mandatory. */ | |
4336 | if (nkeys == -1) | |
4337 | str->writeflag = 1; | |
4338 | else | |
4339 | /* Otherwist, for each key, */ | |
4340 | for (; nkeys--; free (tmpbuf.buf), cp = endofthiskey + 1) { | |
4341 | KEYDEF *kp = (KEYDEF *) cp; | |
4342 | HASHNODE *hp; | |
4343 | ||
4344 | /* It starts with a KEYDEF structure */ | |
4345 | cp += sizeof (KEYDEF); | |
4346 | ||
4347 | /* Find the end of the key. At the end of this for loop we | |
4348 | advance CP to the start of the next key using this variable. */ | |
4349 | endofthiskey = cp + strlen (cp); | |
4350 | kp->str = str; | |
4351 | ||
4352 | /* Expand the key, and enter it into the hash table. */ | |
4353 | tmpbuf = expand_to_temp_buffer (cp, endofthiskey, 0, 0); | |
4354 | tmpbuf.bufp = tmpbuf.buf; | |
4355 | ||
4356 | while (is_hor_space[*tmpbuf.bufp]) | |
4357 | tmpbuf.bufp++; | |
4358 | if (!is_idstart[*tmpbuf.bufp] | |
4359 | || tmpbuf.bufp == tmpbuf.buf + tmpbuf.length) { | |
4360 | str->writeflag = 1; | |
4361 | continue; | |
4362 | } | |
4363 | ||
4364 | hp = lookup (tmpbuf.bufp, -1, -1); | |
4365 | if (hp == NULL) { | |
4366 | kp->chain = 0; | |
4367 | install (tmpbuf.bufp, -1, T_PCSTRING, (int) kp, -1); | |
4368 | } | |
4369 | else if (hp->type == T_PCSTRING) { | |
4370 | kp->chain = hp->value.keydef; | |
4371 | hp->value.keydef = kp; | |
4372 | } | |
4373 | else | |
4374 | str->writeflag = 1; | |
4375 | } | |
4376 | } | |
4377 | /* This output_line_command serves to switch us back to the current | |
4378 | input file in case some of these strings get output (which will | |
4379 | result in line commands for the header file being output). */ | |
4380 | output_line_command (&instack[indepth], op, 0, enter_file); | |
4381 | } | |
4382 | ||
4383 | /* Called from rescan when it hits a key for strings. Mark them all */ | |
4384 | /* used and clean up. */ | |
4385 | static void | |
4386 | pcstring_used (hp) | |
4387 | HASHNODE *hp; | |
4388 | { | |
4389 | KEYDEF *kp, *tmp; | |
4390 | ||
4391 | for (kp = hp->value.keydef; kp; kp = kp->chain) | |
4392 | kp->str->writeflag = 1; | |
4393 | delete_macro (hp); | |
4394 | } | |
4395 | ||
4396 | /* Write the output, interspersing precompiled strings in their */ | |
4397 | /* appropriate places. */ | |
4398 | static void | |
4399 | write_output () | |
4400 | { | |
4401 | STRINGDEF *next_string; | |
4402 | U_CHAR *cur_buf_loc; | |
4403 | int line_command_len = 80; | |
4404 | char *line_command = xmalloc (line_command_len); | |
4405 | int len; | |
4406 | ||
4407 | /* In each run through the loop, either cur_buf_loc == */ | |
4408 | /* next_string_loc, in which case we print a series of strings, or */ | |
4409 | /* it is less than next_string_loc, in which case we write some of */ | |
4410 | /* the buffer. */ | |
4411 | cur_buf_loc = outbuf.buf; | |
4412 | next_string = stringlist; | |
4413 | ||
4414 | while (cur_buf_loc < outbuf.bufp || next_string) { | |
4415 | if (next_string | |
4416 | && cur_buf_loc - outbuf.buf == next_string->output_mark) { | |
4417 | if (next_string->writeflag) { | |
4418 | len = strlen (next_string->filename); | |
4419 | if (len > line_command_len) | |
4420 | line_command = xrealloc (line_command, | |
4421 | line_command_len *= 2); | |
4422 | sprintf (line_command, "\n# %d \"%s\"\n", | |
4423 | next_string->lineno, next_string->filename); | |
4424 | write (fileno (stdout), line_command, | |
4425 | strlen (line_command)); | |
4426 | write (fileno (stdout), | |
4427 | next_string->contents, next_string->len); | |
4428 | } | |
4429 | next_string = next_string->chain; | |
4430 | } | |
4431 | else { | |
4432 | len = (next_string | |
4433 | ? (next_string->output_mark | |
4434 | - (cur_buf_loc - outbuf.buf)) | |
4435 | : outbuf.bufp - cur_buf_loc); | |
4436 | ||
4437 | write (fileno (stdout), cur_buf_loc, len); | |
4438 | cur_buf_loc += len; | |
4439 | } | |
4440 | } | |
4441 | } | |
4442 | ||
4443 | /* Pass a directive through to the output file. | |
4444 | BUF points to the contents of the directive, as a continguous string. | |
4445 | LIMIT points to the first character past the end of the directive. | |
4446 | KEYWORD is the keyword-table entry for the directive. */ | |
4447 | ||
4448 | static void | |
4449 | pass_thru_directive (buf, limit, op, keyword) | |
4450 | U_CHAR *buf, *limit; | |
4451 | FILE_BUF *op; | |
4452 | struct directive *keyword; | |
4453 | { | |
4454 | register unsigned keyword_length = keyword->length; | |
4455 | ||
4456 | check_expand (op, 1 + keyword_length + (limit - buf)); | |
4457 | *op->bufp++ = '#'; | |
4458 | bcopy (keyword->name, op->bufp, keyword_length); | |
4459 | op->bufp += keyword_length; | |
4460 | if (limit != buf && buf[0] != ' ') | |
4461 | *op->bufp++ = ' '; | |
4462 | bcopy (buf, op->bufp, limit - buf); | |
4463 | op->bufp += (limit - buf); | |
4464 | *op->bufp++ = '\n'; | |
4465 | } | |
4466 | \f | |
4467 | /* The arglist structure is built by do_define to tell | |
4468 | collect_definition where the argument names begin. That | |
4469 | is, for a define like "#define f(x,y,z) foo+x-bar*y", the arglist | |
4470 | would contain pointers to the strings x, y, and z. | |
4471 | Collect_definition would then build a DEFINITION node, | |
4472 | with reflist nodes pointing to the places x, y, and z had | |
4473 | appeared. So the arglist is just convenience data passed | |
4474 | between these two routines. It is not kept around after | |
4475 | the current #define has been processed and entered into the | |
4476 | hash table. */ | |
4477 | ||
4478 | struct arglist { | |
4479 | struct arglist *next; | |
4480 | U_CHAR *name; | |
4481 | int length; | |
4482 | int argno; | |
4483 | }; | |
4484 | ||
4485 | /* Create a DEFINITION node from a #define directive. Arguments are | |
4486 | as for do_define. */ | |
4487 | static MACRODEF | |
4488 | create_definition (buf, limit, op) | |
4489 | U_CHAR *buf, *limit; | |
4490 | FILE_BUF *op; | |
4491 | { | |
4492 | U_CHAR *bp; /* temp ptr into input buffer */ | |
4493 | U_CHAR *symname; /* remember where symbol name starts */ | |
4494 | int sym_length; /* and how long it is */ | |
4495 | int line = instack[indepth].lineno; | |
4496 | char *file = instack[indepth].nominal_fname; | |
4497 | ||
4498 | DEFINITION *defn; | |
4499 | int arglengths = 0; /* Accumulate lengths of arg names | |
4500 | plus number of args. */ | |
4501 | MACRODEF mdef; | |
4502 | ||
4503 | bp = buf; | |
4504 | ||
4505 | while (is_hor_space[*bp]) | |
4506 | bp++; | |
4507 | ||
4508 | symname = bp; /* remember where it starts */ | |
4509 | sym_length = check_macro_name (bp, "macro"); | |
4510 | bp += sym_length; | |
4511 | ||
4512 | /* Lossage will occur if identifiers or control keywords are broken | |
4513 | across lines using backslash. This is not the right place to take | |
4514 | care of that. */ | |
4515 | ||
4516 | if (*bp == '(') { | |
4517 | struct arglist *arg_ptrs = NULL; | |
4518 | int argno = 0; | |
4519 | ||
4520 | bp++; /* skip '(' */ | |
4521 | SKIP_WHITE_SPACE (bp); | |
4522 | ||
4523 | /* Loop over macro argument names. */ | |
4524 | while (*bp != ')') { | |
4525 | struct arglist *temp; | |
4526 | ||
4527 | temp = (struct arglist *) alloca (sizeof (struct arglist)); | |
4528 | temp->name = bp; | |
4529 | temp->next = arg_ptrs; | |
4530 | temp->argno = argno++; | |
4531 | arg_ptrs = temp; | |
4532 | ||
4533 | if (!is_idstart[*bp]) | |
4534 | pedwarn ("parameter name starts with a digit in `#define'"); | |
4535 | ||
4536 | /* Find the end of the arg name. */ | |
4537 | while (is_idchar[*bp]) { | |
4538 | bp++; | |
4539 | } | |
4540 | temp->length = bp - temp->name; | |
4541 | arglengths += temp->length + 2; | |
4542 | SKIP_WHITE_SPACE (bp); | |
4543 | if (temp->length == 0 || (*bp != ',' && *bp != ')')) { | |
4544 | error ("badly punctuated parameter list in `#define'"); | |
4545 | goto nope; | |
4546 | } | |
4547 | if (*bp == ',') { | |
4548 | bp++; | |
4549 | SKIP_WHITE_SPACE (bp); | |
4550 | } | |
4551 | if (bp >= limit) { | |
4552 | error ("unterminated parameter list in `#define'"); | |
4553 | goto nope; | |
4554 | } | |
4555 | { | |
4556 | struct arglist *otemp; | |
4557 | ||
4558 | for (otemp = temp->next; otemp != NULL; otemp = otemp->next) | |
4559 | if (temp->length == otemp->length && | |
4560 | strncmp(temp->name, otemp->name, temp->length) == 0) { | |
4561 | U_CHAR *name; | |
4562 | ||
4563 | name = (U_CHAR *) alloca(temp->length + 1); | |
4564 | (void) strncpy(name, temp->name, temp->length); | |
4565 | name[temp->length] = '\0'; | |
4566 | error ("duplicate argument name `%s' in `#define'", name); | |
4567 | goto nope; | |
4568 | } | |
4569 | } | |
4570 | } | |
4571 | ||
4572 | ++bp; /* skip paren */ | |
4573 | /* Skip exactly one space or tab if any. */ | |
4574 | if (bp < limit && (*bp == ' ' || *bp == '\t')) ++bp; | |
4575 | /* now everything from bp before limit is the definition. */ | |
4576 | defn = collect_expansion (bp, limit, argno, arg_ptrs); | |
4577 | ||
4578 | /* Now set defn->args.argnames to the result of concatenating | |
4579 | the argument names in reverse order | |
4580 | with comma-space between them. */ | |
4581 | defn->args.argnames = (U_CHAR *) xmalloc (arglengths + 1); | |
4582 | { | |
4583 | struct arglist *temp; | |
4584 | int i = 0; | |
4585 | for (temp = arg_ptrs; temp; temp = temp->next) { | |
4586 | bcopy (temp->name, &defn->args.argnames[i], temp->length); | |
4587 | i += temp->length; | |
4588 | if (temp->next != 0) { | |
4589 | defn->args.argnames[i++] = ','; | |
4590 | defn->args.argnames[i++] = ' '; | |
4591 | } | |
4592 | } | |
4593 | defn->args.argnames[i] = 0; | |
4594 | } | |
4595 | } else { | |
4596 | /* simple expansion or empty definition; gobble it */ | |
4597 | if (is_hor_space[*bp]) | |
4598 | ++bp; /* skip exactly one blank/tab char */ | |
4599 | /* now everything from bp before limit is the definition. */ | |
4600 | defn = collect_expansion (bp, limit, -1, 0); | |
4601 | defn->args.argnames = (U_CHAR *) ""; | |
4602 | } | |
4603 | ||
4604 | defn->line = line; | |
4605 | defn->file = file; | |
4606 | ||
4607 | /* OP is null if this is a predefinition */ | |
4608 | defn->predefined = !op; | |
4609 | mdef.defn = defn; | |
4610 | mdef.symnam = symname; | |
4611 | mdef.symlen = sym_length; | |
4612 | ||
4613 | return mdef; | |
4614 | ||
4615 | nope: | |
4616 | mdef.defn = 0; | |
4617 | return mdef; | |
4618 | } | |
4619 | ||
4620 | /* Process a #define command. | |
4621 | BUF points to the contents of the #define command, as a continguous string. | |
4622 | LIMIT points to the first character past the end of the definition. | |
4623 | KEYWORD is the keyword-table entry for #define. */ | |
4624 | ||
4625 | static int | |
4626 | do_define (buf, limit, op, keyword) | |
4627 | U_CHAR *buf, *limit; | |
4628 | FILE_BUF *op; | |
4629 | struct directive *keyword; | |
4630 | { | |
4631 | int hashcode; | |
4632 | MACRODEF mdef; | |
4633 | ||
4634 | /* If this is a precompiler run (with -pcp) pass thru #define commands. */ | |
4635 | if (pcp_outfile && op) | |
4636 | pass_thru_directive (buf, limit, op, keyword); | |
4637 | ||
4638 | mdef = create_definition (buf, limit, op); | |
4639 | if (mdef.defn == 0) | |
4640 | goto nope; | |
4641 | ||
4642 | hashcode = hashf (mdef.symnam, mdef.symlen, HASHSIZE); | |
4643 | ||
4644 | { | |
4645 | HASHNODE *hp; | |
4646 | if ((hp = lookup (mdef.symnam, mdef.symlen, hashcode)) != NULL) { | |
4647 | int ok = 0; | |
4648 | /* Redefining a precompiled key is ok. */ | |
4649 | if (hp->type == T_PCSTRING) | |
4650 | ok = 1; | |
4651 | /* Redefining a macro is ok if the definitions are the same. */ | |
4652 | else if (hp->type == T_MACRO) | |
4653 | ok = ! compare_defs (mdef.defn, hp->value.defn); | |
4654 | /* Redefining a constant is ok with -D. */ | |
4655 | else if (hp->type == T_CONST) | |
4656 | ok = ! done_initializing; | |
4657 | /* Print the warning if it's not ok. */ | |
4658 | if (!ok) { | |
4659 | U_CHAR *msg; /* what pain... */ | |
4660 | ||
4661 | /* If we are passing through #define and #undef directives, do | |
4662 | that for this re-definition now. */ | |
4663 | if (debug_output && op) | |
4664 | pass_thru_directive (buf, limit, op, keyword); | |
4665 | ||
4666 | msg = (U_CHAR *) alloca (mdef.symlen + 22); | |
4667 | *msg = '`'; | |
4668 | bcopy (mdef.symnam, msg + 1, mdef.symlen); | |
4669 | strcpy ((char *) (msg + mdef.symlen + 1), "' redefined"); | |
4670 | pedwarn (msg); | |
4671 | if (hp->type == T_MACRO) | |
4672 | pedwarn_with_file_and_line (hp->value.defn->file, hp->value.defn->line, | |
4673 | "this is the location of the previous definition"); | |
4674 | } | |
4675 | /* Replace the old definition. */ | |
4676 | hp->type = T_MACRO; | |
4677 | hp->value.defn = mdef.defn; | |
4678 | } else { | |
4679 | /* If we are passing through #define and #undef directives, do | |
4680 | that for this new definition now. */ | |
4681 | if (debug_output && op) | |
4682 | pass_thru_directive (buf, limit, op, keyword); | |
4683 | install (mdef.symnam, mdef.symlen, T_MACRO, mdef.defn, hashcode); | |
4684 | } | |
4685 | } | |
4686 | ||
4687 | return 0; | |
4688 | ||
4689 | nope: | |
4690 | ||
4691 | return 1; | |
4692 | } | |
4693 | \f | |
4694 | /* Check a purported macro name SYMNAME, and yield its length. | |
4695 | USAGE is the kind of name this is intended for. */ | |
4696 | ||
4697 | static int | |
4698 | check_macro_name (symname, usage) | |
4699 | U_CHAR *symname; | |
4700 | char *usage; | |
4701 | { | |
4702 | U_CHAR *p; | |
4703 | int sym_length; | |
4704 | ||
4705 | for (p = symname; is_idchar[*p]; p++) | |
4706 | ; | |
4707 | sym_length = p - symname; | |
4708 | if (sym_length == 0) | |
4709 | error ("invalid %s name", usage); | |
4710 | else if (!is_idstart[*symname]) { | |
4711 | U_CHAR *msg; /* what pain... */ | |
4712 | msg = (U_CHAR *) alloca (sym_length + 1); | |
4713 | bcopy (symname, msg, sym_length); | |
4714 | msg[sym_length] = 0; | |
4715 | error ("invalid %s name `%s'", usage, msg); | |
4716 | } else { | |
4717 | if (! strncmp (symname, "defined", 7) && sym_length == 7) | |
4718 | error ("invalid %s name `defined'", usage); | |
4719 | } | |
4720 | return sym_length; | |
4721 | } | |
4722 | ||
4723 | /* | |
4724 | * return zero if two DEFINITIONs are isomorphic | |
4725 | */ | |
4726 | static int | |
4727 | compare_defs (d1, d2) | |
4728 | DEFINITION *d1, *d2; | |
4729 | { | |
4730 | register struct reflist *a1, *a2; | |
4731 | register U_CHAR *p1 = d1->expansion; | |
4732 | register U_CHAR *p2 = d2->expansion; | |
4733 | int first = 1; | |
4734 | ||
4735 | if (d1->nargs != d2->nargs) | |
4736 | return 1; | |
4737 | if (strcmp ((char *)d1->args.argnames, (char *)d2->args.argnames)) | |
4738 | return 1; | |
4739 | for (a1 = d1->pattern, a2 = d2->pattern; a1 && a2; | |
4740 | a1 = a1->next, a2 = a2->next) { | |
4741 | if (!((a1->nchars == a2->nchars && ! strncmp (p1, p2, a1->nchars)) | |
4742 | || ! comp_def_part (first, p1, a1->nchars, p2, a2->nchars, 0)) | |
4743 | || a1->argno != a2->argno | |
4744 | || a1->stringify != a2->stringify | |
4745 | || a1->raw_before != a2->raw_before | |
4746 | || a1->raw_after != a2->raw_after) | |
4747 | return 1; | |
4748 | first = 0; | |
4749 | p1 += a1->nchars; | |
4750 | p2 += a2->nchars; | |
4751 | } | |
4752 | if (a1 != a2) | |
4753 | return 1; | |
4754 | if (comp_def_part (first, p1, d1->length - (p1 - d1->expansion), | |
4755 | p2, d2->length - (p2 - d2->expansion), 1)) | |
4756 | return 1; | |
4757 | return 0; | |
4758 | } | |
4759 | ||
4760 | /* Return 1 if two parts of two macro definitions are effectively different. | |
4761 | One of the parts starts at BEG1 and has LEN1 chars; | |
4762 | the other has LEN2 chars at BEG2. | |
4763 | Any sequence of whitespace matches any other sequence of whitespace. | |
4764 | FIRST means these parts are the first of a macro definition; | |
4765 | so ignore leading whitespace entirely. | |
4766 | LAST means these parts are the last of a macro definition; | |
4767 | so ignore trailing whitespace entirely. */ | |
4768 | ||
4769 | static int | |
4770 | comp_def_part (first, beg1, len1, beg2, len2, last) | |
4771 | int first; | |
4772 | U_CHAR *beg1, *beg2; | |
4773 | int len1, len2; | |
4774 | int last; | |
4775 | { | |
4776 | register U_CHAR *end1 = beg1 + len1; | |
4777 | register U_CHAR *end2 = beg2 + len2; | |
4778 | if (first) { | |
4779 | while (beg1 != end1 && is_space[*beg1]) beg1++; | |
4780 | while (beg2 != end2 && is_space[*beg2]) beg2++; | |
4781 | } | |
4782 | if (last) { | |
4783 | while (beg1 != end1 && is_space[end1[-1]]) end1--; | |
4784 | while (beg2 != end2 && is_space[end2[-1]]) end2--; | |
4785 | } | |
4786 | while (beg1 != end1 && beg2 != end2) { | |
4787 | if (is_space[*beg1] && is_space[*beg2]) { | |
4788 | while (beg1 != end1 && is_space[*beg1]) beg1++; | |
4789 | while (beg2 != end2 && is_space[*beg2]) beg2++; | |
4790 | } else if (*beg1 == *beg2) { | |
4791 | beg1++; beg2++; | |
4792 | } else break; | |
4793 | } | |
4794 | return (beg1 != end1) || (beg2 != end2); | |
4795 | } | |
4796 | \f | |
4797 | /* Read a replacement list for a macro with parameters. | |
4798 | Build the DEFINITION structure. | |
4799 | Reads characters of text starting at BUF until END. | |
4800 | ARGLIST specifies the formal parameters to look for | |
4801 | in the text of the definition; NARGS is the number of args | |
4802 | in that list, or -1 for a macro name that wants no argument list. | |
4803 | MACRONAME is the macro name itself (so we can avoid recursive expansion) | |
4804 | and NAMELEN is its length in characters. | |
4805 | ||
4806 | Note that comments and backslash-newlines have already been deleted | |
4807 | from the argument. */ | |
4808 | ||
4809 | /* Leading and trailing Space, Tab, etc. are converted to markers | |
4810 | Newline Space, Newline Tab, etc. | |
4811 | Newline Space makes a space in the final output | |
4812 | but is discarded if stringified. (Newline Tab is similar but | |
4813 | makes a Tab instead.) | |
4814 | ||
4815 | If there is no trailing whitespace, a Newline Space is added at the end | |
4816 | to prevent concatenation that would be contrary to the standard. */ | |
4817 | ||
4818 | static DEFINITION * | |
4819 | collect_expansion (buf, end, nargs, arglist) | |
4820 | U_CHAR *buf, *end; | |
