]>
Commit | Line | Data |
---|---|---|
5f8037c4 RS |
1 | /* Protoize program - Written by Ron Guilmette at the Microelectronics |
2 | and Computer Technology Corporation (MCC). The author's current | |
3 | E-mail address is <rfg@ncd.com>. | |
4 | ||
5 | Copyright (C) 1989, 1991 Free Software Foundation, Inc. | |
6 | ||
7 | This file is part of GNU CC. | |
8 | ||
9 | GNU CC is free software; you can redistribute it and/or modify | |
10 | it under the terms of the GNU General Public License as published by | |
11 | the Free Software Foundation; either version 2, or (at your option) | |
12 | any later version. | |
13 | ||
14 | GNU CC is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
20 | along with GNU CC; see the file COPYING. If not, write to | |
21 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ | |
22 | ||
23 | /* Any reasonable C++ compiler should have all of the same features | |
24 | as __STDC__ plus more, so make sure that __STDC__ is defined if | |
25 | __cplusplus is defined. */ | |
26 | ||
27 | #if defined(__cplusplus) && !defined(__STDC__) | |
28 | #define __STDC__ 1 | |
29 | #endif /* defined(__cplusplus) && !defined(__STDC__) */ | |
30 | ||
a2b22788 RS |
31 | #if defined(__GNUC__) || defined (__GNUG__) |
32 | #define VOLATILE volatile | |
33 | #else | |
34 | #define VOLATILE | |
35 | #endif | |
5f8037c4 RS |
36 | |
37 | /* It is incorrect to include config.h here, because this file is being | |
38 | compiled for the target, and hence definitions concerning only the host | |
a2b22788 RS |
39 | do not apply. Instead we include tconfig.h, which is the xm file |
40 | *for the target*, which also includes the tm.h file *for the target*. | |
41 | This way, we get definitions of POSIX, USG and VMS for the target. */ | |
5f8037c4 | 42 | |
a2b22788 RS |
43 | #include "tconfig.h" |
44 | ||
45 | /* The documentation says to use TARGET_POSIX, so check for it as well | |
46 | as POSIX. */ | |
47 | #if defined(TARGET_POSIX) && !defined(POSIX) | |
48 | #define POSIX | |
49 | #endif | |
50 | ||
51 | /* If the user asked for POSIX via _POSIX_SOURCE, turn on POSIX code. */ | |
52 | #if defined(_POSIX_SOURCE) && !defined(POSIX) | |
53 | #define POSIX | |
54 | #endif | |
5f8037c4 RS |
55 | |
56 | /* On a POSIX system, request definition of all the POSIX facilities. */ | |
a2b22788 | 57 | #if defined(POSIX) && !defined(_POSIX_SOURCE) |
5f8037c4 RS |
58 | #define _POSIX_SOURCE |
59 | #endif | |
60 | ||
61 | #include <stdio.h> | |
62 | #include <ctype.h> | |
63 | #include <errno.h> | |
64 | #include <sys/types.h> | |
65 | #include <sys/stat.h> | |
66 | #include <sys/dir.h> | |
67 | #include <sys/wait.h> | |
68 | #include <setjmp.h> | |
69 | #include "gvarargs.h" | |
70 | #include "getopt.h" | |
71 | ||
5f8037c4 RS |
72 | #ifndef PATH_MAX /* <limits.h> defines this on most POSIX systems. */ |
73 | #include <sys/param.h> | |
74 | #define PATH_MAX MAXPATHLEN | |
75 | /* Sometimes <sys/param.h> defines these macros. */ | |
76 | #undef CHAR_BIT | |
77 | #undef CHAR_MAX | |
78 | #undef CHAR_MIN | |
79 | #undef CLK_TCK | |
80 | #undef INT_MAX | |
81 | #undef INT_MIN | |
82 | #undef LONG_MAX | |
83 | #undef LONG_MIN | |
84 | #undef SCHAR_MAX | |
85 | #undef SCHAR_MIN | |
86 | #undef SHRT_MAX | |
87 | #undef SHRT_MIN | |
88 | #undef UCHAR_MAX | |
89 | #undef UINT_MAX | |
90 | #undef ULONG_MAX | |
91 | #undef USHRT_MAX | |
92 | #endif | |
93 | ||
94 | extern int errno; | |
a2b22788 RS |
95 | extern char *sys_errlist[]; |
96 | extern char *version_string; | |
5f8037c4 | 97 | |
a2b22788 | 98 | #if defined(POSIX) |
5f8037c4 RS |
99 | |
100 | #include <stdlib.h> | |
101 | #include <unistd.h> | |
102 | #include <signal.h> | |
103 | #include <fcntl.h> | |
104 | #include <string.h> | |
105 | ||
106 | /* Systems which are compatible only with POSIX 1003.1-1988 (but *not* | |
107 | with POSIX 1003.1-1990), e.g. Ultrix 4.2, might not have | |
108 | const qualifiers in the prototypes in the system include files. | |
109 | Unfortunately, this can lead to GCC issuing lots of warnings for | |
110 | calls to the following functions. To eliminate these warnings we | |
111 | provide the following #defines. */ | |
112 | ||
113 | #if (_POSIX_VERSION < 199009) | |
114 | ||
115 | #define access(file,flag) access((char *)file, flag) | |
116 | #define stat(file,pkt) stat((char *)file, pkt) | |
117 | #define execvp(prog,argv) execvp((char *)prog, (char **)argv) | |
118 | #define link(file1, file2) link((char *)file1, (char *)file2) | |
119 | #define unlink(file) unlink((char *)file) | |
120 | #define open(file, mode, flag) open((char *)file, mode, flag) | |
121 | #define chmod(file, mode) chmod((char *)file, mode) | |
122 | ||
123 | #endif /* (_POSIX_VERSION < 199009) */ | |
124 | ||
a2b22788 | 125 | #else /* !defined(POSIX) */ |
5f8037c4 RS |
126 | |
127 | #define R_OK 4 /* Test for Read permission */ | |
128 | #define W_OK 2 /* Test for Write permission */ | |
129 | #define X_OK 1 /* Test for eXecute permission */ | |
130 | #define F_OK 0 /* Test for existence of File */ | |
131 | ||
132 | #define O_RDONLY 0 | |
133 | #define O_WRONLY 1 | |
134 | ||
135 | /* Virtually every UN*X system now in common use (except for pre-4.3-tahoe | |
136 | BSD systems) now provides getcwd as called for by POSIX. Allow for | |
137 | the few exceptions to the general rule here. */ | |
138 | ||
139 | #if !(defined (USG) || defined (VMS)) | |
140 | extern char *getwd (char *); | |
141 | #define getcwd(buf,len) getwd(buf) | |
142 | #else /* (defined (USG) || defined (VMS)) */ | |
143 | extern char *getcwd (char *, size_t); | |
144 | #endif /* (defined (USG) || defined (VMS)) */ | |
145 | ||
a2b22788 RS |
146 | /* Declaring stat or __flsbuf with a prototype |
147 | causes conflicts with system headers on some systems. */ | |
5f8037c4 | 148 | |
a2b22788 RS |
149 | #ifndef abort |
150 | extern VOLATILE void abort (void); | |
151 | #endif | |
152 | extern int kill(int, int); | |
5f8037c4 RS |
153 | extern int creat (const char *, int); |
154 | extern int open (const char *, int, ...); | |
a2b22788 | 155 | #if 0 /* These conflict with stdio.h on some systems. */ |
5f8037c4 RS |
156 | extern int fprintf (FILE *, const char *, ...); |
157 | extern int printf (const char *, ...); | |
a2b22788 | 158 | #endif /* 0 */ |
5f8037c4 RS |
159 | extern void exit (int); |
160 | extern void *malloc (size_t); | |
161 | extern void *realloc (void *, size_t); | |
162 | extern void free (void *); | |
163 | extern int read (int, void *, size_t); | |
164 | extern int write (int, const void *, size_t); | |
165 | extern int close (int); | |
166 | extern int link (const char *, const char *); | |
167 | extern int unlink (const char *); | |
168 | extern int fflush (FILE *); | |
169 | extern int atoi (const char *); | |
170 | extern int access (const char *, int); | |
171 | extern int puts (const char *); | |
172 | extern int fputs (const char *, FILE *); | |
173 | extern int fputc (int, FILE *); | |
174 | extern int execvp (const char *, char *const *); | |
175 | extern int setjmp (jmp_buf); | |
176 | extern void longjmp (jmp_buf, int); | |
177 | ||
178 | extern char * strcat(char *, const char *); | |
179 | extern int strcmp(const char *, const char *); | |
180 | extern char * strcpy(char *, const char *); | |
a2b22788 RS |
181 | #if 0 /* size_t from sys/types.h may fail to match GCC. |
182 | If so, we would get a warning from this. | |
183 | So do without the prototype. */ | |
5f8037c4 | 184 | extern size_t strlen(const char *); |
a2b22788 | 185 | #endif |
5f8037c4 RS |
186 | extern int strncmp(const char *, const char *, size_t); |
187 | extern char * strncpy(char *, const char *, size_t); | |
188 | extern char * strrchr(const char *, int); | |
189 | ||
190 | #if !(defined (USG) || defined (VMS)) | |
191 | extern int vfork (void); | |
192 | #define fork vfork | |
a2b22788 RS |
193 | #else |
194 | extern int fork (void); | |
5f8037c4 RS |
195 | #endif /* !(defined (USG) || defined (VMS)) */ |
196 | ||
a2b22788 | 197 | #endif /* !defined(POSIX) */ |
5f8037c4 RS |
198 | |
199 | /* Look for these where the `const' qualifier is intentionally cast aside. */ | |
200 | ||
201 | #define NONCONST | |
202 | ||
203 | /* Aliases for pointers to void. | |
204 | These were made to facilitate compilation with other compilers. */ | |
205 | ||
206 | typedef void * pointer_type; | |
207 | typedef const void * const_pointer_type; | |
208 | ||
209 | /* Define a STRINGIFY macro that's right for ANSI C. */ | |
210 | ||
211 | #define STRINGIFY(STRING) #STRING | |
212 | ||
213 | /* POSIX systems will not have definitions for WIFEXITED or WEXITSTATUS. | |
214 | Define them correctly and so that they work for all environments. */ | |
215 | ||
216 | #undef WIFEXITED | |
217 | #define WIFEXITED(status_word) ((*((int *)&status_word) & 0xff) == 0x00) | |
218 | ||
219 | #undef WEXITSTATUS | |
220 | #define WEXITSTATUS(status_word) ((*((int *)&status_word) & 0xff00) >> 8) | |
221 | ||
222 | /* Define a default place to find the SYSCALLS.X file. */ | |
223 | ||
224 | #ifndef STD_PROTO_DIR | |
225 | #define STD_PROTO_DIR "/usr/local/lib" | |
226 | #endif /* !defined(STD_PROTO_DIR) */ | |
227 | ||
5f8037c4 RS |
228 | /* Suffix of aux_info files. */ |
229 | ||
230 | static const char * const aux_info_suffix = ".X"; | |
231 | ||
a2b22788 | 232 | /* String to attach to filenames for saved versions of original files. */ |
5f8037c4 RS |
233 | |
234 | static const char * const save_suffix = ".save"; | |
235 | ||
236 | #ifndef UNPROTOIZE | |
237 | ||
238 | /* File name of the file which contains descriptions of standard system | |
239 | routines. Note that we never actually do anything with this file per se, | |
240 | but we do read in its corresponding aux_info file. */ | |
241 | ||
242 | static const char * const syscalls_filename = "SYSCALLS.c"; | |
243 | ||
244 | /* Default place to find the above file. */ | |
245 | ||
246 | static const char * const default_syscalls_dir = STD_PROTO_DIR; | |
247 | ||
a2b22788 | 248 | /* Variable to hold the complete absolutized filename of the SYSCALLS.c.X |
5f8037c4 RS |
249 | file. */ |
250 | ||
a2b22788 | 251 | static char * syscalls_absolute_filename; |
5f8037c4 RS |
252 | |
253 | #endif /* !defined(UNPROTOIZE) */ | |
254 | ||
255 | /* Type of the structure that holds information about macro unexpansions. */ | |
256 | ||
257 | struct unexpansion_struct { | |
258 | const char *expanded; | |
259 | const char *contracted; | |
260 | }; | |
261 | typedef struct unexpansion_struct unexpansion; | |
262 | ||
263 | /* A table of conversions that may need to be made for some (stupid) older | |
264 | operating systems where these types are preprocessor macros rather than | |
265 | typedefs (as they really ought to be). | |
266 | ||
267 | WARNING: The contracted forms must be as small (or smaller) as the | |
268 | expanded forms, or else havoc will ensue. */ | |
269 | ||
270 | static const unexpansion unexpansions[] = { | |
271 | { "struct _iobuf", "FILE" }, | |
272 | { 0, 0 } | |
273 | }; | |
274 | ||
275 | /* The number of "primary" slots in the hash tables for filenames and for | |
276 | function names. This can be as big or as small as you like, except that | |
277 | it must be a power of two. */ | |
278 | ||
279 | #define HASH_TABLE_SIZE (1 << 9) | |
280 | ||
281 | /* Bit mask to use when computing hash values. */ | |
282 | ||
283 | static const int hash_mask = (HASH_TABLE_SIZE - 1); | |
284 | ||
285 | /* Make a table of default system include directories | |
286 | just as it is done in cccp.c. */ | |
287 | ||
288 | #ifndef STANDARD_INCLUDE_DIR | |
289 | #define STANDARD_INCLUDE_DIR "/usr/include" | |
290 | #endif | |
291 | ||
292 | #ifndef LOCAL_INCLUDE_DIR | |
293 | #define LOCAL_INCLUDE_DIR "/usr/local/include" | |
294 | #endif | |
295 | ||
296 | struct default_include { char *fname; int cplusplus; } include_defaults[] | |
297 | #ifdef INCLUDE_DEFAULTS | |
298 | = INCLUDE_DEFAULTS; | |
299 | #else | |
300 | = { | |
301 | /* Pick up GNU C++ specific include files. */ | |
302 | { GPLUSPLUS_INCLUDE_DIR, 1}, | |
303 | { GCC_INCLUDE_DIR, 0}, | |
304 | #ifdef CROSS_COMPILE | |
305 | /* For cross-compilation, this dir name is generated | |
306 | automatically in Makefile.in. */ | |
307 | { CROSS_INCLUDE_DIR, 0 }, | |
308 | #else /* not CROSS_COMPILE */ | |
309 | { LOCAL_INCLUDE_DIR, 0}, | |
310 | /* Some systems have an extra dir of include files. */ | |
311 | #ifdef SYSTEM_INCLUDE_DIR | |
312 | { SYSTEM_INCLUDE_DIR, 0}, | |
313 | #endif | |
314 | #ifndef NO_STANDARD_INCLUDE_DIR | |
315 | { STANDARD_INCLUDE_DIR, 0}, | |
316 | #endif | |
317 | #endif /* not CROSS_COMPILE */ | |
318 | { 0, 0} | |
319 | }; | |
320 | #endif /* no INCLUDE_DEFAULTS */ | |
321 | ||
322 | /* Datatype for lists of directories or filenames. */ | |
323 | struct string_list | |
324 | { | |
325 | char *name; | |
326 | struct string_list *next; | |
327 | }; | |
328 | ||
329 | /* List of directories in which files should be converted. */ | |
330 | ||
331 | struct string_list *directory_list; | |
332 | ||
333 | /* List of file names which should not be converted. | |
334 | A file is excluded if the end of its name, following a /, | |
335 | matches one of the names in this list. */ | |
336 | ||
337 | struct string_list *exclude_list; | |
338 | ||
339 | /* The name of the other style of variable-number-of-parameters functions | |
340 | (i.e. the style that we want to leave unconverted because we don't yet | |
341 | know how to convert them to this style. This string is used in warning | |
342 | messages. */ | |
343 | ||
344 | /* Also define here the string that we can search for in the parameter lists | |
345 | taken from the .X files which will unambiguously indicate that we have | |
346 | found a varargs style function. */ | |
347 | ||
348 | #ifdef UNPROTOIZE | |
349 | static const char * const other_var_style = "stdarg"; | |
350 | #else /* !defined(UNPROTOIZE) */ | |
351 | static const char * const other_var_style = "varargs"; | |
a2b22788 RS |
352 | /* Note that this is a string containing the expansion of va_alist. |
353 | But in `main' we discard all but the first token. */ | |
354 | static const char *varargs_style_indicator = STRINGIFY(va_alist); | |
5f8037c4 RS |
355 | #endif /* !defined(UNPROTOIZE) */ |
356 | ||
357 | /* The following two types are used to create hash tables. In this program, | |
358 | there are two hash tables which are used to store and quickly lookup two | |
359 | different classes of strings. The first type of strings stored in the | |
a2b22788 | 360 | first hash table are absolute filenames of files which protoize needs to |
5f8037c4 RS |
361 | know about. The second type of strings (stored in the second hash table) |
362 | are function names. It is this second class of strings which really | |
363 | inspired the use of the hash tables, because there may be a lot of them. */ | |
364 | ||
365 | typedef struct hash_table_entry_struct hash_table_entry; | |
366 | ||
367 | /* Do some typedefs so that we don't have to write "struct" so often. */ | |
368 | ||
369 | typedef struct def_dec_info_struct def_dec_info; | |
370 | typedef struct file_info_struct file_info; | |
371 | typedef struct f_list_chain_item_struct f_list_chain_item; | |
372 | ||
373 | /* In the struct below, note that the "_info" field has two different uses | |
a2b22788 RS |
374 | depending on the type of hash table we are in (i.e. either the filenames |
375 | hash table or the function names hash table). In the filenames hash table | |
5f8037c4 | 376 | the info fields of the entries point to the file_info struct which is |
a2b22788 | 377 | associated with each filename (1 per filename). In the function names |
5f8037c4 RS |
378 | hash table, the info field points to the head of a singly linked list of |
379 | def_dec_info entries which are all defs or decs of the function whose | |
380 | name is pointed to by the "symbol" field. Keeping all of the defs/decs | |
381 | for a given function name on a special list specifically for that function | |
382 | name makes it quick and easy to find out all of the important information | |
383 | about a given (named) function. */ | |
384 | ||
385 | struct hash_table_entry_struct { | |
386 | hash_table_entry * hash_next; /* -> to secondary entries */ | |
387 | const char * symbol; /* -> to the hashed string */ | |
388 | union { | |
389 | const def_dec_info * _ddip; | |
390 | file_info * _fip; | |
391 | } _info; | |
392 | }; | |
393 | #define ddip _info._ddip | |
394 | #define fip _info._fip | |
395 | ||
396 | /* Define a type specifically for our two hash tables. */ | |
397 | ||
398 | typedef hash_table_entry hash_table[HASH_TABLE_SIZE]; | |
399 | ||
400 | /* The following struct holds all of the important information about any | |
a2b22788 | 401 | single filename (e.g. file) which we need to know about. */ |
5f8037c4 RS |
402 | |
403 | struct file_info_struct { | |
404 | const hash_table_entry * hash_entry; /* -> to associated hash entry */ | |
405 | const def_dec_info * defs_decs; /* -> to chain of defs/decs */ | |
406 | time_t mtime; /* Time of last modification. */ | |
407 | }; | |
408 | ||
409 | /* Due to the possibility that functions may return pointers to functions, | |
410 | (which may themselves have their own parameter lists) and due to the | |
411 | fact that returned pointers-to-functions may be of type "pointer-to- | |
412 | function-returning-pointer-to-function" (ad nauseum) we have to keep | |
413 | an entire chain of ANSI style formal parameter lists for each function. | |
414 | ||
415 | Normally, for any given function, there will only be one formals list | |
416 | on the chain, but you never know. | |
417 | ||
418 | Note that the head of each chain of formals lists is pointed to by the | |
419 | `f_list_chain' field of the corresponding def_dec_info record. | |
420 | ||
421 | For any given chain, the item at the head of the chain is the *leftmost* | |
422 | parameter list seen in the actual C language function declaration. If | |
423 | there are other members of the chain, then these are linked in left-to-right | |
424 | order from the head of the chain. */ | |
425 | ||
426 | struct f_list_chain_item_struct { | |
427 | const f_list_chain_item * chain_next; /* -> to next item on chain */ | |
428 | const char * formals_list; /* -> to formals list string */ | |
429 | }; | |
430 | ||
431 | /* The following struct holds all of the important information about any | |
432 | single function definition or declaration which we need to know about. | |
433 | Note that for unprotoize we don't need to know very much because we | |
434 | never even create records for stuff that we don't intend to convert | |
435 | (like for instance defs and decs which are already in old K&R format | |
436 | and "implicit" function declarations). */ | |
437 | ||
438 | struct def_dec_info_struct { | |
439 | const def_dec_info * next_in_file; /* -> to rest of chain for file */ | |
440 | file_info * file; /* -> file_info for containing file */ | |
441 | int line; /* source line number of def/dec */ | |
442 | const char * ansi_decl; /* -> left end of ansi decl */ | |
443 | hash_table_entry * hash_entry; /* -> hash entry for function name */ | |
444 | unsigned int is_func_def; /* = 0 means this is a declaration */ | |
445 | const def_dec_info * next_for_func; /* -> to rest of chain for func name */ | |
446 | unsigned int f_list_count; /* count of formals lists we expect */ | |
447 | char prototyped; /* = 0 means already prototyped */ | |
448 | #ifndef UNPROTOIZE | |
449 | const f_list_chain_item * f_list_chain; /* -> chain of formals lists */ | |
450 | const def_dec_info * definition; /* -> def/dec containing related def */ | |
451 | char is_static; /* = 0 means visiblilty is "extern" */ | |
452 | char is_implicit; /* != 0 for implicit func decl's */ | |
453 | char written; /* != 0 means written for implicit */ | |
454 | #else /* !defined(UNPROTOIZE) */ | |
455 | const char * formal_names; /* -> to list of names of formals */ | |
456 | const char * formal_decls; /* -> to string of formal declartions */ | |
457 | #endif /* !defined(UNPROTOIZE) */ | |
458 | }; | |
459 | ||
a2b22788 | 460 | /* Pointer to the tail component of the filename by which this program was |
5f8037c4 RS |
461 | invoked. Used everywhere in error and warning messages. */ |
462 | ||
463 | static const char *pname; | |
464 | ||
465 | /* Error counter. Will be non-zero if we should give up at the next convenient | |
466 | stopping point. */ | |
467 | ||
468 | static int errors = 0; | |
469 | ||
470 | /* Option flags. */ | |
471 | /* ??? These comments should say what the flag mean as well as the options | |
472 | that set them. */ | |
473 | ||
474 | static int version_flag = 0; /* set by -V option */ | |
475 | static int quiet_flag = 0; /* set by -q option */ | |
476 | #if 0 | |
477 | static int force_flag = 0; /* set by -f option */ | |
478 | #endif | |
479 | static int nochange_flag = 0; /* set by -n option */ | |
480 | static int nosave_flag = 0; /* set by -N option */ | |
481 | static int keep_flag = 0; /* set by -k option */ | |
482 | static const char ** compile_params = 0; /* set by -c option */ | |
483 | #ifdef UNPROTOIZE | |
484 | static const char *indent_string = " "; /* set by -i option */ | |
485 | #else /* !defined(UNPROTOIZE) */ | |
486 | static int local_flag = 0; /* set by -l option */ | |
487 | static int global_flag = 0; /* set by -g option */ | |
488 | static int cplusplus_flag = 0; /* set by -C option */ | |
489 | static const char* nondefault_syscalls_dir = 0; /* set by -B option */ | |
490 | #endif /* !defined(UNPROTOIZE) */ | |
491 | ||
492 | /* An index into the compile_params array where we should insert the filename | |
