]>
Commit | Line | Data |
---|---|---|
45b966db | 1 | /* CPP Library - lexical analysis. |
818ab71a | 2 | Copyright (C) 2000-2016 Free Software Foundation, Inc. |
45b966db ZW |
3 | Contributed by Per Bothner, 1994-95. |
4 | Based on CCCP program by Paul Rubin, June 1986 | |
5 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
6 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
7 | ||
8 | This program is free software; you can redistribute it and/or modify it | |
9 | under the terms of the GNU General Public License as published by the | |
748086b7 | 10 | Free Software Foundation; either version 3, or (at your option) any |
45b966db ZW |
11 | later version. |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
748086b7 JJ |
19 | along with this program; see the file COPYING3. If not see |
20 | <http://www.gnu.org/licenses/>. */ | |
45b966db ZW |
21 | |
22 | #include "config.h" | |
23 | #include "system.h" | |
45b966db | 24 | #include "cpplib.h" |
4f4e53dd | 25 | #include "internal.h" |
45b966db | 26 | |
93c80368 | 27 | enum spell_type |
f9a0e96c | 28 | { |
93c80368 | 29 | SPELL_OPERATOR = 0, |
93c80368 | 30 | SPELL_IDENT, |
6338b358 | 31 | SPELL_LITERAL, |
93c80368 | 32 | SPELL_NONE |
f9a0e96c ZW |
33 | }; |
34 | ||
93c80368 | 35 | struct token_spelling |
f9a0e96c | 36 | { |
93c80368 NB |
37 | enum spell_type category; |
38 | const unsigned char *name; | |
f9a0e96c ZW |
39 | }; |
40 | ||
8206c799 | 41 | static const unsigned char *const digraph_spellings[] = |
b6baa67d | 42 | { UC"%:", UC"%:%:", UC"<:", UC":>", UC"<%", UC"%>" }; |
93c80368 | 43 | |
b6baa67d KVH |
44 | #define OP(e, s) { SPELL_OPERATOR, UC s }, |
45 | #define TK(e, s) { SPELL_ ## s, UC #e }, | |
8206c799 | 46 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; |
93c80368 NB |
47 | #undef OP |
48 | #undef TK | |
49 | ||
50 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
51 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
f2d5f0cc | 52 | |
6cf87ca4 ZW |
53 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); |
54 | static int skip_line_comment (cpp_reader *); | |
55 | static void skip_whitespace (cpp_reader *, cppchar_t); | |
6cf87ca4 ZW |
56 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); |
57 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); | |
631d0d36 | 58 | static void store_comment (cpp_reader *, cpp_token *); |
6cf87ca4 ZW |
59 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, |
60 | unsigned int, enum cpp_ttype); | |
61 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); | |
62 | static int name_p (cpp_reader *, const cpp_string *); | |
6cf87ca4 ZW |
63 | static tokenrun *next_tokenrun (tokenrun *); |
64 | ||
6cf87ca4 | 65 | static _cpp_buff *new_buff (size_t); |
15dad1d9 | 66 | |
9d10c9a9 | 67 | |
041c3194 | 68 | /* Utility routine: |
9e62c811 | 69 | |
bfb9dc7f ZW |
70 | Compares, the token TOKEN to the NUL-terminated string STRING. |
71 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
041c3194 | 72 | int |
6cf87ca4 | 73 | cpp_ideq (const cpp_token *token, const char *string) |
041c3194 | 74 | { |
bfb9dc7f | 75 | if (token->type != CPP_NAME) |
041c3194 | 76 | return 0; |
bfb9dc7f | 77 | |
9a0c6187 | 78 | return !ustrcmp (NODE_NAME (token->val.node.node), (const uchar *) string); |
15dad1d9 | 79 | } |
1368ee70 | 80 | |
26aea073 NB |
81 | /* Record a note TYPE at byte POS into the current cleaned logical |
82 | line. */ | |
87062813 | 83 | static void |
6cf87ca4 | 84 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) |
0d9f234d | 85 | { |
26aea073 NB |
86 | if (buffer->notes_used == buffer->notes_cap) |
87 | { | |
88 | buffer->notes_cap = buffer->notes_cap * 2 + 200; | |
c3f829c1 GDR |
89 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes, |
90 | buffer->notes_cap); | |
26aea073 | 91 | } |
0d9f234d | 92 | |
26aea073 NB |
93 | buffer->notes[buffer->notes_used].pos = pos; |
94 | buffer->notes[buffer->notes_used].type = type; | |
95 | buffer->notes_used++; | |
0d9f234d NB |
96 | } |
97 | ||
246a2fcb RH |
98 | \f |
99 | /* Fast path to find line special characters using optimized character | |
100 | scanning algorithms. Anything complicated falls back to the slow | |
101 | path below. Since this loop is very hot it's worth doing these kinds | |
102 | of optimizations. | |
103 | ||
104 | One of the paths through the ifdefs should provide | |
105 | ||
106 | const uchar *search_line_fast (const uchar *s, const uchar *end); | |
107 | ||
108 | Between S and END, search for \n, \r, \\, ?. Return a pointer to | |
109 | the found character. | |
110 | ||