4821 | int nargs; | |
4822 | struct arglist *arglist; | |
4823 | { | |
4824 | DEFINITION *defn; | |
4825 | register U_CHAR *p, *limit, *lastp, *exp_p; | |
4826 | struct reflist *endpat = NULL; | |
4827 | /* Pointer to first nonspace after last ## seen. */ | |
4828 | U_CHAR *concat = 0; | |
4829 | /* Pointer to first nonspace after last single-# seen. */ | |
4830 | U_CHAR *stringify = 0; | |
4831 | int maxsize; | |
4832 | int expected_delimiter = '\0'; | |
4833 | ||
4834 | /* Scan thru the replacement list, ignoring comments and quoted | |
4835 | strings, picking up on the macro calls. It does a linear search | |
4836 | thru the arg list on every potential symbol. Profiling might say | |
4837 | that something smarter should happen. */ | |
4838 | ||
4839 | if (end < buf) | |
4840 | abort (); | |
4841 | ||
4842 | /* Find the beginning of the trailing whitespace. */ | |
4843 | /* Find end of leading whitespace. */ | |
4844 | limit = end; | |
4845 | p = buf; | |
4846 | while (p < limit && is_space[limit[-1]]) limit--; | |
4847 | while (p < limit && is_space[*p]) p++; | |
4848 | ||
4849 | /* Allocate space for the text in the macro definition. | |
4850 | Leading and trailing whitespace chars need 2 bytes each. | |
4851 | Each other input char may or may not need 1 byte, | |
4852 | so this is an upper bound. | |
4853 | The extra 2 are for invented trailing newline-marker and final null. */ | |
4854 | maxsize = (sizeof (DEFINITION) | |
4855 | + 2 * (end - limit) + 2 * (p - buf) | |
4856 | + (limit - p) + 3); | |
4857 | defn = (DEFINITION *) xcalloc (1, maxsize); | |
4858 | ||
4859 | defn->nargs = nargs; | |
4860 | exp_p = defn->expansion = (U_CHAR *) defn + sizeof (DEFINITION); | |
4861 | lastp = exp_p; | |
4862 | ||
4863 | p = buf; | |
4864 | ||
4865 | /* Convert leading whitespace to Newline-markers. */ | |
4866 | while (p < limit && is_space[*p]) { | |
4867 | *exp_p++ = '\n'; | |
4868 | *exp_p++ = *p++; | |
4869 | } | |
4870 | ||
4871 | if (limit - p >= 2 && p[0] == '#' && p[1] == '#') { | |
4872 | error ("`##' at start of macro definition"); | |
4873 | p += 2; | |
4874 | } | |
4875 | ||
4876 | /* Process the main body of the definition. */ | |
4877 | while (p < limit) { | |
4878 | int skipped_arg = 0; | |
4879 | register U_CHAR c = *p++; | |
4880 | ||
4881 | *exp_p++ = c; | |
4882 | ||
4883 | if (!traditional) { | |
4884 | switch (c) { | |
4885 | case '\'': | |
4886 | case '\"': | |
4887 | if (expected_delimiter != '\0') { | |
4888 | if (c == expected_delimiter) | |
4889 | expected_delimiter = '\0'; | |
4890 | } else | |
4891 | expected_delimiter = c; | |
4892 | break; | |
4893 | ||
4894 | /* Special hack: if a \# is written in the #define | |
4895 | include a # in the definition. This is useless for C code | |
4896 | but useful for preprocessing other things. */ | |
4897 | ||
4898 | case '\\': | |
4899 | /* \# quotes a # even outside of strings. */ | |
4900 | if (p < limit && *p == '#' && !expected_delimiter) { | |
4901 | exp_p--; | |
4902 | *exp_p++ = *p++; | |
4903 | } else if (p < limit && expected_delimiter) { | |
4904 | /* In a string, backslash goes through | |
4905 | and makes next char ordinary. */ | |
4906 | *exp_p++ = *p++; | |
4907 | } | |
4908 | break; | |
4909 | ||
4910 | case '#': | |
4911 | /* # is ordinary inside a string. */ | |
4912 | if (expected_delimiter) | |
4913 | break; | |
4914 | if (p < limit && *p == '#') { | |
4915 | /* ##: concatenate preceding and following tokens. */ | |
4916 | /* Take out the first #, discard preceding whitespace. */ | |
4917 | exp_p--; | |
4918 | while (exp_p > lastp && is_hor_space[exp_p[-1]]) | |
4919 | --exp_p; | |
4920 | /* Skip the second #. */ | |
4921 | p++; | |
4922 | /* Discard following whitespace. */ | |
4923 | SKIP_WHITE_SPACE (p); | |
4924 | concat = p; | |
4925 | if (p == limit) | |
4926 | error ("`##' at end of macro definition"); | |
4927 | } else { | |
4928 | /* Single #: stringify following argument ref. | |
4929 | Don't leave the # in the expansion. */ | |
4930 | exp_p--; | |
4931 | SKIP_WHITE_SPACE (p); | |
4932 | if (p == limit || ! is_idstart[*p] || nargs <= 0) | |
4933 | error ("`#' operator is not followed by a macro argument name"); | |
4934 | else | |
4935 | stringify = p; | |
4936 | } | |
4937 | break; | |
4938 | } | |
4939 | } else { | |
4940 | /* In -traditional mode, recognize arguments inside strings and | |
4941 | and character constants, and ignore special properties of #. | |
4942 | Arguments inside strings are considered "stringified", but no | |
4943 | extra quote marks are supplied. */ | |
4944 | switch (c) { | |
4945 | case '\'': | |
4946 | case '\"': | |
4947 | if (expected_delimiter != '\0') { | |
4948 | if (c == expected_delimiter) | |
4949 | expected_delimiter = '\0'; | |
4950 | } else | |
4951 | expected_delimiter = c; | |
4952 | break; | |
4953 | ||
4954 | case '\\': | |
4955 | /* Backslash quotes delimiters and itself, but not macro args. */ | |
4956 | if (expected_delimiter != 0 && p < limit | |
4957 | && (*p == expected_delimiter || *p == '\\')) { | |
4958 | *exp_p++ = *p++; | |
4959 | continue; | |
4960 | } | |
4961 | break; | |
4962 | ||
4963 | case '/': | |
4964 | if (expected_delimiter != '\0') /* No comments inside strings. */ | |
4965 | break; | |
4966 | if (*p == '*') { | |
4967 | /* If we find a comment that wasn't removed by handle_directive, | |
4968 | this must be -traditional. So replace the comment with | |
4969 | nothing at all. */ | |
4970 | exp_p--; | |
4971 | p += 1; | |
4972 | while (p < limit && !(p[-2] == '*' && p[-1] == '/')) | |
4973 | p++; | |
4974 | #if 0 | |
4975 | /* Mark this as a concatenation-point, as if it had been ##. */ | |
4976 | concat = p; | |
4977 | #endif | |
4978 | } | |
4979 | break; | |
4980 | } | |
4981 | } | |
4982 | ||
4983 | /* Handle the start of a symbol. */ | |
4984 | if (is_idchar[c] && nargs > 0) { | |
4985 | U_CHAR *id_beg = p - 1; | |
4986 | int id_len; | |
4987 | ||
4988 | --exp_p; | |
4989 | while (p != limit && is_idchar[*p]) p++; | |
4990 | id_len = p - id_beg; | |
4991 | ||
4992 | if (is_idstart[c]) { | |
4993 | register struct arglist *arg; | |
4994 | ||
4995 | for (arg = arglist; arg != NULL; arg = arg->next) { | |
4996 | struct reflist *tpat; | |
4997 | ||
4998 | if (arg->name[0] == c | |
4999 | && arg->length == id_len | |
5000 | && strncmp (arg->name, id_beg, id_len) == 0) { | |
5001 | if (expected_delimiter && warn_stringify) { | |
5002 | if (traditional) { | |
5003 | warning ("macro argument `%.*s' is stringified.", | |
5004 | id_len, arg->name); | |
5005 | } else { | |
5006 | warning ("macro arg `%.*s' would be stringified with -traditional.", | |
5007 | id_len, arg->name); | |
5008 | } | |
5009 | } | |
5010 | /* If ANSI, don't actually substitute inside a string. */ | |
5011 | if (!traditional && expected_delimiter) | |
5012 | break; | |
5013 | /* make a pat node for this arg and append it to the end of | |
5014 | the pat list */ | |
5015 | tpat = (struct reflist *) xmalloc (sizeof (struct reflist)); | |
5016 | tpat->next = NULL; | |
5017 | tpat->raw_before = concat == id_beg; | |
5018 | tpat->raw_after = 0; | |
5019 | tpat->stringify = (traditional ? expected_delimiter != '\0' | |
5020 | : stringify == id_beg); | |
5021 | ||
5022 | if (endpat == NULL) | |
5023 | defn->pattern = tpat; | |
5024 | else | |
5025 | endpat->next = tpat; | |
5026 | endpat = tpat; | |
5027 | ||
5028 | tpat->argno = arg->argno; | |
5029 | tpat->nchars = exp_p - lastp; | |
5030 | { | |
5031 | register U_CHAR *p1 = p; | |
5032 | SKIP_WHITE_SPACE (p1); | |
5033 | if (p1 + 2 <= limit && p1[0] == '#' && p1[1] == '#') | |
5034 | tpat->raw_after = 1; | |
5035 | } | |
5036 | lastp = exp_p; /* place to start copying from next time */ | |
5037 | skipped_arg = 1; | |
5038 | break; | |
5039 | } | |
5040 | } | |
5041 | } | |
5042 | ||
5043 | /* If this was not a macro arg, copy it into the expansion. */ | |
5044 | if (! skipped_arg) { | |
5045 | register U_CHAR *lim1 = p; | |
5046 | p = id_beg; | |
5047 | while (p != lim1) | |
5048 | *exp_p++ = *p++; | |
5049 | if (stringify == id_beg) | |
5050 | error ("`#' operator should be followed by a macro argument name"); | |
5051 | } | |
5052 | } | |
5053 | } | |
5054 | ||
5055 | if (limit < end) { | |
5056 | /* Convert trailing whitespace to Newline-markers. */ | |
5057 | while (limit < end && is_space[*limit]) { | |
5058 | *exp_p++ = '\n'; | |
5059 | *exp_p++ = *limit++; | |
5060 | } | |
5061 | } else if (!traditional) { | |
5062 | /* There is no trailing whitespace, so invent some. */ | |
5063 | *exp_p++ = '\n'; | |
5064 | *exp_p++ = ' '; | |
5065 | } | |
5066 | ||
5067 | *exp_p = '\0'; | |
5068 | ||
5069 | defn->length = exp_p - defn->expansion; | |
5070 | ||
5071 | /* Crash now if we overrun the allocated size. */ | |
5072 | if (defn->length + 1 > maxsize) | |
5073 | abort (); | |
5074 | ||
5075 | #if 0 | |
5076 | /* This isn't worth the time it takes. */ | |
5077 | /* give back excess storage */ | |
5078 | defn->expansion = (U_CHAR *) xrealloc (defn->expansion, defn->length + 1); | |
5079 | #endif | |
5080 | ||
5081 | return defn; | |
5082 | } | |
5083 | \f | |
5084 | static int | |
5085 | do_assert (buf, limit, op, keyword) | |
5086 | U_CHAR *buf, *limit; | |
5087 | FILE_BUF *op; | |
5088 | struct directive *keyword; | |
5089 | { | |
5090 | U_CHAR *bp; /* temp ptr into input buffer */ | |
5091 | U_CHAR *symname; /* remember where symbol name starts */ | |
5092 | int sym_length; /* and how long it is */ | |
5093 | struct arglist *tokens = NULL; | |
5094 | ||
5095 | if (pedantic && done_initializing && !instack[indepth].system_header_p) | |
5096 | pedwarn ("ANSI C does not allow `#assert'"); | |
5097 | ||
5098 | bp = buf; | |
5099 | ||
5100 | while (is_hor_space[*bp]) | |
5101 | bp++; | |
5102 | ||
5103 | symname = bp; /* remember where it starts */ | |
5104 | sym_length = check_macro_name (bp, "assertion"); | |
5105 | bp += sym_length; | |
5106 | /* #define doesn't do this, but we should. */ | |
5107 | SKIP_WHITE_SPACE (bp); | |
5108 | ||
5109 | /* Lossage will occur if identifiers or control tokens are broken | |
5110 | across lines using backslash. This is not the right place to take | |
5111 | care of that. */ | |
5112 | ||
5113 | if (*bp != '(') { | |
5114 | error ("missing token-sequence in `#assert'"); | |
5115 | return 1; | |
5116 | } | |
5117 | ||
5118 | { | |
5119 | int error_flag = 0; | |
5120 | ||
5121 | bp++; /* skip '(' */ | |
5122 | SKIP_WHITE_SPACE (bp); | |
5123 | ||
5124 | tokens = read_token_list (&bp, limit, &error_flag); | |
5125 | if (error_flag) | |
5126 | return 1; | |
5127 | if (tokens == 0) { | |
5128 | error ("empty token-sequence in `#assert'"); | |
5129 | return 1; | |
5130 | } | |
5131 | ||
5132 | ++bp; /* skip paren */ | |
5133 | SKIP_WHITE_SPACE (bp); | |
5134 | } | |
5135 | ||
5136 | /* If this name isn't already an assertion name, make it one. | |
5137 | Error if it was already in use in some other way. */ | |
5138 | ||
5139 | { | |
5140 | ASSERTION_HASHNODE *hp; | |
5141 | int hashcode = hashf (symname, sym_length, ASSERTION_HASHSIZE); | |
5142 | struct tokenlist_list *value | |
5143 | = (struct tokenlist_list *) xmalloc (sizeof (struct tokenlist_list)); | |
5144 | ||
5145 | hp = assertion_lookup (symname, sym_length, hashcode); | |
5146 | if (hp == NULL) { | |
5147 | if (sym_length == 7 && ! strncmp (symname, "defined", sym_length)) | |
5148 | error ("`defined' redefined as assertion"); | |
5149 | hp = assertion_install (symname, sym_length, hashcode); | |
5150 | } | |
5151 | ||
5152 | /* Add the spec'd token-sequence to the list of such. */ | |
5153 | value->tokens = tokens; | |
5154 | value->next = hp->value; | |
5155 | hp->value = value; | |
5156 | } | |
5157 | ||
5158 | return 0; | |
5159 | } | |
5160 | \f | |
5161 | static int | |
5162 | do_unassert (buf, limit, op, keyword) | |
5163 | U_CHAR *buf, *limit; | |
5164 | FILE_BUF *op; | |
5165 | struct directive *keyword; | |
5166 | { | |
5167 | U_CHAR *bp; /* temp ptr into input buffer */ | |
5168 | U_CHAR *symname; /* remember where symbol name starts */ | |
5169 | int sym_length; /* and how long it is */ | |
5170 | ||
5171 | struct arglist *tokens = NULL; | |
5172 | int tokens_specified = 0; | |
5173 | ||
5174 | if (pedantic && done_initializing && !instack[indepth].system_header_p) | |
5175 | pedwarn ("ANSI C does not allow `#unassert'"); | |
5176 | ||
5177 | bp = buf; | |
5178 | ||
5179 | while (is_hor_space[*bp]) | |
5180 | bp++; | |
5181 | ||
5182 | symname = bp; /* remember where it starts */ | |
5183 | sym_length = check_macro_name (bp, "assertion"); | |
5184 | bp += sym_length; | |
5185 | /* #define doesn't do this, but we should. */ | |
5186 | SKIP_WHITE_SPACE (bp); | |
5187 | ||
5188 | /* Lossage will occur if identifiers or control tokens are broken | |
5189 | across lines using backslash. This is not the right place to take | |
5190 | care of that. */ | |
5191 | ||
5192 | if (*bp == '(') { | |
5193 | int error_flag = 0; | |
5194 | ||
5195 | bp++; /* skip '(' */ | |
5196 | SKIP_WHITE_SPACE (bp); | |
5197 | ||
5198 | tokens = read_token_list (&bp, limit, &error_flag); | |
5199 | if (error_flag) | |
5200 | return 1; | |
5201 | if (tokens == 0) { | |
5202 | error ("empty token list in `#unassert'"); | |
5203 | return 1; | |
5204 | } | |
5205 | ||
5206 | tokens_specified = 1; | |
5207 | ||
5208 | ++bp; /* skip paren */ | |
5209 | SKIP_WHITE_SPACE (bp); | |
5210 | } | |
5211 | ||
5212 | { | |
5213 | ASSERTION_HASHNODE *hp; | |
5214 | int hashcode = hashf (symname, sym_length, ASSERTION_HASHSIZE); | |
5215 | struct tokenlist_list *tail, *prev; | |
5216 | ||
5217 | hp = assertion_lookup (symname, sym_length, hashcode); | |
5218 | if (hp == NULL) | |
5219 | return 1; | |
5220 | ||
5221 | /* If no token list was specified, then eliminate this assertion | |
5222 | entirely. */ | |
5223 | if (! tokens_specified) { | |
5224 | struct tokenlist_list *next; | |
5225 | for (tail = hp->value; tail; tail = next) { | |
5226 | next = tail->next; | |
5227 | free_token_list (tail->tokens); | |
5228 | free (tail); | |
5229 | } | |
5230 | delete_assertion (hp); | |
5231 | } else { | |
5232 | /* If a list of tokens was given, then delete any matching list. */ | |
5233 | ||
5234 | tail = hp->value; | |
5235 | prev = 0; | |
5236 | while (tail) { | |
5237 | struct tokenlist_list *next = tail->next; | |
5238 | if (compare_token_lists (tail->tokens, tokens)) { | |
5239 | if (prev) | |
5240 | prev->next = next; | |
5241 | else | |
5242 | hp->value = tail->next; | |
5243 | free_token_list (tail->tokens); | |
5244 | free (tail); | |
5245 | } else { | |
5246 | prev = tail; | |
5247 | } | |
5248 | tail = next; | |
5249 | } | |
5250 | } | |
5251 | } | |
5252 | ||
5253 | return 0; | |
5254 | } | |
5255 | \f | |
5256 | /* Test whether there is an assertion named NAME | |
5257 | and optionally whether it has an asserted token list TOKENS. | |
5258 | NAME is not null terminated; its length is SYM_LENGTH. | |
5259 | If TOKENS_SPECIFIED is 0, then don't check for any token list. */ | |
5260 | ||
5261 | int | |
5262 | check_assertion (name, sym_length, tokens_specified, tokens) | |
5263 | U_CHAR *name; | |
5264 | int sym_length; | |
5265 | int tokens_specified; | |
5266 | struct arglist *tokens; | |
5267 | { | |
5268 | ASSERTION_HASHNODE *hp; | |
5269 | int hashcode = hashf (name, sym_length, ASSERTION_HASHSIZE); | |
5270 | ||
5271 | if (pedantic && !instack[indepth].system_header_p) | |
5272 | pedwarn ("ANSI C does not allow testing assertions"); | |
5273 | ||
5274 | hp = assertion_lookup (name, sym_length, hashcode); | |
5275 | if (hp == NULL) | |
5276 | /* It is not an assertion; just return false. */ | |
5277 | return 0; | |
5278 | ||
5279 | /* If no token list was specified, then value is 1. */ | |
5280 | if (! tokens_specified) | |
5281 | return 1; | |
5282 | ||
5283 | { | |
5284 | struct tokenlist_list *tail; | |
5285 | ||
5286 | tail = hp->value; | |
5287 | ||
5288 | /* If a list of tokens was given, | |
5289 | then succeed if the assertion records a matching list. */ | |
5290 | ||
5291 | while (tail) { | |
5292 | if (compare_token_lists (tail->tokens, tokens)) | |
5293 | return 1; | |
5294 | tail = tail->next; | |
5295 | } | |
5296 | ||
5297 | /* Fail if the assertion has no matching list. */ | |
5298 | return 0; | |
5299 | } | |
5300 | } | |
5301 | ||
5302 | /* Compare two lists of tokens for equality including order of tokens. */ | |
5303 | ||
5304 | static int | |
5305 | compare_token_lists (l1, l2) | |
5306 | struct arglist *l1, *l2; | |
5307 | { | |
5308 | while (l1 && l2) { | |
5309 | if (l1->length != l2->length) | |
5310 | return 0; | |
5311 | if (strncmp (l1->name, l2->name, l1->length)) | |
5312 | return 0; | |
5313 | l1 = l1->next; | |
5314 | l2 = l2->next; | |
5315 | } | |
5316 | ||
5317 | /* Succeed if both lists end at the same time. */ | |
5318 | return l1 == l2; | |
5319 | } | |
5320 | \f | |
5321 | /* Read a space-separated list of tokens ending in a close parenthesis. | |
5322 | Return a list of strings, in the order they were written. | |
5323 | (In case of error, return 0 and store -1 in *ERROR_FLAG.) | |
5324 | Parse the text starting at *BPP, and update *BPP. | |
5325 | Don't parse beyond LIMIT. */ | |
5326 | ||
5327 | static struct arglist * | |
5328 | read_token_list (bpp, limit, error_flag) | |
5329 | U_CHAR **bpp; | |
5330 | U_CHAR *limit; | |
5331 | int *error_flag; | |
5332 | { | |
5333 | struct arglist *token_ptrs = 0; | |
5334 | U_CHAR *bp = *bpp; | |
5335 | int depth = 1; | |
5336 | ||
5337 | *error_flag = 0; | |
5338 | ||
5339 | /* Loop over the assertion value tokens. */ | |
5340 | while (depth > 0) { | |
5341 | struct arglist *temp; | |
5342 | int eofp = 0; | |
5343 | U_CHAR *beg = bp; | |
5344 | ||
5345 | /* Find the end of the token. */ | |
5346 | if (*bp == '(') { | |
5347 | bp++; | |
5348 | depth++; | |
5349 | } else if (*bp == ')') { | |
5350 | depth--; | |
5351 | if (depth == 0) | |
5352 | break; | |
5353 | bp++; | |
5354 | } else if (*bp == '"' || *bp == '\'') | |
5355 | bp = skip_quoted_string (bp, limit, 0, 0, 0, &eofp); | |
5356 | else | |
5357 | while (! is_hor_space[*bp] && *bp != '(' && *bp != ')' | |
5358 | && *bp != '"' && *bp != '\'' && bp != limit) | |
5359 | bp++; | |
5360 | ||