493 | parameter when we are ready to exec the C compiler. A zero value indicates | |
494 | that we have not yet called munge_compile_params(). */ | |
495 | ||
496 | static int filename_index = 0; | |
497 | ||
498 | /* Count of command line arguments which were "filename" arguments. */ | |
499 | ||
a2b22788 | 500 | static int n_base_source_files = 0; |
5f8037c4 RS |
501 | |
502 | /* Points to a malloc'ed list of pointers to all of the filenames of base | |
503 | source files which were specified on the command line. */ | |
504 | ||
a2b22788 | 505 | static const char **base_source_filenames; |
5f8037c4 RS |
506 | |
507 | /* Line number of the line within the current aux_info file that we | |
508 | are currently processing. Used for error messages in case the prototypes | |
509 | info file is corrupted somehow. */ | |
510 | ||
511 | static int current_aux_info_lineno; | |
512 | ||
513 | /* Pointer to the name of the source file currently being converted. */ | |
514 | ||
a2b22788 | 515 | static const char *convert_filename; |
5f8037c4 RS |
516 | |
517 | /* Pointer to relative root string (taken from aux_info file) which indicates | |
518 | where directory the user was in when he did the compilation step that | |
519 | produced the containing aux_info file. */ | |
520 | ||
a2b22788 | 521 | static const char *invocation_filename; |
5f8037c4 RS |
522 | |
523 | /* Pointer to the base of the input buffer that holds the original text for the | |
524 | source file currently being converted. */ | |
525 | ||
526 | static const char *orig_text_base; | |
527 | ||
528 | /* Pointer to the byte just beyond the end of the input buffer that holds the | |
529 | original text for the source file currently being converted. */ | |
530 | ||
531 | static const char *orig_text_limit; | |
532 | ||
533 | /* Pointer to the base of the input buffer that holds the cleaned text for the | |
534 | source file currently being converted. */ | |
535 | ||
536 | static const char *clean_text_base; | |
537 | ||
538 | /* Pointer to the byte just beyond the end of the input buffer that holds the | |
539 | cleaned text for the source file currently being converted. */ | |
540 | ||
541 | static const char *clean_text_limit; | |
542 | ||
543 | /* Pointer to the last byte in the cleaned text buffer that we have already | |
544 | (virtually) copied to the output buffer (or decided to ignore). */ | |
545 | ||
546 | static const char * clean_read_ptr; | |
547 | ||
548 | /* Pointer to the base of the output buffer that holds the replacement text | |
549 | for the source file currently being converted. */ | |
550 | ||
551 | static char *repl_text_base; | |
552 | ||
553 | /* Pointer to the byte just beyond the end of the output buffer that holds the | |
554 | replacement text for the source file currently being converted. */ | |
555 | ||
556 | static char *repl_text_limit; | |
557 | ||
558 | /* Pointer to the last byte which has been stored into the output buffer. | |
559 | The next byte to be stored should be stored just past where this points | |
560 | to. */ | |
561 | ||
562 | static char * repl_write_ptr; | |
563 | ||
564 | /* Pointer into the cleaned text buffer for the source file we are currently | |
565 | converting. This points to the first character of the line that we last | |
566 | did a "seek_to_line()" to (see below). */ | |
567 | ||
568 | static const char *last_known_line_start; | |
569 | ||
570 | /* Number of the line (in the cleaned text buffer) that we last did a | |
571 | "seek_to_line()" to. Will be one if we just read a new source file | |
572 | into the cleaned text buffer. */ | |
573 | ||
574 | static int last_known_line_number; | |
575 | ||
a2b22788 | 576 | /* The filenames hash table. */ |
5f8037c4 | 577 | |
a2b22788 | 578 | static hash_table filename_primary; |
5f8037c4 RS |
579 | |
580 | /* The function names hash table. */ | |
581 | ||
582 | static hash_table function_name_primary; | |
583 | ||
584 | /* The place to keep the recovery address which is used only in cases where | |
585 | we get hopelessly confused by something in the cleaned original text. */ | |
586 | ||
587 | static jmp_buf source_confusion_recovery; | |
588 | ||
a2b22788 | 589 | /* A pointer to the current directory filename (used by abspath). */ |
5f8037c4 RS |
590 | |
591 | static char *cwd_buffer; | |
592 | ||
593 | /* A place to save the read pointer until we are sure that an individual | |
594 | attempt at editing will succeed. */ | |
595 | ||
596 | static const char * saved_clean_read_ptr; | |
597 | ||
598 | /* A place to save the write pointer until we are sure that an individual | |
599 | attempt at editing will succeed. */ | |
600 | ||
601 | static char * saved_repl_write_ptr; | |
602 | ||
603 | /* Forward declaration. */ | |
604 | ||
a2b22788 | 605 | static const char *shortpath (const char *cwd, const char *filename); |
5f8037c4 RS |
606 | \f |
607 | /* Allocate some space, but check that the allocation was successful. */ | |
608 | ||
609 | static pointer_type | |
610 | xmalloc (size_t byte_count) | |
611 | { | |
612 | pointer_type rv; | |
613 | ||
614 | if ((rv = malloc (byte_count)) == NULL) | |
615 | { | |
616 | fprintf (stderr, "\n%s: fatal error: can't allocate %u more bytes of memory\n", | |
a2b22788 | 617 | pname, byte_count); |
5f8037c4 RS |
618 | exit (1); |
619 | return 0; /* avoid warnings */ | |
620 | } | |
621 | else | |
622 | return rv; | |
623 | } | |
624 | ||
625 | /* Reallocate some space, but check that the reallocation was successful. */ | |
626 | ||
627 | static pointer_type | |
628 | xrealloc (pointer_type old_space, size_t byte_count) | |
629 | { | |
630 | pointer_type rv; | |
631 | ||
632 | if ((rv = realloc (old_space, byte_count)) == NULL) | |
633 | { | |
634 | fprintf (stderr, "\n%s: fatal error: can't allocate %u more bytes of memory\n", | |
a2b22788 | 635 | pname, byte_count); |
5f8037c4 RS |
636 | exit (1); |
637 | return 0; /* avoid warnings */ | |
638 | } | |
639 | else | |
640 | return rv; | |
641 | } | |
642 | ||
643 | /* Deallocate the area pointed to by an arbitrary pointer, but first, strip | |
644 | the `const' qualifier from it and also make sure that the pointer value | |
645 | is non-null. */ | |
646 | ||
647 | static void | |
648 | xfree (const_pointer_type p) | |
649 | { | |
650 | if (p) | |
651 | free ((NONCONST pointer_type) p); | |
652 | } | |
653 | ||
654 | /* Make a copy of a string INPUT with size SIZE. */ | |
655 | ||
656 | static char * | |
a2b22788 | 657 | savestring (const char *input, int size) |
5f8037c4 RS |
658 | { |
659 | char *output = (char *) xmalloc (size + 1); | |
660 | strcpy (output, input); | |
661 | return output; | |
662 | } | |
663 | ||
664 | /* More 'friendly' abort that prints the line and file. | |
665 | config.h can #define abort fancy_abort if you like that sort of thing. */ | |
666 | ||
667 | void | |
668 | fancy_abort () | |
669 | { | |
670 | fprintf (stderr, "%s: internal abort\n", pname); | |
671 | exit (1); | |
672 | } | |
673 | \f | |
674 | /* Make a duplicate of a given string in a newly allocated area. */ | |
675 | ||
676 | static char * | |
677 | dupstr (const char *s) | |
678 | { | |
679 | return strcpy ((char *) xmalloc (strlen (s) + 1), s); | |
680 | } | |
681 | ||
682 | /* Make a duplicate of the first N bytes of a given string in a newly | |
683 | allocated area. */ | |
684 | ||
685 | static char * | |
686 | dupnstr (const char *s, size_t n) | |
687 | { | |
688 | char *ret_val = strncpy ((char *) xmalloc (n + 1), s, n); | |
689 | ||
690 | ret_val[n] = '\0'; | |
691 | return ret_val; | |
692 | } | |
693 | ||
694 | /* Return a pointer to the first occurance of s2 within s1 or NULL if s2 | |
695 | does not occur within s1. Assume neither s1 nor s2 are null pointers. */ | |
696 | ||
697 | static const char * | |
698 | substr (const char *s1, const char *const s2) | |
699 | { | |
700 | for (; *s1 ; s1++) | |
701 | { | |
702 | const char *p1; | |
703 | const char *p2; | |
a2b22788 | 704 | int c; |
5f8037c4 RS |
705 | |
706 | for (p1 = s1, p2 = s2; c = *p2; p1++, p2++) | |
707 | if (*p1 != c) | |
708 | goto outer; | |
709 | return s1; | |
710 | outer: | |
711 | ; | |
712 | } | |
713 | return 0; | |
714 | } | |
715 | \f | |
716 | /* Get setup to recover in case the edit we are about to do goes awry. */ | |
717 | ||
718 | void | |
719 | save_pointers (void) | |
720 | { | |
721 | saved_clean_read_ptr = clean_read_ptr; | |
722 | saved_repl_write_ptr = repl_write_ptr; | |
723 | } | |
724 | ||
725 | /* Call this routine to recover our previous state whenever something looks | |
726 | too confusing in the source code we are trying to edit. */ | |
727 | ||
728 | void | |
729 | restore_pointers (void) | |
730 | { | |
731 | clean_read_ptr = saved_clean_read_ptr; | |
732 | repl_write_ptr = saved_repl_write_ptr; | |
733 | } | |
734 | ||
735 | /* Return true if the given character is a legal identifier character. */ | |
736 | ||
737 | inline static int | |
738 | is_id_char (char ch) | |
739 | { | |
740 | return (isalnum (ch) || (ch == '_') || (ch == '$')); | |
741 | } | |
742 | ||
743 | /* Give a message indicating the proper way to invoke this program and then | |
744 | exit with non-zero status. */ | |
745 | ||
746 | static void | |
747 | usage (void) | |
748 | { | |
749 | #ifdef UNPROTOIZE | |
a2b22788 RS |
750 | fprintf (stderr, "%s: usage '%s [ -VqfnkN ] [ -i <istring> ] [ filename ... ]'\n", |
751 | pname, pname); | |
5f8037c4 | 752 | #else /* !defined(UNPROTOIZE) */ |
a2b22788 RS |
753 | fprintf (stderr, "%s: usage '%s [ -VqfnkNlgC ] [ -B <diname> ] [ filename ... ]'\n", |
754 | pname, pname); | |
5f8037c4 | 755 | #endif /* !defined(UNPROTOIZE) */ |
5f8037c4 RS |
756 | exit (1); |
757 | } | |
758 | ||
a2b22788 | 759 | /* Return true if the given filename (assumed to be an absolute filename) |
5f8037c4 RS |
760 | designates a file residing anywhere beneath any one of the "system" |
761 | include directories. */ | |
762 | ||
763 | static int | |
764 | in_system_include_dir (const char *path) | |
765 | { | |
766 | struct default_include *p; | |
767 | ||
768 | if (path[0] != '/') | |
a2b22788 | 769 | abort (); /* Must be an absolutized filename. */ |
5f8037c4 RS |
770 | |
771 | for (p = include_defaults; p->fname; p++) | |
772 | if (!strncmp (path, p->fname, strlen (p->fname)) | |
773 | && path[strlen (p->fname)] == '/') | |
774 | return 1; | |
775 | return 0; | |
776 | } | |
777 | \f | |
778 | #if 0 | |
a2b22788 | 779 | /* Return true if the given filename designates a file that the user has |
5f8037c4 RS |
780 | read access to and for which the user has write access to the containing |
781 | directory. */ | |
782 | ||
783 | static int | |
784 | file_could_be_converted (const char *path) | |
785 | { | |
786 | char *const dir_name = (char *) alloca (strlen (path) + 1); | |
787 | ||
788 | if (access (path, R_OK)) | |
789 | return 0; | |
790 | ||
791 | { | |
792 | char *dir_last_slash; | |
793 | ||
794 | strcpy (dir_name, path); | |
795 | dir_last_slash = strrchr (dir_name, '/'); | |
796 | if (dir_last_slash) | |
797 | *dir_last_slash = '\0'; | |
798 | else | |
a2b22788 | 799 | abort (); /* Should have been an absolutized filename. */ |
5f8037c4 RS |
800 | } |
801 | ||
802 | if (access (path, W_OK)) | |
803 | return 0; | |
804 | ||
805 | return 1; | |
806 | } | |
807 | ||
a2b22788 | 808 | /* Return true if the given filename designates a file that we are allowed |
5f8037c4 RS |
809 | to modify. Files which we should not attempt to modify are (a) "system" |
810 | include files, and (b) files which the user doesn't have write access to, | |
811 | and (c) files which reside in directories which the user doesn't have | |
812 | write access to. Unless requested to be quiet, give warnings about | |
813 | files that we will not try to convert for one reason or another. An | |
814 | exception is made for "system" include files, which we never try to | |
815 | convert and for which we don't issue the usual warnings. */ | |
816 | ||
817 | static int | |
818 | file_normally_convertable (const char *path) | |
819 | { | |
820 | char *const dir_name = alloca (strlen (path) + 1); | |
821 | ||
822 | if (in_system_include_dir (path)) | |
823 | return 0; | |
824 | ||
825 | { | |
826 | char *dir_last_slash; | |
827 | ||
828 | strcpy (dir_name, path); | |
829 | dir_last_slash = strrchr (dir_name, '/'); | |
830 | if (dir_last_slash) | |
831 | *dir_last_slash = '\0'; | |
832 | else | |
a2b22788 | 833 | abort (); /* Should have been an absolutized filename. */ |
5f8037c4 RS |
834 | } |
835 | ||
836 | if (access (path, R_OK)) | |
837 | { | |
838 | if (!quiet_flag) | |
839 | fprintf (stderr, "%s: warning: no read access for file `%s'\n", | |
a2b22788 | 840 | pname, shortpath (NULL, path)); |
5f8037c4 RS |
841 | return 0; |
842 | } | |
843 | ||
844 | if (access (path, W_OK)) | |
845 | { | |
846 | if (!quiet_flag) | |
847 | fprintf (stderr, "%s: warning: no write access for file `%s'\n", | |
a2b22788 | 848 | pname, shortpath (NULL, path)); |
5f8037c4 RS |
849 | return 0; |
850 | } | |
851 | ||
852 | if (access (dir_name, W_OK)) | |
853 | { | |
854 | if (!quiet_flag) | |
855 | fprintf (stderr, "%s: warning: no write access for dir containing `%s'\n", | |
a2b22788 | 856 | pname, shortpath (NULL, path)); |
5f8037c4 RS |
857 | return 0; |
858 | } | |
859 | ||
860 | return 1; | |
861 | } | |
862 | #endif /* 0 */ | |
863 | \f | |
864 | #ifndef UNPROTOIZE | |
865 | ||
866 | /* Return true if the given file_info struct refers to the special SYSCALLS.c.X | |
867 | file. Return false otherwise. */ | |
868 | ||
869 | static int | |
870 | is_syscalls_file (const file_info *fi_p) | |
871 | { | |
872 | return (substr (fi_p->hash_entry->symbol, syscalls_filename) != NULL); | |
873 | } | |
874 | ||
875 | #endif /* !defined(UNPROTOIZE) */ | |
876 | ||
877 | /* Check to see if this file will need to have anything done to it on this | |
878 | run. If there is nothing in the given file which both needs conversion | |
879 | and for which we have the necessary stuff to do the conversion, return | |
880 | false. Otherwise, return true. | |
881 | ||
882 | Note that (for protoize) it is only valid to call this function *after* | |
883 | the connections between declarations and definitions have all been made | |
884 | by connect_defs_and_decs(). */ | |
885 | ||
886 | static int | |
887 | needs_to_be_converted (const file_info *file_p) | |
888 | { | |
889 | const def_dec_info *ddp; | |
890 | ||
891 | #ifndef UNPROTOIZE | |
892 | ||
893 | if (is_syscalls_file (file_p)) | |
894 | return 0; | |
895 | ||
896 | #endif /* !defined(UNPROTOIZE) */ | |
897 | ||
898 | for (ddp = file_p->defs_decs; ddp; ddp = ddp->next_in_file) | |
899 | ||
900 | if ( | |
901 | ||
902 | #ifndef UNPROTOIZE | |
903 | ||
904 | /* ... and if we a protoizing and this function is in old style ... */ | |
905 | !ddp->prototyped | |
906 | /* ... and if this a definition or is a decl with an associated def ... */ | |
907 | && (ddp->is_func_def || (!ddp->is_func_def && ddp->definition)) | |
908 | ||
909 | #else /* defined(UNPROTOIZE) */ | |
910 | ||
911 | /* ... and if we are unprotoizing and this function is in new style ... */ | |
912 | ddp->prototyped | |
913 | ||
914 | #endif /* defined(UNPROTOIZE) */ | |
915 | ) | |
916 | /* ... then the containing file needs converting. */ | |
917 | return -1; | |
918 | return 0; | |
919 | } | |
920 | ||
921 | /* Return 1 if the file name NAME is in a directory | |
922 | that should be converted. */ | |
923 | ||
924 | static int | |
925 | directory_specified_p (const char *name) | |
926 | { | |
927 | struct string_list *p; | |
928 | ||
929 | for (p = directory_list; p; p = p->next) | |
930 | if (!strncmp (name, p->name, strlen (p->name)) | |
931 | && name[strlen (p->name)] == '/') | |
932 | return 1; | |
933 | ||
934 | return 0; | |
935 | } | |
936 | ||
937 | /* Return 1 if the file named NAME should be excluded from conversion. */ | |
938 | ||
939 | static int | |
940 | file_excluded_p (const char *name) | |
941 | { | |
942 | struct string_list *p; | |
943 | int len = strlen (name); | |
944 | ||
945 | for (p = exclude_list; p; p = p->next) | |
946 | if (!strcmp (name + len - strlen (p->name), p->name) | |
947 | && name[len - strlen (p->name) - 1] == '/') | |
948 | return 1; | |
949 | ||
950 | return 0; | |
951 | } | |
952 | ||
953 | /* Construct a new element of a string_list. | |
954 | STRING is the new element value, and REST holds the remaining elements. */ | |
955 | ||
956 | static struct string_list * | |
957 | string_list_cons (char *string, struct string_list *rest) | |
958 | { | |
959 | struct string_list *temp = xmalloc (sizeof (struct string_list)); | |
960 | temp->next = rest; | |
961 | temp->name = string; | |
962 | return temp; | |
963 | } | |
964 | \f | |
965 | /* ??? The GNU convention for mentioning function args in its comments | |
966 | is to capitalize them. So change "hash_tab_p" to HASH_TAB_P below. | |
967 | Likewise for all the other functions. */ | |
968 | ||
969 | /* Given a hash table, apply some function to each node in the table. The | |
970 | table to traverse is given as the "hash_tab_p" argument, and the | |
971 | function to be applied to each node in the table is given as "func" | |
972 | argument. */ | |
973 | ||
974 | static void | |
975 | visit_each_hash_node (const hash_table_entry *hash_tab_p, void (*func) (const hash_table_entry *)) | |
976 | { | |
977 | const hash_table_entry *primary; | |
978 | ||
979 | for (primary = hash_tab_p; primary < &hash_tab_p[HASH_TABLE_SIZE]; primary++) | |
980 | if (primary->symbol) | |
981 | { | |
982 | hash_table_entry *second; | |
983 | ||
984 | (*func)(primary); | |
985 | for (second = primary->hash_next; second; second = second->hash_next) | |
986 | (*func) (second); | |
987 | } | |
988 | } | |
989 | ||
990 | /* Initialize all of the fields of a new hash table entry, pointed | |
991 | to by the "p" parameter. Note that the space to hold the entry | |
992 | is assumed to have already been allocated before this routine is | |
993 | called. */ | |
994 | ||
995 | static hash_table_entry * | |
996 | add_symbol (hash_table_entry *p, const char *s) | |
997 | { | |
998 | p->hash_next = NULL; | |
999 | p->symbol = dupstr (s); | |
1000 | p->ddip = NULL; | |
1001 | p->fip = NULL; | |
1002 | return p; | |
1003 | } | |
1004 | ||
a2b22788 | 1005 | /* Look for a particular function name or filename in the particular |
5f8037c4 RS |
1006 | hash table indicated by "hash_tab_p". If the name is not in the |
1007 | given hash table, add it. Either way, return a pointer to the | |
1008 | hash table entry for the given name. */ | |
1009 | ||
1010 | static hash_table_entry * | |
1011 | lookup (hash_table_entry *hash_tab_p, const char *search_symbol) | |
1012 | { | |
1013 | int hash_value = 0; | |
1014 | const char *search_symbol_char_p = search_symbol; | |
1015 | hash_table_entry *p; | |
1016 | ||
1017 | while (*search_symbol_char_p) | |
1018 | hash_value += *search_symbol_char_p++; | |
1019 | hash_value &= hash_mask; | |
1020 | p = &hash_tab_p[hash_value]; | |
1021 | if (! p->symbol) | |
1022 | return add_symbol (p, search_symbol); | |
1023 | if (!strcmp (p->symbol, search_symbol)) | |
1024 | return p; | |
1025 | while (p->hash_next) | |
1026 | { | |
1027 | p = p->hash_next; | |
1028 | if (!strcmp (p->symbol, search_symbol)) | |
1029 | return p; | |
1030 | } | |
1031 | p->hash_next = (hash_table_entry *) xmalloc (sizeof (hash_table_entry)); | |
1032 | p = p->hash_next; | |
1033 | return add_symbol (p, search_symbol); | |
1034 | } | |
1035 | \f | |
1036 | /* Throw a def/dec record on the junk heap. | |
1037 | ||
1038 | Also, since we are not using this record anymore, free up all of the | |
1039 | stuff it pointed to. */ | |
1040 | ||
1041 | inline static void | |
1042 | free_def_dec (def_dec_info *p) | |
1043 | { | |
1044 | xfree (p->ansi_decl); | |
1045 | ||
1046 | #ifndef UNPROTOIZE | |
1047 | { | |
1048 | const f_list_chain_item * curr; | |
1049 | const f_list_chain_item * next; | |
1050 | ||
1051 | for (curr = p->f_list_chain; curr; curr = next) | |
1052 | { | |
1053 | next = curr->chain_next; | |
1054 | xfree (curr); | |
1055 | } | |
1056 | } | |
1057 | #endif /* !defined(UNPROTOIZE) */ | |
1058 | ||
1059 | xfree (p); | |
1060 | } | |
1061 | ||
1062 | /* Unexpand as many macro symbol as we can find. | |
1063 | ||
1064 | If the given line must be unexpanded, make a copy of it in the heap and | |
1065 | return a pointer to the unexpanded copy. Otherwise return NULL. */ | |
1066 | ||
1067 | static char * | |
1068 | unexpand_if_needed (const char *aux_info_line) | |
1069 | { | |
1070 | static char *line_buf = 0; | |
1071 | static int line_buf_size = 0; | |
1072 | const unexpansion* unexp_p; | |
1073 | int got_unexpanded = 0; | |
1074 | const char *s; | |
1075 | char *copy_p = line_buf; | |
1076 | ||
1077 | if (line_buf == 0) | |
1078 | { | |
1079 | line_buf_size = 1024; | |
1080 | line_buf = (char *) xmalloc (line_buf_size); | |
1081 | } | |
1082 | ||
1083 | copy_p = line_buf; | |
1084 | ||
1085 | /* Make a copy of the input string in line_buf, expanding as necessary. */ | |
1086 | ||
1087 | for (s = aux_info_line; *s != '\n'; ) | |
1088 | { | |
1089 | for (unexp_p = unexpansions; unexp_p->expanded; unexp_p++) | |
1090 | { | |
1091 | const char *in_p = unexp_p->expanded; | |
1092 | size_t len = strlen (in_p); | |
1093 | ||
1094 | if (*s == *in_p && !strncmp (s, in_p, len) && !is_id_char (s[len])) | |
1095 | { | |
1096 | int size = strlen (unexp_p->contracted); | |
1097 | got_unexpanded = 1; | |
1098 | if (copy_p + size - line_buf >= line_buf_size) | |
1099 | { | |
1100 | int offset = copy_p - line_buf; | |