111 | Note that the last character of the buffer is *always* a newline, | |
112 | as forced by _cpp_convert_input. This fact can be used to avoid | |
113 | explicitly looking for the end of the buffer. */ | |
114 | ||
115 | /* Configure gives us an ifdef test. */ | |
116 | #ifndef WORDS_BIGENDIAN | |
117 | #define WORDS_BIGENDIAN 0 | |
118 | #endif | |
119 | ||
120 | /* We'd like the largest integer that fits into a register. There's nothing | |
121 | in <stdint.h> that gives us that. For most hosts this is unsigned long, | |
122 | but MS decided on an LLP64 model. Thankfully when building with GCC we | |
123 | can get the "real" word size. */ | |
124 | #ifdef __GNUC__ | |
125 | typedef unsigned int word_type __attribute__((__mode__(__word__))); | |
126 | #else | |
127 | typedef unsigned long word_type; | |
128 | #endif | |
129 | ||
130 | /* The code below is only expecting sizes 4 or 8. | |
131 | Die at compile-time if this expectation is violated. */ | |
132 | typedef char check_word_type_size | |
133 | [(sizeof(word_type) == 8 || sizeof(word_type) == 4) * 2 - 1]; | |
134 | ||
135 | /* Return X with the first N bytes forced to values that won't match one | |
136 | of the interesting characters. Note that NUL is not interesting. */ | |
137 | ||
138 | static inline word_type | |
139 | acc_char_mask_misalign (word_type val, unsigned int n) | |
140 | { | |
141 | word_type mask = -1; | |
142 | if (WORDS_BIGENDIAN) | |
143 | mask >>= n * 8; | |
144 | else | |
145 | mask <<= n * 8; | |
146 | return val & mask; | |
147 | } | |
148 | ||
149 | /* Return X replicated to all byte positions within WORD_TYPE. */ | |
150 | ||
151 | static inline word_type | |
152 | acc_char_replicate (uchar x) | |
153 | { | |
154 | word_type ret; | |
155 | ||
156 | ret = (x << 24) | (x << 16) | (x << 8) | x; | |
157 | if (sizeof(word_type) == 8) | |
158 | ret = (ret << 16 << 16) | ret; | |
159 | return ret; | |
160 | } | |
161 | ||
162 | /* Return non-zero if some byte of VAL is (probably) C. */ | |
163 | ||
164 | static inline word_type | |
165 | acc_char_cmp (word_type val, word_type c) | |
166 | { | |
167 | #if defined(__GNUC__) && defined(__alpha__) | |
168 | /* We can get exact results using a compare-bytes instruction. | |
169 | Get (val == c) via (0 >= (val ^ c)). */ | |
170 | return __builtin_alpha_cmpbge (0, val ^ c); | |
171 | #else | |
172 | word_type magic = 0x7efefefeU; | |
173 | if (sizeof(word_type) == 8) | |
174 | magic = (magic << 16 << 16) | 0xfefefefeU; | |
175 | magic |= 1; | |
176 | ||
177 | val ^= c; | |
178 | return ((val + magic) ^ ~val) & ~magic; | |
179 | #endif | |
180 | } | |
181 | ||
182 | /* Given the result of acc_char_cmp is non-zero, return the index of | |
183 | the found character. If this was a false positive, return -1. */ | |
184 | ||
185 | static inline int | |
186 | acc_char_index (word_type cmp ATTRIBUTE_UNUSED, | |
187 | word_type val ATTRIBUTE_UNUSED) | |
188 | { | |
189 | #if defined(__GNUC__) && defined(__alpha__) && !WORDS_BIGENDIAN | |
190 | /* The cmpbge instruction sets *bits* of the result corresponding to | |
191 | matches in the bytes with no false positives. */ | |
192 | return __builtin_ctzl (cmp); | |
193 | #else | |
194 | unsigned int i; | |
195 | ||
196 | /* ??? It would be nice to force unrolling here, | |
197 | and have all of these constants folded. */ | |
198 | for (i = 0; i < sizeof(word_type); ++i) | |
199 | { | |
200 | uchar c; | |
201 | if (WORDS_BIGENDIAN) | |
202 | c = (val >> (sizeof(word_type) - i - 1) * 8) & 0xff; | |
203 | else | |
204 | c = (val >> i * 8) & 0xff; | |
205 | ||
206 | if (c == '\n' || c == '\r' || c == '\\' || c == '?') | |
207 | return i; | |
208 | } | |
209 | ||
210 | return -1; | |
211 | #endif | |
212 | } | |
213 | ||
214 | /* A version of the fast scanner using bit fiddling techniques. | |
215 | ||
216 | For 32-bit words, one would normally perform 16 comparisons and | |
217 | 16 branches. With this algorithm one performs 24 arithmetic | |
218 | operations and one branch. Whether this is faster with a 32-bit | |
219 | word size is going to be somewhat system dependent. | |
220 | ||
221 | For 64-bit words, we eliminate twice the number of comparisons | |
222 | and branches without increasing the number of arithmetic operations. | |
223 | It's almost certainly going to be a win with 64-bit word size. */ | |
224 | ||
225 | static const uchar * search_line_acc_char (const uchar *, const uchar *) | |
226 | ATTRIBUTE_UNUSED; | |
227 | ||
228 | static const uchar * | |
229 | search_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
230 | { | |