5361 | temp = (struct arglist *) xmalloc (sizeof (struct arglist)); | |
5362 | temp->name = (U_CHAR *) xmalloc (bp - beg + 1); | |
5363 | bcopy (beg, temp->name, bp - beg); | |
5364 | temp->name[bp - beg] = 0; | |
5365 | temp->next = token_ptrs; | |
5366 | token_ptrs = temp; | |
5367 | temp->length = bp - beg; | |
5368 | ||
5369 | SKIP_WHITE_SPACE (bp); | |
5370 | ||
5371 | if (bp >= limit) { | |
5372 | error ("unterminated token sequence in `#assert' or `#unassert'"); | |
5373 | *error_flag = -1; | |
5374 | return 0; | |
5375 | } | |
5376 | } | |
5377 | *bpp = bp; | |
5378 | ||
5379 | /* We accumulated the names in reverse order. | |
5380 | Now reverse them to get the proper order. */ | |
5381 | { | |
5382 | register struct arglist *prev = 0, *this, *next; | |
5383 | for (this = token_ptrs; this; this = next) { | |
5384 | next = this->next; | |
5385 | this->next = prev; | |
5386 | prev = this; | |
5387 | } | |
5388 | return prev; | |
5389 | } | |
5390 | } | |
5391 | ||
5392 | static void | |
5393 | free_token_list (tokens) | |
5394 | struct arglist *tokens; | |
5395 | { | |
5396 | while (tokens) { | |
5397 | struct arglist *next = tokens->next; | |
5398 | free (tokens->name); | |
5399 | free (tokens); | |
5400 | tokens = next; | |
5401 | } | |
5402 | } | |
5403 | \f | |
5404 | /* | |
5405 | * Install a name in the assertion hash table. | |
5406 | * | |
5407 | * If LEN is >= 0, it is the length of the name. | |
5408 | * Otherwise, compute the length by scanning the entire name. | |
5409 | * | |
5410 | * If HASH is >= 0, it is the precomputed hash code. | |
5411 | * Otherwise, compute the hash code. | |
5412 | */ | |
5413 | static ASSERTION_HASHNODE * | |
5414 | assertion_install (name, len, hash) | |
5415 | U_CHAR *name; | |
5416 | int len; | |
5417 | int hash; | |
5418 | { | |
5419 | register ASSERTION_HASHNODE *hp; | |
5420 | register int i, bucket; | |
5421 | register U_CHAR *p, *q; | |
5422 | ||
5423 | i = sizeof (ASSERTION_HASHNODE) + len + 1; | |
5424 | hp = (ASSERTION_HASHNODE *) xmalloc (i); | |
5425 | bucket = hash; | |
5426 | hp->bucket_hdr = &assertion_hashtab[bucket]; | |
5427 | hp->next = assertion_hashtab[bucket]; | |
5428 | assertion_hashtab[bucket] = hp; | |
5429 | hp->prev = NULL; | |
5430 | if (hp->next != NULL) | |
5431 | hp->next->prev = hp; | |
5432 | hp->length = len; | |
5433 | hp->value = 0; | |
5434 | hp->name = ((U_CHAR *) hp) + sizeof (ASSERTION_HASHNODE); | |
5435 | p = hp->name; | |
5436 | q = name; | |
5437 | for (i = 0; i < len; i++) | |
5438 | *p++ = *q++; | |
5439 | hp->name[len] = 0; | |
5440 | return hp; | |
5441 | } | |
5442 | ||
5443 | /* | |
5444 | * find the most recent hash node for name name (ending with first | |
5445 | * non-identifier char) installed by install | |
5446 | * | |
5447 | * If LEN is >= 0, it is the length of the name. | |
5448 | * Otherwise, compute the length by scanning the entire name. | |
5449 | * | |
5450 | * If HASH is >= 0, it is the precomputed hash code. | |
5451 | * Otherwise, compute the hash code. | |
5452 | */ | |
5453 | static ASSERTION_HASHNODE * | |
5454 | assertion_lookup (name, len, hash) | |
5455 | U_CHAR *name; | |
5456 | int len; | |
5457 | int hash; | |
5458 | { | |
5459 | register U_CHAR *bp; | |
5460 | register ASSERTION_HASHNODE *bucket; | |
5461 | ||
5462 | bucket = assertion_hashtab[hash]; | |
5463 | while (bucket) { | |
5464 | if (bucket->length == len && strncmp (bucket->name, name, len) == 0) | |
5465 | return bucket; | |
5466 | bucket = bucket->next; | |
5467 | } | |
5468 | return NULL; | |
5469 | } | |
5470 | ||
5471 | static void | |
5472 | delete_assertion (hp) | |
5473 | ASSERTION_HASHNODE *hp; | |
5474 | { | |
5475 | ||
5476 | if (hp->prev != NULL) | |
5477 | hp->prev->next = hp->next; | |
5478 | if (hp->next != NULL) | |
5479 | hp->next->prev = hp->prev; | |
5480 | ||
5481 | /* make sure that the bucket chain header that | |
5482 | the deleted guy was on points to the right thing afterwards. */ | |
5483 | if (hp == *hp->bucket_hdr) | |
5484 | *hp->bucket_hdr = hp->next; | |
5485 | ||
5486 | free (hp); | |
5487 | } | |
5488 | \f | |
5489 | /* | |
5490 | * interpret #line command. Remembers previously seen fnames | |
5491 | * in its very own hash table. | |
5492 | */ | |
5493 | #define FNAME_HASHSIZE 37 | |
5494 | ||
5495 | static int | |
5496 | do_line (buf, limit, op, keyword) | |
5497 | U_CHAR *buf, *limit; | |
5498 | FILE_BUF *op; | |
5499 | struct directive *keyword; | |
5500 | { | |
5501 | register U_CHAR *bp; | |
5502 | FILE_BUF *ip = &instack[indepth]; | |
5503 | FILE_BUF tem; | |
5504 | int new_lineno; | |
5505 | enum file_change_code file_change = same_file; | |
5506 | ||
5507 | /* Expand any macros. */ | |
5508 | tem = expand_to_temp_buffer (buf, limit, 0, 0); | |
5509 | ||
5510 | /* Point to macroexpanded line, which is null-terminated now. */ | |
5511 | bp = tem.buf; | |
5512 | SKIP_WHITE_SPACE (bp); | |
5513 | ||
5514 | if (!isdigit (*bp)) { | |
5515 | error ("invalid format `#line' command"); | |
5516 | return 0; | |
5517 | } | |
5518 | ||
5519 | /* The Newline at the end of this line remains to be processed. | |
5520 | To put the next line at the specified line number, | |
5521 | we must store a line number now that is one less. */ | |
5522 | new_lineno = atoi (bp) - 1; | |
5523 | ||
5524 | /* skip over the line number. */ | |
5525 | while (isdigit (*bp)) | |
5526 | bp++; | |
5527 | ||
5528 | #if 0 /* #line 10"foo.c" is supposed to be allowed. */ | |
5529 | if (*bp && !is_space[*bp]) { | |
5530 | error ("invalid format `#line' command"); | |
5531 | return; | |
5532 | } | |
5533 | #endif | |
5534 | ||
5535 | SKIP_WHITE_SPACE (bp); | |
5536 | ||
5537 | if (*bp == '\"') { | |
5538 | static HASHNODE *fname_table[FNAME_HASHSIZE]; | |
5539 | HASHNODE *hp, **hash_bucket; | |
5540 | U_CHAR *fname; | |
5541 | int fname_length; | |
5542 | ||
5543 | fname = ++bp; | |
5544 | ||
5545 | while (*bp && *bp != '\"') | |
5546 | bp++; | |
5547 | if (*bp != '\"') { | |
5548 | error ("invalid format `#line' command"); | |
5549 | return 0; | |
5550 | } | |
5551 | ||
5552 | fname_length = bp - fname; | |
5553 | ||
5554 | bp++; | |
5555 | SKIP_WHITE_SPACE (bp); | |
5556 | if (*bp) { | |
5557 | if (*bp == '1') | |
5558 | file_change = enter_file; | |
5559 | else if (*bp == '2') | |
5560 | file_change = leave_file; | |
5561 | else { | |
5562 | error ("invalid format `#line' command"); | |
5563 | return 0; | |
5564 | } | |
5565 | ||
5566 | bp++; | |
5567 | SKIP_WHITE_SPACE (bp); | |
5568 | if (*bp) { | |
5569 | error ("invalid format `#line' command"); | |
5570 | return 0; | |
5571 | } | |
5572 | } | |
5573 | ||
5574 | hash_bucket = | |
5575 | &fname_table[hashf (fname, fname_length, FNAME_HASHSIZE)]; | |
5576 | for (hp = *hash_bucket; hp != NULL; hp = hp->next) | |
5577 | if (hp->length == fname_length && | |
5578 | strncmp (hp->value.cpval, fname, fname_length) == 0) { | |
5579 | ip->nominal_fname = hp->value.cpval; | |
5580 | break; | |
5581 | } | |
5582 | if (hp == 0) { | |
5583 | /* Didn't find it; cons up a new one. */ | |
5584 | hp = (HASHNODE *) xcalloc (1, sizeof (HASHNODE) + fname_length + 1); | |
5585 | hp->next = *hash_bucket; | |
5586 | *hash_bucket = hp; | |
5587 | ||
5588 | hp->length = fname_length; | |
5589 | ip->nominal_fname = hp->value.cpval = ((char *) hp) + sizeof (HASHNODE); | |
5590 | bcopy (fname, hp->value.cpval, fname_length); | |
5591 | } | |
5592 | } else if (*bp) { | |
5593 | error ("invalid format `#line' command"); | |
5594 | return 0; | |
5595 | } | |
5596 | ||
5597 | ip->lineno = new_lineno; | |
5598 | output_line_command (ip, op, 0, file_change); | |
5599 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
5600 | return 0; | |
5601 | } | |
5602 | ||
5603 | /* | |
5604 | * remove the definition of a symbol from the symbol table. | |
5605 | * according to un*x /lib/cpp, it is not an error to undef | |
5606 | * something that has no definitions, so it isn't one here either. | |
5607 | */ | |
5608 | ||
5609 | static int | |
5610 | do_undef (buf, limit, op, keyword) | |
5611 | U_CHAR *buf, *limit; | |
5612 | FILE_BUF *op; | |
5613 | struct directive *keyword; | |
5614 | { | |
5615 | int sym_length; | |
5616 | HASHNODE *hp; | |
5617 | U_CHAR *orig_buf = buf; | |
5618 | ||
5619 | /* If this is a precompiler run (with -pcp) pass thru #undef commands. */ | |
5620 | if (pcp_outfile && op) | |
5621 | pass_thru_directive (buf, limit, op, keyword); | |
5622 | ||
5623 | SKIP_WHITE_SPACE (buf); | |
5624 | sym_length = check_macro_name (buf, "macro"); | |
5625 | ||
5626 | while ((hp = lookup (buf, sym_length, -1)) != NULL) { | |
5627 | /* If we are generating additional info for debugging (with -g) we | |
5628 | need to pass through all effective #undef commands. */ | |
5629 | if (debug_output && op) | |
5630 | pass_thru_directive (orig_buf, limit, op, keyword); | |
5631 | if (hp->type != T_MACRO) | |
5632 | warning ("undefining `%s'", hp->name); | |
5633 | delete_macro (hp); | |
5634 | } | |
5635 | ||
5636 | if (pedantic) { | |
5637 | buf += sym_length; | |
5638 | SKIP_WHITE_SPACE (buf); | |
5639 | if (buf != limit) | |
5640 | pedwarn ("garbage after `#undef' directive"); | |
5641 | } | |
5642 | return 0; | |
5643 | } | |
5644 | \f | |
5645 | /* | |
5646 | * Report a fatal error detected by the program we are processing. | |
5647 | * Use the text of the line in the error message, then terminate. | |
5648 | * (We use error() because it prints the filename & line#.) | |
5649 | */ | |
5650 | ||
5651 | static int | |
5652 | do_error (buf, limit, op, keyword) | |
5653 | U_CHAR *buf, *limit; | |
5654 | FILE_BUF *op; | |
5655 | struct directive *keyword; | |
5656 | { | |
5657 | int length = limit - buf; | |
5658 | char *copy = (char *) xmalloc (length + 1); | |
5659 | bcopy (buf, copy, length); | |
5660 | copy[length] = 0; | |
5661 | SKIP_WHITE_SPACE (copy); | |
5662 | error ("#error %s", copy); | |
5663 | exit (FAILURE_EXIT_CODE); | |
5664 | /* NOTREACHED */ | |
5665 | return 0; | |
5666 | } | |
5667 | ||
5668 | /* | |
5669 | * Report a warning detected by the program we are processing. | |
5670 | * Use the text of the line in the warning message, then continue. | |
5671 | * (We use error() because it prints the filename & line#.) | |
5672 | */ | |
5673 | ||
5674 | static int | |
5675 | do_warning (buf, limit, op, keyword) | |
5676 | U_CHAR *buf, *limit; | |
5677 | FILE_BUF *op; | |
5678 | struct directive *keyword; | |
5679 | { | |
5680 | int length = limit - buf; | |
5681 | char *copy = (char *) xmalloc (length + 1); | |
5682 | bcopy (buf, copy, length); | |
5683 | copy[length] = 0; | |
5684 | SKIP_WHITE_SPACE (copy); | |
5685 | error ("#warning %s", copy); | |
5686 | return 0; | |
5687 | } | |
5688 | ||
5689 | /* Remember the name of the current file being read from so that we can | |
5690 | avoid ever including it again. */ | |
5691 | ||
5692 | static int | |
5693 | do_once () | |
5694 | { | |
5695 | int i; | |
5696 | FILE_BUF *ip = NULL; | |
5697 | ||
5698 | for (i = indepth; i >= 0; i--) | |
5699 | if (instack[i].fname != NULL) { | |
5700 | ip = &instack[i]; | |
5701 | break; | |
5702 | } | |
5703 | ||
5704 | if (ip != NULL) { | |
5705 | struct file_name_list *new; | |
5706 | ||
5707 | new = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
5708 | new->next = dont_repeat_files; | |
5709 | dont_repeat_files = new; | |
5710 | new->fname = savestring (ip->fname); | |
5711 | new->control_macro = 0; | |
5712 | } | |
5713 | return 0; | |
5714 | } | |
5715 | ||
5716 | /* #ident has already been copied to the output file, so just ignore it. */ | |
5717 | ||
5718 | static int | |
5719 | do_ident (buf, limit) | |
5720 | U_CHAR *buf, *limit; | |
5721 | { | |
5722 | /* Allow #ident in system headers, since that's not user's fault. */ | |
5723 | if (pedantic && !instack[indepth].system_header_p) | |
5724 | pedwarn ("ANSI C does not allow `#ident'"); | |
5725 | return 0; | |
5726 | } | |
5727 | ||
5728 | /* #pragma and its argument line have already been copied to the output file. | |
5729 | Here just check for recognized pragmas. */ | |
5730 | ||
5731 | static int | |
5732 | do_pragma (buf, limit) | |
5733 | U_CHAR *buf, *limit; | |
5734 | { | |
5735 | while (*buf == ' ' || *buf == '\t') | |
5736 | buf++; | |
5737 | if (!strncmp (buf, "once", 4)) { | |
5738 | warning ("`#pragma once' is obsolete"); | |
5739 | do_once (); | |
5740 | } | |
5741 | return 0; | |
5742 | } | |
5743 | ||
5744 | #if 0 | |
5745 | /* This was a fun hack, but #pragma seems to start to be useful. | |
5746 | By failing to recognize it, we pass it through unchanged to cc1. */ | |
5747 | ||
5748 | /* | |
5749 | * the behavior of the #pragma directive is implementation defined. | |
5750 | * this implementation defines it as follows. | |
5751 | */ | |
5752 | ||
5753 | static int | |
5754 | do_pragma () | |
5755 | { | |
5756 | close (0); | |
5757 | if (open ("/dev/tty", O_RDONLY, 0666) != 0) | |
5758 | goto nope; | |
5759 | close (1); | |
5760 | if (open ("/dev/tty", O_WRONLY, 0666) != 1) | |
5761 | goto nope; | |
5762 | execl ("/usr/games/hack", "#pragma", 0); | |
5763 | execl ("/usr/games/rogue", "#pragma", 0); | |
5764 | execl ("/usr/new/emacs", "-f", "hanoi", "9", "-kill", 0); | |
5765 | execl ("/usr/local/emacs", "-f", "hanoi", "9", "-kill", 0); | |
5766 | nope: | |
5767 | fatal ("You are in a maze of twisty compiler features, all different"); | |
5768 | } | |
5769 | #endif | |
5770 | ||
5771 | /* Just ignore #sccs, on systems where we define it at all. */ | |
5772 | ||
5773 | static int | |
5774 | do_sccs () | |
5775 | { | |
5776 | if (pedantic) | |
5777 | pedwarn ("ANSI C does not allow `#sccs'"); | |
5778 | return 0; | |
5779 | } | |
5780 | \f | |
5781 | /* | |
5782 | * handle #if command by | |
5783 | * 1) inserting special `defined' keyword into the hash table | |
5784 | * that gets turned into 0 or 1 by special_symbol (thus, | |
5785 | * if the luser has a symbol called `defined' already, it won't | |
5786 | * work inside the #if command) | |
5787 | * 2) rescan the input into a temporary output buffer | |
5788 | * 3) pass the output buffer to the yacc parser and collect a value | |
5789 | * 4) clean up the mess left from steps 1 and 2. | |
5790 | * 5) call conditional_skip to skip til the next #endif (etc.), | |
5791 | * or not, depending on the value from step 3. | |
5792 | */ | |
5793 | ||
5794 | static int | |
5795 | do_if (buf, limit, op, keyword) | |
5796 | U_CHAR *buf, *limit; | |
5797 | FILE_BUF *op; | |
5798 | struct directive *keyword; | |
5799 | { | |
5800 | int value; | |
5801 | FILE_BUF *ip = &instack[indepth]; | |
5802 | ||
5803 | value = eval_if_expression (buf, limit - buf); | |
5804 | conditional_skip (ip, value == 0, T_IF, 0); | |
5805 | return 0; | |
5806 | } | |
5807 | ||
5808 | /* | |
5809 | * handle a #elif directive by not changing if_stack either. | |
5810 | * see the comment above do_else. | |
5811 | */ | |
5812 | ||
5813 | static int | |
5814 | do_elif (buf, limit, op, keyword) | |
5815 | U_CHAR *buf, *limit; | |
5816 | FILE_BUF *op; | |
5817 | struct directive *keyword; | |
5818 | { | |
5819 | int value; | |
5820 | FILE_BUF *ip = &instack[indepth]; | |
5821 | ||
5822 | if (if_stack == instack[indepth].if_stack) { | |
5823 | error ("`#elif' not within a conditional"); | |
5824 | return 0; | |
5825 | } else { | |
5826 | if (if_stack->type != T_IF && if_stack->type != T_ELIF) { | |
5827 | error ("`#elif' after `#else'"); | |
5828 | fprintf (stderr, " (matches line %d", if_stack->lineno); | |
5829 | if (if_stack->fname != NULL && ip->fname != NULL && | |
5830 | strcmp (if_stack->fname, ip->nominal_fname) != 0) | |
5831 | fprintf (stderr, ", file %s", if_stack->fname); | |
5832 | fprintf (stderr, ")\n"); | |
5833 | } | |
5834 | if_stack->type = T_ELIF; | |
5835 | } | |
5836 | ||
5837 | if (if_stack->if_succeeded) | |
5838 | skip_if_group (ip, 0); | |
5839 | else { | |
5840 | value = eval_if_expression (buf, limit - buf); | |
5841 | if (value == 0) | |
5842 | skip_if_group (ip, 0); | |
5843 | else { | |
5844 | ++if_stack->if_succeeded; /* continue processing input */ | |
5845 | output_line_command (ip, op, 1, same_file); | |
5846 | } | |
5847 | } | |
5848 | return 0; | |
5849 | } | |
5850 | ||
5851 | /* | |
5852 | * evaluate a #if expression in BUF, of length LENGTH, | |
5853 | * then parse the result as a C expression and return the value as an int. | |
5854 | */ | |
5855 | static int | |
5856 | eval_if_expression (buf, length) | |
5857 | U_CHAR *buf; | |
5858 | int length; | |
5859 | { | |
5860 | FILE_BUF temp_obuf; | |
5861 | HASHNODE *save_defined; | |
5862 | int value; | |
5863 | ||
5864 | save_defined = install ("defined", -1, T_SPEC_DEFINED, 0, -1); | |
5865 | pcp_inside_if = 1; | |
5866 | temp_obuf = expand_to_temp_buffer (buf, buf + length, 0, 1); | |
5867 | pcp_inside_if = 0; | |
5868 | delete_macro (save_defined); /* clean up special symbol */ | |
5869 | ||
5870 | value = parse_c_expression (temp_obuf.buf); | |
5871 | ||
5872 | free (temp_obuf.buf); | |
5873 | ||
5874 | return value; | |
5875 | } | |
5876 | ||
5877 | /* | |
5878 | * routine to handle ifdef/ifndef. Try to look up the symbol, | |
5879 | * then do or don't skip to the #endif/#else/#elif depending | |
5880 | * on what directive is actually being processed. | |
5881 | */ | |
5882 | ||
5883 | static int | |
5884 | do_xifdef (buf, limit, op, keyword) | |
5885 | U_CHAR *buf, *limit; | |
5886 | FILE_BUF *op; | |
5887 | struct directive *keyword; | |
5888 | { | |
5889 | int skip; | |
5890 | FILE_BUF *ip = &instack[indepth]; | |
5891 | U_CHAR *end; | |
5892 | int start_of_file = 0; | |
5893 | U_CHAR *control_macro = 0; | |
5894 | ||
5895 | /* Detect a #ifndef at start of file (not counting comments). */ | |
5896 | if (ip->fname != 0 && keyword->type == T_IFNDEF) { | |
5897 | U_CHAR *p = ip->buf; | |
5898 | while (p != directive_start) { | |
5899 | char c = *p++; | |
5900 | switch (c) { | |
5901 | case ' ': | |
5902 | case '\t': | |
5903 | case '\n': | |
5904 | break; | |
5905 | case '/': | |
5906 | if (p != ip->bufp && *p == '*') { | |
5907 | /* Skip this comment. */ | |
5908 | int junk; | |
5909 | U_CHAR *save_bufp = ip->bufp; | |
5910 | ip->bufp = p + 1; | |
5911 | p = skip_to_end_of_comment (ip, &junk); | |
5912 | ip->bufp = save_bufp; | |