1101 | line_buf_size *= 2; | |
1102 | line_buf_size += size; | |
1103 | line_buf = (char *) xrealloc (line_buf, line_buf_size); | |
1104 | copy_p = line_buf + offset; | |
1105 | } | |
1106 | strcpy (copy_p, unexp_p->contracted); | |
1107 | copy_p += size; | |
1108 | ||
1109 | /* Assume the there will not be another replacement required | |
1110 | within the text just replaced. */ | |
1111 | ||
1112 | s += len; | |
1113 | goto continue_outer; | |
1114 | } | |
1115 | } | |
1116 | if (copy_p - line_buf == line_buf_size) | |
1117 | { | |
1118 | int offset = copy_p - line_buf; | |
1119 | line_buf_size *= 2; | |
1120 | line_buf = (char *) xrealloc (line_buf, line_buf_size); | |
1121 | copy_p = line_buf + offset; | |
1122 | } | |
1123 | *copy_p++ = *s++; | |
1124 | continue_outer: ; | |
1125 | } | |
1126 | if (copy_p + 2 - line_buf >= line_buf_size) | |
1127 | { | |
1128 | int offset = copy_p - line_buf; | |
1129 | line_buf_size *= 2; | |
1130 | line_buf = (char *) xrealloc (line_buf, line_buf_size); | |
1131 | copy_p = line_buf + offset; | |
1132 | } | |
1133 | *copy_p++ = '\n'; | |
1134 | *copy_p++ = '\0'; | |
1135 | ||
1136 | return (got_unexpanded ? dupstr (line_buf) : 0); | |
1137 | } | |
1138 | \f | |
a2b22788 RS |
1139 | /* Return the absolutized filename for the given relative |
1140 | filename. Note that if that filename is already absolute, it may | |
5f8037c4 RS |
1141 | still be returned in a modified form because this routine also |
1142 | eliminates redundant slashes and single dots and eliminates double | |
a2b22788 RS |
1143 | dots to get a shortest possible filename from the given input |
1144 | filename. The absolutization of relative filenames is made by | |
1145 | assuming that the given filename is to be taken as relative to | |
5f8037c4 RS |
1146 | the first argument (cwd) or to the current directory if cwd is |
1147 | NULL. */ | |
1148 | ||
1149 | static char * | |
a2b22788 | 1150 | abspath (const char *cwd, const char *rel_filename) |
5f8037c4 RS |
1151 | { |
1152 | /* Setup the current working directory as needed. */ | |
1153 | const char *cwd2 = (cwd) ? cwd : cwd_buffer; | |
1154 | char *const abs_buffer | |
a2b22788 | 1155 | = (char *) alloca (strlen (cwd2) + strlen (rel_filename) + 1); |
5f8037c4 RS |
1156 | char *endp = abs_buffer; |
1157 | char *outp, *inp; | |
1158 | ||
a2b22788 | 1159 | /* Copy the filename (possibly preceeded by the current working |
5f8037c4 RS |
1160 | directory name) into the absolutization buffer. */ |
1161 | ||
1162 | { | |
1163 | const char *src_p; | |
1164 | ||
a2b22788 | 1165 | if (rel_filename[0] != '/') |
5f8037c4 RS |
1166 | { |
1167 | src_p = cwd2; | |
1168 | while (*endp++ = *src_p++) | |
1169 | continue; | |
1170 | *(endp-1) = '/'; /* overwrite null */ | |
1171 | } | |
a2b22788 | 1172 | src_p = rel_filename; |
5f8037c4 RS |
1173 | while (*endp++ = *src_p++) |
1174 | continue; | |
1175 | if (endp[-1] == '/') | |
1176 | *endp = '\0'; | |
1177 | } | |
1178 | ||
1179 | /* Now make a copy of abs_buffer into abs_buffer, shortening the | |
a2b22788 | 1180 | filename (by taking out slashes and dots) as we go. */ |
5f8037c4 RS |
1181 | |
1182 | outp = inp = abs_buffer; | |
1183 | *outp++ = *inp++; /* copy first slash */ | |
1184 | for (;;) | |
1185 | { | |
1186 | if (!inp[0]) | |
1187 | break; | |
1188 | else if (inp[0] == '/' && outp[-1] == '/') | |
1189 | { | |
1190 | inp++; | |
1191 | continue; | |
1192 | } | |
1193 | else if (inp[0] == '.' && outp[-1] == '/') | |
1194 | { | |
1195 | if (!inp[1]) | |
1196 | break; | |
1197 | else if (inp[1] == '/') | |
1198 | { | |
1199 | inp += 2; | |
1200 | continue; | |
1201 | } | |
1202 | else if ((inp[1] == '.') && (inp[2] == 0 || inp[2] == '/')) | |
1203 | { | |
1204 | inp += (inp[2] == '/') ? 3 : 2; | |
1205 | outp -= 2; | |
1206 | while (outp >= abs_buffer && *outp != '/') | |
1207 | outp--; | |
1208 | if (outp < abs_buffer) | |
1209 | { | |
1210 | /* Catch cases like /.. where we try to backup to a | |
1211 | point above the absolute root of the logical file | |
1212 | system. */ | |
1213 | ||
a2b22788 RS |
1214 | fprintf (stderr, "%s: invalid file name: %s\n", |
1215 | pname, rel_filename); | |
5f8037c4 RS |
1216 | exit (1); |
1217 | } | |
1218 | *++outp = '\0'; | |
1219 | continue; | |
1220 | } | |
1221 | } | |
1222 | *outp++ = *inp++; | |
1223 | } | |
1224 | ||
1225 | /* On exit, make sure that there is a trailing null, and make sure that | |
1226 | the last character of the returned string is *not* a slash. */ | |
1227 | ||
1228 | *outp = '\0'; | |
1229 | if (outp[-1] == '/') | |
1230 | *--outp = '\0'; | |
1231 | ||
1232 | /* Make a copy (in the heap) of the stuff left in the absolutization | |
1233 | buffer and return a pointer to the copy. */ | |
1234 | ||
1235 | return dupstr (abs_buffer); | |
1236 | } | |
1237 | \f | |
a2b22788 | 1238 | /* Given a filename (and possibly a directory name from which the filename |
5f8037c4 | 1239 | is relative) return a string which is the shortest possible |
a2b22788 | 1240 | equivalent for the corresponding full (absolutized) filename. The |
5f8037c4 | 1241 | shortest possible equivalent may be constructed by converting the |
a2b22788 RS |
1242 | absolutized filename to be a relative filename (i.e. relative to |
1243 | the actual current working directory). However if a relative filename | |
1244 | is longer, then the full absolute filename is returned. | |
5f8037c4 RS |
1245 | |
1246 | KNOWN BUG: | |
1247 | ||
a2b22788 RS |
1248 | Note that "simple-minded" conversion of any given type of filename (either |
1249 | relative or absolute) may not result in a valid equivalent filename if any | |
1250 | subpart of the original filename is actually a symbolic link. */ | |
5f8037c4 RS |
1251 | |
1252 | static const char * | |
a2b22788 | 1253 | shortpath (const char *cwd, const char *filename) |
5f8037c4 RS |
1254 | { |
1255 | char *rel_buffer; | |
1256 | char *rel_buf_p; | |
1257 | char *cwd_p = cwd_buffer; | |
1258 | char *path_p; | |
1259 | int unmatched_slash_count = 0; | |
1260 | ||
a2b22788 | 1261 | path_p = abspath (cwd, filename); |
5f8037c4 RS |
1262 | rel_buf_p = rel_buffer = (char *) xmalloc (strlen (path_p) + 1); |
1263 | ||
1264 | while (*cwd_p && (*cwd_p == *path_p)) | |
1265 | { | |
1266 | cwd_p++; | |
1267 | path_p++; | |
1268 | } | |
1269 | if (!*cwd_p) /* whole pwd matched */ | |
1270 | { | |
1271 | if (!*path_p) /* input *is* the current path! */ | |
1272 | return "."; | |
1273 | else | |
1274 | return ++path_p; | |
1275 | } | |
1276 | else | |
1277 | { | |
1278 | if (*path_p) | |
1279 | { | |
1280 | --cwd_p; | |
1281 | --path_p; | |
1282 | while (*cwd_p != '/') /* backup to last slash */ | |
1283 | { | |
1284 | --cwd_p; | |
1285 | --path_p; | |
1286 | } | |
1287 | cwd_p++; | |
1288 | path_p++; | |
1289 | unmatched_slash_count++; | |
1290 | } | |
1291 | while (*cwd_p) | |
1292 | if (*cwd_p++ == '/') | |
1293 | unmatched_slash_count++; | |
1294 | while (unmatched_slash_count--) | |
1295 | { | |
1296 | *rel_buf_p++ = '.'; | |
1297 | *rel_buf_p++ = '.'; | |
1298 | *rel_buf_p++ = '/'; | |
1299 | } | |
1300 | while (*rel_buf_p++ = *path_p++) | |
1301 | continue; | |
1302 | --rel_buf_p; | |
1303 | if (*(rel_buf_p-1) == '/') | |
1304 | *--rel_buf_p = '\0'; | |
a2b22788 RS |
1305 | if (strlen (rel_buffer) > strlen (filename)) |
1306 | strcpy (rel_buffer, filename); | |
5f8037c4 RS |
1307 | return rel_buffer; |
1308 | } | |
1309 | /* NOTREACHED */ | |
1310 | return 0; /* Prevent warnings for old versions of GCC. */ | |
1311 | } | |
1312 | \f | |
a2b22788 | 1313 | /* Lookup the given filename in the hash table for filenames. If it is a |
5f8037c4 | 1314 | new one, then the hash table info pointer will be null. In this case, |
a2b22788 | 1315 | we create a new file_info record to go with the filename, and we initialize |
5f8037c4 RS |
1316 | that record with some reasonable values. */ |
1317 | ||
1318 | static file_info * | |
a2b22788 RS |
1319 | find_file (char *filename, int do_not_stat) |
1320 | /* FILENAME was const, but that causes a warning on AIX when calling stat. | |
1321 | That is probably a bug in AIX, but might as well avoid the warning. */ | |
5f8037c4 RS |
1322 | { |
1323 | hash_table_entry *hash_entry_p; | |
1324 | ||
a2b22788 | 1325 | hash_entry_p = lookup (filename_primary, filename); |
5f8037c4 RS |
1326 | if (hash_entry_p->fip) |
1327 | return hash_entry_p->fip; | |
1328 | else | |
1329 | { | |
1330 | struct stat stat_buf; | |
1331 | file_info *file_p = (file_info *) xmalloc (sizeof (file_info)); | |
1332 | ||
1333 | /* If we cannot get status on any given source file, give a warning | |
1334 | and then just set its time of last modification to infinity. */ | |
1335 | ||
1336 | if (do_not_stat) | |
1337 | stat_buf.st_mtime = (time_t) 0; | |
1338 | else | |
1339 | { | |
a2b22788 | 1340 | if (stat (filename, &stat_buf) == -1) |
5f8037c4 RS |
1341 | { |
1342 | fprintf (stderr, "%s: error: can't get status of `%s': %s\n", | |
a2b22788 | 1343 | pname, shortpath (NULL, filename), sys_errlist[errno]); |
5f8037c4 RS |
1344 | stat_buf.st_mtime = (time_t) -1; |
1345 | } | |
1346 | } | |
1347 | ||
1348 | hash_entry_p->fip = file_p; | |
1349 | file_p->hash_entry = hash_entry_p; | |
1350 | file_p->defs_decs = NULL; | |
1351 | file_p->mtime = stat_buf.st_mtime; | |
1352 | return file_p; | |
1353 | } | |
1354 | } | |
1355 | ||
1356 | /* Generate a fatal error because some part of the aux_info file is | |
1357 | messed up. */ | |
1358 | ||
1359 | static void | |
1360 | aux_info_corrupted (void) | |
1361 | { | |
1362 | fprintf (stderr, "\n%s: fatal error: aux info file corrupted at line %d\n", | |
a2b22788 | 1363 | pname, current_aux_info_lineno); |
5f8037c4 RS |
1364 | exit (1); |
1365 | } | |
1366 | ||
1367 | /* ??? This comment is vague. Say what the condition is for. */ | |
1368 | /* Check to see that a condition is true. This is kind of like an assert(). */ | |
1369 | ||
1370 | inline static void | |
1371 | check_aux_info (int cond) | |
1372 | { | |
1373 | if (! cond) | |
1374 | aux_info_corrupted (); | |
1375 | } | |
1376 | ||
1377 | /* Given a pointer to the closing right parenthesis for a particular formals | |
1378 | list (in a aux_info file) find the corresponding left parenthesis and | |
1379 | return a pointer to it. */ | |
1380 | ||
1381 | static const char * | |
1382 | find_corresponding_lparen (const char *p) | |
1383 | { | |
1384 | const char *q; | |
1385 | int paren_depth; | |
1386 | ||
1387 | for (paren_depth = 1, q = p-1; paren_depth; q--) | |
1388 | { | |
1389 | switch (*q) | |
1390 | { | |
1391 | case ')': | |
1392 | paren_depth++; | |
1393 | break; | |
1394 | case '(': | |
1395 | paren_depth--; | |
1396 | break; | |
1397 | } | |
1398 | } | |
1399 | return ++q; | |
1400 | } | |
1401 | \f | |
1402 | /* Given a line from an aux info file, and a time at which the aux info | |
1403 | file it came from was created, check to see if the item described in | |
1404 | the line comes from a file which has been modified since the aux info | |
1405 | file was created. If so, return non-zero, else return zero. */ | |
1406 | ||
1407 | static int | |
1408 | referenced_file_is_newer (const char *l, time_t aux_info_mtime) | |
1409 | { | |
1410 | const char *p; | |
1411 | file_info *fi_p; | |
1412 | char *filename; | |
1413 | ||
1414 | check_aux_info (l[0] == '/'); | |
1415 | check_aux_info (l[1] == '*'); | |
1416 | check_aux_info (l[2] == ' '); | |
1417 | ||
1418 | { | |
1419 | const char *filename_start = p = l + 3; | |
1420 | ||
1421 | while (*p != ':') | |
1422 | p++; | |
1423 | filename = (char *) alloca ((size_t) (p - filename_start) + 1); | |
1424 | strncpy (filename, filename_start, (size_t) (p - filename_start)); | |
1425 | filename[p-filename_start] = '\0'; | |
1426 | } | |
1427 | ||
1428 | /* Call find_file to find the file_info record associated with the file | |
1429 | which contained this particular def or dec item. Note that this call | |
1430 | may cause a new file_info record to be created if this is the first time | |
1431 | that we have ever known about this particular file. */ | |
1432 | ||
a2b22788 | 1433 | fi_p = find_file (abspath (invocation_filename, filename), 0); |
5f8037c4 RS |
1434 | |
1435 | return (fi_p->mtime > aux_info_mtime); | |
1436 | } | |
1437 | \f | |
1438 | /* Given a line of info from the aux_info file, create a new | |
1439 | def_dec_info record to remember all of the important information about | |
1440 | a function definition or declaration. | |
1441 | ||
1442 | Link this record onto the list of such records for the particular file in | |
1443 | which it occured in proper (descending) line number order (for now). | |
1444 | ||
1445 | If there is an identical record already on the list for the file, throw | |
1446 | this one away. Doing so takes care of the (useless and troublesome) | |
1447 | duplicates which are bound to crop up due to multiple inclusions of any | |
1448 | given individual header file. | |
1449 | ||
1450 | Finally, link the new def_dec record onto the list of such records | |
1451 | pertaining to this particular function name. */ | |
1452 | ||
1453 | static void | |
1454 | save_def_or_dec (const char *l, int is_syscalls) | |
1455 | { | |
1456 | const char *p; | |
1457 | const char *semicolon_p; | |
1458 | def_dec_info *def_dec_p = (def_dec_info *) xmalloc (sizeof (def_dec_info)); | |
1459 | ||
1460 | #ifndef UNPROTOIZE | |
1461 | def_dec_p->written = 0; | |
1462 | #endif /* !defined(UNPROTOIZE) */ | |
1463 | ||
1464 | /* Start processing the line by picking off 5 pieces of information from | |
1465 | the left hand end of the line. These are filename, line number, | |
1466 | new/old/implicit flag (new = ANSI prototype format), definition or | |
1467 | declaration flag, and extern/static flag). */ | |
1468 | ||
1469 | check_aux_info (l[0] == '/'); | |
1470 | check_aux_info (l[1] == '*'); | |
1471 | check_aux_info (l[2] == ' '); | |
1472 | ||
1473 | { | |
1474 | const char *filename_start = p = l + 3; | |
1475 | char *filename; | |
1476 | ||
1477 | while (*p != ':') | |
1478 | p++; | |
1479 | filename = (char *) alloca ((size_t) (p - filename_start) + 1); | |
1480 | strncpy (filename, filename_start, (size_t) (p - filename_start)); | |
1481 | filename[p-filename_start] = '\0'; | |
1482 | ||
1483 | /* Call find_file to find the file_info record associated with the file | |
1484 | which contained this particular def or dec item. Note that this call | |
1485 | may cause a new file_info record to be created if this is the first time | |
1486 | that we have ever known about this particular file. | |
1487 | ||
a2b22788 | 1488 | Note that we started out by forcing all of the base source file names |
5f8037c4 | 1489 | (i.e. the names of the aux_info files with the .X stripped off) into the |
a2b22788 RS |
1490 | filenames hash table, and we simultaneously setup file_info records for |
1491 | all of these base file names (even if they may be useless later). | |
1492 | The file_info records for all of these "base" file names (properly) | |
5f8037c4 RS |
1493 | act as file_info records for the "original" (i.e. un-included) files |
1494 | which were submitted to gcc for compilation (when the -fgen-aux-info | |
1495 | option was used). */ | |
1496 | ||
a2b22788 | 1497 | def_dec_p->file = find_file (abspath (invocation_filename, filename), is_syscalls); |
5f8037c4 RS |
1498 | } |
1499 | ||
1500 | { | |
1501 | const char *line_number_start = ++p; | |
1502 | char line_number[10]; | |
1503 | ||
1504 | while (*p != ':') | |
1505 | p++; | |
1506 | strncpy (line_number, line_number_start, (size_t) (p - line_number_start)); | |
1507 | line_number[p-line_number_start] = '\0'; | |
1508 | def_dec_p->line = atoi (line_number); | |
1509 | } | |
1510 | ||
1511 | /* Check that this record describes a new-style, old-style, or implicit | |
1512 | definition or declaration. */ | |
1513 | ||
1514 | p++; /* Skip over the `:'. */ | |
1515 | check_aux_info ((*p == 'N') || (*p == 'O') || (*p == 'I')); | |
1516 | ||
1517 | /* Is this a new style (ANSI prototyped) definition or declaration? */ | |
1518 | ||
1519 | def_dec_p->prototyped = (*p == 'N'); | |
1520 | ||
1521 | #ifndef UNPROTOIZE | |
1522 | ||
1523 | /* Is this an implicit declaration? */ | |
1524 | ||
1525 | def_dec_p->is_implicit = (*p == 'I'); | |
1526 | ||
1527 | #endif /* !defined(UNPROTOIZE) */ | |
1528 | ||
1529 | p++; | |
1530 | ||
1531 | check_aux_info ((*p == 'C') || (*p == 'F')); | |
1532 | ||
1533 | /* Is this item a function definition (F) or a declaration (C). Note that | |
1534 | we treat item taken from the syscalls file as though they were function | |
1535 | definitions regardless of what the stuff in the file says. */ | |
1536 | ||
1537 | def_dec_p->is_func_def = ((*p++ == 'F') || is_syscalls); | |
1538 | ||
1539 | #ifndef UNPROTOIZE | |
1540 | def_dec_p->definition = 0; /* Fill this in later if protoizing. */ | |
1541 | #endif /* !defined(UNPROTOIZE) */ | |
1542 | ||
1543 | check_aux_info (*p++ == ' '); | |
1544 | check_aux_info (*p++ == '*'); | |
1545 | check_aux_info (*p++ == '/'); | |
1546 | check_aux_info (*p++ == ' '); | |
1547 | ||
1548 | #ifdef UNPROTOIZE | |
1549 | check_aux_info ((!strncmp (p, "static", 6)) || (!strncmp (p, "extern", 6))); | |
1550 | #else /* !defined(UNPROTOIZE) */ | |
1551 | if (!strncmp (p, "static", 6)) | |
1552 | def_dec_p->is_static = -1; | |
1553 | else if (!strncmp (p, "extern", 6)) | |
1554 | def_dec_p->is_static = 0; | |
1555 | else | |
1556 | check_aux_info (0); /* Didn't find either `extern' or `static'. */ | |
1557 | #endif /* !defined(UNPROTOIZE) */ | |
1558 | ||
1559 | { | |
1560 | const char *ansi_start = p; | |
1561 | ||
1562 | p += 6; /* Pass over the "static" or "extern". */ | |
1563 | ||
1564 | /* We are now past the initial stuff. Search forward from here to find | |
1565 | the terminating semicolon that should immediately follow the entire | |
1566 | ANSI format function declaration. */ | |
1567 | ||
1568 | while (*++p != ';') | |
1569 | continue; | |
1570 | ||
1571 | semicolon_p = p; | |
1572 | ||
1573 | /* Make a copy of the ansi declaration part of the line from the aux_info | |
1574 | file. */ | |
1575 | ||
1576 | def_dec_p->ansi_decl | |
1577 | = dupnstr (ansi_start, (size_t) ((semicolon_p+1) - ansi_start)); | |
1578 | } | |
1579 | ||
1580 | /* Backup and point at the final right paren of the final argument list. */ | |
1581 | ||
1582 | p--; | |
1583 | ||
1584 | /* Now isolate a whole set of formal argument lists, one-by-one. Normally, | |
1585 | there will only be one list to isolate, but there could be more. */ | |
1586 | ||
1587 | def_dec_p->f_list_count = 0; | |
1588 | ||
1589 | #ifndef UNPROTOIZE | |
1590 | def_dec_p->f_list_chain = NULL; | |
1591 | #endif /* !defined(UNPROTOIZE) */ | |
1592 | ||
1593 | for (;;) | |
1594 | { | |
1595 | const char *left_paren_p = find_corresponding_lparen (p); | |
1596 | #ifndef UNPROTOIZE | |
1597 | { | |
1598 | f_list_chain_item *cip = | |
1599 | (f_list_chain_item *) xmalloc (sizeof (f_list_chain_item)); | |
1600 | ||
1601 | cip->formals_list | |
1602 | = dupnstr (left_paren_p + 1, (size_t) (p - (left_paren_p+1))); | |
1603 | ||
1604 | /* Add the new chain item at the head of the current list. */ | |
1605 | ||
1606 | cip->chain_next = def_dec_p->f_list_chain; | |
1607 | def_dec_p->f_list_chain = cip; | |
1608 | } | |
1609 | #endif /* !defined(UNPROTOIZE) */ | |
1610 | def_dec_p->f_list_count++; | |
1611 | ||
1612 | p = left_paren_p - 2; | |
1613 | ||
1614 | /* p must now point either to another right paren, or to the last | |
1615 | character of the name of the function that was declared/defined. | |
1616 | If p points to another right paren, then this indicates that we | |
1617 | are dealing with multiple formals lists. In that case, there | |
1618 | really should be another right paren preceeding this right paren. */ | |
1619 | ||
1620 | if (*p != ')') | |
1621 | break; | |
1622 | else | |
1623 | check_aux_info (*--p == ')'); | |
1624 | } | |
1625 | ||
1626 | ||
1627 | { | |
1628 | const char *past_fn = p + 1; | |
1629 | ||
1630 | check_aux_info (*past_fn == ' '); | |
1631 | ||
1632 | /* Scan leftwards over the identifier that names the function. */ | |
1633 | ||
1634 | while (is_id_char (*p)) | |
1635 | p--; | |
1636 | p++; | |
1637 | ||
1638 | /* p now points to the leftmost character of the function name. */ | |
1639 | ||
1640 | { | |
1641 | char fn_string[past_fn - p + 1]; | |
1642 | ||
1643 | strncpy (fn_string, p, (size_t) (past_fn - p)); | |
1644 | fn_string[past_fn-p] = '\0'; | |
1645 | def_dec_p->hash_entry = lookup (function_name_primary, fn_string); | |
1646 | } | |
1647 | } | |
1648 | ||
1649 | /* Look at all of the defs and decs for this function name that we have | |
1650 | collected so far. If there is already one which is at the same | |
1651 | line number in the same file, then we can discard this new def_dec_info | |
1652 | record. | |
1653 | ||
1654 | As an extra assurance that any such pair of (nominally) identical | |
1655 | function declarations are in fact identical, we also compare the | |
1656 | ansi_decl parts of the lines from the aux_info files just to be on | |
1657 | the safe side. | |
1658 | ||
1659 | This comparison will fail if (for instance) the user was playing | |
1660 | messy games with the preprocessor which ultimately causes one | |
1661 | function declaration in one header file to look differently when | |
1662 | that file is included by two (or more) other files. */ | |
1663 | ||
1664 | { | |
1665 | const def_dec_info *other; | |
1666 | ||
1667 | for (other = def_dec_p->hash_entry->ddip; other; other = other->next_for_func) | |
1668 | { | |
1669 | if (def_dec_p->line == other->line && def_dec_p->file == other->file) | |
1670 | { | |
1671 | if (strcmp (def_dec_p->ansi_decl, other->ansi_decl)) | |
1672 | { | |
1673 | fprintf (stderr, "%s: error: declaration of function `%s' at %s(%d) takes different forms\n", | |
1674 | pname, | |
1675 | def_dec_p->hash_entry->symbol, | |
1676 | def_dec_p->file->hash_entry->symbol, | |
1677 | def_dec_p->line); | |
1678 | exit (1); | |
1679 | } | |
1680 | free_def_dec (def_dec_p); | |
1681 | return; | |
1682 | } | |
1683 | } | |
1684 | } | |
1685 | ||
1686 | #ifdef UNPROTOIZE | |
1687 | ||
1688 | /* If we are doing unprotoizing, we must now setup the pointers that will | |
1689 | point to the K&R name list and to the K&R argument declarations list. | |
1690 | ||
1691 | Note that if this is only a function declaration, then we should not | |
1692 | expect to find any K&R style formals list following the ANSI-style | |
1693 | formals list. This is because GCC knows that such information is | |
1694 | useless in the case of function declarations (function definitions | |
1695 | are a different story however). | |
1696 | ||
1697 | Since we are unprotoizing, we don't need any such lists anyway. | |
1698 | All we plan to do is to delete all characters between ()'s in any | |
1699 | case. */ | |
1700 | ||
1701 | def_dec_p->formal_names = NULL; | |
1702 | def_dec_p->formal_decls = NULL; | |
1703 | ||
1704 | if (def_dec_p->is_func_def) | |
1705 | { | |
1706 | p = semicolon_p; | |
1707 | check_aux_info (*++p == ' '); | |
1708 | check_aux_info (*++p == '/'); | |
1709 | check_aux_info (*++p == '*'); | |
1710 | check_aux_info (*++p == ' '); | |
1711 | check_aux_info (*++p == '('); | |
1712 | ||
1713 | { | |
1714 | const char *kr_names_start = ++p; /* Point just inside '('. */ | |
1715 | ||
1716 | while (*p++ != ')') | |
1717 | continue; | |
1718 | p--; /* point to closing right paren */ | |
1719 | ||
1720 | /* Make a copy of the K&R parameter names list. */ | |
1721 | ||
1722 | def_dec_p->formal_names | |
1723 | = dupnstr (kr_names_start, (size_t) (p - kr_names_start)); | |
1724 | } | |
1725 | ||
1726 | check_aux_info (*++p == ' '); | |
1727 | p++; | |
1728 | ||
1729 | /* p now points to the first character of the K&R style declarations | |
1730 | list (if there is one) or to the star-slash combination that ends | |
1731 | the comment in which such lists get embedded. */ | |
1732 | ||
1733 | /* Make a copy of the K&R formal decls list and set the def_dec record | |
1734 | to point to it. */ | |
1735 | ||
1736 | if (*p == '*') /* Are there no K&R declarations? */ | |
1737 | { | |
1738 | check_aux_info (*++p == '/'); | |
1739 | def_dec_p->formal_decls = ""; | |
1740 | } | |
1741 | else | |
1742 | { | |
1743 | const char *kr_decls_start = p; | |
1744 | ||
1745 | while (p[0] != '*' || p[1] != '/') | |
1746 | p++; | |
1747 | p--; | |
1748 | ||
1749 | check_aux_info (*p == ' '); | |
1750 | ||
1751 | def_dec_p->formal_decls | |
1752 | = dupnstr (kr_decls_start, (size_t) (p - kr_decls_start)); | |
1753 | } | |
1754 | ||
1755 | /* Handle a special case. If we have a function definition marked as | |
1756 | being in "old" style, and if it's formal names list is empty, then | |
1757 | it may actually have the string "void" in its real formals list | |
1758 | in the original source code. Just to make sure, we will get setup | |
1759 | to convert such things anyway. | |
1760 | ||
1761 | This kludge only needs to be here because of an insurmountable | |
1762 | problem with generating .X files. */ | |
1763 | ||
1764 | if (!def_dec_p->prototyped && !*def_dec_p->formal_names) | |
1765 | def_dec_p->prototyped = 1; | |
1766 | } | |
1767 | ||
1768 | /* Since we are unprotoizing, if this item is already in old (K&R) style, | |
1769 | we can just ignore it. If that is true, throw away the itme now. */ | |
1770 | ||
1771 | if (!def_dec_p->prototyped) | |
1772 | { | |
1773 | free_def_dec (def_dec_p); | |
1774 | return; | |
1775 | } | |
1776 | ||
1777 | #endif /* defined(UNPROTOIZE) */ | |
1778 | ||
1779 | /* Add this record to the head of the list of records pertaining to this | |
1780 | particular function name. */ | |
1781 | ||
1782 | def_dec_p->next_for_func = def_dec_p->hash_entry->ddip; | |
1783 | def_dec_p->hash_entry->ddip = def_dec_p; | |
1784 | ||
1785 | /* Add this new def_dec_info record to the sorted list of def_dec_info | |
1786 | records for this file. Note that we don't have to worry about duplicates | |
1787 | (caused by multiple inclusions of header files) here because we have | |
1788 | already eliminated duplicates above. */ | |
1789 | ||
1790 | if (!def_dec_p->file->defs_decs) | |
1791 | { | |
1792 | def_dec_p->file->defs_decs = def_dec_p; | |
1793 | def_dec_p->next_in_file = NULL; | |
1794 | } | |
1795 | else | |
1796 | { | |
1797 | int line = def_dec_p->line; | |
1798 | const def_dec_info *prev = NULL; | |
1799 | const def_dec_info *curr = def_dec_p->file->defs_decs; | |
1800 | const def_dec_info *next = curr->next_in_file; | |
1801 | ||
1802 | while (next && (line < curr->line)) | |
1803 | { | |
1804 | prev = curr; | |
1805 | curr = next; | |
1806 | next = next->next_in_file; | |
1807 | } | |
1808 | if (line >= curr->line) | |
1809 | { | |
1810 | def_dec_p->next_in_file = curr; | |
1811 | if (prev) | |
1812 | ((NONCONST def_dec_info *) prev)->next_in_file = def_dec_p; | |
1813 | else | |
1814 | def_dec_p->file->defs_decs = def_dec_p; | |
1815 | } | |
1816 | else /* assert (next == NULL); */ | |
1817 | { | |
1818 | ((NONCONST def_dec_info *) curr)->next_in_file = def_dec_p; | |
1819 | /* assert (next == NULL); */ | |
1820 | def_dec_p->next_in_file = next; | |
1821 | } | |
1822 | } | |
1823 | } | |
1824 | \f | |
1825 | /* Rewrite the options list used to recompile base source files. All we are | |
1826 | really doing here is removing -g, -O, -S, -c, and -o options, and then | |
1827 | adding a final group of options like '-fgen-aux-info -S -o /dev/null'. */ | |
1828 | ||
1829 | static void | |
1830 | munge_compile_params (const char *params_list) | |
1831 | { | |
1832 | const char *temp_params[strlen (params_list) + 10]; | |
1833 | int param_count = 0; | |
1834 | const char *param; | |
1835 | ||
1836 | temp_params[param_count++] = "gcc"; | |
1837 | for (;;) | |
1838 | { | |
1839 | while (isspace (*params_list)) | |
1840 | params_list++; | |
1841 | if (!*params_list) | |
1842 | break; | |
1843 | param = params_list; | |
1844 | while (*params_list && !isspace (*params_list)) | |
1845 | params_list++; | |
1846 | if (param[0] != '-') | |
1847 | temp_params[param_count++] | |
1848 | = dupnstr (param, (size_t) (params_list - param)); | |
1849 | else | |
1850 | { | |
1851 | switch (param[1]) | |
1852 | { | |
1853 | case 'g': | |
1854 | case 'O': | |
1855 | case 'S': | |
1856 | case 'c': | |
1857 | break; /* Don't copy these. */ | |
1858 | case 'o': | |
1859 | while (isspace (*params_list)) | |
1860 | params_list++; | |
1861 | while (*params_list && !isspace (*params_list)) | |
1862 | params_list++; | |
1863 | break; | |
1864 | default: | |
1865 | temp_params[param_count++] | |
1866 | = dupnstr (param, (size_t) (params_list - param)); | |
1867 | } | |
1868 | } | |
1869 | if (!*params_list) | |
1870 | break; | |
1871 | } | |
1872 | temp_params[param_count++] = "-fgen-aux-info"; | |
1873 | temp_params[param_count++] = "-S"; | |
1874 | temp_params[param_count++] = "-o"; | |
1875 | temp_params[param_count++] = "/dev/null"; | |
1876 | ||
1877 | /* Leave room for the filename argument and a terminating null pointer. */ | |
1878 | ||
1879 | temp_params[filename_index = param_count++] = NULL; | |
1880 | temp_params[param_count++] = NULL; | |
1881 | ||
1882 | /* Make a copy of the compile_params in heap space. */ | |
1883 | ||
1884 | compile_params = xmalloc (sizeof (char *) * (param_count+1)); | |
1885 | memcpy (compile_params, temp_params, sizeof (char *) * param_count); | |
1886 | } | |
1887 | ||
1888 | /* Do a recompilation for the express purpose of generating a new aux_info | |
1889 | file to go with a specific base source file. */ | |
1890 | ||
1891 | static int | |
a2b22788 | 1892 | gen_aux_info_file (const char *base_filename) |
5f8037c4 RS |
1893 | { |
1894 | int child_pid; | |
1895 | ||
1896 | if (!filename_index) | |
1897 | munge_compile_params (""); | |
1898 | ||
a2b22788 | 1899 | compile_params[filename_index] = shortpath (NULL, base_filename); |
5f8037c4 RS |
1900 | |
1901 | if (!quiet_flag) | |
1902 | fprintf (stderr, "%s: compiling `%s'\n", | |
a2b22788 | 1903 | pname, compile_params[filename_index]); |
5f8037c4 RS |
1904 | |
1905 | if (child_pid = fork ()) | |
1906 | { | |
1907 | if (child_pid == -1) | |
1908 | { | |
1909 | fprintf (stderr, "%s: error: could not fork process: %s\n", | |
a2b22788 | 1910 | pname, sys_errlist[errno]); |
5f8037c4 RS |
1911 | return 0; |
1912 | } | |
1913 | ||
1914 | #if 0 | |
1915 | /* Print out the command line that the other process is now executing. */ | |
1916 | ||
1917 | if (!quiet_flag) | |
1918 | { | |
1919 | const char **arg; | |
1920 | ||
1921 | fputs ("\t", stderr); | |
1922 | for (arg = compile_params; *arg; arg++) | |
1923 | { | |
1924 | fputs (*arg, stderr); | |
1925 | fputc (' ', stderr); | |
1926 | } | |
1927 | fputc ('\n', stderr); | |
1928 | fflush (stderr); | |
1929 | } | |
1930 | #endif /* 0 */ | |
1931 | ||
1932 | { | |
1933 | int wait_status; | |
1934 | ||
1935 | if (wait (&wait_status) == -1) | |
1936 | { | |
1937 | fprintf (stderr, "%s: error: wait for process failed: %s\n", | |
a2b22788 | 1938 | pname, sys_errlist[errno]); |
5f8037c4 RS |
1939 | return 0; |
1940 | } | |
1941 | if (!WIFEXITED (wait_status)) | |
1942 | { | |
1943 | kill (child_pid, 9); | |
1944 | return 0; | |
1945 | } | |
1946 | return (WEXITSTATUS (wait_status) == 0) ? 1 : 0; | |
1947 | } | |
1948 | } | |
1949 | else | |
1950 | { | |
1951 | if (execvp (compile_params[0], (char *const *) compile_params)) | |
1952 | { | |
1953 | fprintf (stderr, "%s: error: execvp returned: %s\n", | |
a2b22788 | 1954 | pname, sys_errlist[errno]); |
5f8037c4 RS |
1955 | exit (errno); |
1956 | } | |
1957 | return 1; /* Never executed. */ | |
1958 | } | |
1959 | } | |
1960 | \f | |
1961 | /* Read in all of the information contained in a single aux_info file. | |
1962 | Save all of the important stuff for later. */ | |
1963 | ||
1964 | static void | |
a2b22788 | 1965 | process_aux_info_file (const char *base_source_filename, int keep_it, int is_syscalls) |
5f8037c4 | 1966 | { |
a2b22788 RS |
1967 | char *const aux_info_filename |
1968 | = (char *) alloca (strlen (base_source_filename) | |
5f8037c4 RS |
1969 | + strlen (aux_info_suffix) + 1); |
1970 | char *aux_info_base; | |
1971 | char *aux_info_limit; | |
1972 | const char *aux_info_second_line; | |
1973 | time_t aux_info_mtime; | |
1974 | size_t aux_info_size; | |
1975 | ||
a2b22788 | 1976 | /* Construct the aux_info filename from the base source filename. */ |
5f8037c4 | 1977 | |
a2b22788 RS |
1978 | strcpy (aux_info_filename, base_source_filename); |
1979 | strcat (aux_info_filename, aux_info_suffix); | |
5f8037c4 RS |
1980 | |
1981 | /* Check that the aux_info file exists and is readable. If it does not | |
1982 | exist, try to create it (once only). */ | |
1983 | ||
1984 | start_over: ; | |
1985 | ||
1986 | { | |
1987 | int retries = 0; | |
1988 | ||
1989 | retry: | |
a2b22788 | 1990 | if (access (aux_info_filename, R_OK) == -1) |
5f8037c4 RS |
1991 | { |
1992 | if (errno == ENOENT && retries == 0) | |
1993 | { | |
1994 | if (is_syscalls) | |
1995 | { | |
1996 | fprintf (stderr, "%s: warning: missing SYSCALLS file `%s'\n", | |
a2b22788 | 1997 | pname, aux_info_filename); |
5f8037c4 RS |
1998 | return; |
1999 | } | |
a2b22788 | 2000 | if (!gen_aux_info_file (base_source_filename)) |
5f8037c4 RS |
2001 | return; |
2002 | retries++; | |
2003 | goto retry; | |
2004 | } | |
2005 | else | |
2006 | { | |
2007 | fprintf (stderr, "%s: error: can't read aux info file `%s': %s\n", | |
a2b22788 RS |
2008 | pname, shortpath (NULL, aux_info_filename), |
2009 | sys_errlist[errno]); | |
5f8037c4 RS |
2010 | errors++; |
2011 | return; | |
2012 | } | |
2013 | } | |
2014 | } | |
2015 | ||
2016 | { | |
2017 | struct stat stat_buf; | |
2018 | ||
2019 | /* Get some status information about this aux_info file. */ | |
2020 | ||
a2b22788 | 2021 | if (stat (aux_info_filename, &stat_buf) == -1) |
5f8037c4 RS |
2022 | { |
2023 | fprintf (stderr, "%s: error: can't get status of aux info file `%s': %s\n", | |
a2b22788 RS |
2024 | pname, shortpath (NULL, aux_info_filename), |
2025 | sys_errlist[errno]); | |
5f8037c4 RS |
2026 | errors++; |
2027 | return; | |
2028 | } | |
2029 | ||
2030 | /* Check on whether or not this aux_info file is zero length. If it is, | |
2031 | then just ignore it and return. */ | |
2032 | ||
2033 | if ((aux_info_size = stat_buf.st_size) == 0) | |
2034 | return; | |
2035 | ||
2036 | /* Get the date/time of last modification for this aux_info file and | |
2037 | remember it. We will have to check that any source files that it | |
2038 | contains information about are at least this old or older. */ | |
2039 | ||
2040 | aux_info_mtime = stat_buf.st_mtime; | |
2041 | } | |
2042 | ||
2043 | { | |
2044 | int aux_info_file; | |
2045 | ||
2046 | /* Open the aux_info file. */ | |
2047 | ||
a2b22788 | 2048 | if ((aux_info_file = open (aux_info_filename, O_RDONLY, 0444 )) == -1) |
5f8037c4 RS |
2049 | { |
2050 | fprintf (stderr, "%s: error: can't open aux info file `%s' for reading: %s\n", | |
a2b22788 RS |
2051 | pname, shortpath (NULL, aux_info_filename), |
2052 | sys_errlist[errno]); | |
5f8037c4 RS |
2053 | return; |
2054 | } | |
2055 | ||
2056 | /* Allocate space to hold the aux_info file in memory. */ | |
2057 | ||
2058 | aux_info_base = xmalloc (aux_info_size + 1); | |
2059 | aux_info_limit = aux_info_base + aux_info_size; | |
2060 | *aux_info_limit = '\0'; | |
2061 | ||
2062 | /* Read the aux_info file into memory. */ | |
2063 | ||
2064 | if (read (aux_info_file, aux_info_base, aux_info_size) != aux_info_size) | |
2065 | { | |
2066 | fprintf (stderr, "%s: error: while reading aux info file `%s': %s\n", | |
a2b22788 RS |
2067 | pname, shortpath (NULL, aux_info_filename), |
2068 | sys_errlist[errno]); | |
5f8037c4 RS |
2069 | free (aux_info_base); |
2070 | close (aux_info_file); | |
2071 | return; | |
2072 | } | |
2073 | ||
2074 | /* Close the aux info file. */ | |
2075 | ||
2076 | if (close (aux_info_file)) | |
2077 | { | |
2078 | fprintf (stderr, "%s: error: while closing aux info file `%s': %s\n", | |
a2b22788 RS |
2079 | pname, shortpath (NULL, aux_info_filename), |
2080 | sys_errlist[errno]); | |
5f8037c4 RS |
2081 | free (aux_info_base); |
2082 | close (aux_info_file); | |
2083 | return; | |
2084 | } | |
2085 | } | |
2086 | ||
2087 | /* Delete the aux_info file (unless requested not to). If the deletion | |
2088 | fails for some reason, don't even worry about it. */ | |
2089 | ||
2090 | if (!keep_it) | |
a2b22788 | 2091 | if (unlink (aux_info_filename) == -1) |
5f8037c4 | 2092 | fprintf (stderr, "%s: error: can't delete aux info file `%s': %s\n", |
a2b22788 RS |
2093 | pname, shortpath (NULL, aux_info_filename), |
2094 | sys_errlist[errno]); | |
5f8037c4 RS |
2095 | |
2096 | /* Save a pointer into the first line of the aux_info file which | |
a2b22788 | 2097 | contains the filename of the directory from which the compiler |
5f8037c4 RS |
2098 | was invoked when the associated source file was compiled. |
2099 | This information is used later to help create complete | |
a2b22788 | 2100 | filenames out of the (potentially) relative filenames in |
5f8037c4 RS |
2101 | the aux_info file. */ |
2102 | ||
2103 | { | |
2104 | char *p = aux_info_base; | |
2105 | ||
2106 | while (*p != ':') | |
2107 | p++; | |
2108 | p++; | |
2109 | while (*p == ' ') | |
2110 | p++; | |
a2b22788 | 2111 | invocation_filename = p; /* Save a pointer to first byte of path. */ |
5f8037c4 RS |
2112 | while (*p != ' ') |
2113 | p++; | |
2114 | *p++ = '/'; | |
2115 | *p++ = '\0'; | |
2116 | while (*p++ != '\n') | |
2117 | continue; | |
2118 | aux_info_second_line = p; | |
2119 | } | |
2120 | ||
2121 | ||
2122 | { | |
2123 | const char *aux_info_p; | |
2124 | ||
2125 | /* Do a pre-pass on the lines in the aux_info file, making sure that all | |
2126 | of the source files referenced in there are at least as old as this | |
2127 | aux_info file itself. If not, go back and regenerate the aux_info | |
2128 | file anew. Don't do any of this for the syscalls file. */ | |
2129 | ||
2130 | if (!is_syscalls) | |
2131 | { | |
2132 | current_aux_info_lineno = 2; | |
2133 | ||
2134 | for (aux_info_p = aux_info_second_line; *aux_info_p; ) | |
2135 | { | |
2136 | if (referenced_file_is_newer (aux_info_p, aux_info_mtime)) | |
2137 | { | |
2138 | free (aux_info_base); | |
a2b22788 | 2139 | if (unlink (aux_info_filename) == -1) |
5f8037c4 RS |
2140 | { |
2141 | fprintf (stderr, "%s: error: can't delete file `%s': %s\n", | |
a2b22788 RS |
2142 | pname, shortpath (NULL, aux_info_filename), |
2143 | sys_errlist[errno]); | |
5f8037c4 RS |
2144 | return; |
2145 | } | |
2146 | goto start_over; | |
2147 | } | |
2148 | ||
2149 | /* Skip over the rest of this line to start of next line. */ | |
2150 | ||
2151 | while (*aux_info_p != '\n') | |
2152 | aux_info_p++; | |
2153 | aux_info_p++; | |
2154 | current_aux_info_lineno++; | |
2155 | } | |
2156 | } | |
2157 | ||
2158 | /* Now do the real pass on the aux_info lines. Save their information in | |
2159 | the in-core data base. */ | |
2160 | ||
2161 | current_aux_info_lineno = 2; | |
2162 | ||
2163 | for (aux_info_p = aux_info_second_line; *aux_info_p;) | |
2164 | { | |
2165 | char *unexpanded_line = unexpand_if_needed (aux_info_p); | |
2166 | ||
2167 | if (unexpanded_line) | |
2168 | { | |
2169 | save_def_or_dec (unexpanded_line, is_syscalls); | |
2170 | free (unexpanded_line); | |
2171 | } | |
2172 | else | |
2173 | save_def_or_dec (aux_info_p, is_syscalls); | |
2174 | ||
2175 | /* Skip over the rest of this line and get to start of next line. */ | |
2176 | ||
2177 | while (*aux_info_p != '\n') | |
2178 | aux_info_p++; | |
2179 | aux_info_p++; | |
2180 | current_aux_info_lineno++; | |
2181 | } | |
2182 | } | |
2183 | ||
2184 | free (aux_info_base); | |
2185 | } | |
2186 | \f | |
2187 | #ifndef UNPROTOIZE | |
2188 | ||
2189 | /* Check an individual filename for a .c suffix. If the filename has this | |
2190 | suffix, rename the file such that its suffix is changed to .C. This | |
2191 | function implements the -C option. */ | |
2192 | ||
2193 | static void | |
2194 | rename_c_file (const hash_table_entry *hp) | |
2195 | { | |
a2b22788 RS |
2196 | const char *filename = hp->symbol; |
2197 | int last_char_index = strlen (filename) - 1; | |
2198 | char *const new_filename = (char *) alloca (strlen (filename) + 1); | |
5f8037c4 RS |
2199 | |
2200 | /* Note that we don't care here if the given file was converted or not. It | |
2201 | is possible that the given file was *not* converted, simply because there | |
2202 | was nothing in it which actually required conversion. Even in this case, | |
2203 | we want to do the renaming. Note that we only rename files with the .c | |
2204 | suffix. */ | |
2205 | ||
a2b22788 | 2206 | if (filename[last_char_index] != 'c' || filename[last_char_index-1] != '.') |
5f8037c4 RS |
2207 | return; |
2208 | ||
a2b22788 RS |
2209 | strcpy (new_filename, filename); |
2210 | new_filename[last_char_index] = 'C'; | |
5f8037c4 | 2211 | |
a2b22788 | 2212 | if (link (filename, new_filename) == -1) |
5f8037c4 RS |
2213 | { |
2214 | fprintf (stderr, "%s: warning: can't link file `%s' to `%s': %s\n", | |
a2b22788 RS |
2215 | pname, shortpath (NULL, filename), |
2216 | shortpath (NULL, new_filename), sys_errlist[errno]); | |
5f8037c4 RS |
2217 | errors++; |
2218 | return; | |
2219 | } | |
2220 | ||
a2b22788 | 2221 | if (unlink (filename) == -1) |
5f8037c4 RS |
2222 | { |
2223 | fprintf (stderr, "%s: warning: can't delete file `%s': %s\n", | |
a2b22788 | 2224 | pname, shortpath (NULL, filename), sys_errlist[errno]); |
5f8037c4 RS |
2225 | errors++; |
2226 | return; | |
2227 | } | |
2228 | } | |
2229 | ||
2230 | #endif /* !defined(UNPROTOIZE) */ | |
2231 | \f | |
2232 | /* Take the list of definitions and declarations attached to a particular | |
2233 | file_info node and reverse the order of the list. This should get the | |
2234 | list into an order such that the item with the lowest associated line | |
2235 | number is nearest the head of the list. When these lists are originally | |
2236 | built, they are in the opposite order. We want to traverse them in | |
2237 | normal line number order later (i.e. lowest to highest) so reverse the | |
2238 | order here. */ | |
2239 | ||
2240 | static void | |
2241 | reverse_def_dec_list (const hash_table_entry *hp) | |
2242 | { | |
2243 | file_info *file_p = hp->fip; | |
2244 | const def_dec_info *prev = NULL; | |
2245 | const def_dec_info *current = file_p->defs_decs; | |
2246 | ||
2247 | if (!( current = file_p->defs_decs)) | |
2248 | return; /* no list to reverse */ | |
2249 | ||
2250 | prev = current; | |
2251 | if (! (current = current->next_in_file)) | |
2252 | return; /* can't reverse a single list element */ | |
2253 | ||
2254 | ((NONCONST def_dec_info *) prev)->next_in_file = NULL; | |
2255 | ||
2256 | while (current) | |
2257 | { | |
2258 | const def_dec_info *next = current->next_in_file; | |
2259 | ||
2260 | ((NONCONST def_dec_info *) current)->next_in_file = prev; | |
2261 | prev = current; | |
2262 | current = next; | |
2263 | } | |
2264 | ||
2265 | file_p->defs_decs = prev; | |
2266 | } | |
2267 | ||
2268 | #ifndef UNPROTOIZE | |
2269 | ||
2270 | /* Find the (only?) extern definition for a particular function name, starting | |