5913 | } | |
5914 | break; | |
5915 | default: | |
5916 | goto fail; | |
5917 | } | |
5918 | } | |
5919 | /* If we get here, this conditional is the beginning of the file. */ | |
5920 | start_of_file = 1; | |
5921 | fail: ; | |
5922 | } | |
5923 | ||
5924 | /* Discard leading and trailing whitespace. */ | |
5925 | SKIP_WHITE_SPACE (buf); | |
5926 | while (limit != buf && is_hor_space[limit[-1]]) limit--; | |
5927 | ||
5928 | /* Find the end of the identifier at the beginning. */ | |
5929 | for (end = buf; is_idchar[*end]; end++); | |
5930 | ||
5931 | if (end == buf) { | |
5932 | skip = (keyword->type == T_IFDEF); | |
5933 | if (! traditional) | |
5934 | pedwarn (end == limit ? "`#%s' with no argument" | |
5935 | : "`#%s' argument starts with punctuation", | |
5936 | keyword->name); | |
5937 | } else { | |
5938 | HASHNODE *hp; | |
5939 | ||
5940 | if (pedantic && buf[0] >= '0' && buf[0] <= '9') | |
5941 | pedwarn ("`#%s' argument starts with a digit", keyword->name); | |
5942 | else if (end != limit && !traditional) | |
5943 | pedwarn ("garbage at end of `#%s' argument", keyword->name); | |
5944 | ||
5945 | hp = lookup (buf, end-buf, -1); | |
5946 | ||
5947 | if (pcp_outfile) { | |
5948 | /* Output a precondition for this macro. */ | |
5949 | if (hp && hp->value.defn->predefined) | |
5950 | fprintf(pcp_outfile, "#define %s\n", hp->name); | |
5951 | else { | |
5952 | U_CHAR *cp = buf; | |
5953 | fprintf(pcp_outfile, "#undef "); | |
5954 | while (is_idchar[*cp]) /* Ick! */ | |
5955 | fputc (*cp++, pcp_outfile); | |
5956 | putc ('\n', pcp_outfile); | |
5957 | } | |
5958 | } | |
5959 | ||
5960 | skip = (hp == NULL) ^ (keyword->type == T_IFNDEF); | |
5961 | if (start_of_file && !skip) { | |
5962 | control_macro = (U_CHAR *) xmalloc (end - buf + 1); | |
5963 | bcopy (buf, control_macro, end - buf); | |
5964 | control_macro[end - buf] = 0; | |
5965 | } | |
5966 | } | |
5967 | ||
5968 | conditional_skip (ip, skip, T_IF, control_macro); | |
5969 | return 0; | |
5970 | } | |
5971 | ||
5972 | /* Push TYPE on stack; then, if SKIP is nonzero, skip ahead. | |
5973 | If this is a #ifndef starting at the beginning of a file, | |
5974 | CONTROL_MACRO is the macro name tested by the #ifndef. | |
5975 | Otherwise, CONTROL_MACRO is 0. */ | |
5976 | ||
5977 | static void | |
5978 | conditional_skip (ip, skip, type, control_macro) | |
5979 | FILE_BUF *ip; | |
5980 | int skip; | |
5981 | enum node_type type; | |
5982 | U_CHAR *control_macro; | |
5983 | { | |
5984 | IF_STACK_FRAME *temp; | |
5985 | ||
5986 | temp = (IF_STACK_FRAME *) xcalloc (1, sizeof (IF_STACK_FRAME)); | |
5987 | temp->fname = ip->nominal_fname; | |
5988 | temp->lineno = ip->lineno; | |
5989 | temp->next = if_stack; | |
5990 | temp->control_macro = control_macro; | |
5991 | if_stack = temp; | |
5992 | ||
5993 | if_stack->type = type; | |
5994 | ||
5995 | if (skip != 0) { | |
5996 | skip_if_group (ip, 0); | |
5997 | return; | |
5998 | } else { | |
5999 | ++if_stack->if_succeeded; | |
6000 | output_line_command (ip, &outbuf, 1, same_file); | |
6001 | } | |
6002 | } | |
6003 | ||
6004 | /* | |
6005 | * skip to #endif, #else, or #elif. adjust line numbers, etc. | |
6006 | * leaves input ptr at the sharp sign found. | |
6007 | * If ANY is nonzero, return at next directive of any sort. | |
6008 | */ | |
6009 | static void | |
6010 | skip_if_group (ip, any) | |
6011 | FILE_BUF *ip; | |
6012 | int any; | |
6013 | { | |
6014 | register U_CHAR *bp = ip->bufp, *cp; | |
6015 | register U_CHAR *endb = ip->buf + ip->length; | |
6016 | struct directive *kt; | |
6017 | IF_STACK_FRAME *save_if_stack = if_stack; /* don't pop past here */ | |
6018 | U_CHAR *beg_of_line = bp; | |
6019 | register int ident_length; | |
6020 | U_CHAR *ident, *after_ident; | |
6021 | ||
6022 | while (bp < endb) { | |
6023 | switch (*bp++) { | |
6024 | case '/': /* possible comment */ | |
6025 | if (*bp == '\\' && bp[1] == '\n') | |
6026 | newline_fix (bp); | |
6027 | if (*bp == '*' | |
6028 | || ((cplusplus || objc) && *bp == '/')) { | |
6029 | ip->bufp = ++bp; | |
6030 | bp = skip_to_end_of_comment (ip, &ip->lineno); | |
6031 | } | |
6032 | break; | |
6033 | case '\"': | |
6034 | case '\'': | |
6035 | bp = skip_quoted_string (bp - 1, endb, ip->lineno, &ip->lineno, 0, 0); | |
6036 | break; | |
6037 | case '\\': | |
6038 | /* Char after backslash loses its special meaning. */ | |
6039 | if (bp < endb) { | |
6040 | if (*bp == '\n') | |
6041 | ++ip->lineno; /* But do update the line-count. */ | |
6042 | bp++; | |
6043 | } | |
6044 | break; | |
6045 | case '\n': | |
6046 | ++ip->lineno; | |
6047 | beg_of_line = bp; | |
6048 | break; | |
6049 | case '#': | |
6050 | ip->bufp = bp - 1; | |
6051 | ||
6052 | /* # keyword: a # must be first nonblank char on the line */ | |
6053 | if (beg_of_line == 0) | |
6054 | break; | |
6055 | /* Scan from start of line, skipping whitespace, comments | |
6056 | and backslash-newlines, and see if we reach this #. | |
6057 | If not, this # is not special. */ | |
6058 | bp = beg_of_line; | |
6059 | while (1) { | |
6060 | if (is_hor_space[*bp]) | |
6061 | bp++; | |
6062 | else if (*bp == '\\' && bp[1] == '\n') | |
6063 | bp += 2; | |
6064 | else if (*bp == '/' && bp[1] == '*') { | |
6065 | bp += 2; | |
6066 | while (!(*bp == '*' && bp[1] == '/')) | |
6067 | bp++; | |
6068 | bp += 2; | |
6069 | } else if ((cplusplus || objc) && *bp == '/' && bp[1] == '/') { | |
6070 | bp += 2; | |
6071 | while (*bp++ != '\n') ; | |
6072 | } | |
6073 | else break; | |
6074 | } | |
6075 | if (bp != ip->bufp) { | |
6076 | bp = ip->bufp + 1; /* Reset bp to after the #. */ | |
6077 | break; | |
6078 | } | |
6079 | ||
6080 | bp = ip->bufp + 1; /* Point after the '#' */ | |
6081 | ||
6082 | /* Skip whitespace and \-newline. */ | |
6083 | while (1) { | |
6084 | if (is_hor_space[*bp]) | |
6085 | bp++; | |
6086 | else if (*bp == '\\' && bp[1] == '\n') | |
6087 | bp += 2; | |
6088 | else if (*bp == '/' && bp[1] == '*') { | |
6089 | bp += 2; | |
6090 | while (!(*bp == '*' && bp[1] == '/')) { | |
6091 | if (*bp == '\n') | |
6092 | ip->lineno++; | |
6093 | bp++; | |
6094 | } | |
6095 | bp += 2; | |
6096 | } else if ((cplusplus || objc) && *bp == '/' && bp[1] == '/') { | |
6097 | bp += 2; | |
6098 | while (*bp++ != '\n') ; | |
6099 | } | |
6100 | else break; | |
6101 | } | |
6102 | ||
6103 | cp = bp; | |
6104 | ||
6105 | /* Now find end of directive name. | |
6106 | If we encounter a backslash-newline, exchange it with any following | |
6107 | symbol-constituents so that we end up with a contiguous name. */ | |
6108 | ||
6109 | while (1) { | |
6110 | if (is_idchar[*bp]) | |
6111 | bp++; | |
6112 | else { | |
6113 | if (*bp == '\\' && bp[1] == '\n') | |
6114 | name_newline_fix (bp); | |
6115 | if (is_idchar[*bp]) | |
6116 | bp++; | |
6117 | else break; | |
6118 | } | |
6119 | } | |
6120 | ident_length = bp - cp; | |
6121 | ident = cp; | |
6122 | after_ident = bp; | |
6123 | ||
6124 | /* A line of just `#' becomes blank. */ | |
6125 | ||
6126 | if (ident_length == 0 && *after_ident == '\n') { | |
6127 | continue; | |
6128 | } | |
6129 | ||
6130 | if (ident_length == 0 || !is_idstart[*ident]) { | |
6131 | U_CHAR *p = ident; | |
6132 | while (is_idchar[*p]) { | |
6133 | if (*p < '0' || *p > '9') | |
6134 | break; | |
6135 | p++; | |
6136 | } | |
6137 | /* Handle # followed by a line number. */ | |
6138 | if (p != ident && !is_idchar[*p]) { | |
6139 | if (pedantic) | |
6140 | pedwarn ("`#' followed by integer"); | |
6141 | continue; | |
6142 | } | |
6143 | ||
6144 | /* Avoid error for `###' and similar cases unless -pedantic. */ | |
6145 | if (p == ident) { | |
6146 | while (*p == '#' || is_hor_space[*p]) p++; | |
6147 | if (*p == '\n') { | |
6148 | if (pedantic && !lang_asm) | |
6149 | pedwarn ("invalid preprocessor directive"); | |
6150 | continue; | |
6151 | } | |
6152 | } | |
6153 | ||
6154 | if (!lang_asm && pedantic) | |
6155 | pedwarn ("invalid preprocessor directive name"); | |
6156 | continue; | |
6157 | } | |
6158 | ||
6159 | for (kt = directive_table; kt->length >= 0; kt++) { | |
6160 | IF_STACK_FRAME *temp; | |
6161 | if (ident_length == kt->length | |
6162 | && strncmp (cp, kt->name, kt->length) == 0) { | |
6163 | /* If we are asked to return on next directive, do so now. */ | |
6164 | if (any) | |
6165 | return; | |
6166 | ||
6167 | switch (kt->type) { | |
6168 | case T_IF: | |
6169 | case T_IFDEF: | |
6170 | case T_IFNDEF: | |
6171 | temp = (IF_STACK_FRAME *) xcalloc (1, sizeof (IF_STACK_FRAME)); | |
6172 | temp->next = if_stack; | |
6173 | if_stack = temp; | |
6174 | temp->lineno = ip->lineno; | |
6175 | temp->fname = ip->nominal_fname; | |
6176 | temp->type = kt->type; | |
6177 | break; | |
6178 | case T_ELSE: | |
6179 | case T_ENDIF: | |
6180 | if (pedantic && if_stack != save_if_stack) | |
6181 | validate_else (bp); | |
6182 | case T_ELIF: | |
6183 | if (if_stack == instack[indepth].if_stack) { | |
6184 | error ("`#%s' not within a conditional", kt->name); | |
6185 | break; | |
6186 | } | |
6187 | else if (if_stack == save_if_stack) | |
6188 | return; /* found what we came for */ | |
6189 | ||
6190 | if (kt->type != T_ENDIF) { | |
6191 | if (if_stack->type == T_ELSE) | |
6192 | error ("`#else' or `#elif' after `#else'"); | |
6193 | if_stack->type = kt->type; | |
6194 | break; | |
6195 | } | |
6196 | ||
6197 | temp = if_stack; | |
6198 | if_stack = if_stack->next; | |
6199 | free (temp); | |
6200 | break; | |
6201 | } | |
6202 | break; | |
6203 | } | |
6204 | } | |
6205 | /* Don't let erroneous code go by. */ | |
6206 | if (kt->length < 0 && !lang_asm && pedantic) | |
6207 | pedwarn ("invalid preprocessor directive name"); | |
6208 | } | |
6209 | } | |
6210 | ip->bufp = bp; | |
6211 | /* after this returns, rescan will exit because ip->bufp | |
6212 | now points to the end of the buffer. | |
6213 | rescan is responsible for the error message also. */ | |
6214 | } | |
6215 | ||
6216 | /* | |
6217 | * handle a #else directive. Do this by just continuing processing | |
6218 | * without changing if_stack ; this is so that the error message | |
6219 | * for missing #endif's etc. will point to the original #if. It | |
6220 | * is possible that something different would be better. | |
6221 | */ | |
6222 | ||
6223 | static int | |
6224 | do_else (buf, limit, op, keyword) | |
6225 | U_CHAR *buf, *limit; | |
6226 | FILE_BUF *op; | |
6227 | struct directive *keyword; | |
6228 | { | |
6229 | FILE_BUF *ip = &instack[indepth]; | |
6230 | ||
6231 | if (pedantic) { | |
6232 | SKIP_WHITE_SPACE (buf); | |
6233 | if (buf != limit) | |
6234 | pedwarn ("text following `#else' violates ANSI standard"); | |
6235 | } | |
6236 | ||
6237 | if (if_stack == instack[indepth].if_stack) { | |
6238 | error ("`#else' not within a conditional"); | |
6239 | return 0; | |
6240 | } else { | |
6241 | /* #ifndef can't have its special treatment for containing the whole file | |
6242 | if it has a #else clause. */ | |
6243 | if_stack->control_macro = 0; | |
6244 | ||
6245 | if (if_stack->type != T_IF && if_stack->type != T_ELIF) { | |
6246 | error ("`#else' after `#else'"); | |
6247 | fprintf (stderr, " (matches line %d", if_stack->lineno); | |
6248 | if (strcmp (if_stack->fname, ip->nominal_fname) != 0) | |
6249 | fprintf (stderr, ", file %s", if_stack->fname); | |
6250 | fprintf (stderr, ")\n"); | |
6251 | } | |
6252 | if_stack->type = T_ELSE; | |
6253 | } | |
6254 | ||
6255 | if (if_stack->if_succeeded) | |
6256 | skip_if_group (ip, 0); | |
6257 | else { | |
6258 | ++if_stack->if_succeeded; /* continue processing input */ | |
6259 | output_line_command (ip, op, 1, same_file); | |
6260 | } | |
6261 | return 0; | |
6262 | } | |
6263 | ||
6264 | /* | |
6265 | * unstack after #endif command | |
6266 | */ | |
6267 | ||
6268 | static int | |
6269 | do_endif (buf, limit, op, keyword) | |
6270 | U_CHAR *buf, *limit; | |
6271 | FILE_BUF *op; | |
6272 | struct directive *keyword; | |
6273 | { | |
6274 | if (pedantic) { | |
6275 | SKIP_WHITE_SPACE (buf); | |
6276 | if (buf != limit) | |
6277 | pedwarn ("text following `#endif' violates ANSI standard"); | |
6278 | } | |
6279 | ||
6280 | if (if_stack == instack[indepth].if_stack) | |
6281 | error ("unbalanced `#endif'"); | |
6282 | else { | |
6283 | IF_STACK_FRAME *temp = if_stack; | |
6284 | if_stack = if_stack->next; | |
6285 | if (temp->control_macro != 0) { | |
6286 | /* This #endif matched a #ifndef at the start of the file. | |
6287 | See if it is at the end of the file. */ | |
6288 | FILE_BUF *ip = &instack[indepth]; | |
6289 | U_CHAR *p = ip->bufp; | |
6290 | U_CHAR *ep = ip->buf + ip->length; | |
6291 | ||
6292 | while (p != ep) { | |
6293 | U_CHAR c = *p++; | |
6294 | switch (c) { | |
6295 | case ' ': | |
6296 | case '\t': | |
6297 | case '\n': | |
6298 | break; | |
6299 | case '/': | |
6300 | if (p != ep && *p == '*') { | |
6301 | /* Skip this comment. */ | |
6302 | int junk; | |
6303 | U_CHAR *save_bufp = ip->bufp; | |
6304 | ip->bufp = p + 1; | |
6305 | p = skip_to_end_of_comment (ip, &junk); | |
6306 | ip->bufp = save_bufp; | |
6307 | } | |
6308 | break; | |
6309 | default: | |
6310 | goto fail; | |
6311 | } | |
6312 | } | |
6313 | /* If we get here, this #endif ends a #ifndef | |
6314 | that contains all of the file (aside from whitespace). | |
6315 | Arrange not to include the file again | |
6316 | if the macro that was tested is defined. */ | |
6317 | if (indepth != 0) | |
6318 | record_control_macro (ip->fname, temp->control_macro); | |
6319 | fail: ; | |
6320 | } | |
6321 | free (temp); | |
6322 | output_line_command (&instack[indepth], op, 1, same_file); | |
6323 | } | |
6324 | return 0; | |
6325 | } | |
6326 | ||
6327 | /* When an #else or #endif is found while skipping failed conditional, | |
6328 | if -pedantic was specified, this is called to warn about text after | |
6329 | the command name. P points to the first char after the command name. */ | |
6330 | ||
6331 | static void | |
6332 | validate_else (p) | |
6333 | register U_CHAR *p; | |
6334 | { | |
6335 | /* Advance P over whitespace and comments. */ | |
6336 | while (1) { | |
6337 | if (*p == '\\' && p[1] == '\n') | |
6338 | p += 2; | |
6339 | if (is_hor_space[*p]) | |
6340 | p++; | |
6341 | else if (*p == '/') { | |
6342 | if (p[1] == '\\' && p[2] == '\n') | |
6343 | newline_fix (p + 1); | |
6344 | if (p[1] == '*') { | |
6345 | p += 2; | |
6346 | /* Don't bother warning about unterminated comments | |
6347 | since that will happen later. Just be sure to exit. */ | |
6348 | while (*p) { | |
6349 | if (p[1] == '\\' && p[2] == '\n') | |
6350 | newline_fix (p + 1); | |
6351 | if (*p == '*' && p[1] == '/') { | |
6352 | p += 2; | |
6353 | break; | |
6354 | } | |
6355 | p++; | |
6356 | } | |
6357 | } | |
6358 | else if ((cplusplus || objc) && p[1] == '/') { | |
6359 | p += 2; | |
6360 | while (*p && *p++ != '\n') ; | |
6361 | } | |
6362 | } else break; | |
6363 | } | |
6364 | if (*p && *p != '\n') | |
6365 | pedwarn ("text following `#else' or `#endif' violates ANSI standard"); | |
6366 | } | |
6367 | \f | |
6368 | /* | |
6369 | * Skip a comment, assuming the input ptr immediately follows the | |
6370 | * initial slash-star. Bump line counter as necessary. | |
6371 | * (The canonical line counter is &ip->lineno). | |
6372 | * Don't use this routine (or the next one) if bumping the line | |
6373 | * counter is not sufficient to deal with newlines in the string. | |
6374 | */ | |
6375 | static U_CHAR * | |
6376 | skip_to_end_of_comment (ip, line_counter) | |
6377 | register FILE_BUF *ip; | |
6378 | int *line_counter; /* place to remember newlines, or NULL */ | |
6379 | { | |
6380 | register U_CHAR *limit = ip->buf + ip->length; | |
6381 | register U_CHAR *bp = ip->bufp; | |
6382 | FILE_BUF *op = &outbuf; /* JF */ | |
6383 | int output = put_out_comments && !line_counter; | |
6384 | ||
6385 | /* JF this line_counter stuff is a crock to make sure the | |
6386 | comment is only put out once, no matter how many times | |
6387 | the comment is skipped. It almost works */ | |
6388 | if (output) { | |
6389 | *op->bufp++ = '/'; | |
6390 | *op->bufp++ = '*'; | |
6391 | } | |
6392 | if ((cplusplus || objc) && bp[-1] == '/') { | |
6393 | if (output) { | |
6394 | while (bp < limit) | |
6395 | if ((*op->bufp++ = *bp++) == '\n') { | |
6396 | bp--; | |
6397 | break; | |
6398 | } | |
6399 | op->bufp[-1] = '*'; | |
6400 | *op->bufp++ = '/'; | |
6401 | *op->bufp++ = '\n'; | |
6402 | } else { | |
6403 | while (bp < limit) { | |
6404 | if (*bp++ == '\n') { | |
6405 | bp--; | |
6406 | break; | |
6407 | } | |
6408 | } | |
6409 | } | |
6410 | ip->bufp = bp; | |
6411 | return bp; | |
6412 | } | |
6413 | while (bp < limit) { | |
6414 | if (output) | |
6415 | *op->bufp++ = *bp; | |
6416 | switch (*bp++) { | |
6417 | case '/': | |
6418 | if (warn_comments && bp < limit && *bp == '*') | |
6419 | warning ("`/*' within comment"); | |
6420 | break; | |
6421 | case '\n': | |
6422 | if (line_counter != NULL) | |
6423 | ++*line_counter; | |
6424 | if (output) | |
6425 | ++op->lineno; | |
6426 | break; | |
6427 | case '*': | |
6428 | if (*bp == '\\' && bp[1] == '\n') | |
6429 | newline_fix (bp); | |
6430 | if (*bp == '/') { | |
6431 | if (output) | |
6432 | *op->bufp++ = '/'; | |
6433 | ip->bufp = ++bp; | |
6434 | return bp; | |
6435 | } | |
6436 | break; | |
6437 | } | |
6438 | } | |
6439 | ip->bufp = bp; | |
6440 | return bp; | |
6441 | } | |
6442 | ||
6443 | /* | |
6444 | * Skip over a quoted string. BP points to the opening quote. | |
6445 | * Returns a pointer after the closing quote. Don't go past LIMIT. | |
6446 | * START_LINE is the line number of the starting point (but it need | |
6447 | * not be valid if the starting point is inside a macro expansion). | |
6448 | * | |
6449 | * The input stack state is not changed. | |
6450 | * | |
6451 | * If COUNT_NEWLINES is nonzero, it points to an int to increment | |
6452 | * for each newline passed. | |
6453 | * | |
6454 | * If BACKSLASH_NEWLINES_P is nonzero, store 1 thru it | |
6455 | * if we pass a backslash-newline. | |
6456 | * | |
6457 | * If EOFP is nonzero, set *EOFP to 1 if the string is unterminated. | |
6458 | */ | |
6459 | static U_CHAR * | |
6460 | skip_quoted_string (bp, limit, start_line, count_newlines, backslash_newlines_p, eofp) | |
6461 | register U_CHAR *bp; | |