2271 | from the head of the linked list of entries for the given name. If we | |
2272 | cannot find an extern definition for the given function name, issue a | |
2273 | warning and scrounge around for the next best thing, i.e. an extern | |
2274 | function declaration with a prototype attached to it. Note that we only | |
2275 | allow such substitutions for extern declarations and never for static | |
2276 | declarations. That's because the only reason we allow them at all is | |
2277 | to let un-prototyped function declarations for system-supplied library | |
2278 | functions get their prototypes from our own extra SYSCALLS.c.X file which | |
2279 | contains all of the correct prototypes for system functions. */ | |
2280 | ||
2281 | static const def_dec_info * | |
2282 | find_extern_def (const def_dec_info *head, const def_dec_info *user) | |
2283 | { | |
2284 | const def_dec_info *dd_p; | |
2285 | const def_dec_info *extern_def_p = NULL; | |
2286 | int conflict_noted = 0; | |
2287 | ||
2288 | /* Don't act too stupid here. Somebody may try to convert an entire system | |
2289 | in one swell fwoop (rather than one program at a time, as should be done) | |
2290 | and in that case, we may find that there are multiple extern definitions | |
2291 | of a given function name in the entire set of source files that we are | |
2292 | converting. If however one of these definitions resides in exactly the | |
2293 | same source file as the reference we are trying to satisfy then in that | |
2294 | case it would be stupid for us to fail to realize that this one definition | |
2295 | *must* be the precise one we are looking for. | |
2296 | ||
2297 | To make sure that we don't miss an opportunity to make this "same file" | |
2298 | leap of faith, we do a prescan of the list of records relating to the | |
2299 | given function name, and we look (on this first scan) *only* for a | |
2300 | definition of the function which is in the same file as the reference | |
2301 | we are currently trying to satisfy. */ | |
2302 | ||
2303 | for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) | |
2304 | if (dd_p->is_func_def && !dd_p->is_static && dd_p->file == user->file) | |
2305 | return dd_p; | |
2306 | ||
2307 | /* Now, since we have not found a definition in the same file as the | |
2308 | reference, we scan the list again and consider all possibilities from | |
2309 | all files. Here we may get conflicts with the things listed in the | |
2310 | SYSCALLS.c.X file, but if that happens it only means that the source | |
2311 | code being converted contains its own definition of a function which | |
2312 | could have been supplied by libc.a. In such cases, we should avoid | |
2313 | issuing the normal warning, and defer to the definition given in the | |
2314 | user's own code. */ | |
2315 | ||
2316 | for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) | |
2317 | if (dd_p->is_func_def && !dd_p->is_static) | |
2318 | { | |
2319 | if (!extern_def_p) /* Previous definition? */ | |
2320 | extern_def_p = dd_p; /* Remember the first definition found. */ | |
2321 | else | |
2322 | { | |
2323 | /* Ignore definition just found if it came from SYSCALLS.c.X. */ | |
2324 | ||
2325 | if (is_syscalls_file (dd_p->file)) | |
2326 | continue; | |
2327 | ||
2328 | /* Quietly replace the definition previously found with the one | |
2329 | just found if the previous one was from SYSCALLS.c.X. */ | |
2330 | ||
2331 | if (is_syscalls_file (extern_def_p->file)) | |
2332 | { | |
2333 | extern_def_p = dd_p; | |
2334 | continue; | |
2335 | } | |
2336 | ||
2337 | /* If we get here, then there is a conflict between two function | |
2338 | declarations for the same function, both of which came from the | |
2339 | user's own code. */ | |
2340 | ||
2341 | if (!conflict_noted) /* first time we noticed? */ | |
2342 | { | |
2343 | conflict_noted = 1; | |
2344 | fprintf (stderr, "%s: error: conflicting extern definitions of '%s'\n", | |
a2b22788 | 2345 | pname, head->hash_entry->symbol); |
5f8037c4 RS |
2346 | if (!quiet_flag) |
2347 | { | |
2348 | fprintf (stderr, "%s: declarations of '%s' will not be converted\n", | |
a2b22788 | 2349 | pname, head->hash_entry->symbol); |
5f8037c4 | 2350 | fprintf (stderr, "%s: conflict list for '%s' follows:\n", |
a2b22788 | 2351 | pname, head->hash_entry->symbol); |
5f8037c4 | 2352 | fprintf (stderr, "%s: %s(%d): %s\n", |
a2b22788 RS |
2353 | pname, |
2354 | shortpath (NULL, extern_def_p->file->hash_entry->symbol), | |
2355 | extern_def_p->line, extern_def_p->ansi_decl); | |
5f8037c4 RS |
2356 | } |
2357 | } | |
2358 | if (!quiet_flag) | |
2359 | fprintf (stderr, "%s: %s(%d): %s\n", | |
a2b22788 RS |
2360 | pname, |
2361 | shortpath (NULL, dd_p->file->hash_entry->symbol), | |
2362 | dd_p->line, dd_p->ansi_decl); | |
5f8037c4 RS |
2363 | } |
2364 | } | |
2365 | ||
2366 | /* We want to err on the side of caution, so if we found multiple conflicting | |
2367 | definitions for the same function, treat this as being that same as if we | |
2368 | had found no definitions (i.e. return NULL). */ | |
2369 | ||
2370 | if (conflict_noted) | |
2371 | return NULL; | |
2372 | ||
2373 | if (!extern_def_p) | |
2374 | { | |
2375 | /* We have no definitions for this function so do the next best thing. | |
2376 | Search for an extern declaration already in prototype form. */ | |
2377 | ||
2378 | for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) | |
2379 | if (!dd_p->is_func_def && !dd_p->is_static && dd_p->prototyped) | |
2380 | { | |
2381 | extern_def_p = dd_p; /* save a pointer to the definition */ | |
2382 | if (!quiet_flag) | |
2383 | fprintf (stderr, "%s: warning: using formals list from %s(%d) for function `%s'\n", | |
a2b22788 RS |
2384 | pname, |
2385 | shortpath (NULL, dd_p->file->hash_entry->symbol), | |
2386 | dd_p->line, dd_p->hash_entry->symbol); | |
5f8037c4 RS |
2387 | break; |
2388 | } | |
2389 | ||
2390 | /* Gripe about unprototyped function declarations that we found no | |
2391 | corresponding definition (or other source of prototype information) | |
2392 | for. | |
2393 | ||
2394 | Gripe even if the unprototyped declaration we are worried about | |
2395 | exists in a file in one of the "system" include directories. We | |
2396 | can gripe about these because we should have at least found a | |
2397 | corresponding (pseudo) definition in the SYSCALLS.c.X file. If we | |
2398 | didn't, then that means that the SYSCALLS.c.X file is missing some | |
2399 | needed prototypes for this particular system. That is worth telling | |
2400 | the user about! */ | |
2401 | ||
2402 | if (!extern_def_p) | |
2403 | { | |
2404 | const char *file = user->file->hash_entry->symbol; | |
2405 | ||
2406 | if (!quiet_flag) | |
2407 | if (in_system_include_dir (file)) | |
2408 | { | |
2409 | /* Why copy this string into `needed' at all? | |
2410 | Why not just use user->ansi_decl without copying? */ | |
2411 | char needed[strlen (user->ansi_decl) + 1]; | |
2412 | char *p; | |
2413 | ||
2414 | strcpy (needed, user->ansi_decl); | |
2415 | p = (NONCONST char *) substr (needed, user->hash_entry->symbol) | |
2416 | + strlen (user->hash_entry->symbol) + 2; | |
2417 | strcpy (p, "???);"); | |
2418 | ||
a2b22788 RS |
2419 | fprintf (stderr, "%s: %d: `%s' used but missing from SYSCALLS\n", |
2420 | shortpath (NULL, file), user->line, | |
2421 | needed+7); /* Don't print "extern " */ | |
5f8037c4 RS |
2422 | } |
2423 | else | |
a2b22788 RS |
2424 | fprintf (stderr, "%s: %d: warning: no extern definition for `%s'\n", |
2425 | shortpath (NULL, file), user->line, | |
2426 | user->hash_entry->symbol); | |
5f8037c4 RS |
2427 | } |
2428 | } | |
2429 | return extern_def_p; | |
2430 | } | |
2431 | \f | |
2432 | /* Find the (only?) static definition for a particular function name in a | |
2433 | given file. Here we get the function-name and the file info indirectly | |
2434 | from the def_dec_info record pointer which is passed in. */ | |
2435 | ||
2436 | static const def_dec_info * | |
2437 | find_static_definition (const def_dec_info *user) | |
2438 | { | |
2439 | const def_dec_info *head = user->hash_entry->ddip; | |
2440 | const def_dec_info *dd_p; | |
2441 | int num_static_defs = 0; | |
2442 | const def_dec_info *static_def_p = NULL; | |
2443 | ||
2444 | for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) | |
2445 | if (dd_p->is_func_def && dd_p->is_static && (dd_p->file == user->file)) | |
2446 | { | |
2447 | static_def_p = dd_p; /* save a pointer to the definition */ | |
2448 | num_static_defs++; | |
2449 | } | |
2450 | if (num_static_defs == 0) | |
2451 | { | |
2452 | if (!quiet_flag) | |
2453 | fprintf (stderr, "%s: warning: no static definition for `%s' in file `%s'\n", | |
a2b22788 RS |
2454 | pname, head->hash_entry->symbol, |
2455 | shortpath (NULL, user->file->hash_entry->symbol)); | |
5f8037c4 RS |
2456 | } |
2457 | else if (num_static_defs > 1) | |
2458 | { | |
2459 | fprintf (stderr, "%s: error: multiple static defs of `%s' in file `%s'\n", | |
a2b22788 RS |
2460 | pname, head->hash_entry->symbol, |
2461 | shortpath (NULL, user->file->hash_entry->symbol)); | |
5f8037c4 RS |
2462 | return NULL; |
2463 | } | |
2464 | return static_def_p; | |
2465 | } | |
2466 | ||
2467 | /* Find good prototype style formal argument lists for all of the function | |
2468 | declarations which didn't have them before now. | |
2469 | ||
2470 | To do this we consider each function name one at a time. For each function | |
2471 | name, we look at the items on the linked list of def_dec_info records for | |
2472 | that particular name. | |
2473 | ||
2474 | Somewhere on this list we should find one (and only one) def_dec_info | |
2475 | record which represents the actual function definition, and this record | |
2476 | should have a nice formal argument list already associated with it. | |
2477 | ||
2478 | Thus, all we have to do is to connect up all of the other def_dec_info | |
2479 | records for this particular function name to the special one which has | |
2480 | the full-blown formals list. | |
2481 | ||
2482 | Of course it is a little more complicated than just that. See below for | |
2483 | more details. */ | |
2484 | ||
2485 | static void | |
2486 | connect_defs_and_decs (const hash_table_entry *hp) | |
2487 | { | |
2488 | const def_dec_info *dd_p; | |
2489 | const def_dec_info *extern_def_p = NULL; | |
2490 | int first_extern_reference = 1; | |
2491 | ||
2492 | /* Traverse the list of definitions and declarations for this particular | |
2493 | function name. For each item on the list, if it is a function | |
2494 | definition (either old style or new style) then GCC has already been | |
2495 | kind enough to produce a prototype for us, and it is associated with | |
2496 | the item already, so declare the item as its own associated "definition". | |
2497 | ||
2498 | Also, for each item which is only a function declaration, but which | |
2499 | nonetheless has its own prototype already (obviously supplied by the user) | |
2500 | declare the item as it's own definition. | |
2501 | ||
2502 | Note that when/if there are multiple user-supplied prototypes already | |
2503 | present for multiple declarations of any given function, these multiple | |
2504 | prototypes *should* all match exactly with one another and with the | |
2505 | prototype for the actual function definition. We don't check for this | |
2506 | here however, since we assume that the compiler must have already done | |
2507 | this consistancy checking when it was creating the .X files. */ | |
2508 | ||
2509 | for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) | |
2510 | if (dd_p->prototyped) | |
2511 | ((NONCONST def_dec_info *) dd_p)->definition = dd_p; | |
2512 | ||
2513 | /* Traverse the list of definitions and declarations for this particular | |
2514 | function name. For each item on the list, if it is an extern function | |
2515 | declaration and if it has no associated definition yet, go try to find | |
2516 | the matching extern definition for the declaration. | |
2517 | ||
2518 | When looking for the matching function definition, warn the user if we | |
2519 | fail to find one. | |
2520 | ||
2521 | If we find more that one function definition also issue a warning. | |
2522 | ||
2523 | Do the search for the matching definition only once per unique function | |
2524 | name (and only when absolutely needed) so that we can avoid putting out | |
2525 | redundant warning messages, and so that we will only put out warning | |
2526 | messages when there is actually a reference (i.e. a declaration) for | |
2527 | which we need to find a matching definition. */ | |
2528 | ||
2529 | for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) | |
2530 | if (!dd_p->is_func_def && !dd_p->is_static && !dd_p->definition) | |
2531 | { | |
2532 | if (first_extern_reference) | |
2533 | { | |
2534 | extern_def_p = find_extern_def (hp->ddip, dd_p); | |
2535 | first_extern_reference = 0; | |
2536 | } | |
2537 | ((NONCONST def_dec_info *) dd_p)->definition = extern_def_p; | |
2538 | } | |
2539 | ||
2540 | /* Traverse the list of definitions and declarations for this particular | |
2541 | function name. For each item on the list, if it is a static function | |
2542 | declaration and if it has no associated definition yet, go try to find | |
2543 | the matching static definition for the declaration within the same file. | |
2544 | ||
2545 | When looking for the matching function definition, warn the user if we | |
2546 | fail to find one in the same file with the declaration, and refuse to | |
2547 | convert this kind of cross-file static function declaration. After all, | |
2548 | this is stupid practice and should be discouraged. | |
2549 | ||
2550 | We don't have to worry about the possibility that there is more than one | |
2551 | matching function definition in the given file because that would have | |
2552 | been flagged as an error by the compiler. | |
2553 | ||
2554 | Do the search for the matching definition only once per unique | |
2555 | function-name/source-file pair (and only when absolutely needed) so that | |
2556 | we can avoid putting out redundant warning messages, and so that we will | |
2557 | only put out warning messages when there is actually a reference (i.e. a | |
2558 | declaration) for which we actually need to find a matching definition. */ | |
2559 | ||
2560 | for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) | |
2561 | if (!dd_p->is_func_def && dd_p->is_static && !dd_p->definition) | |
2562 | { | |
2563 | const def_dec_info *dd_p2; | |
2564 | const def_dec_info *static_def; | |
2565 | ||
2566 | /* We have now found a single static declaration for which we need to | |
2567 | find a matching definition. We want to minimize the work (and the | |
2568 | number of warnings), so we will find an appropriate (matching) | |
2569 | static definition for this declaration, and then distribute it | |
2570 | (as the definition for) any and all other static declarations | |
2571 | for this function name which occur within the same file, and which | |
2572 | do not already have definitions. | |
2573 | ||
2574 | Note that a trick is used here to prevent subsequent attempts to | |
2575 | call find_static_definition() for a given function-name & file | |
2576 | if the first such call returns NULL. Essentially, we convert | |
2577 | these NULL return values to -1, and put the -1 into the definition | |
2578 | field for each other static declaration from the same file which | |
2579 | does not already have an associated definition. | |
2580 | This makes these other static declarations look like they are | |
2581 | actually defined already when the outer loop here revisits them | |
2582 | later on. Thus, the outer loop will skip over them. Later, we | |
2583 | turn the -1's back to NULL's. */ | |
2584 | ||
2585 | ((NONCONST def_dec_info *) dd_p)->definition = | |
2586 | (static_def = find_static_definition (dd_p)) | |
2587 | ? static_def | |
2588 | : (const def_dec_info *) -1; | |
2589 | ||
2590 | for (dd_p2 = dd_p->next_for_func; dd_p2; dd_p2 = dd_p2->next_for_func) | |
2591 | if (!dd_p2->is_func_def && dd_p2->is_static | |
2592 | && !dd_p2->definition && (dd_p2->file == dd_p->file)) | |
2593 | ((NONCONST def_dec_info *)dd_p2)->definition = dd_p->definition; | |
2594 | } | |
2595 | ||
2596 | /* Convert any dummy (-1) definitions we created in the step above back to | |
2597 | NULL's (as they should be). */ | |
2598 | ||
2599 | for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) | |
2600 | if (dd_p->definition == (def_dec_info *) -1) | |
2601 | ((NONCONST def_dec_info *) dd_p)->definition = NULL; | |
2602 | } | |
2603 | ||
2604 | #endif /* !defined(UNPROTOIZE) */ | |
2605 | ||
2606 | /* Give a pointer into the clean text buffer, return a number which is the | |
2607 | original source line number that the given pointer points into. */ | |
2608 | ||
2609 | static int | |
2610 | identify_lineno (const char *clean_p) | |
2611 | { | |
2612 | int line_num = 1; | |
2613 | const char *scan_p; | |
2614 | ||
2615 | for (scan_p = clean_text_base; scan_p <= clean_p; scan_p++) | |
2616 | if (*scan_p == '\n') | |
2617 | line_num++; | |
2618 | return line_num; | |
2619 | } | |
2620 | ||
2621 | /* Issue an error message and give up on doing this particular edit. */ | |
2622 | ||
2623 | static void | |
2624 | declare_source_confusing (const char *clean_p) | |
2625 | { | |
2626 | if (!quiet_flag) | |
2627 | { | |
2628 | if (clean_p == 0) | |
a2b22788 RS |
2629 | fprintf (stderr, "%s: %d: warning: source too confusing\n", |
2630 | shortpath (NULL, convert_filename), last_known_line_number); | |
5f8037c4 | 2631 | else |
a2b22788 RS |
2632 | fprintf (stderr, "%s: %d: warning: source too confusing\n", |
2633 | shortpath (NULL, convert_filename), | |
2634 | identify_lineno (clean_p)); | |
5f8037c4 RS |
2635 | } |
2636 | longjmp (source_confusion_recovery, 1); | |
2637 | } | |
2638 | ||
2639 | /* Check that a condition which is expected to be true in the original source | |
2640 | code is in fact true. If not, issue an error message and give up on | |
2641 | converting this particular source file. */ | |
2642 | ||
2643 | inline static void | |
2644 | check_source (int cond, const char *clean_p) | |
2645 | { | |
2646 | if (!cond) | |
2647 | declare_source_confusing (clean_p); | |
2648 | } | |
2649 | ||
2650 | /* If we think of the in-core cleaned text buffer as a memory mapped | |
2651 | file (with the variable last_known_line_start acting as sort of a | |
2652 | file pointer) then we can imagine doing "seeks" on the buffer. The | |
2653 | following routine implements a kind of "seek" operation for the in-core | |
2654 | (cleaned) copy of the source file. When finished, it returns a pointer to | |
2655 | the start of a given (numbered) line in the cleaned text buffer. | |
2656 | ||
2657 | Note that protoize only has to "seek" in the forward direction on the | |
2658 | in-core cleaned text file buffers, and it never needs to back up. | |
2659 | ||
2660 | This routine is made a little bit faster by remembering the line number | |
2661 | (and pointer value) supplied (and returned) from the previous "seek". | |
2662 | This prevents us from always having to start all over back at the top | |
2663 | of the in-core cleaned buffer again. */ | |
2664 | ||
2665 | static const char * | |
2666 | seek_to_line (int n) | |
2667 | { | |
2668 | if (n < last_known_line_number) | |
2669 | abort (); | |
2670 | ||
2671 | while (n > last_known_line_number) | |
2672 | { | |
2673 | while (*last_known_line_start != '\n') | |
2674 | check_source (++last_known_line_start < clean_text_limit, 0); | |
2675 | last_known_line_start++; | |
2676 | last_known_line_number++; | |
2677 | } | |
2678 | return last_known_line_start; | |
2679 | } | |
2680 | ||
2681 | /* Given a pointer to a character in the cleaned text buffer, return a pointer | |
2682 | to the next non-whitepace character which follows it. */ | |
2683 | ||
2684 | static const char * | |
2685 | forward_to_next_token_char (const char *ptr) | |
2686 | { | |
2687 | for (++ptr; isspace (*ptr); check_source (++ptr < clean_text_limit, 0)) | |
2688 | continue; | |
2689 | return ptr; | |
2690 | } | |
2691 | ||
2692 | /* Copy a chunk of text of length `len' and starting at `str' to the current | |
2693 | output buffer. Note that all attempts to add stuff to the current output | |
2694 | buffer ultimately go through here. */ | |
2695 | ||
2696 | static void | |
2697 | output_bytes (const char *str, size_t len) | |
2698 | { | |
2699 | if ((repl_write_ptr + 1) + len >= repl_text_limit) | |
2700 | { | |
2701 | size_t new_size = (repl_text_limit - repl_text_base) << 1; | |
2702 | char *new_buf = (char *) xrealloc (repl_text_base, new_size); | |
2703 | ||
2704 | repl_write_ptr = new_buf + (repl_write_ptr - repl_text_base); | |
2705 | repl_text_base = new_buf; | |
2706 | repl_text_limit = new_buf + new_size; | |
2707 | } | |
2708 | memcpy (repl_write_ptr + 1, str, len); | |
2709 | repl_write_ptr += len; | |
2710 | } | |
2711 | ||
2712 | /* Copy all bytes (except the trailing null) of a null terminated string to | |
2713 | the current output buffer. */ | |
2714 | ||
2715 | static void | |
2716 | output_string (const char *str) | |
2717 | { | |
2718 | output_bytes (str, strlen (str)); | |
2719 | } | |
2720 | ||
2721 | /* Copy some characters from the original text buffer to the current output | |
2722 | buffer. | |
2723 | ||
2724 | This routine takes a pointer argument `p' which is assumed to be a pointer | |
2725 | into the cleaned text buffer. The bytes which are copied are the `original' | |
2726 | equivalents for the set of bytes between the last value of `clean_read_ptr' | |
2727 | and the argument value `p'. | |
2728 | ||
2729 | The set of bytes copied however, comes *not* from the cleaned text buffer, | |
2730 | but rather from the direct counterparts of these bytes within the original | |
2731 | text buffer. | |
2732 | ||
2733 | Thus, when this function is called, some bytes from the original text | |
2734 | buffer (which may include original comments and preprocessing directives) | |
2735 | will be copied into the output buffer. | |
2736 | ||
2737 | Note that the request implide when this routine is called includes the | |
2738 | byte pointed to by the argument pointer `p'. */ | |
2739 | ||
2740 | static void | |
2741 | output_up_to (const char *p) | |
2742 | { | |
2743 | size_t copy_length = (size_t) (p - clean_read_ptr); | |
2744 | const char *copy_start = orig_text_base+(clean_read_ptr-clean_text_base)+1; | |
2745 | ||
2746 | if (copy_length == 0) | |
2747 | return; | |
2748 | ||
2749 | output_bytes (copy_start, copy_length); | |
2750 | clean_read_ptr = p; | |
2751 | } | |
2752 | ||
2753 | /* Given a pointer to a def_dec_info record which represents some form of | |
2754 | definition of a function (perhaps a real definition, or in lieu of that | |
2755 | perhaps just a declaration with a full prototype) return true if this | |
2756 | function is one which we should avoid converting. Return false | |
2757 | otherwise. */ | |
2758 | ||
2759 | static int | |
2760 | other_variable_style_function (const char *ansi_header) | |
2761 | { | |
2762 | #ifdef UNPROTOIZE | |
2763 | ||
2764 | /* See if we have a stdarg function, or a function which has stdarg style | |
2765 | parameters or a stdarg style return type. */ | |
2766 | ||
2767 | return (int) substr (ansi_header, "..."); | |
2768 | ||
2769 | #else /* !defined(UNPROTOIZE) */ | |
2770 | ||
2771 | /* See if we have a varargs function, or a function which has varargs style | |
2772 | parameters or a varargs style return type. */ | |
2773 | ||
2774 | const char *p; | |
2775 | int len = strlen (varargs_style_indicator); | |
2776 | ||
2777 | for (p = ansi_header; p; ) | |
2778 | { | |
2779 | const char *candidate; | |
2780 | ||
2781 | if ((candidate = substr (p, varargs_style_indicator)) == 0) | |
2782 | return 0; | |
2783 | else | |
2784 | if (!is_id_char (candidate[-1]) && !is_id_char (candidate[len])) | |
2785 | return 1; | |
2786 | else | |
2787 | p = candidate + 1; | |
2788 | } | |
2789 | return 0; | |
2790 | #endif /* !defined(UNPROTOIZE) */ | |
2791 | } | |
2792 | ||
2793 | /* Do the editing operation specifically for a function "declaration". Note | |
2794 | that editing for function "definitions" are handled in a separate routine | |
2795 | below. */ | |
2796 | ||
2797 | static void | |
2798 | edit_fn_declaration (const def_dec_info *def_dec_p, | |
2799 | const char *volatile clean_text_p) | |
2800 | { | |
2801 | const char *start_formals; | |
2802 | const char *end_formals; | |
2803 | const char *function_to_edit = def_dec_p->hash_entry->symbol; | |
2804 | size_t func_name_len = strlen (function_to_edit); | |
2805 | const char *end_of_fn_name; | |
2806 | ||
2807 | #ifndef UNPROTOIZE | |
2808 | ||
2809 | const f_list_chain_item *this_f_list_chain_item; | |
2810 | const def_dec_info *definition = def_dec_p->definition; | |
2811 | ||
2812 | /* If we are protoizing, and if we found no corresponding definition for | |
2813 | this particular function declaration, then just leave this declaration | |
2814 | exactly as it is. */ | |
2815 | ||
2816 | if (!definition) | |
2817 | return; | |
2818 | ||
2819 | /* If we are protoizing, and if the corresponding definition that we found | |
2820 | for this particular function declaration defined an old style varargs | |
2821 | function, then we want to issue a warning and just leave this function | |
2822 | declaration unconverted. */ | |
2823 | ||
2824 | if (other_variable_style_function (definition->ansi_decl)) | |
2825 | { | |
2826 | if (!quiet_flag) | |
a2b22788 RS |
2827 | fprintf (stderr, "%s: %d: warning: varargs function declaration not converted\n", |
2828 | shortpath (NULL, def_dec_p->file->hash_entry->symbol), | |
2829 | def_dec_p->line); | |
5f8037c4 RS |
2830 | return; |
2831 | } | |
2832 | ||
2833 | #endif /* !defined(UNPROTOIZE) */ | |
2834 | ||
2835 | /* Setup here to recover from confusing source code detected during this | |
2836 | particular "edit". */ | |
2837 | ||
2838 | save_pointers (); | |
2839 | if (setjmp (source_confusion_recovery)) | |
2840 | { | |
2841 | restore_pointers (); | |
2842 | fprintf (stderr, "%s: declaration of function `%s' not converted\n", | |
a2b22788 | 2843 | pname, function_to_edit); |
5f8037c4 RS |
2844 | return; |
2845 | } | |
2846 | ||
2847 | /* We are editing a function declaration. The line number we did a seek to | |
2848 | contains the comma or semicolon which follows the declaration. Our job | |
2849 | now is to scan backwards looking for the function name. This name *must* | |
2850 | be followed by open paren (ignoring whitespace, of course). We need to | |
2851 | replace everything between that open paren and the corresponding closing | |
2852 | paren. If we are protoizing, we need to insert the prototype-style | |
2853 | formals lists. If we are unprotoizing, we need to just delete everything | |
2854 | between the pairs of opening and closing parens. */ | |
2855 | ||
2856 | /* First move up to the end of the line. */ | |
2857 | ||
2858 | while (*clean_text_p != '\n') | |
2859 | check_source (++clean_text_p < clean_text_limit, 0); | |
2860 | clean_text_p--; /* Point to just before the newline character. */ | |
2861 | ||
2862 | /* Now we can scan backwards for the function name. */ | |
2863 | ||
2864 | do | |
2865 | { | |
2866 | for (;;) | |
2867 | { | |
2868 | /* Scan leftwards until we find some character which can be | |
2869 | part of an identifier. */ | |
2870 | ||
2871 | while (!is_id_char (*clean_text_p)) | |
2872 | check_source (--clean_text_p > clean_read_ptr, 0); | |
2873 | ||
2874 | /* Scan backwards until we find a char that cannot be part of an | |
2875 | identifier. */ | |
2876 | ||
2877 | while (is_id_char (*clean_text_p)) | |
2878 | check_source (--clean_text_p > clean_read_ptr, 0); | |
2879 | ||
2880 | /* Having found an "id break", see if the following id is the one | |
2881 | that we are looking for. If so, then exit from this loop. */ | |
2882 | ||
2883 | if (!strncmp (clean_text_p+1, function_to_edit, func_name_len)) | |
2884 | { | |
2885 | char ch = *(clean_text_p + 1 + func_name_len); | |
2886 | ||
2887 | /* Must also check to see that the name in the source text | |
2888 | ends where it should (in order to prevent bogus matches | |
2889 | on similar but longer identifiers. */ | |
2890 | ||
2891 | if (! is_id_char (ch)) | |
2892 | break; /* exit from loop */ | |
2893 | } | |
2894 | } | |
2895 | ||
2896 | /* We have now found the first perfect match for the function name in | |
2897 | our backward search. This may or may not be the actual function | |
2898 | name at the start of the actual function declaration (i.e. we could | |
2899 | have easily been mislead). We will try to avoid getting fooled too | |
2900 | often by looking forward for the open paren which should follow the | |
2901 | identifier we just found. We ignore whitespace while hunting. If | |
2902 | the next non-whitespace byte we see is *not* an open left paren, | |
2903 | then we must assume that we have been fooled and we start over | |
2904 | again accordingly. Note that there is no guarrantee, that even if | |
2905 | we do see the open paren, that we are in the right place. | |
2906 | Programmers do the strangest things sometimes! */ | |
2907 | ||
2908 | end_of_fn_name = clean_text_p + strlen (def_dec_p->hash_entry->symbol); | |
2909 | start_formals = forward_to_next_token_char (end_of_fn_name); | |
2910 | } | |
2911 | while (*start_formals != '('); | |
2912 | ||
2913 | /* start_of_formals now points to the opening left paren which immediately | |
2914 | follows the name of the function. */ | |
2915 | ||
2916 | /* Note that there may be several formals lists which need to be modified | |
2917 | due to the possibility that the return type of this function is a | |
2918 | pointer-to-function type. If there are several formals lists, we | |
2919 | convert them in left-to-right order here. */ | |
2920 | ||
2921 | #ifndef UNPROTOIZE | |
2922 | this_f_list_chain_item = definition->f_list_chain; | |
2923 | #endif /* !defined(UNPROTOIZE) */ | |
2924 | ||
2925 | for (;;) | |
2926 | { | |
2927 | { | |
2928 | int depth; | |
2929 | ||
2930 | end_formals = start_formals + 1; | |
2931 | depth = 1; | |
2932 | for (; depth; check_source (++end_formals < clean_text_limit, 0)) | |
2933 | { | |
2934 | switch (*end_formals) | |
2935 | { | |
2936 | case '(': | |
2937 | depth++; | |
2938 | break; | |
2939 | case ')': | |
2940 | depth--; | |
2941 | break; | |
2942 | } | |
2943 | } | |
2944 | end_formals--; | |
2945 | } | |
2946 | ||
2947 | /* end_formals now points to the closing right paren of the formals | |
2948 | list whose left paren is pointed to by start_formals. */ | |
2949 | ||
2950 | /* Now, if we are protoizing, we insert the new ANSI-style formals list | |
2951 | attached to the associated definition of this function. If however | |
2952 | we are unprotoizing, then we simply delete any formals list which | |
2953 | may be present. */ | |
2954 | ||
2955 | output_up_to (start_formals); | |
2956 | #ifndef UNPROTOIZE | |
2957 | if (this_f_list_chain_item) | |
2958 | { | |
2959 | output_string (this_f_list_chain_item->formals_list); | |
2960 | this_f_list_chain_item = this_f_list_chain_item->chain_next; | |
2961 | } | |
2962 | else | |
2963 | { | |
2964 | if (!quiet_flag) | |
2965 | fprintf (stderr, "%s: warning: too many parameter lists in declaration of `%s'\n", | |
a2b22788 | 2966 | pname, def_dec_p->hash_entry->symbol); |
5f8037c4 RS |
2967 | check_source (0, end_formals); /* leave the declaration intact */ |
2968 | } | |
2969 | #endif /* !defined(UNPROTOIZE) */ | |
2970 | clean_read_ptr = end_formals - 1; | |
2971 | ||
2972 | /* Now see if it looks like there may be another formals list associated | |
2973 | with the function declaration that we are converting (following the | |
2974 | formals list that we just converted. */ | |
2975 | ||
2976 | { | |
2977 | const char *another_r_paren = forward_to_next_token_char (end_formals); | |
2978 | ||
2979 | if ((*another_r_paren != ')') | |
2980 | || (*(start_formals = forward_to_next_token_char (another_r_paren)) != '(')) | |
2981 | { | |
2982 | #ifndef UNPROTOIZE | |
2983 | if (this_f_list_chain_item) | |
2984 | { | |
2985 | if (!quiet_flag) | |
2986 | fprintf (stderr, "\n%s: warning: too few parameter lists in declaration of `%s'\n", | |
a2b22788 | 2987 | pname, def_dec_p->hash_entry->symbol); |
5f8037c4 RS |
2988 | check_source (0, start_formals); /* leave the decl intact */ |
2989 | } | |
2990 | #endif /* !defined(UNPROTOIZE) */ | |
2991 | break; | |
2992 | ||
2993 | } | |
2994 | } | |
2995 | ||
2996 | /* There does appear to be yet another formals list, so loop around | |
2997 | again, and convert it also. */ | |
2998 | } | |
2999 | } | |
3000 | ||
3001 | /* Edit a whole group of formals lists, starting with the rightmost one | |
3002 | from some set of formals lists. This routine is called once (from the | |
3003 | outside) for each function declaration which is converted. It is | |
3004 | recursive however, and it calls itself once for each remaining formal | |
3005 | list that lies to the left of the one it was originally called to work | |
3006 | on. Thus, a whole set gets done in right-to-left order. | |
3007 | ||
3008 | This routine returns non-zero if it thinks that it should not be trying | |
3009 | to convert this particular function definition (because the name of the | |
3010 | function doesn't match the one expected). */ | |
3011 | ||
3012 | static int | |
3013 | edit_formals_lists (const char *end_formals, unsigned f_list_count, const def_dec_info *def_dec_p) | |
3014 | { | |
3015 | const char *start_formals; | |
3016 | int depth; | |
3017 | ||
3018 | start_formals = end_formals - 1; | |
3019 | depth = 1; | |
3020 | for (; depth; check_source (--start_formals > clean_read_ptr, 0)) | |
3021 | { | |
3022 | switch (*start_formals) | |
3023 | { | |
3024 | case '(': | |
3025 | depth--; | |
3026 | break; | |
3027 | case ')': | |
3028 | depth++; | |
3029 | break; | |
3030 | } | |
3031 | } | |
3032 | start_formals++; | |
3033 | ||
3034 | /* start_formals now points to the opening left paren of the formals list. */ | |
3035 | ||
3036 | f_list_count--; | |
3037 | ||
3038 | if (f_list_count) | |
3039 | { | |
3040 | const char *next_end; | |
3041 | ||
3042 | /* There should be more formal lists to the left of here. */ | |
3043 | ||
3044 | next_end = start_formals - 1; | |
3045 | check_source (next_end > clean_read_ptr, 0); | |
3046 | while (isspace (*next_end)) | |
3047 | check_source (--next_end > clean_read_ptr, 0); | |
3048 | check_source (*next_end == ')', next_end); | |
3049 | check_source (--next_end > clean_read_ptr, 0); | |
3050 | check_source (*next_end == ')', next_end); | |
3051 | if (edit_formals_lists (next_end, f_list_count, def_dec_p)) | |
3052 | return 1; | |
3053 | } | |
3054 | ||
3055 | /* Check that the function name in the header we are working on is the same | |
3056 | as the one we would expect to find. If not, issue a warning and return | |
3057 | non-zero. */ | |
3058 | ||
3059 | if (f_list_count == 0) | |
3060 | { | |
3061 | const char *expected = def_dec_p->hash_entry->symbol; | |
3062 | const char *func_name_start; | |
3063 | const char *func_name_limit; | |
3064 | size_t func_name_len; | |
3065 | ||
3066 | for (func_name_limit = start_formals-1; isspace (*func_name_limit); ) | |
3067 | check_source (--func_name_limit > clean_read_ptr, 0); | |
3068 | ||
3069 | for (func_name_start = func_name_limit++; | |
3070 | is_id_char (*func_name_start); | |
3071 | func_name_start--) | |
3072 | check_source (func_name_start > clean_read_ptr, 0); | |
3073 | func_name_start++; | |
3074 | func_name_len = func_name_limit - func_name_start; | |
3075 | if (func_name_len == 0) | |
3076 | check_source (0, func_name_start); | |
3077 | if (func_name_len != strlen (expected) | |
a2b22788 | 3078 | || strncmp (func_name_start, expected, func_name_len)) |
5f8037c4 | 3079 | { |
a2b22788 RS |
3080 | fprintf (stderr, "%s: %d: warning: found `%s' but expected `%s'\n", |
3081 | shortpath (NULL, def_dec_p->file->hash_entry->symbol), | |
3082 | identify_lineno (func_name_start), | |
3083 | dupnstr (func_name_start, func_name_len), | |
3084 | expected); | |
5f8037c4 RS |
3085 | return 1; |
3086 | } | |
3087 | } | |
3088 | ||
3089 | output_up_to (start_formals); | |
3090 | ||
3091 | #ifdef UNPROTOIZE | |
3092 | if (f_list_count == 0) | |
3093 | output_string (def_dec_p->formal_names); | |
3094 | #else /* !defined(UNPROTOIZE) */ | |
3095 | { | |
3096 | unsigned f_list_depth; | |
3097 | const f_list_chain_item *flci_p = def_dec_p->f_list_chain; | |
3098 | ||
3099 | /* At this point, the current value of f_list count says how many | |
3100 | links we have to follow through the f_list_chain to get to the | |
3101 | particular formals list that we need to output next. */ | |
3102 | ||
3103 | for (f_list_depth = 0; f_list_depth < f_list_count; f_list_depth++) | |
3104 | flci_p = flci_p->chain_next; | |
3105 | output_string (flci_p->formals_list); | |
3106 | } | |
3107 | #endif /* !defined(UNPROTOIZE) */ | |
3108 | ||
3109 | clean_read_ptr = end_formals - 1; | |
3110 | return 0; | |
3111 | } | |
3112 | ||
3113 | /* Given a pointer to a byte in the clean text buffer which points to the | |
3114 | beginning of a line that contains a "follower" token for a function | |
3115 | definition header, do whatever is necessary to find the right closing | |
3116 | paren for the rightmost formals list of the function definition header. | |
3117 | */ | |
3118 | ||
3119 | static const char * | |
3120 | find_rightmost_formals_list (const char *clean_text_p) | |
3121 | { | |
3122 | const char *end_formals; | |
3123 | ||
3124 | /* We are editing a function definition. The line number we did a seek | |
3125 | to contains the first token which immediately follows the entire set of | |
3126 | formals lists which are part of this particular function definition | |
3127 | header. | |
3128 | ||
3129 | Our job now is to scan leftwards in the clean text looking for the | |
3130 | right-paren which is at the end of the function header's rightmost | |
3131 | formals list. | |
3132 | ||
3133 | If we ignore whitespace, this right paren should be the first one we | |
3134 | see which is (ignoring whitespace) immediately followed either by the | |
3135 | open curly-brace beginning the function body or by an alphabetic | |
3136 | character (in the case where the function definition is in old (K&R) | |
3137 | style and there are some declarations of formal parameters). */ | |
3138 | ||
3139 | /* It is possible that the right paren we are looking for is on the | |
3140 | current line (together with its following token). Just in case that | |
3141 | might be true, we start out here by skipping down to the right end of | |
3142 | the current line before starting our scan. */ | |
3143 | ||
3144 | for (end_formals = clean_text_p; *end_formals != '\n'; end_formals++) | |
3145 | continue; | |
3146 | end_formals--; | |
3147 | ||
3148 | /* Now scan backwards while looking for the right end of the rightmost | |
3149 | formals list associated with this function definition. */ | |
3150 | ||
3151 | for (;;) | |
3152 | { | |
3153 | char ch; | |
3154 | const char *l_brace_p; | |
3155 | ||
3156 | /* Look leftward and try to find a right-paren. */ | |
3157 | ||
3158 | while (*end_formals != ')') | |
3159 | { | |
3160 | if (isspace (*end_formals)) | |
3161 | while (isspace (*end_formals)) | |
3162 | check_source (--end_formals > clean_read_ptr, 0); | |
3163 | else | |
3164 | check_source (--end_formals > clean_read_ptr, 0); | |
3165 | } | |
3166 | ||
3167 | ch = *(l_brace_p = forward_to_next_token_char (end_formals)); | |
3168 | ||
3169 | #ifdef UNPROTOIZE | |
3170 | ||
3171 | /* Since we are unprotoizing an ANSI-style (prototyped) function | |
3172 | definition, there had better not be anything (except whitespace) | |
3173 | between the end of the ANSI formals list and the beginning of the | |
3174 | function body (i.e. the '{'). */ | |
3175 | ||
3176 | check_source (ch == '{', l_brace_p); | |
3177 | break; | |
3178 | ||
3179 | #else /* !defined(UNPROTOIZE) */ | |
3180 | ||
3181 | /* Since it is possible that we found a right paren before the starting | |
3182 | '{' of the body which IS NOT the one at the end of the real K&R | |
3183 | formals list (say for instance, we found one embedded inside one of | |
3184 | the old K&R formal parameter declarations) we have to check to be | |
3185 | sure that this is in fact the right paren that we were looking for. | |
3186 | ||
3187 | The one we were looking for *must* be followed by either a '{' or | |
3188 | by an alphabetic character, while others *cannot* legally be followed | |
3189 | by such characters. */ | |
3190 | ||
3191 | if ((ch == '{') || isalpha (ch)) | |
3192 | break; | |
3193 | ||
3194 | /* At this point, we have found a right paren, but we know that it is | |
3195 | not the one we were looking for, so backup one character and keep | |
3196 | looking. */ | |
3197 | ||
3198 | check_source (--end_formals > clean_read_ptr, 0); | |
3199 | ||
3200 | #endif /* !defined(UNPROTOIZE) */ | |
3201 | ||
3202 | } | |
3203 | ||
3204 | return end_formals; | |
3205 | } | |
3206 | ||
3207 | #ifndef UNPROTOIZE | |
3208 | ||
3209 | /* Insert into the output file a totally new declaration for a function | |
3210 | which (up until now) was being called from within the current block | |
3211 | without having been declared at any point such that the declaration | |
3212 | was visible (i.e. in scope) at the point of the call. | |
3213 | ||
3214 | We need to add in explicit declarations for all such function calls | |
3215 | in order to get the full benefit of prototype-based function call | |
3216 | parameter type checking. */ | |
3217 | ||
3218 | static void | |
3219 | add_local_decl (const def_dec_info *def_dec_p, const char *clean_text_p) | |
3220 | { | |
3221 | const char *start_of_block; | |
3222 | const char *function_to_edit = def_dec_p->hash_entry->symbol; | |
3223 | ||
3224 | /* Don't insert new local explicit declarations unless explicitly requested | |
3225 | to do so. */ | |
3226 | ||
3227 | if (!local_flag) | |
3228 | return; | |
3229 | ||
3230 | /* Setup here to recover from confusing source code detected during this | |
3231 | particular "edit". */ | |
3232 | ||
3233 | save_pointers (); | |
3234 | if (setjmp (source_confusion_recovery)) | |
3235 | { | |
3236 | restore_pointers (); | |
3237 | fprintf (stderr, "%s: local declaration for function `%s' not inserted\n", | |
a2b22788 | 3238 | pname, function_to_edit); |
5f8037c4 RS |
3239 | return; |
3240 | } | |
3241 | ||
3242 | /* We have already done a seek to the start of the line which should | |
3243 | contain *the* open curly brace which begins the block in which we need | |
3244 | to insert an explicit function declaration (to replace the implicit one). | |
3245 | ||
3246 | Now we scan that line, starting from the left, until we find the | |
3247 | open curly brace we are looking for. Note that there may actually be | |
3248 | multiple open curly braces on the given line, but we will be happy | |
3249 | with the leftmost one no matter what. */ | |
3250 | ||
3251 | start_of_block = clean_text_p; | |