6462 | register U_CHAR *limit; | |
6463 | int start_line; | |
6464 | int *count_newlines; | |
6465 | int *backslash_newlines_p; | |
6466 | int *eofp; | |
6467 | { | |
6468 | register U_CHAR c, match; | |
6469 | ||
6470 | match = *bp++; | |
6471 | while (1) { | |
6472 | if (bp >= limit) { | |
6473 | error_with_line (line_for_error (start_line), | |
6474 | "unterminated string or character constant"); | |
6475 | if (eofp) | |
6476 | *eofp = 1; | |
6477 | break; | |
6478 | } | |
6479 | c = *bp++; | |
6480 | if (c == '\\') { | |
6481 | while (*bp == '\\' && bp[1] == '\n') { | |
6482 | if (backslash_newlines_p) | |
6483 | *backslash_newlines_p = 1; | |
6484 | if (count_newlines) | |
6485 | ++*count_newlines; | |
6486 | bp += 2; | |
6487 | } | |
6488 | if (*bp == '\n' && count_newlines) { | |
6489 | if (backslash_newlines_p) | |
6490 | *backslash_newlines_p = 1; | |
6491 | ++*count_newlines; | |
6492 | } | |
6493 | bp++; | |
6494 | } else if (c == '\n') { | |
6495 | if (traditional) { | |
6496 | /* Unterminated strings and character constants are 'legal'. */ | |
6497 | bp--; /* Don't consume the newline. */ | |
6498 | if (eofp) | |
6499 | *eofp = 1; | |
6500 | break; | |
6501 | } | |
6502 | if (match == '\'') { | |
6503 | error_with_line (line_for_error (start_line), | |
6504 | "unterminated character constant"); | |
6505 | bp--; | |
6506 | if (eofp) | |
6507 | *eofp = 1; | |
6508 | break; | |
6509 | } | |
6510 | if (traditional) { /* Unterminated strings are 'legal'. */ | |
6511 | if (eofp) | |
6512 | *eofp = 1; | |
6513 | break; | |
6514 | } | |
6515 | /* If not traditional, then allow newlines inside strings. */ | |
6516 | if (count_newlines) | |
6517 | ++*count_newlines; | |
6518 | } else if (c == match) | |
6519 | break; | |
6520 | } | |
6521 | return bp; | |
6522 | } | |
6523 | ||
6524 | /* Skip across a group of balanced parens, starting from IP->bufp. | |
6525 | IP->bufp is updated. Use this with IP->bufp pointing at an open-paren. | |
6526 | ||
6527 | This does not handle newlines, because it's used for the arg of #if, | |
6528 | where there aren't any newlines. Also, bacslash-newline can't appear. */ | |
6529 | ||
6530 | static U_CHAR * | |
6531 | skip_paren_group (ip) | |
6532 | register FILE_BUF *ip; | |
6533 | { | |
6534 | U_CHAR *limit = ip->buf + ip->length; | |
6535 | U_CHAR *p = ip->bufp; | |
6536 | int depth = 0; | |
6537 | int lines_dummy = 0; | |
6538 | ||
6539 | while (p != limit) { | |
6540 | int c = *p++; | |
6541 | switch (c) { | |
6542 | case '(': | |
6543 | depth++; | |
6544 | break; | |
6545 | ||
6546 | case ')': | |
6547 | depth--; | |
6548 | if (depth == 0) | |
6549 | return ip->bufp = p; | |
6550 | break; | |
6551 | ||
6552 | case '/': | |
6553 | if (*p == '*') { | |
6554 | ip->bufp = p; | |
6555 | p = skip_to_end_of_comment (ip, &lines_dummy); | |
6556 | p = ip->bufp; | |
6557 | } | |
6558 | ||
6559 | case '"': | |
6560 | case '\'': | |
6561 | { | |
6562 | int eofp = 0; | |
6563 | p = skip_quoted_string (p - 1, limit, 0, 0, 0, &eofp); | |
6564 | if (eofp) | |
6565 | return ip->bufp = p; | |
6566 | } | |
6567 | break; | |
6568 | } | |
6569 | } | |
6570 | ||
6571 | ip->bufp = p; | |
6572 | return p; | |
6573 | } | |
6574 | \f | |
6575 | /* | |
6576 | * write out a #line command, for instance, after an #include file. | |
6577 | * If CONDITIONAL is nonzero, we can omit the #line if it would | |
6578 | * appear to be a no-op, and we can output a few newlines instead | |
6579 | * if we want to increase the line number by a small amount. | |
6580 | * FILE_CHANGE says whether we are entering a file, leaving, or neither. | |
6581 | */ | |
6582 | ||
6583 | static void | |
6584 | output_line_command (ip, op, conditional, file_change) | |
6585 | FILE_BUF *ip, *op; | |
6586 | int conditional; | |
6587 | enum file_change_code file_change; | |
6588 | { | |
6589 | int len; | |
6590 | char line_cmd_buf[500]; | |
6591 | ||
6592 | if (no_line_commands | |
6593 | || ip->fname == NULL | |
6594 | || no_output) { | |
6595 | op->lineno = ip->lineno; | |
6596 | return; | |
6597 | } | |
6598 | ||
6599 | if (conditional) { | |
6600 | if (ip->lineno == op->lineno) | |
6601 | return; | |
6602 | ||
6603 | /* If the inherited line number is a little too small, | |
6604 | output some newlines instead of a #line command. */ | |
6605 | if (ip->lineno > op->lineno && ip->lineno < op->lineno + 8) { | |
6606 | check_expand (op, 10); | |
6607 | while (ip->lineno > op->lineno) { | |
6608 | *op->bufp++ = '\n'; | |
6609 | op->lineno++; | |
6610 | } | |
6611 | return; | |
6612 | } | |
6613 | } | |
6614 | ||
6615 | #ifdef OUTPUT_LINE_COMMANDS | |
6616 | sprintf (line_cmd_buf, "#line %d \"%s\"", ip->lineno, ip->nominal_fname); | |
6617 | #else | |
6618 | sprintf (line_cmd_buf, "# %d \"%s\"", ip->lineno, ip->nominal_fname); | |
6619 | #endif | |
6620 | if (file_change != same_file) | |
6621 | strcat (line_cmd_buf, file_change == enter_file ? " 1" : " 2"); | |
6622 | /* Tell cc1 if following text comes from a system header file. */ | |
6623 | if (ip->system_header_p) | |
6624 | strcat (line_cmd_buf, " 3"); | |
6625 | len = strlen (line_cmd_buf); | |
6626 | line_cmd_buf[len++] = '\n'; | |
6627 | check_expand (op, len + 1); | |
6628 | if (op->bufp > op->buf && op->bufp[-1] != '\n') | |
6629 | *op->bufp++ = '\n'; | |
6630 | bcopy (line_cmd_buf, op->bufp, len); | |
6631 | op->bufp += len; | |
6632 | op->lineno = ip->lineno; | |
6633 | } | |
6634 | \f | |
6635 | /* This structure represents one parsed argument in a macro call. | |
6636 | `raw' points to the argument text as written (`raw_length' is its length). | |
6637 | `expanded' points to the argument's macro-expansion | |
6638 | (its length is `expand_length'). | |
6639 | `stringified_length' is the length the argument would have | |
6640 | if stringified. | |
6641 | `use_count' is the number of times this macro arg is substituted | |
6642 | into the macro. If the actual use count exceeds 10, | |
6643 | the value stored is 10. | |
6644 | `free1' and `free2', if nonzero, point to blocks to be freed | |
6645 | when the macro argument data is no longer needed. */ | |
6646 | ||
6647 | struct argdata { | |
6648 | U_CHAR *raw, *expanded; | |
6649 | int raw_length, expand_length; | |
6650 | int stringified_length; | |
6651 | U_CHAR *free1, *free2; | |
6652 | char newlines; | |
6653 | char comments; | |
6654 | char use_count; | |
6655 | }; | |
6656 | ||
6657 | /* Expand a macro call. | |
6658 | HP points to the symbol that is the macro being called. | |
6659 | Put the result of expansion onto the input stack | |
6660 | so that subsequent input by our caller will use it. | |
6661 | ||
6662 | If macro wants arguments, caller has already verified that | |
6663 | an argument list follows; arguments come from the input stack. */ | |
6664 | ||
6665 | static void | |
6666 | macroexpand (hp, op) | |
6667 | HASHNODE *hp; | |
6668 | FILE_BUF *op; | |
6669 | { | |
6670 | int nargs; | |
6671 | DEFINITION *defn = hp->value.defn; | |
6672 | register U_CHAR *xbuf; | |
6673 | int xbuf_len; | |
6674 | int start_line = instack[indepth].lineno; | |
6675 | ||
6676 | CHECK_DEPTH (return;); | |
6677 | ||
6678 | /* it might not actually be a macro. */ | |
6679 | if (hp->type != T_MACRO) { | |
6680 | special_symbol (hp, op); | |
6681 | return; | |
6682 | } | |
6683 | ||
6684 | /* This macro is being used inside a #if, which means it must be */ | |
6685 | /* recorded as a precondition. */ | |
6686 | if (pcp_inside_if && pcp_outfile && defn->predefined) | |
6687 | dump_single_macro (hp, pcp_outfile); | |
6688 | ||
6689 | nargs = defn->nargs; | |
6690 | ||
6691 | if (nargs >= 0) { | |
6692 | register int i; | |
6693 | struct argdata *args; | |
6694 | char *parse_error = 0; | |
6695 | ||
6696 | args = (struct argdata *) alloca ((nargs + 1) * sizeof (struct argdata)); | |
6697 | ||
6698 | for (i = 0; i < nargs; i++) { | |
6699 | args[i].raw = args[i].expanded = (U_CHAR *) ""; | |
6700 | args[i].raw_length = args[i].expand_length | |
6701 | = args[i].stringified_length = 0; | |
6702 | args[i].free1 = args[i].free2 = 0; | |
6703 | args[i].use_count = 0; | |
6704 | } | |
6705 | ||
6706 | /* Parse all the macro args that are supplied. I counts them. | |
6707 | The first NARGS args are stored in ARGS. | |
6708 | The rest are discarded. */ | |
6709 | i = 0; | |
6710 | do { | |
6711 | /* Discard the open-parenthesis or comma before the next arg. */ | |
6712 | ++instack[indepth].bufp; | |
6713 | parse_error | |
6714 | = macarg ((i < nargs || (nargs == 0 && i == 0)) ? &args[i] : 0); | |
6715 | if (parse_error) { | |
6716 | error_with_line (line_for_error (start_line), parse_error); | |
6717 | break; | |
6718 | } | |
6719 | i++; | |
6720 | } while (*instack[indepth].bufp != ')'); | |
6721 | ||
6722 | /* If we got one arg but it was just whitespace, call that 0 args. */ | |
6723 | if (i == 1) { | |
6724 | register U_CHAR *bp = args[0].raw; | |
6725 | register U_CHAR *lim = bp + args[0].raw_length; | |
6726 | while (bp != lim && is_space[*bp]) bp++; | |
6727 | if (bp == lim) | |
6728 | i = 0; | |
6729 | } | |
6730 | ||
6731 | if (nargs == 0 && i > 0) | |
6732 | error ("arguments given to macro `%s'", hp->name); | |
6733 | else if (i < nargs) { | |
6734 | /* traditional C allows foo() if foo wants one argument. */ | |
6735 | if (nargs == 1 && i == 0 && traditional) | |
6736 | ; | |
6737 | else if (i == 0) | |
6738 | error ("macro `%s' used without args", hp->name); | |
6739 | else if (i == 1) | |
6740 | error ("macro `%s' used with just one arg", hp->name); | |
6741 | else | |
6742 | error ("macro `%s' used with only %d args", hp->name, i); | |
6743 | } else if (i > nargs) | |
6744 | error ("macro `%s' used with too many (%d) args", hp->name, i); | |
6745 | ||
6746 | /* Swallow the closeparen. */ | |
6747 | ++instack[indepth].bufp; | |
6748 | ||
6749 | /* If macro wants zero args, we parsed the arglist for checking only. | |
6750 | Read directly from the macro definition. */ | |
6751 | if (nargs == 0) { | |
6752 | xbuf = defn->expansion; | |
6753 | xbuf_len = defn->length; | |
6754 | } else { | |
6755 | register U_CHAR *exp = defn->expansion; | |
6756 | register int offset; /* offset in expansion, | |
6757 | copied a piece at a time */ | |
6758 | register int totlen; /* total amount of exp buffer filled so far */ | |
6759 | ||
6760 | register struct reflist *ap; | |
6761 | ||
6762 | /* Macro really takes args. Compute the expansion of this call. */ | |
6763 | ||
6764 | /* Compute length in characters of the macro's expansion. | |
6765 | Also count number of times each arg is used. */ | |
6766 | xbuf_len = defn->length; | |
6767 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
6768 | if (ap->stringify) | |
6769 | xbuf_len += args[ap->argno].stringified_length; | |
6770 | else if (ap->raw_before || ap->raw_after || traditional) | |
6771 | xbuf_len += args[ap->argno].raw_length; | |
6772 | else | |
6773 | xbuf_len += args[ap->argno].expand_length; | |
6774 | ||
6775 | if (args[ap->argno].use_count < 10) | |
6776 | args[ap->argno].use_count++; | |
6777 | } | |
6778 | ||
6779 | xbuf = (U_CHAR *) xmalloc (xbuf_len + 1); | |
6780 | ||
6781 | /* Generate in XBUF the complete expansion | |
6782 | with arguments substituted in. | |
6783 | TOTLEN is the total size generated so far. | |
6784 | OFFSET is the index in the definition | |
6785 | of where we are copying from. */ | |
6786 | offset = totlen = 0; | |
6787 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
6788 | register struct argdata *arg = &args[ap->argno]; | |
6789 | ||
6790 | for (i = 0; i < ap->nchars; i++) | |
6791 | xbuf[totlen++] = exp[offset++]; | |
6792 | ||
6793 | if (ap->stringify != 0) { | |
6794 | int arglen = arg->raw_length; | |
6795 | int escaped = 0; | |
6796 | int in_string = 0; | |
6797 | int c; | |
6798 | i = 0; | |
6799 | while (i < arglen | |
6800 | && (c = arg->raw[i], is_space[c])) | |
6801 | i++; | |
6802 | while (i < arglen | |
6803 | && (c = arg->raw[arglen - 1], is_space[c])) | |
6804 | arglen--; | |
6805 | if (!traditional) | |
6806 | xbuf[totlen++] = '\"'; /* insert beginning quote */ | |
6807 | for (; i < arglen; i++) { | |
6808 | c = arg->raw[i]; | |
6809 | ||
6810 | /* Special markers Newline Space | |
6811 | generate nothing for a stringified argument. */ | |
6812 | if (c == '\n' && arg->raw[i+1] != '\n') { | |
6813 | i++; | |
6814 | continue; | |
6815 | } | |
6816 | ||
6817 | /* Internal sequences of whitespace are replaced by one space | |
6818 | except within an string or char token. */ | |
6819 | if (! in_string | |
6820 | && (c == '\n' ? arg->raw[i+1] == '\n' : is_space[c])) { | |
6821 | while (1) { | |
6822 | /* Note that Newline Space does occur within whitespace | |
6823 | sequences; consider it part of the sequence. */ | |
6824 | if (c == '\n' && is_space[arg->raw[i+1]]) | |
6825 | i += 2; | |
6826 | else if (c != '\n' && is_space[c]) | |
6827 | i++; | |
6828 | else break; | |
6829 | c = arg->raw[i]; | |
6830 | } | |
6831 | i--; | |
6832 | c = ' '; | |
6833 | } | |
6834 | ||
6835 | if (escaped) | |
6836 | escaped = 0; | |
6837 | else { | |
6838 | if (c == '\\') | |
6839 | escaped = 1; | |
6840 | if (in_string) { | |
6841 | if (c == in_string) | |
6842 | in_string = 0; | |
6843 | } else if (c == '\"' || c == '\'') | |
6844 | in_string = c; | |
6845 | } | |
6846 | ||
6847 | /* Escape these chars */ | |
6848 | if (c == '\"' || (in_string && c == '\\')) | |
6849 | xbuf[totlen++] = '\\'; | |
6850 | if (isprint (c)) | |
6851 | xbuf[totlen++] = c; | |
6852 | else { | |
6853 | sprintf ((char *) &xbuf[totlen], "\\%03o", (unsigned int) c); | |
6854 | totlen += 4; | |
6855 | } | |
6856 | } | |
6857 | if (!traditional) | |
6858 | xbuf[totlen++] = '\"'; /* insert ending quote */ | |
6859 | } else if (ap->raw_before || ap->raw_after || traditional) { | |
6860 | U_CHAR *p1 = arg->raw; | |
6861 | U_CHAR *l1 = p1 + arg->raw_length; | |
6862 | if (ap->raw_before) { | |
6863 | while (p1 != l1 && is_space[*p1]) p1++; | |
6864 | while (p1 != l1 && is_idchar[*p1]) | |
6865 | xbuf[totlen++] = *p1++; | |
6866 | /* Delete any no-reexpansion marker that follows | |
6867 | an identifier at the beginning of the argument | |
6868 | if the argument is concatenated with what precedes it. */ | |
6869 | if (p1[0] == '\n' && p1[1] == '-') | |
6870 | p1 += 2; | |
6871 | } | |
6872 | if (ap->raw_after) { | |
6873 | /* Arg is concatenated after: delete trailing whitespace, | |
6874 | whitespace markers, and no-reexpansion markers. */ | |
6875 | while (p1 != l1) { | |
6876 | if (is_space[l1[-1]]) l1--; | |
6877 | else if (l1[-1] == '-') { | |
6878 | U_CHAR *p2 = l1 - 1; | |
6879 | /* If a `-' is preceded by an odd number of newlines then it | |
6880 | and the last newline are a no-reexpansion marker. */ | |
6881 | while (p2 != p1 && p2[-1] == '\n') p2--; | |
6882 | if ((l1 - 1 - p2) & 1) { | |
6883 | l1 -= 2; | |
6884 | } | |
6885 | else break; | |
6886 | } | |
6887 | else break; | |
6888 | } | |
6889 | } | |
6890 | bcopy (p1, xbuf + totlen, l1 - p1); | |
6891 | totlen += l1 - p1; | |
6892 | } else { | |
6893 | bcopy (arg->expanded, xbuf + totlen, arg->expand_length); | |
6894 | totlen += arg->expand_length; | |
6895 | /* If a macro argument with newlines is used multiple times, | |
6896 | then only expand the newlines once. This avoids creating output | |
6897 | lines which don't correspond to any input line, which confuses | |
6898 | gdb and gcov. */ | |
6899 | if (arg->use_count > 1 && arg->newlines > 0) { | |
6900 | /* Don't bother doing delete_newlines for subsequent | |
6901 | uses of arg. */ | |
6902 | arg->use_count = 1; | |
6903 | arg->expand_length | |
6904 | = delete_newlines (arg->expanded, arg->expand_length); | |
6905 | } | |
6906 | } | |
6907 | ||
6908 | if (totlen > xbuf_len) | |
6909 | abort (); | |
6910 | } | |
6911 | ||
6912 | /* if there is anything left of the definition | |
6913 | after handling the arg list, copy that in too. */ | |
6914 | ||
6915 | for (i = offset; i < defn->length; i++) | |
6916 | xbuf[totlen++] = exp[i]; | |
6917 | ||
6918 | xbuf[totlen] = 0; | |
6919 | xbuf_len = totlen; | |
6920 | ||
6921 | for (i = 0; i < nargs; i++) { | |
6922 | if (args[i].free1 != 0) | |
6923 | free (args[i].free1); | |
6924 | if (args[i].free2 != 0) | |
6925 | free (args[i].free2); | |
6926 | } | |
6927 | } | |
6928 | } else { | |
6929 | xbuf = defn->expansion; | |
6930 | xbuf_len = defn->length; | |
6931 | } | |
6932 | ||
6933 | /* Now put the expansion on the input stack | |
6934 | so our caller will commence reading from it. */ | |
6935 | { | |
6936 | register FILE_BUF *ip2; | |
6937 | ||
6938 | ip2 = &instack[++indepth]; | |
6939 | ||
6940 | ip2->fname = 0; | |
6941 | ip2->nominal_fname = 0; | |
6942 | ip2->lineno = 0; | |
6943 | ip2->buf = xbuf; | |
6944 | ip2->length = xbuf_len; | |
6945 | ip2->bufp = xbuf; | |
6946 | ip2->free_ptr = (nargs > 0) ? xbuf : 0; | |
6947 | ip2->macro = hp; | |
6948 | ip2->if_stack = if_stack; | |
6949 | ip2->system_header_p = 0; | |
6950 | ||
6951 | /* Recursive macro use sometimes works traditionally. | |
6952 | #define foo(x,y) bar(x(y,0), y) | |
6953 | foo(foo, baz) */ | |
6954 | ||
6955 | if (!traditional) | |
6956 | hp->type = T_DISABLED; | |
6957 | } | |
6958 | } | |
6959 | \f | |
6960 | /* | |
6961 | * Parse a macro argument and store the info on it into *ARGPTR. | |
6962 | * Return nonzero to indicate a syntax error. | |
6963 | */ | |
6964 | ||
6965 | static char * | |
6966 | macarg (argptr) | |
6967 | register struct argdata *argptr; | |
6968 | { | |
6969 | FILE_BUF *ip = &instack[indepth]; | |
6970 | int paren = 0; | |
6971 | int newlines = 0; | |
6972 | int comments = 0; | |
6973 | ||
6974 | /* Try to parse as much of the argument as exists at this | |
6975 | input stack level. */ | |
6976 | U_CHAR *bp = macarg1 (ip->bufp, ip->buf + ip->length, | |
6977 | &paren, &newlines, &comments); | |
6978 | ||
6979 | /* If we find the end of the argument at this level, | |
6980 | set up *ARGPTR to point at it in the input stack. */ | |
6981 | if (!(ip->fname != 0 && (newlines != 0 || comments != 0)) | |
6982 | && bp != ip->buf + ip->length) { | |
6983 | if (argptr != 0) { | |
6984 | argptr->raw = ip->bufp; | |
6985 | argptr->raw_length = bp - ip->bufp; | |
6986 | argptr->newlines = newlines; | |