3252 | while (*start_of_block != '{' && *start_of_block != '\n') | |
3253 | check_source (++start_of_block < clean_text_limit, 0); | |
3254 | ||
3255 | /* Note that the line from the original source could possibly | |
3256 | contain *no* open curly braces! This happens if the line contains | |
3257 | a macro call which expands into a chunk of text which includes a | |
3258 | block (and that block's associated open and close curly braces). | |
3259 | In cases like this, we give up, issue a warning, and do nothing. */ | |
3260 | ||
3261 | if (*start_of_block != '{') | |
3262 | { | |
3263 | if (!quiet_flag) | |
3264 | fprintf (stderr, | |
a2b22788 RS |
3265 | "\n%s: %d: warning: can't add declaration of `%s' into macro call\n", |
3266 | def_dec_p->file->hash_entry->symbol, def_dec_p->line, | |
3267 | def_dec_p->hash_entry->symbol); | |
5f8037c4 RS |
3268 | return; |
3269 | } | |
3270 | ||
3271 | /* Figure out what a nice (pretty) indentation would be for the new | |
3272 | declaration we are adding. In order to do this, we must scan forward | |
3273 | from the '{' until we find the first line which starts with some | |
3274 | non-whitespace characters (i.e. real "token" material). */ | |
3275 | ||
3276 | { | |
3277 | const char *ep = forward_to_next_token_char (start_of_block) - 1; | |
3278 | const char *sp; | |
3279 | ||
3280 | /* Now we have ep pointing at the rightmost byte of some existing indent | |
3281 | stuff. At least that is the hope. | |
3282 | ||
3283 | We can now just scan backwards and find the left end of the existing | |
3284 | indentation string, and then copy it to the output buffer. */ | |
3285 | ||
3286 | for (sp = ep; isspace (*sp) && *sp != '\n'; sp--) | |
3287 | continue; | |
3288 | ||
3289 | /* Now write out the open { which began this block, and any following | |
3290 | trash up to and including the last byte of the existing indent that | |
3291 | we just found. */ | |
3292 | ||
3293 | output_up_to (ep); | |
3294 | ||
3295 | /* Now we go ahead and insert the new declaration at this point. | |
3296 | ||
3297 | If the definition of the given function is in the same file that we | |
3298 | are currently editing, and if its full ANSI declaration normally | |
3299 | would start with the keyword `extern', suppress the `extern'. */ | |
3300 | ||
3301 | { | |
3302 | const char *decl = def_dec_p->definition->ansi_decl; | |
3303 | ||
3304 | if ((*decl == 'e') && (def_dec_p->file == def_dec_p->definition->file)) | |
3305 | decl += 7; | |
3306 | output_string (decl); | |
3307 | } | |
3308 | ||
3309 | /* Finally, write out a new indent string, just like the preceeding one | |
3310 | that we found. This will typically include a newline as the first | |
3311 | character of the indent string. */ | |
3312 | ||
3313 | output_bytes (sp, (size_t) (ep - sp) + 1); | |
3314 | } | |
3315 | } | |
3316 | ||
3317 | /* Given a pointer to a file_info record, and a pointer to the beginning | |
3318 | of a line (in the clean text buffer) which is assumed to contain the | |
3319 | first "follower" token for the first function definition header in the | |
3320 | given file, find a good place to insert some new global function | |
3321 | declarations (which will replace scattered and imprecise implicit ones) | |
3322 | and then insert the new explicit declaration at that point in the file. */ | |
3323 | ||
3324 | static void | |
3325 | add_global_decls (const file_info *file_p, const char *clean_text_p) | |
3326 | { | |
3327 | const def_dec_info *dd_p; | |
3328 | const char *scan_p; | |
3329 | ||
3330 | /* Setup here to recover from confusing source code detected during this | |
3331 | particular "edit". */ | |
3332 | ||
3333 | save_pointers (); | |
3334 | if (setjmp (source_confusion_recovery)) | |
3335 | { | |
3336 | restore_pointers (); | |
3337 | fprintf (stderr, "%s: global declarations for file `%s' not inserted\n", | |
a2b22788 | 3338 | pname, shortpath (NULL, file_p->hash_entry->symbol)); |
5f8037c4 RS |
3339 | return; |
3340 | } | |
3341 | ||
3342 | /* Start by finding a good location for adding the new explicit function | |
3343 | declarations. To do this, we scan backwards, ignoring whitespace | |
3344 | and comments and other junk until we find either a semicolon, or until | |
3345 | we hit the beginning of the file. */ | |
3346 | ||
3347 | scan_p = find_rightmost_formals_list (clean_text_p); | |
3348 | for (;; --scan_p) | |
3349 | { | |
3350 | if (scan_p < clean_text_base) | |
3351 | break; | |
3352 | check_source (scan_p > clean_read_ptr, 0); | |
3353 | if (*scan_p == ';') | |
3354 | break; | |
3355 | } | |
3356 | ||
3357 | /* scan_p now points either to a semicolon, or to just before the start | |
3358 | of the whole file. */ | |
3359 | ||
3360 | /* Now scan forward for the first non-whitespace character. In theory, | |
3361 | this should be the first character of the following function definition | |
3362 | header. We will put in the added declarations just prior to that. */ | |
3363 | ||
3364 | scan_p++; | |
3365 | while (isspace (*scan_p)) | |
3366 | scan_p++; | |
3367 | scan_p--; | |
3368 | ||
3369 | output_up_to (scan_p); | |
3370 | ||
3371 | /* Now write out full prototypes for all of the things that had been | |
3372 | implicitly declared in this file (but only those for which we were | |
3373 | actually able to find unique matching definitions). Avoid duplicates | |
3374 | by marking things that we write out as we go. */ | |
3375 | ||
3376 | { | |
3377 | int some_decls_added = 0; | |
3378 | ||
3379 | for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) | |
3380 | if (dd_p->is_implicit && dd_p->definition && !dd_p->definition->written) | |
3381 | { | |
3382 | const char *decl = dd_p->definition->ansi_decl; | |
3383 | ||
3384 | /* If the function for which we are inserting a declaration is | |
3385 | actually defined later in the same file, then suppress the | |
3386 | leading `extern' keyword (if there is one). */ | |
3387 | ||
3388 | if (*decl == 'e' && (dd_p->file == dd_p->definition->file)) | |
3389 | decl += 7; | |
3390 | ||
3391 | output_string ("\n"); | |
3392 | output_string (decl); | |
3393 | some_decls_added = 1; | |
3394 | ((NONCONST def_dec_info *) dd_p->definition)->written = 1; | |
3395 | } | |
3396 | if (some_decls_added) | |
3397 | output_string ("\n\n"); | |
3398 | } | |
3399 | ||
3400 | /* Unmark all of the definitions that we just marked. */ | |
3401 | ||
3402 | for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) | |
3403 | if (dd_p->definition) | |
3404 | ((NONCONST def_dec_info *) dd_p->definition)->written = 0; | |
3405 | } | |
3406 | ||
3407 | #endif /* !defined(UNPROTOIZE) */ | |
3408 | ||
3409 | /* Do the editing operation specifically for a function "definition". Note | |
3410 | that editing operations for function "declarations" are handled by a | |
3411 | separate routine above. */ | |
3412 | ||
3413 | static void | |
3414 | edit_fn_definition (const def_dec_info *def_dec_p, const char *clean_text_p) | |
3415 | { | |
3416 | const char *end_formals; | |
3417 | const char *function_to_edit = def_dec_p->hash_entry->symbol; | |
3418 | ||
3419 | /* Setup here to recover from confusing source code detected during this | |
3420 | particular "edit". */ | |
3421 | ||
3422 | save_pointers (); | |
3423 | if (setjmp (source_confusion_recovery)) | |
3424 | { | |
3425 | restore_pointers (); | |
3426 | fprintf (stderr, "%s: definition of function `%s' not converted\n", | |
a2b22788 | 3427 | pname, function_to_edit); |
5f8037c4 RS |
3428 | return; |
3429 | } | |
3430 | ||
3431 | end_formals = find_rightmost_formals_list (clean_text_p); | |
3432 | ||
3433 | /* end_of_formals now points to the closing right paren of the rightmost | |
3434 | formals list which is actually part of the `header' of the function | |
3435 | definition that we are converting. */ | |
3436 | ||
3437 | /* If the header of this function definition looks like it declares a | |
3438 | function with a variable number of arguments, and if the way it does | |
3439 | that is different from that way we would like it (i.e. varargs vs. | |
3440 | stdarg) then issue a warning and leave the header unconverted. */ | |
3441 | ||
3442 | if (other_variable_style_function (def_dec_p->ansi_decl)) | |
3443 | { | |
3444 | if (!quiet_flag) | |
a2b22788 RS |
3445 | fprintf (stderr, "%s: %d: warning: definition of %s not converted\n", |
3446 | shortpath (NULL, def_dec_p->file->hash_entry->symbol), | |
3447 | identify_lineno (end_formals), | |
3448 | other_var_style); | |
5f8037c4 RS |
3449 | output_up_to (end_formals); |
3450 | return; | |
3451 | } | |
3452 | ||
3453 | if (edit_formals_lists (end_formals, def_dec_p->f_list_count, def_dec_p)) | |
3454 | { | |
3455 | restore_pointers (); | |
3456 | fprintf (stderr, "%s: definition of function `%s' not converted\n", | |
a2b22788 | 3457 | pname, function_to_edit); |
5f8037c4 RS |
3458 | return; |
3459 | } | |
3460 | ||
3461 | /* Have to output the last right paren because this never gets flushed by | |
3462 | edit_formals_list. */ | |
3463 | ||
3464 | output_up_to (end_formals); | |
3465 | ||
3466 | #ifdef UNPROTOIZE | |
3467 | { | |
3468 | const char *decl_p; | |
3469 | const char *semicolon_p; | |
3470 | const char *limit_p; | |
3471 | const char *scan_p; | |
3472 | int had_newlines = 0; | |
3473 | ||
3474 | /* Now write out the K&R style formal declarations, one per line. */ | |
3475 | ||
3476 | decl_p = def_dec_p->formal_decls; | |
3477 | limit_p = decl_p + strlen (decl_p); | |
3478 | for (;decl_p < limit_p; decl_p = semicolon_p + 2) | |
3479 | { | |
3480 | for (semicolon_p = decl_p; *semicolon_p != ';'; semicolon_p++) | |
3481 | continue; | |
3482 | output_string ("\n"); | |
3483 | output_string (indent_string); | |
3484 | output_bytes (decl_p, (size_t) ((semicolon_p + 1) - decl_p)); | |
3485 | } | |
3486 | ||
3487 | /* If there are no newlines between the end of the formals list and the | |
3488 | start of the body, we should insert one now. */ | |
3489 | ||
3490 | for (scan_p = end_formals+1; *scan_p != '{'; ) | |
3491 | { | |
3492 | if (*scan_p == '\n') | |
3493 | { | |
3494 | had_newlines = 1; | |
3495 | break; | |
3496 | } | |
3497 | check_source (++scan_p < clean_text_limit, 0); | |
3498 | } | |
3499 | if (!had_newlines) | |
3500 | output_string ("\n"); | |
3501 | } | |
3502 | #else /* !defined(UNPROTOIZE) */ | |
3503 | /* If we are protoizing, there may be some flotsum & jetsum (like comments | |
3504 | and preprocessing directives) after the old formals list but before | |
3505 | the following { and we would like to preserve that stuff while effectively | |
3506 | deleting the existing K&R formal parameter declarations. We do so here | |
3507 | in a rather tricky way. Basically, we white out any stuff *except* | |
3508 | the comments/pp-directives in the original text buffer, then, if there | |
3509 | is anything in this area *other* than whitespace, we output it. */ | |
3510 | { | |
3511 | const char *end_formals_orig; | |
3512 | const char *start_body; | |
3513 | const char *start_body_orig; | |
3514 | const char *scan; | |
3515 | const char *scan_orig; | |
3516 | int have_flotsum = 0; | |
3517 | int have_newlines = 0; | |
3518 | ||
3519 | for (start_body = end_formals + 1; *start_body != '{';) | |
3520 | check_source (++start_body < clean_text_limit, 0); | |
3521 | ||
3522 | end_formals_orig = orig_text_base + (end_formals - clean_text_base); | |
3523 | start_body_orig = orig_text_base + (start_body - clean_text_base); | |
3524 | scan = end_formals + 1; | |
3525 | scan_orig = end_formals_orig + 1; | |
3526 | for (; scan < start_body; scan++, scan_orig++) | |
3527 | { | |
3528 | if (*scan == *scan_orig) | |
3529 | { | |
3530 | have_newlines |= (*scan_orig == '\n'); | |
3531 | /* Leave identical whitespace alone. */ | |
3532 | if (!isspace (*scan_orig)) | |
3533 | *((NONCONST char *)scan_orig) = ' '; /* identical - so whiteout */ | |
3534 | } | |
3535 | else | |
3536 | have_flotsum = 1; | |
3537 | } | |
3538 | if (have_flotsum) | |
3539 | output_bytes (end_formals_orig + 1, | |
3540 | (size_t) (start_body_orig - end_formals_orig) - 1); | |
3541 | else | |
3542 | if (have_newlines) | |
3543 | output_string ("\n"); | |
3544 | else | |
3545 | output_string (" "); | |
3546 | clean_read_ptr = start_body - 1; | |
3547 | } | |
3548 | #endif /* !defined(UNPROTOIZE) */ | |
3549 | } | |
3550 | ||
3551 | /* Clean up the clean text buffer. Do this by converting comments and | |
3552 | preprocessor directives into spaces. Also convert line continuations | |
3553 | into whitespace. Also, whiteout string and character literals. */ | |
3554 | ||
3555 | static void | |
3556 | do_cleaning (char *new_clean_text_base, char *new_clean_text_limit) | |
3557 | { | |
3558 | char *scan_p; | |
3559 | int non_whitespace_since_newline = 0; | |
3560 | ||
3561 | for (scan_p = new_clean_text_base; scan_p < new_clean_text_limit; scan_p++) | |
3562 | { | |
3563 | switch (*scan_p) | |
3564 | { | |
3565 | case '/': /* Handle comments. */ | |
3566 | if (scan_p[1] != '*') | |
3567 | goto regular; | |
3568 | non_whitespace_since_newline = 1; | |
3569 | scan_p[0] = ' '; | |
3570 | scan_p[1] = ' '; | |
3571 | scan_p += 2; | |
3572 | while (scan_p[1] != '/' || scan_p[0] != '*') | |
3573 | { | |
3574 | if (!isspace (*scan_p)) | |
3575 | *scan_p = ' '; | |
3576 | if (++scan_p >= new_clean_text_limit) | |
3577 | abort (); | |
3578 | } | |
3579 | *scan_p++ = ' '; | |
3580 | *scan_p = ' '; | |
3581 | break; | |
3582 | ||
3583 | case '#': /* Handle pp directives. */ | |
3584 | if (non_whitespace_since_newline) | |
3585 | goto regular; | |
3586 | *scan_p = ' '; | |
3587 | while (scan_p[1] != '\n' || scan_p[0] == '\\') | |
3588 | { | |
3589 | if (!isspace (*scan_p)) | |
3590 | *scan_p = ' '; | |
3591 | if (++scan_p >= new_clean_text_limit) | |
3592 | abort (); | |
3593 | } | |
3594 | *scan_p++ = ' '; | |
3595 | break; | |
3596 | ||
3597 | case '\'': /* Handle character literals. */ | |
3598 | non_whitespace_since_newline = 1; | |
3599 | while (scan_p[1] != '\'' || scan_p[0] == '\\') | |
3600 | { | |
3601 | if (scan_p[0] == '\\' && !isspace (scan_p[1])) | |
3602 | scan_p[1] = ' '; | |
3603 | if (!isspace (*scan_p)) | |
3604 | *scan_p = ' '; | |
3605 | if (++scan_p >= new_clean_text_limit) | |
3606 | abort (); | |
3607 | } | |
3608 | *scan_p++ = ' '; | |
3609 | break; | |
3610 | ||
3611 | case '"': /* Handle string literals. */ | |
3612 | non_whitespace_since_newline = 1; | |
3613 | while (scan_p[1] != '"' || scan_p[0] == '\\') | |
3614 | { | |
3615 | if (scan_p[0] == '\\' && !isspace (scan_p[1])) | |
3616 | scan_p[1] = ' '; | |
3617 | if (!isspace (*scan_p)) | |
3618 | *scan_p = ' '; | |
3619 | if (++scan_p >= new_clean_text_limit) | |
3620 | abort (); | |
3621 | } | |
3622 | *scan_p++ = ' '; | |
3623 | break; | |
3624 | ||
3625 | case '\\': /* Handle line continuations. */ | |
3626 | if (scan_p[1] != '\n') | |
3627 | goto regular; | |
3628 | *scan_p = ' '; | |
3629 | break; | |
3630 | ||
3631 | case '\n': | |
3632 | non_whitespace_since_newline = 0; /* Reset. */ | |
3633 | break; | |
3634 | ||
3635 | case ' ': | |
3636 | case '\v': | |
3637 | case '\t': | |
3638 | case '\r': | |
3639 | case '\f': | |
3640 | case '\b': | |
3641 | break; /* Whitespace characters. */ | |
3642 | ||
3643 | default: | |
3644 | regular: | |
3645 | non_whitespace_since_newline = 1; | |
3646 | break; | |
3647 | } | |
3648 | } | |
3649 | } | |
3650 | ||
3651 | /* Given a pointer to the closing right parenthesis for a particular formals | |
3652 | list (in the clean text buffer) find the corresponding left parenthesis | |
3653 | and return a pointer to it. */ | |
3654 | ||
3655 | static const char * | |
3656 | careful_find_l_paren (const char *p) | |
3657 | { | |
3658 | const char *q; | |
3659 | int paren_depth; | |
3660 | ||
3661 | for (paren_depth = 1, q = p-1; paren_depth; check_source (--q >= clean_text_base, 0)) | |
3662 | { | |
3663 | switch (*q) | |
3664 | { | |
3665 | case ')': | |
3666 | paren_depth++; | |
3667 | break; | |
3668 | case '(': | |
3669 | paren_depth--; | |
3670 | break; | |
3671 | } | |
3672 | } | |
3673 | return ++q; | |
3674 | } | |
3675 | ||
3676 | /* Scan the clean text buffer for cases of function definitions that we | |
3677 | don't really know about because they were preprocessed out when the | |
3678 | aux info files were created. | |
3679 | ||
3680 | In this version of protoize/unprotoize we just give a warning for each | |
3681 | one found. A later version may be able to at least unprotoize such | |
3682 | missed items. | |
3683 | ||
3684 | Note that we may easily find all function definitions simply by | |
3685 | looking for places where there is a left paren which is (ignoring | |
3686 | whitespace) immediately followed by either a left-brace or by an | |
3687 | upper or lower case letter. Whenever we find this combination, we | |
3688 | have also found a function definition header. | |
3689 | ||
3690 | Finding function *declarations* using syntactic clues is much harder. | |
3691 | I will probably try to do this in a later version though. */ | |
3692 | ||
3693 | static void | |
3694 | scan_for_missed_items (const file_info *file_p) | |
3695 | { | |
3696 | static const char *scan_p; | |
3697 | const char *limit = clean_text_limit - 3; | |
3698 | static const char *backup_limit; | |
3699 | ||
3700 | backup_limit = clean_text_base - 1; | |
3701 | ||
3702 | for (scan_p = clean_text_base; scan_p < limit; scan_p++) | |
3703 | { | |
3704 | if (*scan_p == ')') | |
3705 | { | |
3706 | static const char *last_r_paren; | |
3707 | const char *ahead_p; | |
3708 | ||
3709 | last_r_paren = scan_p; | |
3710 | ||
3711 | for (ahead_p = scan_p + 1; isspace (*ahead_p); ) | |
3712 | check_source (++ahead_p < limit, limit); | |
3713 | ||
3714 | scan_p = ahead_p - 1; | |
3715 | ||
3716 | if (isalpha (*ahead_p) || *ahead_p == '{') | |
3717 | { | |
3718 | const char *last_l_paren; | |
3719 | const int lineno = identify_lineno (ahead_p); | |
3720 | ||
3721 | if (setjmp (source_confusion_recovery)) | |
3722 | continue; | |
3723 | ||
3724 | /* We know we have a function definition header. Now skip | |
3725 | leftwards over all of its associated formals lists. */ | |
3726 | ||
3727 | do | |
3728 | { | |
3729 | last_l_paren = careful_find_l_paren (last_r_paren); | |
3730 | for (last_r_paren = last_l_paren-1; isspace (*last_r_paren); ) | |
3731 | check_source (--last_r_paren >= backup_limit, backup_limit); | |
3732 | } | |
3733 | while (*last_r_paren == ')'); | |
3734 | ||
3735 | if (is_id_char (*last_r_paren)) | |
3736 | { | |
3737 | const char *id_limit = last_r_paren + 1; | |
3738 | const char *id_start; | |
3739 | size_t id_length; | |
3740 | const def_dec_info *dd_p; | |
3741 | ||
3742 | for (id_start = id_limit-1; is_id_char (*id_start); ) | |
3743 | check_source (--id_start >= backup_limit, backup_limit); | |
3744 | id_start++; | |
3745 | backup_limit = id_start; | |
3746 | if ((id_length = (size_t) (id_limit - id_start)) == 0) | |
3747 | goto not_missed; | |
3748 | ||
3749 | { | |
3750 | char func_name[id_length + 1]; | |
3751 | static const char * const stmt_keywords[] | |
3752 | = { "if", "while", "for", "switch", "return", 0 }; | |
3753 | const char * const *stmt_keyword; | |
3754 | ||
3755 | strncpy (func_name, id_start, id_length); | |
3756 | func_name[id_length] = '\0'; | |
3757 | ||
3758 | /* We must check here to see if we are actually looking at | |
3759 | a statement rather than an actual function call. */ | |
3760 | ||
3761 | for (stmt_keyword = stmt_keywords; *stmt_keyword; stmt_keyword++) | |
3762 | if (!strcmp (func_name, *stmt_keyword)) | |
3763 | goto not_missed; | |
3764 | ||
3765 | #if 0 | |
3766 | fprintf (stderr, "%s: found definition of `%s' at %s(%d)\n", | |
3767 | pname, | |
3768 | func_name, | |
3769 | shortpath (NULL, file_p->hash_entry->symbol), | |
3770 | identify_lineno (id_start)); | |
3771 | #endif /* 0 */ | |
3772 | /* We really should check for a match of the function name | |
3773 | here also, but why bother. */ | |
3774 | ||
3775 | for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) | |
3776 | if (dd_p->is_func_def && dd_p->line == lineno) | |
3777 | goto not_missed; | |
3778 | ||
3779 | /* If we make it here, then we did not know about this | |
3780 | function definition. */ | |
3781 | ||
a2b22788 | 3782 | fprintf (stderr, "%s: %d: warning: `%s' was #if 0\n", |
5f8037c4 | 3783 | shortpath (NULL, file_p->hash_entry->symbol), |
a2b22788 | 3784 | identify_lineno (id_start), func_name); |
5f8037c4 RS |
3785 | fprintf (stderr, "%s: function definition not converted\n", |
3786 | pname); | |
3787 | } | |
3788 | not_missed: ; | |
3789 | } | |
3790 | } | |
3791 | } | |
3792 | } | |
3793 | } | |
3794 | ||
3795 | /* Do all editing operations for a single source file (either a "base" file | |
3796 | or an "include" file). To do this we read the file into memory, keep a | |
3797 | virgin copy there, make another cleaned in-core copy of the original file | |
3798 | (i.e. one in which all of the comments and preprocessor directives have | |
3799 | been replaced with whitespace), then use these two in-core copies of the | |
3800 | file to make a new edited in-core copy of the file. Finally, rename the | |
3801 | original file (as a way of saving it), and then write the edited version | |
3802 | of the file from core to a disk file of the same name as the original. | |
3803 | ||
3804 | Note that the trick of making a copy of the original sans comments & | |
3805 | preprocessor directives make the editing a whole lot easier. */ | |
3806 | ||
3807 | static void | |
3808 | edit_file (const hash_table_entry *hp) | |
3809 | { | |
3810 | struct stat stat_buf; | |
3811 | const file_info *file_p = hp->fip; | |
3812 | char *new_orig_text_base; | |
3813 | char *new_orig_text_limit; | |
3814 | char *new_clean_text_base; | |
3815 | char *new_clean_text_limit; | |
3816 | size_t orig_size; | |
3817 | size_t repl_size; | |
3818 | int first_definition_in_file; | |
3819 | ||
3820 | /* If we are not supposed to be converting this file, or if there is | |
3821 | nothing in there which needs converting, just skip this file. */ | |
3822 | ||
3823 | if (!needs_to_be_converted (file_p)) | |
3824 | return; | |
3825 | ||
a2b22788 | 3826 | convert_filename = file_p->hash_entry->symbol; |
5f8037c4 RS |
3827 | |
3828 | /* Convert a file if it is in a directory where we want conversion | |
3829 | and the file is not excluded. */ | |
3830 | ||
a2b22788 RS |
3831 | if (!directory_specified_p (convert_filename) |
3832 | || file_excluded_p (convert_filename)) | |
5f8037c4 RS |
3833 | { |
3834 | if (!quiet_flag | |
3835 | #ifdef UNPROTOIZE | |
3836 | /* Don't even mention "system" include files unless we are | |
3837 | protoizing. If we are protoizing, we mention these as a | |
3838 | gentile way of prodding the user to convert his "system" | |
3839 | include files to prototype format. */ | |
a2b22788 | 3840 | && !in_system_include_dir (convert_filename) |
5f8037c4 RS |
3841 | #endif /* defined(UNPROTOIZE) */ |
3842 | ) | |
3843 | fprintf (stderr, "%s: file `%s' not converted\n", | |
a2b22788 | 3844 | pname, shortpath (NULL, convert_filename)); |
5f8037c4 RS |
3845 | return; |
3846 | } | |
3847 | ||
3848 | /* Let the user know what we are up to. */ | |
3849 | ||
3850 | if (nochange_flag) | |
a2b22788 | 3851 | puts (shortpath (NULL, convert_filename)); |
5f8037c4 RS |
3852 | else |
3853 | { | |
3854 | fprintf (stderr, "%s: converting file `%s'\n", | |
a2b22788 | 3855 | pname, shortpath (NULL, convert_filename)); |
5f8037c4 RS |
3856 | fflush (stderr); |
3857 | } | |
3858 | ||
3859 | /* Find out the size (in bytes) of the original file. */ | |
3860 | ||
a2b22788 RS |
3861 | /* The cast avoids an erroneous warning on AIX. */ |
3862 | if (stat ((char *)convert_filename, &stat_buf) == -1) | |
5f8037c4 RS |
3863 | { |
3864 | fprintf (stderr, "%s: error: can't get status for file `%s': %s\n", | |
a2b22788 | 3865 | pname, shortpath (NULL, convert_filename), sys_errlist[errno]); |
5f8037c4 RS |
3866 | return; |
3867 | } | |
3868 | orig_size = stat_buf.st_size; | |
3869 | ||
3870 | /* Allocate a buffer to hold the original text. */ | |
3871 | ||
3872 | orig_text_base = new_orig_text_base = (char *) xmalloc (orig_size + 2); | |
3873 | orig_text_limit = new_orig_text_limit = new_orig_text_base + orig_size; | |
3874 | ||
3875 | /* Allocate a buffer to hold the cleaned-up version of the original text. */ | |
3876 | ||
3877 | clean_text_base = new_clean_text_base = (char *) xmalloc (orig_size + 2); | |
3878 | clean_text_limit = new_clean_text_limit = new_clean_text_base + orig_size; | |
3879 | clean_read_ptr = clean_text_base - 1; | |
3880 | ||
3881 | /* Allocate a buffer that will hopefully be large enough to hold the entire | |
3882 | converted output text. As an initial guess for the maximum size of the | |
3883 | output buffer, use 125% of the size of the original + some extra. This | |
3884 | buffer can be expanded later as needed. */ | |
3885 | ||
3886 | repl_size = orig_size + (orig_size >> 2) + 4096; | |
3887 | repl_text_base = (char *) xmalloc (repl_size + 2); | |
3888 | repl_text_limit = repl_text_base + repl_size - 1; | |
3889 | repl_write_ptr = repl_text_base - 1; | |
3890 | ||
3891 | { | |
3892 | int input_file; | |
3893 | ||
3894 | /* Open the file to be converted in READ ONLY mode. */ | |
3895 | ||
a2b22788 | 3896 | if ((input_file = open (convert_filename, O_RDONLY, 0444)) == -1) |
5f8037c4 RS |
3897 | { |
3898 | fprintf (stderr, "%s: error: can't open file `%s' for reading: %s\n", | |
a2b22788 RS |
3899 | pname, shortpath (NULL, convert_filename), |
3900 | sys_errlist[errno]); | |
5f8037c4 RS |
3901 | return; |
3902 | } | |
3903 | ||
3904 | /* Read the entire original source text file into the original text buffer | |
3905 | in one swell fwoop. Then figure out where the end of the text is and | |
3906 | make sure that it ends with a newline followed by a null. */ | |
3907 | ||
3908 | if (read (input_file, new_orig_text_base, orig_size) != orig_size) | |
3909 | { | |
3910 | close (input_file); | |
3911 | fprintf (stderr, "\n%s: error: while reading input file `%s': %s\n", | |
a2b22788 RS |
3912 | pname, shortpath (NULL, convert_filename), |
3913 | sys_errlist[errno]); | |
5f8037c4 RS |
3914 | return; |
3915 | } | |
3916 | ||
3917 | close (input_file); | |
3918 | } | |
3919 | ||
3920 | if (orig_size == 0 || orig_text_limit[-1] != '\n') | |
3921 | { | |
3922 | *new_orig_text_limit++ = '\n'; | |
3923 | orig_text_limit++; | |
3924 | } | |
3925 | ||
3926 | /* Create the cleaned up copy of the original text. */ | |
3927 | ||
3928 | memcpy (new_clean_text_base, orig_text_base, | |
3929 | (size_t) (orig_text_limit - orig_text_base)); | |
3930 | do_cleaning (new_clean_text_base, new_clean_text_limit); | |
3931 | ||
3932 | #if 0 | |
3933 | { | |
3934 | int clean_file; | |
3935 | size_t clean_size = orig_text_limit - orig_text_base; | |
a2b22788 | 3936 | char *const clean_filename = (char *) alloca (strlen (convert_filename) + 6 + 1); |
5f8037c4 RS |
3937 | |
3938 | /* Open (and create) the clean file. */ | |
3939 | ||
a2b22788 RS |
3940 | strcpy (clean_filename, convert_filename); |
3941 | strcat (clean_filename, ".clean"); | |
3942 | if ((clean_file = creat (clean_filename, 0666)) == -1) | |
5f8037c4 RS |
3943 | { |
3944 | fprintf (stderr, "%s: error: can't create/open clean file `%s': %s\n", | |
a2b22788 RS |
3945 | pname, shortpath (NULL, clean_filename), |
3946 | sys_errlist[errno]); | |
5f8037c4 RS |
3947 | return; |
3948 | } | |
3949 | ||
3950 | /* Write the clean file. */ | |
3951 | ||
3952 | if (write (clean_file, new_clean_text_base, clean_size) != clean_size) | |
3953 | fprintf (stderr, "%s: error: while writing file `%s': %s\n", | |
a2b22788 | 3954 | pname, shortpath (NULL, clean_filename), sys_errlist[errno]); |
5f8037c4 RS |
3955 | |
3956 | close (clean_file); | |
3957 | } | |
3958 | #endif /* 0 */ | |
3959 | ||
3960 | /* Do a simplified scan of the input looking for things that were not | |
3961 | mentioned in the aux info files because of the fact that they were | |
3962 | in a region of the source which was preprocessed-out (via #if or | |
3963 | via #ifdef). */ | |
3964 | ||
3965 | scan_for_missed_items (file_p); | |
3966 | ||
3967 | /* Setup to do line-oriented forward seeking in the clean text buffer. */ | |
3968 | ||
3969 | last_known_line_number = 1; | |
3970 | last_known_line_start = clean_text_base; | |
3971 | ||
3972 | /* Now get down to business and make all of the necessary edits. */ | |
3973 | ||
3974 | { | |
3975 | const def_dec_info *def_dec_p; | |
3976 | ||
3977 | first_definition_in_file = 1; | |
3978 | def_dec_p = file_p->defs_decs; | |
3979 | for (; def_dec_p; def_dec_p = def_dec_p->next_in_file) | |
3980 | { | |
3981 | const char *clean_text_p = seek_to_line (def_dec_p->line); | |
3982 | ||
3983 | /* clean_text_p now points to the first character of the line which | |
3984 | contains the `terminator' for the declaration or definition that | |
3985 | we are about to process. */ | |
3986 | ||
3987 | #ifndef UNPROTOIZE | |
3988 | ||
3989 | if (global_flag && def_dec_p->is_func_def && first_definition_in_file) | |
3990 | { | |
3991 | add_global_decls (def_dec_p->file, clean_text_p); | |
3992 | first_definition_in_file = 0; | |
3993 | } | |
3994 | ||
3995 | /* Don't edit this item if it is already in prototype format or if it | |
3996 | is a function declaration and we have found no corresponding | |
3997 | definition. */ | |
3998 | ||
3999 | if (def_dec_p->prototyped | |
4000 | || (!def_dec_p->is_func_def && !def_dec_p->definition)) | |
4001 | continue; | |
4002 | ||
4003 | #endif /* !defined(UNPROTOIZE) */ | |
4004 | ||
4005 | if (def_dec_p->is_func_def) | |
4006 | edit_fn_definition (def_dec_p, clean_text_p); | |
4007 | else | |
4008 | #ifndef UNPROTOIZE | |
4009 | if (def_dec_p->is_implicit) | |
4010 | add_local_decl (def_dec_p, clean_text_p); | |
4011 | else | |
4012 | #endif /* !defined(UNPROTOIZE) */ | |
4013 | edit_fn_declaration (def_dec_p, clean_text_p); | |
4014 | } | |
4015 | } | |
4016 | ||
4017 | /* Finalize things. Output the last trailing part of the original text. */ | |
4018 | ||
4019 | output_up_to (clean_text_limit - 1); | |
4020 | ||
4021 | /* If this is just a test run, stop now and just deallocate the buffers. */ | |
4022 | ||
4023 | if (nochange_flag) | |
4024 | { | |
4025 | free (new_orig_text_base); | |
4026 | free (new_clean_text_base); | |
4027 | free (repl_text_base); | |
4028 | return; | |
4029 | } | |
4030 | ||
4031 | /* Change the name of the original input file. This is just a quick way of | |
4032 | saving the original file. */ | |
4033 | ||
4034 | if (!nosave_flag) | |
4035 | { | |
a2b22788 RS |
4036 | char *new_filename = |
4037 | (char *) xmalloc (strlen (convert_filename) + strlen (save_suffix) + 2); | |
5f8037c4 | 4038 | |
a2b22788 RS |
4039 | strcpy (new_filename, convert_filename); |
4040 | strcat (new_filename, save_suffix); | |
4041 | if (link (convert_filename, new_filename) == -1) | |
5f8037c4 RS |
4042 | { |
4043 | if (errno == EEXIST) | |
4044 | { | |
4045 | if (!quiet_flag) | |
4046 | fprintf (stderr, "%s: warning: file `%s' already saved in `%s'\n", | |
a2b22788 RS |
4047 | pname, |
4048 | shortpath (NULL, convert_filename), | |
4049 | shortpath (NULL, new_filename)); | |
5f8037c4 RS |
4050 | } |
4051 | else | |
4052 | { | |
4053 | fprintf (stderr, "%s: error: can't link file `%s' to `%s': %s\n", | |
a2b22788 RS |
4054 | pname, |
4055 | shortpath (NULL, convert_filename), | |
4056 | shortpath (NULL, new_filename), | |
4057 | sys_errlist[errno]); | |
5f8037c4 RS |
4058 | return; |
4059 | } | |
4060 | } | |
4061 | } | |
4062 | ||
a2b22788 | 4063 | if (unlink (convert_filename) == -1) |
5f8037c4 RS |
4064 | { |
4065 | fprintf (stderr, "%s: error: can't delete file `%s': %s\n", | |
a2b22788 | 4066 | pname, shortpath (NULL, convert_filename), sys_errlist[errno]); |
5f8037c4 RS |
4067 | return; |
4068 | } | |
4069 | ||
4070 | { | |
4071 | int output_file; | |
4072 | ||
4073 | /* Open (and create) the output file. */ | |
4074 | ||
a2b22788 | 4075 | if ((output_file = creat (convert_filename, 0666)) == -1) |
5f8037c4 RS |
4076 | { |
4077 | fprintf (stderr, "%s: error: can't create/open output file `%s': %s\n", | |
a2b22788 RS |
4078 | pname, shortpath (NULL, convert_filename), |
4079 | sys_errlist[errno]); | |
5f8037c4 RS |
4080 | return; |
4081 | } | |
4082 | ||
4083 | /* Write the output file. */ | |
4084 | ||
4085 | { | |
4086 | unsigned int out_size = (repl_write_ptr + 1) - repl_text_base; | |
4087 | ||
4088 | if (write (output_file, repl_text_base, out_size) != out_size) | |
4089 | fprintf (stderr, "%s: error: while writing file `%s': %s\n", | |
a2b22788 RS |
4090 | pname, shortpath (NULL, convert_filename), |
4091 | sys_errlist[errno]); | |
5f8037c4 RS |
4092 | } |
4093 | ||
4094 | close (output_file); | |
4095 | } | |
4096 | ||
4097 | /* Deallocate the conversion buffers. */ | |
4098 | ||
4099 | free (new_orig_text_base); | |
4100 | free (new_clean_text_base); | |
4101 | free (repl_text_base); | |
4102 | ||
4103 | /* Change the mode of the output file to match the original file. */ | |
4104 | ||
a2b22788 RS |
4105 | /* The cast avoids an erroneous warning on AIX. */ |
4106 | if (chmod ((char *)convert_filename, stat_buf.st_mode) == -1) | |
5f8037c4 | 4107 | fprintf (stderr, "%s: error: can't change mode of file `%s': %s\n", |
a2b22788 | 4108 | pname, shortpath (NULL, convert_filename), sys_errlist[errno]); |
5f8037c4 RS |
4109 | |
4110 | /* Note: We would try to change the owner and group of the output file | |
4111 | to match those of the input file here, except that may not be a good | |
4112 | thing to do because it might be misleading. Also, it might not even | |
4113 | be possible to do that (on BSD systems with quotas for instance). */ | |
4114 | } | |
4115 | ||
4116 | /* Do all of the individual steps needed to do the protoization (or | |
4117 | unprotoization) of the files referenced in the aux_info files given | |
4118 | in the command line. */ | |
4119 | ||
4120 | static void | |
4121 | do_processing (void) | |
4122 | { | |
4123 | const char * const *base_pp; | |
a2b22788 RS |
4124 | const char * const * const end_pps |
4125 | = &base_source_filenames[n_base_source_files]; | |
5f8037c4 RS |
4126 | |
4127 | #ifndef UNPROTOIZE | |
4128 | int syscalls_len; | |
4129 | #endif /* !defined(UNPROTOIZE) */ | |
4130 | ||
4131 | /* One-by-one, check (and create if necessary), open, and read all of the | |
4132 | stuff in each aux_info file. After reading each aux_info file, the | |
4133 | aux_info_file just read will be automatically deleted unless the | |
4134 | keep_flag is set. */ | |
4135 | ||
a2b22788 | 4136 | for (base_pp = base_source_filenames; base_pp < end_pps; base_pp++) |
5f8037c4 RS |
4137 | process_aux_info_file (*base_pp, keep_flag, 0); |
4138 | ||
4139 | #ifndef UNPROTOIZE | |
4140 | ||
4141 | /* Also open and read the special SYSCALLS.c aux_info file which gives us | |
4142 | the prototypes for all of the standard system-supplied functions. */ | |
4143 | ||
4144 | if (nondefault_syscalls_dir) | |
4145 | { | |
a2b22788 | 4146 | syscalls_absolute_filename |
5f8037c4 RS |
4147 | = (char *) xmalloc (strlen (nondefault_syscalls_dir) |
4148 | + strlen (syscalls_filename) + 1); | |
a2b22788 | 4149 | strcpy (syscalls_absolute_filename, nondefault_syscalls_dir); |
5f8037c4 RS |
4150 | } |
4151 | else | |
4152 | { | |
a2b22788 | 4153 | syscalls_absolute_filename |
5f8037c4 RS |
4154 | = (char *) xmalloc (strlen (default_syscalls_dir) |
4155 | + strlen (syscalls_filename) + 1); | |
a2b22788 | 4156 | strcpy (syscalls_absolute_filename, default_syscalls_dir); |
5f8037c4 RS |
4157 | } |
4158 | ||
a2b22788 RS |
4159 | syscalls_len = strlen (syscalls_filename); |
4160 | if (*(syscalls_absolute_filename + syscalls_len - 1) != '/') | |
5f8037c4 | 4161 | { |
a2b22788 RS |
4162 | *(syscalls_absolute_filename + syscalls_len++) = '/'; |
4163 | *(syscalls_absolute_filename + syscalls_len) = '\0'; | |
5f8037c4 | 4164 | } |
a2b22788 | 4165 | strcat (syscalls_absolute_filename, syscalls_filename); |
5f8037c4 RS |
4166 | |
4167 | /* Call process_aux_info_file in such a way that it does not try to | |
4168 | delete the SYSCALLS aux_info file. */ | |
4169 | ||
a2b22788 | 4170 | process_aux_info_file (syscalls_absolute_filename, 1, 1); |
5f8037c4 RS |
4171 | |
4172 | #endif /* !defined(UNPROTOIZE) */ | |
4173 | ||
4174 | /* When we first read in all of the information from the aux_info files | |
4175 | we saved in it decending line number order, because that was likely to | |
4176 | be faster. Now however, we want the chains of def & dec records to | |
4177 | appear in ascending line number order as we get further away from the | |
4178 | file_info record that they hang from. The following line causes all of | |
4179 | these lists to be rearranged into ascending line number order. */ | |
4180 | ||
a2b22788 | 4181 | visit_each_hash_node (filename_primary, reverse_def_dec_list); |
5f8037c4 RS |
4182 | |
4183 | #ifndef UNPROTOIZE | |
4184 | ||
4185 | /* Now do the "real" work. The following line causes each declaration record | |
4186 | to be "visited". For each of these nodes, an attempt is made to match | |
4187 | up the function declaration with a corresponding function definition, | |
4188 | which should have a full prototype-format formals list with it. Once | |
4189 | these match-ups are made, the conversion of the function declarations | |
4190 | to prototype format can be made. */ | |
4191 | ||
4192 | visit_each_hash_node (function_name_primary, connect_defs_and_decs); | |
4193 | ||
4194 | #endif /* !defined(UNPROTOIZE) */ | |
4195 | ||
4196 | /* Now convert each file that can be converted (and needs to be). */ | |
4197 | ||
a2b22788 | 4198 | visit_each_hash_node (filename_primary, edit_file); |
5f8037c4 RS |
4199 | |
4200 | #ifndef UNPROTOIZE | |
4201 | ||
4202 | /* If we are working in cplusplus mode, try to rename all .c files to .C | |
4203 | files. Don't panic if some of the renames don't work. */ | |
4204 | ||
4205 | if (cplusplus_flag && !nochange_flag) | |
a2b22788 | 4206 | visit_each_hash_node (filename_primary, rename_c_file); |
5f8037c4 RS |
4207 | |
4208 | #endif /* !defined(UNPROTOIZE) */ | |
4209 | } | |
4210 | \f | |
4211 | static struct option longopts[] = | |
4212 | { | |
4213 | {"version", 0, 0, 'V'}, | |
4214 | {"quiet", 0, 0, 'q'}, | |
4215 | {"silent", 0, 0, 'q'}, | |
4216 | {"force", 0, 0, 'f'}, | |
4217 | {"keep", 0, 0, 'k'}, | |
4218 | {"nosave", 0, 0, 'N'}, | |
4219 | {"nochange", 0, 0, 'n'}, | |
4220 | {"compiler-options", 1, 0, 'c'}, | |
4221 | {"exclude", 1, 0, 'x'}, | |
4222 | {"directory", 1, 0, 'd'}, | |
4223 | #ifdef UNPROTOIZE | |
4224 | {"indent", 1, 0, 'i'}, | |
4225 | #else | |
4226 | {"local", 0, 0, 'l'}, | |
4227 | {"global", 0, 0, 'g'}, | |
4228 | {"c++", 0, 0, 'C'}, | |
4229 | {"syscalls-dir", 1, 0, 'B'}, | |
4230 | #endif | |
4231 | {0, 0, 0, 0} | |
4232 | }; | |
4233 | ||
4234 | int | |
4235 | main (int argc, char **const argv) | |
4236 | { | |
4237 | int longind; | |
4238 | int c; | |
4239 | int size; | |
4240 | ||
4241 | pname = strrchr (argv[0], '/'); | |
4242 | pname = pname ? pname+1 : argv[0]; | |
4243 | ||
4244 | /* Read the working directory, avoiding arbitrary limit. */ | |
4245 | size = 100; | |
4246 | while (1) | |
4247 | { | |
a2b22788 | 4248 | char *value; |
5f8037c4 RS |
4249 | |
4250 | cwd_buffer = (char *) xmalloc (size); | |
4251 | value = getcwd (cwd_buffer, size); | |
a2b22788 | 4252 | if (value != 0 || errno != ERANGE) |
5f8037c4 RS |
4253 | break; |
4254 | free (cwd_buffer); | |
4255 | size *= 2; | |
4256 | } | |
4257 | ||
4258 | /* By default, convert the files in the current directory. */ | |
4259 | directory_list = string_list_cons (cwd_buffer, NULL); | |
4260 | ||
4261 | while ((c = getopt_long (argc, argv, | |
4262 | #ifdef UNPROTOIZE | |
4263 | "c:d:i:knNqVx:", | |
4264 | #else | |
4265 | "B:c:Cd:gklnNqVx:", | |
4266 | #endif | |
4267 | longopts, &longind)) != EOF) | |
4268 | { | |
4269 | if (c == 0) /* Long option. */ | |
4270 | c = longopts[longind].val; | |
4271 | switch (c) | |
4272 | { | |
4273 | case 'd': | |
4274 | directory_list | |
4275 | = string_list_cons (abspath (NULL, optarg), directory_list); | |
4276 | break; | |
4277 | case 'x': | |
4278 | exclude_list = string_list_cons (optarg, exclude_list); | |
4279 | break; | |
4280 | ||
4281 | case 'V': | |
4282 | version_flag = 1; | |
4283 | break; | |
4284 | case 'q': | |
4285 | quiet_flag = 1; | |
4286 | break; | |
4287 | #if 0 | |
4288 | case 'f': | |
4289 | force_flag = 1; | |
4290 | break; | |
4291 | #endif | |
4292 | case 'n': | |
4293 | nochange_flag = 1; | |
4294 | keep_flag = 1; | |
4295 | break; | |
4296 | case 'N': | |
4297 | nosave_flag = 1; | |
4298 | break; | |
4299 | case 'k': | |
4300 | keep_flag = 1; | |
4301 | break; | |
4302 | case 'c': | |
4303 | munge_compile_params (optarg); | |
4304 | break; | |
4305 | #ifdef UNPROTOIZE | |
4306 | case 'i': | |
4307 | indent_string = optarg; | |
4308 | break; | |
4309 | #else /* !defined(UNPROTOIZE) */ | |
4310 | case 'l': | |
4311 | local_flag = 1; | |
4312 | break; | |
4313 | case 'g': | |
4314 | global_flag = 1; | |
4315 | break; | |
4316 | case 'C': | |
4317 | cplusplus_flag = 1; | |
4318 | break; | |
4319 | case 'B': | |
4320 | nondefault_syscalls_dir = optarg; | |
4321 | break; | |
4322 | #endif /* !defined(UNPROTOIZE) */ | |
4323 | default: | |
4324 | usage (); | |
4325 | } | |
4326 | } | |
4327 | ||
a2b22788 | 4328 | n_base_source_files = argc - optind; |
5f8037c4 | 4329 | |
a2b22788 | 4330 | /* Now actually make a list of the base source filenames. */ |
5f8037c4 | 4331 | |
a2b22788 RS |
4332 | base_source_filenames = |
4333 | (const char **) xmalloc ((n_base_source_files + 1) * sizeof (char *)); | |
4334 | n_base_source_files = 0; | |
5f8037c4 RS |
4335 | for (; optind < argc; optind++) |
4336 | { | |
4337 | const char *path = abspath (NULL, argv[optind]); | |
4338 | int len = strlen (path); | |
4339 | ||
4340 | if (path[len-1] == 'c' && path[len-2] == '.') | |
a2b22788 | 4341 | base_source_filenames[n_base_source_files++] = path; |
5f8037c4 RS |
4342 | else |
4343 | { | |
a2b22788 | 4344 | fprintf (stderr, "%s: input file names must have .c suffixes: %s\n", |
5f8037c4 RS |
4345 | pname, shortpath (NULL, path)); |
4346 | errors++; | |
4347 | } | |
4348 | } | |
4349 | ||
4350 | #ifndef UNPROTOIZE | |
4351 | /* We are only interested in the very first identifier token in the | |
4352 | definition of `va_list', so if there is more junk after that first | |
4353 | identifier token, delete it from the `varargs_style_indicator'. */ | |
4354 | { | |
4355 | const char *cp; | |
4356 | ||
4357 | for (cp = varargs_style_indicator; isalnum (*cp) || *cp == '_'; cp++) | |
4358 | continue; | |
4359 | if (*cp != 0) | |
4360 | varargs_style_indicator = savestring (varargs_style_indicator, | |
4361 | cp - varargs_style_indicator); | |
4362 | } | |
4363 | #endif /* !defined(UNPROTOIZE) */ | |
4364 | ||
4365 | if (errors) | |
4366 | usage (); | |
4367 | else | |
4368 | { | |
4369 | if (version_flag) | |
4370 | fprintf (stderr, "%s: %s\n", pname, version_string); | |
4371 | do_processing (); | |
4372 | } | |
4373 | if (errors) | |
4374 | exit (1); | |
4375 | else | |
4376 | exit (0); | |
4377 | return 1; | |
4378 | } |