6987 | } | |
6988 | ip->bufp = bp; | |
6989 | } else { | |
6990 | /* This input stack level ends before the macro argument does. | |
6991 | We must pop levels and keep parsing. | |
6992 | Therefore, we must allocate a temporary buffer and copy | |
6993 | the macro argument into it. */ | |
6994 | int bufsize = bp - ip->bufp; | |
6995 | int extra = newlines; | |
6996 | U_CHAR *buffer = (U_CHAR *) xmalloc (bufsize + extra + 1); | |
6997 | int final_start = 0; | |
6998 | ||
6999 | bcopy (ip->bufp, buffer, bufsize); | |
7000 | ip->bufp = bp; | |
7001 | ip->lineno += newlines; | |
7002 | ||
7003 | while (bp == ip->buf + ip->length) { | |
7004 | if (instack[indepth].macro == 0) { | |
7005 | free (buffer); | |
7006 | return "unterminated macro call"; | |
7007 | } | |
7008 | ip->macro->type = T_MACRO; | |
7009 | if (ip->free_ptr) | |
7010 | free (ip->free_ptr); | |
7011 | ip = &instack[--indepth]; | |
7012 | newlines = 0; | |
7013 | comments = 0; | |
7014 | bp = macarg1 (ip->bufp, ip->buf + ip->length, &paren, | |
7015 | &newlines, &comments); | |
7016 | final_start = bufsize; | |
7017 | bufsize += bp - ip->bufp; | |
7018 | extra += newlines; | |
7019 | buffer = (U_CHAR *) xrealloc (buffer, bufsize + extra + 1); | |
7020 | bcopy (ip->bufp, buffer + bufsize - (bp - ip->bufp), bp - ip->bufp); | |
7021 | ip->bufp = bp; | |
7022 | ip->lineno += newlines; | |
7023 | } | |
7024 | ||
7025 | /* Now, if arg is actually wanted, record its raw form, | |
7026 | discarding comments and duplicating newlines in whatever | |
7027 | part of it did not come from a macro expansion. | |
7028 | EXTRA space has been preallocated for duplicating the newlines. | |
7029 | FINAL_START is the index of the start of that part. */ | |
7030 | if (argptr != 0) { | |
7031 | argptr->raw = buffer; | |
7032 | argptr->raw_length = bufsize; | |
7033 | argptr->free1 = buffer; | |
7034 | argptr->newlines = newlines; | |
7035 | argptr->comments = comments; | |
7036 | if ((newlines || comments) && ip->fname != 0) | |
7037 | argptr->raw_length | |
7038 | = final_start + | |
7039 | discard_comments (argptr->raw + final_start, | |
7040 | argptr->raw_length - final_start, | |
7041 | newlines); | |
7042 | argptr->raw[argptr->raw_length] = 0; | |
7043 | if (argptr->raw_length > bufsize + extra) | |
7044 | abort (); | |
7045 | } | |
7046 | } | |
7047 | ||
7048 | /* If we are not discarding this argument, | |
7049 | macroexpand it and compute its length as stringified. | |
7050 | All this info goes into *ARGPTR. */ | |
7051 | ||
7052 | if (argptr != 0) { | |
7053 | FILE_BUF obuf; | |
7054 | register U_CHAR *buf, *lim; | |
7055 | register int totlen; | |
7056 | ||
7057 | obuf = expand_to_temp_buffer (argptr->raw, | |
7058 | argptr->raw + argptr->raw_length, | |
7059 | 1, 0); | |
7060 | ||
7061 | argptr->expanded = obuf.buf; | |
7062 | argptr->expand_length = obuf.length; | |
7063 | argptr->free2 = obuf.buf; | |
7064 | ||
7065 | buf = argptr->raw; | |
7066 | lim = buf + argptr->raw_length; | |
7067 | ||
7068 | while (buf != lim && is_space[*buf]) | |
7069 | buf++; | |
7070 | while (buf != lim && is_space[lim[-1]]) | |
7071 | lim--; | |
7072 | totlen = traditional ? 0 : 2; /* Count opening and closing quote. */ | |
7073 | while (buf != lim) { | |
7074 | register U_CHAR c = *buf++; | |
7075 | totlen++; | |
7076 | /* Internal sequences of whitespace are replaced by one space | |
7077 | in most cases, but not always. So count all the whitespace | |
7078 | in case we need to keep it all. */ | |
7079 | #if 0 | |
7080 | if (is_space[c]) | |
7081 | SKIP_ALL_WHITE_SPACE (buf); | |
7082 | else | |
7083 | #endif | |
7084 | if (c == '\"' || c == '\\') /* escape these chars */ | |
7085 | totlen++; | |
7086 | else if (!isprint (c)) | |
7087 | totlen += 3; | |
7088 | } | |
7089 | argptr->stringified_length = totlen; | |
7090 | } | |
7091 | return 0; | |
7092 | } | |
7093 | \f | |
7094 | /* Scan text from START (inclusive) up to LIMIT (exclusive), | |
7095 | counting parens in *DEPTHPTR, | |
7096 | and return if reach LIMIT | |
7097 | or before a `)' that would make *DEPTHPTR negative | |
7098 | or before a comma when *DEPTHPTR is zero. | |
7099 | Single and double quotes are matched and termination | |
7100 | is inhibited within them. Comments also inhibit it. | |
7101 | Value returned is pointer to stopping place. | |
7102 | ||
7103 | Increment *NEWLINES each time a newline is passed. | |
7104 | Set *COMMENTS to 1 if a comment is seen. */ | |
7105 | ||
7106 | static U_CHAR * | |
7107 | macarg1 (start, limit, depthptr, newlines, comments) | |
7108 | U_CHAR *start; | |
7109 | register U_CHAR *limit; | |
7110 | int *depthptr, *newlines, *comments; | |
7111 | { | |
7112 | register U_CHAR *bp = start; | |
7113 | ||
7114 | while (bp < limit) { | |
7115 | switch (*bp) { | |
7116 | case '(': | |
7117 | (*depthptr)++; | |
7118 | break; | |
7119 | case ')': | |
7120 | if (--(*depthptr) < 0) | |
7121 | return bp; | |
7122 | break; | |
7123 | case '\\': | |
7124 | /* Traditionally, backslash makes following char not special. */ | |
7125 | if (bp + 1 < limit && traditional) | |
7126 | { | |
7127 | bp++; | |
7128 | /* But count source lines anyway. */ | |
7129 | if (*bp == '\n') | |
7130 | ++*newlines; | |
7131 | } | |
7132 | break; | |
7133 | case '\n': | |
7134 | ++*newlines; | |
7135 | break; | |
7136 | case '/': | |
7137 | if (bp[1] == '\\' && bp[2] == '\n') | |
7138 | newline_fix (bp + 1); | |
7139 | if ((cplusplus || objc) && bp[1] == '/') { | |
7140 | *comments = 1; | |
7141 | bp += 2; | |
7142 | while (bp < limit && *bp++ != '\n') ; | |
7143 | ++*newlines; | |
7144 | break; | |
7145 | } | |
7146 | if (bp[1] != '*' || bp + 1 >= limit) | |
7147 | break; | |
7148 | *comments = 1; | |
7149 | bp += 2; | |
7150 | while (bp + 1 < limit) { | |
7151 | if (bp[0] == '*' | |
7152 | && bp[1] == '\\' && bp[2] == '\n') | |
7153 | newline_fix (bp + 1); | |
7154 | if (bp[0] == '*' && bp[1] == '/') | |
7155 | break; | |
7156 | if (*bp == '\n') ++*newlines; | |
7157 | bp++; | |
7158 | } | |
7159 | break; | |
7160 | case '\'': | |
7161 | case '\"': | |
7162 | { | |
7163 | int quotec; | |
7164 | for (quotec = *bp++; bp + 1 < limit && *bp != quotec; bp++) { | |
7165 | if (*bp == '\\') { | |
7166 | bp++; | |
7167 | if (*bp == '\n') | |
7168 | ++*newlines; | |
7169 | while (*bp == '\\' && bp[1] == '\n') { | |
7170 | bp += 2; | |
7171 | } | |
7172 | } else if (*bp == '\n') { | |
7173 | ++*newlines; | |
7174 | if (quotec == '\'') | |
7175 | break; | |
7176 | } | |
7177 | } | |
7178 | } | |
7179 | break; | |
7180 | case ',': | |
7181 | if ((*depthptr) == 0) | |
7182 | return bp; | |
7183 | break; | |
7184 | } | |
7185 | bp++; | |
7186 | } | |
7187 | ||
7188 | return bp; | |
7189 | } | |
7190 | \f | |
7191 | /* Discard comments and duplicate newlines | |
7192 | in the string of length LENGTH at START, | |
7193 | except inside of string constants. | |
7194 | The string is copied into itself with its beginning staying fixed. | |
7195 | ||
7196 | NEWLINES is the number of newlines that must be duplicated. | |
7197 | We assume that that much extra space is available past the end | |
7198 | of the string. */ | |
7199 | ||
7200 | static int | |
7201 | discard_comments (start, length, newlines) | |
7202 | U_CHAR *start; | |
7203 | int length; | |
7204 | int newlines; | |
7205 | { | |
7206 | register U_CHAR *ibp; | |
7207 | register U_CHAR *obp; | |
7208 | register U_CHAR *limit; | |
7209 | register int c; | |
7210 | ||
7211 | /* If we have newlines to duplicate, copy everything | |
7212 | that many characters up. Then, in the second part, | |
7213 | we will have room to insert the newlines | |
7214 | while copying down. | |
7215 | NEWLINES may actually be too large, because it counts | |
7216 | newlines in string constants, and we don't duplicate those. | |
7217 | But that does no harm. */ | |
7218 | if (newlines > 0) { | |
7219 | ibp = start + length; | |
7220 | obp = ibp + newlines; | |
7221 | limit = start; | |
7222 | while (limit != ibp) | |
7223 | *--obp = *--ibp; | |
7224 | } | |
7225 | ||
7226 | ibp = start + newlines; | |
7227 | limit = start + length + newlines; | |
7228 | obp = start; | |
7229 | ||
7230 | while (ibp < limit) { | |
7231 | *obp++ = c = *ibp++; | |
7232 | switch (c) { | |
7233 | case '\n': | |
7234 | /* Duplicate the newline. */ | |
7235 | *obp++ = '\n'; | |
7236 | break; | |
7237 | ||
7238 | case '\\': | |
7239 | if (*ibp == '\n') { | |
7240 | obp--; | |
7241 | ibp++; | |
7242 | } | |
7243 | break; | |
7244 | ||
7245 | case '/': | |
7246 | if (*ibp == '\\' && ibp[1] == '\n') | |
7247 | newline_fix (ibp); | |
7248 | /* Delete any comment. */ | |
7249 | if ((cplusplus || objc) && ibp[0] == '/') { | |
7250 | obp--; | |
7251 | ibp++; | |
7252 | while (ibp < limit && *ibp++ != '\n') ; | |
7253 | break; | |
7254 | } | |
7255 | if (ibp[0] != '*' || ibp + 1 >= limit) | |
7256 | break; | |
7257 | obp--; | |
7258 | ibp++; | |
7259 | while (ibp + 1 < limit) { | |
7260 | if (ibp[0] == '*' | |
7261 | && ibp[1] == '\\' && ibp[2] == '\n') | |
7262 | newline_fix (ibp + 1); | |
7263 | if (ibp[0] == '*' && ibp[1] == '/') | |
7264 | break; | |
7265 | ibp++; | |
7266 | } | |
7267 | ibp += 2; | |
7268 | break; | |
7269 | ||
7270 | case '\'': | |
7271 | case '\"': | |
7272 | /* Notice and skip strings, so that we don't | |
7273 | think that comments start inside them, | |
7274 | and so we don't duplicate newlines in them. */ | |
7275 | { | |
7276 | int quotec = c; | |
7277 | while (ibp < limit) { | |
7278 | *obp++ = c = *ibp++; | |
7279 | if (c == quotec) | |
7280 | break; | |
7281 | if (c == '\n' && quotec == '\'') | |
7282 | break; | |
7283 | if (c == '\\' && ibp < limit) { | |
7284 | while (*ibp == '\\' && ibp[1] == '\n') | |
7285 | ibp += 2; | |
7286 | *obp++ = *ibp++; | |
7287 | } | |
7288 | } | |
7289 | } | |
7290 | break; | |
7291 | } | |
7292 | } | |
7293 | ||
7294 | return obp - start; | |
7295 | } | |
7296 | \f | |
7297 | /* Delete newlines in the string of length LENGTH at START, except inside | |
7298 | of string constants. The string is copied into itself with its beginning | |
7299 | staying fixed. */ | |
7300 | ||
7301 | static int | |
7302 | delete_newlines (start, length) | |
7303 | U_CHAR *start; | |
7304 | int length; | |
7305 | { | |
7306 | register U_CHAR *ibp; | |
7307 | register U_CHAR *obp; | |
7308 | register U_CHAR *limit; | |
7309 | register int c; | |
7310 | ||
7311 | ibp = start; | |
7312 | limit = start + length; | |
7313 | obp = start; | |
7314 | ||
7315 | while (ibp < limit) { | |
7316 | *obp++ = c = *ibp++; | |
7317 | switch (c) { | |
7318 | case '\n': | |
7319 | /* If this is a NEWLINE NEWLINE, then this is a real newline in the | |
7320 | output. Skip past the newline and its duplicate. */ | |
7321 | if (*ibp == '\n') | |
7322 | { | |
7323 | ibp++; | |
7324 | obp--; | |
7325 | } | |
7326 | break; | |
7327 | ||
7328 | case '\'': | |
7329 | case '\"': | |
7330 | /* Notice and skip strings, so that we don't delete newlines in them. */ | |
7331 | { | |
7332 | int quotec = c; | |
7333 | while (ibp < limit) { | |
7334 | *obp++ = c = *ibp++; | |
7335 | if (c == quotec) | |
7336 | break; | |
7337 | if (c == '\n' && quotec == '\'') | |
7338 | break; | |
7339 | } | |
7340 | } | |
7341 | break; | |
7342 | } | |
7343 | } | |
7344 | ||
7345 | return obp - start; | |
7346 | } | |
7347 | \f | |
7348 | /* | |
7349 | * error - print error message and increment count of errors. | |
7350 | */ | |
7351 | ||
7352 | void | |
7353 | error (msg, arg1, arg2, arg3) | |
7354 | char *msg; | |
7355 | { | |
7356 | int i; | |
7357 | FILE_BUF *ip = NULL; | |
7358 | ||
7359 | print_containing_files (); | |
7360 | ||
7361 | for (i = indepth; i >= 0; i--) | |
7362 | if (instack[i].fname != NULL) { | |
7363 | ip = &instack[i]; | |
7364 | break; | |
7365 | } | |
7366 | ||
7367 | if (ip != NULL) | |
7368 | fprintf (stderr, "%s:%d: ", ip->nominal_fname, ip->lineno); | |
7369 | fprintf (stderr, msg, arg1, arg2, arg3); | |
7370 | fprintf (stderr, "\n"); | |
7371 | errors++; | |
7372 | } | |
7373 | ||
7374 | /* Error including a message from `errno'. */ | |
7375 | ||
7376 | static void | |
7377 | error_from_errno (name) | |
7378 | char *name; | |
7379 | { | |
7380 | int i; | |
7381 | FILE_BUF *ip = NULL; | |
7382 | ||
7383 | print_containing_files (); | |
7384 | ||
7385 | for (i = indepth; i >= 0; i--) | |
7386 | if (instack[i].fname != NULL) { | |
7387 | ip = &instack[i]; | |
7388 | break; | |
7389 | } | |
7390 | ||
7391 | if (ip != NULL) | |
7392 | fprintf (stderr, "%s:%d: ", ip->nominal_fname, ip->lineno); | |
7393 | ||
7394 | if (errno < sys_nerr) | |
7395 | fprintf (stderr, "%s: %s\n", name, sys_errlist[errno]); | |
7396 | else | |
7397 | fprintf (stderr, "%s: undocumented I/O error\n", name); | |
7398 | ||
7399 | errors++; | |
7400 | } | |
7401 | ||
7402 | /* Print error message but don't count it. */ | |
7403 | ||
7404 | void | |
7405 | warning (msg, arg1, arg2, arg3) | |
7406 | char *msg; | |
7407 | { | |
7408 | int i; | |
7409 | FILE_BUF *ip = NULL; | |
7410 | ||
7411 | if (inhibit_warnings) | |
7412 | return; | |
7413 | ||
7414 | if (warnings_are_errors) | |
7415 | errors++; | |
7416 | ||
7417 | print_containing_files (); | |
7418 | ||
7419 | for (i = indepth; i >= 0; i--) | |
7420 | if (instack[i].fname != NULL) { | |
7421 | ip = &instack[i]; | |
7422 | break; | |
7423 | } | |
7424 | ||
7425 | if (ip != NULL) | |
7426 | fprintf (stderr, "%s:%d: ", ip->nominal_fname, ip->lineno); | |
7427 | fprintf (stderr, "warning: "); | |
7428 | fprintf (stderr, msg, arg1, arg2, arg3); | |
7429 | fprintf (stderr, "\n"); | |
7430 | } | |
7431 | ||
7432 | static void | |
7433 | error_with_line (line, msg, arg1, arg2, arg3) | |
7434 | int line; | |
7435 | char *msg; | |
7436 | { | |
7437 | int i; | |
7438 | FILE_BUF *ip = NULL; | |
7439 | ||
7440 | print_containing_files (); | |
7441 | ||
7442 | for (i = indepth; i >= 0; i--) | |
7443 | if (instack[i].fname != NULL) { | |
7444 | ip = &instack[i]; | |
7445 | break; | |
7446 | } | |
7447 | ||
7448 | if (ip != NULL) | |
7449 | fprintf (stderr, "%s:%d: ", ip->nominal_fname, line); | |
7450 | fprintf (stderr, msg, arg1, arg2, arg3); | |
7451 | fprintf (stderr, "\n"); | |
7452 | errors++; | |
7453 | } | |
7454 | ||
7455 | /* print an error message and maybe count it. */ | |
7456 | ||
7457 | void | |
7458 | pedwarn (msg, arg1, arg2, arg3) | |
7459 | char *msg; | |
7460 | { | |
7461 | if (pedantic_errors) | |
7462 | error (msg, arg1, arg2, arg3); | |
7463 | else | |
7464 | warning (msg, arg1, arg2, arg3); | |
7465 | } | |
7466 | ||
7467 | /* Report a warning (or an error if pedantic_errors) | |
7468 | giving specified file name and line number, not current. */ | |
7469 | ||
7470 | static void | |
7471 | pedwarn_with_file_and_line (file, line, msg, arg1, arg2, arg3) | |
7472 | char *file; | |
7473 | int line; | |
7474 | char *msg; | |
7475 | { | |
7476 | int i; | |
7477 | if (!pedantic_errors && inhibit_warnings) | |
7478 | return; | |
7479 | if (file != NULL) | |
7480 | fprintf (stderr, "%s:%d: ", file, line); | |
7481 | if (pedantic_errors || warnings_are_errors) | |
7482 | errors++; | |
7483 | if (!pedantic_errors) | |
7484 | fprintf (stderr, "warning: "); | |
7485 | fprintf (stderr, msg, arg1, arg2, arg3); | |
7486 | fprintf (stderr, "\n"); | |
7487 | } | |
7488 | \f | |
7489 | /* Print the file names and line numbers of the #include | |
7490 | commands which led to the current file. */ | |
7491 | ||
7492 | static void | |
7493 | print_containing_files () | |
7494 | { | |
7495 | FILE_BUF *ip = NULL; | |
7496 | int i; | |
7497 | int first = 1; | |
7498 | ||
7499 | /* If stack of files hasn't changed since we last printed | |
7500 | this info, don't repeat it. */ | |
7501 | if (last_error_tick == input_file_stack_tick) | |
7502 | return; | |
7503 | ||
7504 | for (i = indepth; i >= 0; i--) | |
7505 | if (instack[i].fname != NULL) { | |
7506 | ip = &instack[i]; | |
7507 | break; | |
7508 | } | |
7509 | ||
7510 | /* Give up if we don't find a source file. */ | |
7511 | if (ip == NULL) | |
7512 | return; | |
7513 | ||
7514 | /* Find the other, outer source files. */ | |
7515 | for (i--; i >= 0; i--) | |
7516 | if (instack[i].fname != NULL) { | |
7517 | ip = &instack[i]; | |
7518 | if (first) { | |
7519 | first = 0; | |
7520 | fprintf (stderr, "In file included"); | |
7521 | } else { | |
7522 | fprintf (stderr, ","); | |
7523 | } | |
7524 | ||
7525 | fprintf (stderr, " from %s:%d", ip->nominal_fname, ip->lineno); | |
7526 | } | |
7527 | if (! first) | |
7528 | fprintf (stderr, ":\n"); | |
7529 | ||
7530 | /* Record we have printed the status as of this time. */ | |
7531 | last_error_tick = input_file_stack_tick; | |
7532 | } | |
7533 | \f | |
7534 | /* Return the line at which an error occurred. | |
7535 | The error is not necessarily associated with the current spot | |
7536 | in the input stack, so LINE says where. LINE will have been | |
7537 | copied from ip->lineno for the current input level. | |
7538 | If the current level is for a file, we return LINE. | |
7539 | But if the current level is not for a file, LINE is meaningless. | |
7540 | In that case, we return the lineno of the innermost file. */ | |
7541 | ||
7542 | static int | |
7543 | line_for_error (line) | |
7544 | int line; | |
7545 | { | |
7546 | int i; | |
7547 | int line1 = line; | |
7548 | ||
7549 | for (i = indepth; i >= 0; ) { | |
7550 | if (instack[i].fname != 0) | |
7551 | return line1; | |
7552 | i--; | |
7553 | if (i < 0) | |
7554 | return 0; | |
7555 | line1 = instack[i].lineno; | |
7556 | } | |
7557 | abort (); | |
7558 | /*NOTREACHED*/ | |
7559 | return 0; | |
7560 | } | |
7561 | ||
7562 | /* | |
7563 | * If OBUF doesn't have NEEDED bytes after OPTR, make it bigger. | |
7564 | * | |
7565 | * As things stand, nothing is ever placed in the output buffer to be | |
7566 | * removed again except when it's KNOWN to be part of an identifier, | |
7567 | * so flushing and moving down everything left, instead of expanding, | |
7568 | * should work ok. | |
7569 | */ | |
7570 | ||
7571 | /* You might think void was cleaner for the return type, | |
7572 | but that would get type mismatch in check_expand in strict ANSI. */ | |
7573 | static int | |
7574 | grow_outbuf (obuf, needed) | |
7575 | register FILE_BUF *obuf; | |
7576 | register int needed; | |
7577 | { | |
7578 | register U_CHAR *p; | |
7579 | int minsize; | |
7580 | ||
7581 | if (obuf->length - (obuf->bufp - obuf->buf) > needed) | |
7582 | return 0; | |
7583 | ||
7584 | /* Make it at least twice as big as it is now. */ | |
7585 | obuf->length *= 2; | |
7586 | /* Make it have at least 150% of the free space we will need. */ | |
7587 | minsize = (3 * needed) / 2 + (obuf->bufp - obuf->buf); | |
7588 | if (minsize > obuf->length) | |
7589 | obuf->length = minsize; | |
7590 | ||
7591 | if ((p = (U_CHAR *) xrealloc (obuf->buf, obuf->length)) == NULL) | |
7592 | memory_full (); | |
7593 | ||
7594 | obuf->bufp = p + (obuf->bufp - obuf->buf); | |
7595 | obuf->buf = p; | |
7596 | ||
7597 | return 0; | |
7598 | } | |
7599 | \f | |
7600 | /* Symbol table for macro names and special symbols */ | |
7601 | ||
7602 | /* | |
7603 | * install a name in the main hash table, even if it is already there. | |
7604 | * name stops with first non alphanumeric, except leading '#'. | |
7605 | * caller must check against redefinition if that is desired. | |
7606 | * delete_macro () removes things installed by install () in fifo order. | |
7607 | * this is important because of the `defined' special symbol used | |
7608 | * in #if, and also if pushdef/popdef directives are ever implemented. | |
7609 | * | |
7610 | * If LEN is >= 0, it is the length of the name. | |
7611 | * Otherwise, compute the length by scanning the entire name. | |
7612 | * | |
7613 | * If HASH is >= 0, it is the precomputed hash code. | |
7614 | * Otherwise, compute the hash code. | |
7615 | */ | |
7616 | static HASHNODE * | |
7617 | install (name, len, type, value, hash) | |
7618 | U_CHAR *name; | |
7619 | int len; | |
7620 | enum node_type type; | |
7621 | int value; | |
7622 | int hash; | |
7623 | /* watch out here if sizeof (U_CHAR *) != sizeof (int) */ | |
7624 | { | |
7625 | register HASHNODE *hp; | |
7626 | register int i, bucket; | |
7627 | register U_CHAR *p, *q; | |
7628 | ||
7629 | if (len < 0) { | |
7630 | p = name; | |
7631 | while (is_idchar[*p]) | |
7632 | p++; | |
7633 | len = p - name; | |
7634 | } | |
7635 | ||
7636 | if (hash < 0) | |
7637 | hash = hashf (name, len, HASHSIZE); | |
7638 | ||
7639 | i = sizeof (HASHNODE) + len + 1; | |
7640 | hp = (HASHNODE *) xmalloc (i); | |
7641 | bucket = hash; | |
7642 | hp->bucket_hdr = &hashtab[bucket]; | |
7643 | hp->next = hashtab[bucket]; | |
7644 | hashtab[bucket] = hp; | |
7645 | hp->prev = NULL; | |
7646 | if (hp->next != NULL) | |
7647 | hp->next->prev = hp; | |
7648 | hp->type = type; | |
7649 | hp->length = len; | |
7650 | hp->value.ival = value; | |
7651 | hp->name = ((U_CHAR *) hp) + sizeof (HASHNODE); | |
7652 | p = hp->name; | |
7653 | q = name; | |
7654 | for (i = 0; i < len; i++) | |
7655 | *p++ = *q++; | |
7656 | hp->name[len] = 0; | |
7657 | return hp; | |
7658 | } | |
7659 | ||
7660 | /* | |
7661 | * find the most recent hash node for name name (ending with first | |
7662 | * non-identifier char) installed by install | |
7663 | * | |
7664 | * If LEN is >= 0, it is the length of the name. | |
7665 | * Otherwise, compute the length by scanning the entire name. | |
7666 | * | |
7667 | * If HASH is >= 0, it is the precomputed hash code. | |
7668 | * Otherwise, compute the hash code. | |
7669 | */ | |
7670 | HASHNODE * | |
7671 | lookup (name, len, hash) | |
7672 | U_CHAR *name; | |
7673 | int len; | |
7674 | int hash; | |
7675 | { | |
7676 | register U_CHAR *bp; | |
7677 | register HASHNODE *bucket; | |
7678 | ||
7679 | if (len < 0) { | |
7680 | for (bp = name; is_idchar[*bp]; bp++) ; | |
7681 | len = bp - name; | |
7682 | } | |
7683 | ||
7684 | if (hash < 0) | |
7685 | hash = hashf (name, len, HASHSIZE); | |
7686 | ||
7687 | bucket = hashtab[hash]; | |
7688 | while (bucket) { | |
7689 | if (bucket->length == len && strncmp (bucket->name, name, len) == 0) | |
7690 | return bucket; | |
7691 | bucket = bucket->next; | |
7692 | } | |
7693 | return NULL; | |
7694 | } | |
7695 | ||
7696 | /* | |
7697 | * Delete a hash node. Some weirdness to free junk from macros. | |
7698 | * More such weirdness will have to be added if you define more hash | |
7699 | * types that need it. | |
7700 | */ | |
7701 | ||
7702 | /* Note that the DEFINITION of a macro is removed from the hash table | |
7703 | but its storage is not freed. This would be a storage leak | |
7704 | except that it is not reasonable to keep undefining and redefining | |
7705 | large numbers of macros many times. | |
7706 | In any case, this is necessary, because a macro can be #undef'd | |
7707 | in the middle of reading the arguments to a call to it. | |
7708 | If #undef freed the DEFINITION, that would crash. */ | |
7709 | ||
7710 | static void | |
7711 | delete_macro (hp) | |
7712 | HASHNODE *hp; | |
7713 | { | |
7714 | ||
7715 | if (hp->prev != NULL) | |
7716 | hp->prev->next = hp->next; | |
7717 | if (hp->next != NULL) | |
7718 | hp->next->prev = hp->prev; | |
7719 | ||
7720 | /* make sure that the bucket chain header that | |
7721 | the deleted guy was on points to the right thing afterwards. */ | |
7722 | if (hp == *hp->bucket_hdr) | |
7723 | *hp->bucket_hdr = hp->next; | |
7724 | ||
7725 | #if 0 | |
7726 | if (hp->type == T_MACRO) { | |
7727 | DEFINITION *d = hp->value.defn; | |
7728 | struct reflist *ap, *nextap; | |
7729 | ||
7730 | for (ap = d->pattern; ap != NULL; ap = nextap) { | |
7731 | nextap = ap->next; | |
7732 | free (ap); | |
7733 | } | |
7734 | free (d); | |
7735 | } | |
7736 | #endif | |
7737 | free (hp); | |
7738 | } | |
7739 | ||
7740 | /* | |
7741 | * return hash function on name. must be compatible with the one | |
7742 | * computed a step at a time, elsewhere | |
7743 | */ | |
7744 | static int | |
7745 | hashf (name, len, hashsize) | |
7746 | register U_CHAR *name; | |
7747 | register int len; | |
7748 | int hashsize; | |
7749 | { | |
7750 | register int r = 0; | |
7751 | ||
7752 | while (len--) | |
7753 | r = HASHSTEP (r, *name++); | |
7754 | ||
7755 | return MAKE_POS (r) % hashsize; | |
7756 | } | |
7757 | \f | |
7758 | ||
7759 | /* Dump the definition of a single macro HP to OF. */ | |
7760 | static void | |
7761 | dump_single_macro (hp, of) | |
7762 | register HASHNODE *hp; | |
7763 | FILE *of; | |
7764 | { | |
7765 | register DEFINITION *defn = hp->value.defn; | |
7766 | struct reflist *ap; | |
7767 | int offset; | |
7768 | int concat; | |
7769 | ||
7770 | ||
7771 | /* Print the definition of the macro HP. */ | |
7772 | ||
7773 | fprintf (of, "#define %s", hp->name); | |
7774 | ||
7775 | if (defn->nargs >= 0) { | |
7776 | int i; | |
7777 | ||
7778 | fprintf (of, "("); | |
7779 | for (i = 0; i < defn->nargs; i++) { | |
7780 | dump_arg_n (defn, i, of); | |
7781 | if (i + 1 < defn->nargs) | |
7782 | fprintf (of, ", "); | |
7783 | } | |
7784 | fprintf (of, ")"); | |
7785 | } | |
7786 | ||
7787 | fprintf (of, " "); | |
7788 | ||
7789 | offset = 0; | |
7790 | concat = 0; | |
7791 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
7792 | dump_defn_1 (defn->expansion, offset, ap->nchars, of); | |
7793 | if (ap->nchars != 0) | |
7794 | concat = 0; | |
7795 | offset += ap->nchars; | |
7796 | if (ap->stringify) | |
7797 | fprintf (of, " #"); | |
7798 | if (ap->raw_before && !concat) | |
7799 | fprintf (of, " ## "); | |
7800 | concat = 0; | |
7801 | dump_arg_n (defn, ap->argno, of); | |
7802 | if (ap->raw_after) { | |
7803 | fprintf (of, " ## "); | |
7804 | concat = 1; | |
7805 | } | |
7806 | } | |
7807 | dump_defn_1 (defn->expansion, offset, defn->length - offset, of); | |
7808 | fprintf (of, "\n"); | |
7809 | } | |
7810 | ||
7811 | /* Dump all macro definitions as #defines to stdout. */ | |
7812 | ||
7813 | static void | |
7814 | dump_all_macros () | |
7815 | { | |
7816 | int bucket; | |
7817 | ||
7818 | for (bucket = 0; bucket < HASHSIZE; bucket++) { | |
7819 | register HASHNODE *hp; | |
7820 | ||
7821 | for (hp = hashtab[bucket]; hp; hp= hp->next) { | |
7822 | if (hp->type == T_MACRO) | |
7823 | dump_single_macro (hp, stdout); | |
7824 | } | |
7825 | } | |
7826 | } | |
7827 | ||
7828 | /* Output to OF a substring of a macro definition. | |
7829 | BASE is the beginning of the definition. | |
7830 | Output characters START thru LENGTH. | |
7831 | Discard newlines outside of strings, thus | |
7832 | converting funny-space markers to ordinary spaces. */ | |
7833 | ||
7834 | static void | |
7835 | dump_defn_1 (base, start, length, of) | |
7836 | U_CHAR *base; | |
7837 | int start; | |
7838 | int length; | |
7839 | FILE *of; | |
7840 | { | |
7841 | U_CHAR *p = base + start; | |
7842 | U_CHAR *limit = base + start + length; | |
7843 | ||
7844 | while (p < limit) { | |
7845 | if (*p != '\n') | |
7846 | putc (*p, of); | |
7847 | else if (*p == '\"' || *p =='\'') { | |
7848 | U_CHAR *p1 = skip_quoted_string (p, limit, 0, 0, 0, 0); | |
7849 | fwrite (p, p1 - p, 1, of); | |
7850 | p = p1 - 1; | |
7851 | } | |
7852 | p++; | |
7853 | } | |
7854 | } | |
7855 | ||
7856 | /* Print the name of argument number ARGNUM of macro definition DEFN | |
7857 | to OF. | |
7858 | Recall that DEFN->args.argnames contains all the arg names | |
7859 | concatenated in reverse order with comma-space in between. */ | |
7860 | ||
7861 | static void | |
7862 | dump_arg_n (defn, argnum, of) | |
7863 | DEFINITION *defn; | |
7864 | int argnum; | |
7865 | FILE *of; | |
7866 | { | |
7867 | register U_CHAR *p = defn->args.argnames; | |
7868 | while (argnum + 1 < defn->nargs) { | |
7869 | p = (U_CHAR *) index (p, ' ') + 1; | |
7870 | argnum++; | |
7871 | } | |
7872 | ||
7873 | while (*p && *p != ',') { | |
7874 | putc (*p, of); | |
7875 | p++; | |
7876 | } | |
7877 | } | |
7878 | \f | |
7879 | /* Initialize syntactic classifications of characters. */ | |
7880 | ||
7881 | static void | |
7882 | initialize_char_syntax () | |
7883 | { | |
7884 | register int i; | |
7885 | ||
7886 | /* | |
7887 | * Set up is_idchar and is_idstart tables. These should be | |
7888 | * faster than saying (is_alpha (c) || c == '_'), etc. | |
7889 | * Set up these things before calling any routines tthat | |
7890 | * refer to them. | |
7891 | */ | |
7892 | for (i = 'a'; i <= 'z'; i++) { | |
7893 | is_idchar[i - 'a' + 'A'] = 1; | |
7894 | is_idchar[i] = 1; | |
7895 | is_idstart[i - 'a' + 'A'] = 1; | |
7896 | is_idstart[i] = 1; | |
7897 | } | |
7898 | for (i = '0'; i <= '9'; i++) | |
7899 | is_idchar[i] = 1; | |
7900 | is_idchar['_'] = 1; | |
7901 | is_idstart['_'] = 1; | |
7902 | is_idchar['$'] = dollars_in_ident; | |
7903 | is_idstart['$'] = dollars_in_ident; | |
7904 | ||
7905 | /* horizontal space table */ | |
7906 | is_hor_space[' '] = 1; | |
7907 | is_hor_space['\t'] = 1; | |
7908 | is_hor_space['\v'] = 1; | |
7909 | is_hor_space['\f'] = 1; | |
7910 | is_hor_space['\r'] = 1; | |
7911 | ||
7912 | is_space[' '] = 1; | |
7913 | is_space['\t'] = 1; | |
7914 | is_space['\v'] = 1; | |
7915 | is_space['\f'] = 1; | |
7916 | is_space['\n'] = 1; | |
7917 | is_space['\r'] = 1; | |
7918 | } | |
7919 | ||
7920 | /* Initialize the built-in macros. */ | |
7921 | ||
7922 | static void | |
7923 | initialize_builtins (inp, outp) | |
7924 | FILE_BUF *inp; | |
7925 | FILE_BUF *outp; | |
7926 | { | |
7927 | time_t t; | |
7928 | ||
7929 | t = time (0); | |
7930 | timebuf = localtime (&t); | |
7931 | ||
7932 | install ("__LINE__", -1, T_SPECLINE, 0, -1); | |
7933 | install ("__DATE__", -1, T_DATE, 0, -1); | |
7934 | install ("__FILE__", -1, T_FILE, 0, -1); | |
7935 | install ("__BASE_FILE__", -1, T_BASE_FILE, 0, -1); | |
7936 | install ("__INCLUDE_LEVEL__", -1, T_INCLUDE_LEVEL, 0, -1); | |
7937 | install ("__VERSION__", -1, T_VERSION, 0, -1); | |
7938 | install ("__SIZE_TYPE__", -1, T_SIZE_TYPE, 0, -1); | |
7939 | install ("__PTRDIFF_TYPE__ ", -1, T_PTRDIFF_TYPE, 0, -1); | |
7940 | install ("__WCHAR_TYPE__", -1, T_WCHAR_TYPE, 0, -1); | |
7941 | install ("__TIME__", -1, T_TIME, 0, -1); | |
7942 | if (!traditional) | |
7943 | install ("__STDC__", -1, T_CONST, STDC_VALUE, -1); | |
7944 | if (objc) | |
7945 | install ("__OBJC__", -1, T_CONST, 1, -1); | |
7946 | /* This is supplied using a -D by the compiler driver | |
7947 | so that it is present only when truly compiling with GNU C. */ | |
7948 | /* install ("__GNUC__", -1, T_CONST, 2, -1); */ | |
7949 | ||
7950 | if (debug_output) | |
7951 | { | |
7952 | char directive[2048]; | |
7953 | register struct directive *dp = &directive_table[0]; | |
7954 | ||
7955 | sprintf (directive, " __BASE_FILE__ \"%s\"", | |
7956 | instack[0].nominal_fname); | |
7957 | output_line_command (inp, outp, 0, same_file); | |
7958 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7959 | ||
7960 | sprintf (directive, " __VERSION__ \"%s\"", version_string); | |
7961 | output_line_command (inp, outp, 0, same_file); | |
7962 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7963 | ||
7964 | sprintf (directive, " __SIZE_TYPE__ %s", SIZE_TYPE); | |
7965 | output_line_command (inp, outp, 0, same_file); | |
7966 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7967 | ||
7968 | sprintf (directive, " __PTRDIFF_TYPE__ %s", PTRDIFF_TYPE); | |
7969 | output_line_command (inp, outp, 0, same_file); | |
7970 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7971 | ||
7972 | sprintf (directive, " __WCHAR_TYPE__ %s", WCHAR_TYPE); | |
7973 | output_line_command (inp, outp, 0, same_file); | |
7974 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7975 | ||
7976 | sprintf (directive, " __WCHAR_TYPE__ %s", WCHAR_TYPE); | |
7977 | output_line_command (inp, outp, 0, same_file); | |
7978 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7979 | ||
7980 | sprintf (directive, " __DATE__ \"%s %2d %4d\"", | |
7981 | monthnames[timebuf->tm_mon], | |
7982 | timebuf->tm_mday, timebuf->tm_year + 1900); | |
7983 | output_line_command (inp, outp, 0, same_file); | |
7984 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7985 | ||
7986 | sprintf (directive, " __TIME__ \"%02d:%02d:%02d\"", | |
7987 | timebuf->tm_hour, timebuf->tm_min, timebuf->tm_sec); | |
7988 | output_line_command (inp, outp, 0, same_file); | |
7989 | pass_thru_directive (directive, &directive[strlen (directive)], outp, dp); | |
7990 | ||
7991 | if (!traditional) | |
7992 | { | |
7993 | sprintf (directive, " __STDC__ 1"); | |
7994 | output_line_command (inp, outp, 0, same_file); | |
7995 | pass_thru_directive (directive, &directive[strlen (directive)], | |
7996 | outp, dp); | |
7997 | } | |
7998 | if (objc) | |
7999 | { | |
8000 | sprintf (directive, " __OBJC__ 1"); | |
8001 | output_line_command (inp, outp, 0, same_file); | |
8002 | pass_thru_directive (directive, &directive[strlen (directive)], | |
8003 | outp, dp); | |
8004 | } | |
8005 | } | |
8006 | } | |
8007 | \f | |
8008 | /* | |
8009 | * process a given definition string, for initialization | |
8010 | * If STR is just an identifier, define it with value 1. | |
8011 | * If STR has anything after the identifier, then it should | |
8012 | * be identifier=definition. | |
8013 | */ | |
8014 | ||
8015 | static void | |
8016 | make_definition (str, op) | |
8017 | U_CHAR *str; | |
8018 | FILE_BUF *op; | |
8019 | { | |
8020 | FILE_BUF *ip; | |
8021 | struct directive *kt; | |
8022 | U_CHAR *buf, *p; | |
8023 | ||
8024 | buf = str; | |
8025 | p = str; | |
8026 | if (!is_idstart[*p]) { | |
8027 | error ("malformed option `-D %s'", str); | |
8028 | return; | |
8029 | } | |
8030 | while (is_idchar[*++p]) | |
8031 | ; | |
8032 | if (*p == 0) { | |
8033 | buf = (U_CHAR *) alloca (p - buf + 4); | |
8034 | strcpy ((char *)buf, str); | |
8035 | strcat ((char *)buf, " 1"); | |
8036 | } else if (*p != '=') { | |
8037 | error ("malformed option `-D %s'", str); | |
8038 | return; | |
8039 | } else { | |
8040 | U_CHAR *q; | |
8041 | /* Copy the entire option so we can modify it. */ | |
8042 | buf = (U_CHAR *) alloca (2 * strlen (str) + 1); | |
8043 | strncpy (buf, str, p - str); | |
8044 | /* Change the = to a space. */ | |
8045 | buf[p - str] = ' '; | |
8046 | /* Scan for any backslash-newline and remove it. */ | |
8047 | p++; | |
8048 | q = &buf[p - str]; | |
8049 | while (*p) { | |
8050 | if (*p == '\\' && p[1] == '\n') | |
8051 | p += 2; | |
8052 | /* Change newline chars into newline-markers. */ | |
8053 | else if (*p == '\n') | |
8054 | { | |
8055 | *q++ = '\n'; | |
8056 | *q++ = '\n'; | |
8057 | p++; | |
8058 | } | |
8059 | else | |
8060 | *q++ = *p++; | |
8061 | } | |
8062 | *q = 0; | |
8063 | } | |
8064 | ||
8065 | ip = &instack[++indepth]; | |
8066 | ip->nominal_fname = ip->fname = "*Initialization*"; | |
8067 | ||
8068 | ip->buf = ip->bufp = buf; | |
8069 | ip->length = strlen (buf); | |
8070 | ip->lineno = 1; | |
8071 | ip->macro = 0; | |
8072 | ip->free_ptr = 0; | |
8073 | ip->if_stack = if_stack; | |
8074 | ip->system_header_p = 0; | |
8075 | ||
8076 | for (kt = directive_table; kt->type != T_DEFINE; kt++) | |
8077 | ; | |
8078 | ||
8079 | do_define (buf, buf + strlen (buf) , op, kt); | |
8080 | --indepth; | |
8081 | } | |
8082 | ||
8083 | /* JF, this does the work for the -U option */ | |
8084 | ||
8085 | static void | |
8086 | make_undef (str, op) | |
8087 | U_CHAR *str; | |
8088 | FILE_BUF *op; | |
8089 | { | |
8090 | FILE_BUF *ip; | |
8091 | struct directive *kt; | |
8092 | ||
8093 | ip = &instack[++indepth]; | |
8094 | ip->nominal_fname = ip->fname = "*undef*"; | |
8095 | ||
8096 | ip->buf = ip->bufp = str; | |
8097 | ip->length = strlen (str); | |
8098 | ip->lineno = 1; | |
8099 | ip->macro = 0; | |
8100 | ip->free_ptr = 0; | |
8101 | ip->if_stack = if_stack; | |
8102 | ip->system_header_p = 0; | |
8103 | ||
8104 | for (kt = directive_table; kt->type != T_UNDEF; kt++) | |
8105 | ; | |
8106 | ||
8107 | do_undef (str, str + strlen (str), op, kt); | |
8108 | --indepth; | |
8109 | } | |
8110 | \f | |
8111 | /* Process the string STR as if it appeared as the body of a #assert. | |
8112 | OPTION is the option name for which STR was the argument. */ | |
8113 | ||
8114 | static void | |
8115 | make_assertion (option, str) | |
8116 | char *option; | |
8117 | U_CHAR *str; | |
8118 | { | |
8119 | FILE_BUF *ip; | |
8120 | struct directive *kt; | |
8121 | U_CHAR *buf, *p, *q; | |
8122 | ||
8123 | /* Copy the entire option so we can modify it. */ | |
8124 | buf = (U_CHAR *) alloca (strlen (str) + 1); | |
8125 | strcpy ((char *) buf, str); | |
8126 | /* Scan for any backslash-newline and remove it. */ | |
8127 | p = q = buf; | |
8128 | while (*p) { | |
8129 | if (*p == '\\' && p[1] == '\n') | |
8130 | p += 2; | |
8131 | else | |
8132 | *q++ = *p++; | |
8133 | } | |
8134 | *q = 0; | |
8135 | ||
8136 | p = buf; | |
8137 | if (!is_idstart[*p]) { | |
8138 | error ("malformed option `%s %s'", option, str); | |
8139 | return; | |
8140 | } | |
8141 | while (is_idchar[*++p]) | |
8142 | ; | |
8143 | while (*p == ' ' || *p == '\t') p++; | |
8144 | if (! (*p == 0 || *p == '(')) { | |
8145 | error ("malformed option `%s %s'", option, str); | |
8146 | return; | |
8147 | } | |
8148 | ||
8149 | ip = &instack[++indepth]; | |
8150 | ip->nominal_fname = ip->fname = "*Initialization*"; | |
8151 | ||
8152 | ip->buf = ip->bufp = buf; | |
8153 | ip->length = strlen (buf); | |
8154 | ip->lineno = 1; | |
8155 | ip->macro = 0; | |
8156 | ip->free_ptr = 0; | |
8157 | ip->if_stack = if_stack; | |
8158 | ip->system_header_p = 0; | |
8159 | ||
8160 | for (kt = directive_table; kt->type != T_ASSERT; kt++) | |
8161 | ; | |
8162 | ||
8163 | /* pass NULL as output ptr to do_define since we KNOW it never | |
8164 | does any output.... */ | |
8165 | do_assert (buf, buf + strlen (buf) , NULL, kt); | |
8166 | --indepth; | |
8167 | } | |
8168 | \f | |
8169 | /* Add output to `deps_buffer' for the -M switch. | |
8170 | STRING points to the text to be output. | |
8171 | SIZE is the number of bytes, or 0 meaning output until a null. | |
8172 | Outputting the empty string breaks the line if it is long enough. */ | |
8173 | ||
8174 | static void | |
8175 | deps_output (string, size) | |
8176 | char *string; | |
8177 | unsigned size; | |
8178 | { | |
8179 | if (size == 0) | |
8180 | size = strlen (string); | |
8181 | ||
8182 | #ifndef MAX_OUTPUT_COLUMNS | |
8183 | #define MAX_OUTPUT_COLUMNS 75 | |
8184 | #endif | |
8185 | if (size == 0 && deps_column != 0 | |
8186 | && size + deps_column > MAX_OUTPUT_COLUMNS) { | |
8187 | deps_output ("\\\n ", 0); | |
8188 | deps_column = 0; | |
8189 | } | |
8190 | ||
8191 | if (deps_size + size + 1 > deps_allocated_size) { | |
8192 | deps_allocated_size = deps_size + size + 50; | |
8193 | deps_allocated_size *= 2; | |
8194 | deps_buffer = (char *) xrealloc (deps_buffer, deps_allocated_size); | |
8195 | } | |
8196 | bcopy (string, &deps_buffer[deps_size], size); | |
8197 | deps_size += size; | |
8198 | deps_column += size; | |
8199 | deps_buffer[deps_size] = 0; | |
8200 | } | |
8201 | \f | |
8202 | #if defined(USG) || defined(VMS) | |
8203 | #ifndef BSTRING | |
8204 | ||
8205 | void | |
8206 | bzero (b, length) | |
8207 | register char *b; | |
8208 | register unsigned length; | |
8209 | { | |
8210 | #ifdef VMS | |
8211 | short zero = 0; | |
8212 | long max_str = 65535; | |
8213 | ||
8214 | while (length > max_str) { | |
8215 | (void) LIB$MOVC5 (&zero, &zero, &zero, &max_str, b); | |
8216 | length -= max_str; | |
8217 | b += max_str; | |
8218 | } | |
8219 | (void) LIB$MOVC5 (&zero, &zero, &zero, &length, b); | |
8220 | #else | |
8221 | while (length-- > 0) | |
8222 | *b++ = 0; | |
8223 | #endif /* not VMS */ | |
8224 | } | |
8225 | ||
8226 | void | |
8227 | bcopy (b1, b2, length) | |
8228 | register char *b1; | |
8229 | register char *b2; | |
8230 | register unsigned length; | |
8231 | { | |
8232 | #ifdef VMS | |
8233 | long max_str = 65535; | |
8234 | ||
8235 | while (length > max_str) { | |
8236 | (void) LIB$MOVC3 (&max_str, b1, b2); | |
8237 | length -= max_str; | |
8238 | b1 += max_str; | |
8239 | b2 += max_str; | |
8240 | } | |
8241 | (void) LIB$MOVC3 (&length, b1, b2); | |
8242 | #else | |
8243 | while (length-- > 0) | |
8244 | *b2++ = *b1++; | |
8245 | #endif /* not VMS */ | |
8246 | } | |
8247 | ||
8248 | int | |
8249 | bcmp (b1, b2, length) /* This could be a macro! */ | |
8250 | register char *b1; | |
8251 | register char *b2; | |
8252 | register unsigned length; | |
8253 | { | |
8254 | #ifdef VMS | |
8255 | struct dsc$descriptor_s src1 = {length, DSC$K_DTYPE_T, DSC$K_CLASS_S, b1}; | |
8256 | struct dsc$descriptor_s src2 = {length, DSC$K_DTYPE_T, DSC$K_CLASS_S, b2}; | |
8257 | ||
8258 | return STR$COMPARE (&src1, &src2); | |
8259 | #else | |
8260 | while (length-- > 0) | |
8261 | if (*b1++ != *b2++) | |
8262 | return 1; | |
8263 | ||
8264 | return 0; | |
8265 | #endif /* not VMS */ | |
8266 | } | |
8267 | #endif /* not BSTRING */ | |
8268 | #endif /* USG or VMS */ | |
8269 | ||
8270 | \f | |
8271 | static void | |
8272 | fatal (str, arg) | |
8273 | char *str, *arg; | |
8274 | { | |
8275 | if (deps_file) | |
8276 | unlink (deps_file); | |
8277 | fprintf (stderr, "%s: ", progname); | |
8278 | fprintf (stderr, str, arg); | |
8279 | fprintf (stderr, "\n"); | |
8280 | exit (FAILURE_EXIT_CODE); | |
8281 | } | |
8282 | ||
8283 | /* More 'friendly' abort that prints the line and file. | |
8284 | config.h can #define abort fancy_abort if you like that sort of thing. */ | |
8285 | ||
8286 | void | |
8287 | fancy_abort () | |
8288 | { | |
8289 | fatal ("Internal gcc abort."); | |
8290 | } | |
8291 | ||
8292 | static void | |
8293 | perror_with_name (name) | |
8294 | char *name; | |
8295 | { | |
8296 | fprintf (stderr, "%s: ", progname); | |
8297 | if (errno < sys_nerr) | |
8298 | fprintf (stderr, "%s: %s\n", name, sys_errlist[errno]); | |
8299 | else | |
8300 | fprintf (stderr, "%s: undocumented I/O error\n", name); | |
8301 | errors++; | |
8302 | } | |
8303 | ||
8304 | static void | |
8305 | pfatal_with_name (name) | |
8306 | char *name; | |
8307 | { | |
8308 | perror_with_name (name); | |
8309 | #ifdef VMS | |
8310 | exit (vaxc$errno); | |
8311 | #else | |
8312 | exit (FAILURE_EXIT_CODE); | |
8313 | #endif | |
8314 | } | |
8315 | ||
8316 | \f | |
8317 | static void | |
8318 | memory_full () | |
8319 | { | |
8320 | fatal ("Memory exhausted."); | |
8321 | } | |
8322 | ||
8323 | ||
8324 | char * | |
8325 | xmalloc (size) | |
8326 | unsigned size; | |
8327 | { | |
8328 | register char *ptr = (char *) malloc (size); | |
8329 | if (ptr != 0) return (ptr); | |
8330 | memory_full (); | |
8331 | /*NOTREACHED*/ | |
8332 | return 0; | |
8333 | } | |
8334 | ||
8335 | static char * | |
8336 | xrealloc (old, size) | |
8337 | char *old; | |
8338 | unsigned size; | |
8339 | { | |
8340 | register char *ptr = (char *) realloc (old, size); | |
8341 | if (ptr != 0) return (ptr); | |
8342 | memory_full (); | |
8343 | /*NOTREACHED*/ | |
8344 | return 0; | |
8345 | } | |
8346 | ||
8347 | static char * | |
8348 | xcalloc (number, size) | |
8349 | unsigned number, size; | |
8350 | { | |
8351 | register unsigned total = number * size; | |
8352 | register char *ptr = (char *) malloc (total); | |
8353 | if (ptr != 0) { | |
8354 | if (total > 100) | |
8355 | bzero (ptr, total); | |
8356 | else { | |
8357 | /* It's not too long, so loop, zeroing by longs. | |
8358 | It must be safe because malloc values are always well aligned. */ | |
8359 | register long *zp = (long *) ptr; | |
8360 | register long *zl = (long *) (ptr + total - 4); | |
8361 | register int i = total - 4; | |
8362 | while (zp < zl) | |
8363 | *zp++ = 0; | |
8364 | if (i < 0) | |
8365 | i = 0; | |
8366 | while (i < total) | |
8367 | ptr[i++] = 0; | |
8368 | } | |
8369 | return ptr; | |
8370 | } | |
8371 | memory_full (); | |
8372 | /*NOTREACHED*/ | |
8373 | return 0; | |
8374 | } | |
8375 | ||
8376 | static char * | |
8377 | savestring (input) | |
8378 | char *input; | |
8379 | { | |
8380 | unsigned size = strlen (input); | |
8381 | char *output = xmalloc (size + 1); | |
8382 | strcpy (output, input); | |
8383 | return output; | |
8384 | } | |
8385 | \f | |
8386 | /* Get the file-mode and data size of the file open on FD | |
8387 | and store them in *MODE_POINTER and *SIZE_POINTER. */ | |
8388 | ||
8389 | static int | |
8390 | file_size_and_mode (fd, mode_pointer, size_pointer) | |
8391 | int fd; | |
8392 | int *mode_pointer; | |
8393 | long int *size_pointer; | |
8394 | { | |
8395 | struct stat sbuf; | |
8396 | ||
8397 | if (fstat (fd, &sbuf) < 0) return (-1); | |
8398 | if (mode_pointer) *mode_pointer = sbuf.st_mode; | |
8399 | if (size_pointer) *size_pointer = sbuf.st_size; | |
8400 | return 0; | |
8401 | } | |
8402 | \f | |
8403 | #ifdef VMS | |
8404 | ||
8405 | /* Under VMS we need to fix up the "include" specification | |
8406 | filename so that everything following the 1st slash is | |
8407 | changed into its correct VMS file specification. */ | |
8408 | ||
8409 | static void | |
8410 | hack_vms_include_specification (fname) | |
8411 | char *fname; | |
8412 | { | |
8413 | register char *cp, *cp1, *cp2; | |
8414 | int f, check_filename_before_returning, no_prefix_seen; | |
8415 | char Local[512]; | |
8416 | ||
8417 | check_filename_before_returning = 0; | |
8418 | no_prefix_seen = 0; | |
8419 | ||
8420 | /* Ignore leading "./"s */ | |
8421 | while (fname[0] == '.' && fname[1] == '/') { | |
8422 | strcpy (fname, fname+2); | |
8423 | no_prefix_seen = 1; /* mark this for later */ | |
8424 | } | |
8425 | /* Look for the boundary between the VMS and UNIX filespecs */ | |
8426 | cp = rindex (fname, ']'); /* Look for end of dirspec. */ | |
8427 | if (cp == 0) cp = rindex (fname, '>'); /* ... Ditto */ | |
8428 | if (cp == 0) cp = rindex (fname, ':'); /* Look for end of devspec. */ | |
8429 | if (cp) { | |
8430 | cp++; | |
8431 | } else { | |
8432 | cp = index (fname, '/'); /* Look for the "/" */ | |
8433 | } | |
8434 | ||
8435 | cp2 = Local; /* initialize */ | |
8436 | ||
8437 | /* We are trying to do a number of things here. First of all, we are | |
8438 | trying to hammer the filenames into a standard format, such that later | |
8439 | processing can handle them. | |
8440 | ||
8441 | If the file name contains something like [dir.], then it recognizes this | |
8442 | as a root, and strips the ".]". Later processing will add whatever is | |
8443 | needed to get things working properly. | |
8444 | ||
8445 | If no device is specified, then the first directory name is taken to be | |
8446 | a device name (or a rooted logical). */ | |
8447 | ||
8448 | /* See if we found that 1st slash */ | |
8449 | if (cp == 0) return; /* Nothing to do!!! */ | |
8450 | if (*cp != '/') return; /* Nothing to do!!! */ | |
8451 | /* Point to the UNIX filename part (which needs to be fixed!) */ | |
8452 | cp1 = cp+1; | |
8453 | /* If the directory spec is not rooted, we can just copy | |
8454 | the UNIX filename part and we are done */ | |
8455 | if (((cp - fname) > 1) && ((cp[-1] == ']') || (cp[-1] == '>'))) { | |
8456 | if (cp[-2] != '.') { | |
8457 | /* | |
8458 | * The VMS part ends in a `]', and the preceeding character is not a `.'. | |
8459 | * We strip the `]', and then splice the two parts of the name in the | |
8460 | * usual way. Given the default locations for include files in cccp.c, | |
8461 | * we will only use this code if the user specifies alternate locations | |
8462 | * with the /include (-I) switch on the command line. */ | |
8463 | cp -= 1; /* Strip "]" */ | |
8464 | cp1--; /* backspace */ | |
8465 | } else { | |
8466 | /* | |
8467 | * The VMS part has a ".]" at the end, and this will not do. Later | |
8468 | * processing will add a second directory spec, and this would be a syntax | |
8469 | * error. Thus we strip the ".]", and thus merge the directory specs. | |
8470 | * We also backspace cp1, so that it points to a '/'. This inhibits the | |
8471 | * generation of the 000000 root directory spec (which does not belong here | |
8472 | * in this case). | |
8473 | */ | |
8474 | cp -= 2; /* Strip ".]" */ | |
8475 | cp1--; }; /* backspace */ | |
8476 | } else { | |
8477 | ||
8478 | /* We drop in here if there is no VMS style directory specification yet. | |
8479 | * If there is no device specification either, we make the first dir a | |
8480 | * device and try that. If we do not do this, then we will be essentially | |
8481 | * searching the users default directory (as if they did a #include "asdf.h"). | |
8482 | * | |
8483 | * Then all we need to do is to push a '[' into the output string. Later | |
8484 | * processing will fill this in, and close the bracket. | |
8485 | */ | |
8486 | if(cp[-1] != ':') *cp2++ = ':'; /* dev not in spec. take first dir */ | |
8487 | *cp2++ = '['; /* Open the directory specification */ | |
8488 | } | |
8489 | ||
8490 | /* at this point we assume that we have the device spec, and (at least | |
8491 | the opening "[" for a directory specification. We may have directories | |
8492 | specified already */ | |
8493 | ||
8494 | /* If there are no other slashes then the filename will be | |
8495 | in the "root" directory. Otherwise, we need to add | |
8496 | directory specifications. */ | |
8497 | if (index (cp1, '/') == 0) { | |
8498 | /* Just add "000000]" as the directory string */ | |
8499 | strcpy (cp2, "000000]"); | |
8500 | cp2 += strlen (cp2); | |
8501 | check_filename_before_returning = 1; /* we might need to fool with this later */ | |
8502 | } else { | |
8503 | /* As long as there are still subdirectories to add, do them. */ | |
8504 | while (index (cp1, '/') != 0) { | |
8505 | /* If this token is "." we can ignore it */ | |
8506 | if ((cp1[0] == '.') && (cp1[1] == '/')) { | |
8507 | cp1 += 2; | |
8508 | continue; | |
8509 | } | |
8510 | /* Add a subdirectory spec. Do not duplicate "." */ | |
8511 | if (cp2[-1] != '.' && cp2[-1] != '[' && cp2[-1] != '<') | |
8512 | *cp2++ = '.'; | |
8513 | /* If this is ".." then the spec becomes "-" */ | |
8514 | if ((cp1[0] == '.') && (cp1[1] == '.') && (cp[2] == '/')) { | |
8515 | /* Add "-" and skip the ".." */ | |
8516 | *cp2++ = '-'; | |
8517 | cp1 += 3; | |
8518 | continue; | |
8519 | } | |
8520 | /* Copy the subdirectory */ | |
8521 | while (*cp1 != '/') *cp2++= *cp1++; | |
8522 | cp1++; /* Skip the "/" */ | |
8523 | } | |
8524 | /* Close the directory specification */ | |
8525 | if(cp2[-1] == '.') /* no trailing periods */ | |
8526 | cp2--; | |
8527 | *cp2++ = ']'; | |
8528 | } | |
8529 | /* Now add the filename */ | |
8530 | while (*cp1) *cp2++ = *cp1++; | |
8531 | *cp2 = 0; | |
8532 | /* Now append it to the original VMS spec. */ | |
8533 | strcpy (cp, Local); | |
8534 | ||
8535 | /* If we put a [000000] in the filename, try to open it first. If this fails, | |
8536 | remove the [000000], and return that name. This provides flexibility | |
8537 | to the user in that they can use both rooted and non-rooted logical names | |
8538 | to point to the location of the file. */ | |
8539 | ||
8540 | if (check_filename_before_returning && no_prefix_seen) { | |
8541 | f = open (fname, O_RDONLY, 0666); | |
8542 | if (f >= 0) { | |
8543 | /* The file name is OK as it is, so return it as is. */ | |
8544 | close (f); | |
8545 | return; | |
8546 | } | |
8547 | /* The filename did not work. Try to remove the [000000] from the name, | |
8548 | and return it. */ | |
8549 | cp = index (fname, '['); | |
8550 | cp2 = index (fname, ']') + 1; | |
8551 | strcpy (cp, cp2); /* this gets rid of it */ | |
8552 | } | |
8553 | return; | |
8554 | } | |
8555 | #endif /* VMS */ | |
8556 | \f | |
8557 | #ifdef VMS | |
8558 | ||
8559 | /* These are the read/write replacement routines for | |
8560 | VAX-11 "C". They make read/write behave enough | |
8561 | like their UNIX counterparts that CCCP will work */ | |
8562 | ||
8563 | static int | |
8564 | read (fd, buf, size) | |
8565 | int fd; | |
8566 | char *buf; | |
8567 | int size; | |
8568 | { | |
8569 | #undef read /* Get back the REAL read routine */ | |
8570 | register int i; | |
8571 | register int total = 0; | |
8572 | ||
8573 | /* Read until the buffer is exhausted */ | |
8574 | while (size > 0) { | |
8575 | /* Limit each read to 32KB */ | |
8576 | i = (size > (32*1024)) ? (32*1024) : size; | |
8577 | i = read (fd, buf, i); | |
8578 | if (i <= 0) { | |
8579 | if (i == 0) return (total); | |
8580 | return(i); | |
8581 | } | |
8582 | /* Account for this read */ | |
8583 | total += i; | |
8584 | buf += i; | |
8585 | size -= i; | |
8586 | } | |
8587 | return (total); | |
8588 | } | |
8589 | ||
8590 | static int | |
8591 | write (fd, buf, size) | |
8592 | int fd; | |
8593 | char *buf; | |
8594 | int size; | |
8595 | { | |
8596 | #undef write /* Get back the REAL write routine */ | |
8597 | int i; | |
8598 | int j; | |
8599 | ||
8600 | /* Limit individual writes to 32Kb */ | |
8601 | i = size; | |
8602 | while (i > 0) { | |
8603 | j = (i > (32*1024)) ? (32*1024) : i; | |
8604 | if (write (fd, buf, j) < 0) return (-1); | |
8605 | /* Account for the data written */ | |
8606 | buf += j; | |
8607 | i -= j; | |
8608 | } | |
8609 | return (size); | |
8610 | } | |
8611 | ||
8612 | /* The following wrapper functions supply additional arguments to the VMS | |
8613 | I/O routines to optimize performance with file handling. The arguments | |
8614 | are: | |
8615 | "mbc=16" - Set multi-block count to 16 (use a 8192 byte buffer). | |
8616 | "deq=64" - When extending the file, extend it in chunks of 32Kbytes. | |
8617 | "fop=tef"- Truncate unused portions of file when closing file. | |
8618 | "shr=nil"- Disallow file sharing while file is open. | |
8619 | */ | |
8620 | ||
8621 | static FILE * | |
8622 | freopen (fname, type, oldfile) | |
8623 | char *fname; | |
8624 | char *type; | |
8625 | FILE *oldfile; | |
8626 | { | |
8627 | #undef freopen /* Get back the REAL fopen routine */ | |
8628 | if (strcmp (type, "w") == 0) | |
8629 | return freopen (fname, type, oldfile, "mbc=16", "deq=64", "fop=tef", "shr=nil"); | |
8630 | return freopen (fname, type, oldfile, "mbc=16"); | |
8631 | } | |
8632 | ||
8633 | static FILE * | |
8634 | fopen (fname, type) | |
8635 | char *fname; | |
8636 | char *type; | |
8637 | { | |
8638 | #undef fopen /* Get back the REAL fopen routine */ | |
8639 | if (strcmp (type, "w") == 0) | |
8640 | return fopen (fname, type, "mbc=16", "deq=64", "fop=tef", "shr=nil"); | |
8641 | return fopen (fname, type, "mbc=16"); | |
8642 | } | |
8643 | ||
8644 | static int | |
8645 | open (fname, flags, prot) | |
8646 | char *fname; | |
8647 | int flags; | |
8648 | int prot; | |
8649 | { | |
8650 | #undef open /* Get back the REAL open routine */ | |
8651 | return open (fname, flags, prot, "mbc=16", "deq=64", "fop=tef"); | |
8652 | } | |
8653 | ||
8654 | #endif /* VMS */ |