231 | const word_type repl_nl = acc_char_replicate ('\n'); | |
232 | const word_type repl_cr = acc_char_replicate ('\r'); | |
233 | const word_type repl_bs = acc_char_replicate ('\\'); | |
234 | const word_type repl_qm = acc_char_replicate ('?'); | |
235 | ||
236 | unsigned int misalign; | |
237 | const word_type *p; | |
238 | word_type val, t; | |
239 | ||
240 | /* Align the buffer. Mask out any bytes from before the beginning. */ | |
241 | p = (word_type *)((uintptr_t)s & -sizeof(word_type)); | |
242 | val = *p; | |
243 | misalign = (uintptr_t)s & (sizeof(word_type) - 1); | |
244 | if (misalign) | |
245 | val = acc_char_mask_misalign (val, misalign); | |
246 | ||
247 | /* Main loop. */ | |
248 | while (1) | |
249 | { | |
250 | t = acc_char_cmp (val, repl_nl); | |
251 | t |= acc_char_cmp (val, repl_cr); | |
252 | t |= acc_char_cmp (val, repl_bs); | |
253 | t |= acc_char_cmp (val, repl_qm); | |
254 | ||
255 | if (__builtin_expect (t != 0, 0)) | |
256 | { | |
257 | int i = acc_char_index (t, val); | |
258 | if (i >= 0) | |
259 | return (const uchar *)p + i; | |
260 | } | |
261 | ||
262 | val = *++p; | |
263 | } | |
264 | } | |
265 | ||
d9f069ab | 266 | /* Disable on Solaris 2/x86 until the following problem can be properly |
789d73cb RO |
267 | autoconfed: |
268 | ||
789d73cb RO |
269 | The Solaris 10+ assembler tags objects with the instruction set |
270 | extensions used, so SSE4.2 executables cannot run on machines that | |
271 | don't support that extension. */ | |
272 | ||
1b6b13f3 | 273 | #if (GCC_VERSION >= 4005) && (__GNUC__ >= 5 || !defined(__PIC__)) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__)) |
246a2fcb RH |
274 | |
275 | /* Replicated character data to be shared between implementations. | |
276 | Recall that outside of a context with vector support we can't | |
277 | define compatible vector types, therefore these are all defined | |
278 | in terms of raw characters. */ | |
279 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { | |
280 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
281 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, | |
282 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
283 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, | |
284 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
285 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, | |
286 | { '?', '?', '?', '?', '?', '?', '?', '?', | |
287 | '?', '?', '?', '?', '?', '?', '?', '?' }, | |
288 | }; | |
289 | ||
290 | /* A version of the fast scanner using MMX vectorized byte compare insns. | |
291 | ||
292 | This uses the PMOVMSKB instruction which was introduced with "MMX2", | |
ef230b38 | 293 | which was packaged into SSE1; it is also present in the AMD MMX |
246a2fcb RH |
294 | extension. Mark the function as using "sse" so that we emit a real |
295 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ | |
296 | ||
297 | static const uchar * | |
298 | #ifndef __SSE__ | |
299 | __attribute__((__target__("sse"))) | |
300 | #endif | |
301 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
302 | { | |
303 | typedef char v8qi __attribute__ ((__vector_size__ (8))); | |
304 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); | |
305 | ||
306 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; | |
307 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; | |
308 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; | |
309 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; | |
310 | ||
311 | unsigned int misalign, found, mask; | |
312 | const v8qi *p; | |
313 | v8qi data, t, c; | |
314 | ||
315 | /* Align the source pointer. While MMX doesn't generate unaligned data | |
316 | faults, this allows us to safely scan to the end of the buffer without | |
317 | reading beyond the end of the last page. */ | |
318 | misalign = (uintptr_t)s & 7; | |
319 | p = (const v8qi *)((uintptr_t)s & -8); | |
320 | data = *p; | |
321 | ||
322 | /* Create a mask for the bytes that are valid within the first | |
323 | 16-byte block. The Idea here is that the AND with the mask | |
324 | within the loop is "free", since we need some AND or TEST | |
325 | insn in order to set the flags for the branch anyway. */ | |
326 | mask = -1u << misalign; | |
327 | ||
328 | /* Main loop processing 8 bytes at a time. */ | |
329 | goto start; | |
330 | do | |
331 | { | |
332 | data = *++p; | |
333 | mask = -1; | |
334 | ||
335 | start: | |
336 | t = __builtin_ia32_pcmpeqb(data, repl_nl); | |
337 | c = __builtin_ia32_pcmpeqb(data, repl_cr); | |
338 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
339 | c = __builtin_ia32_pcmpeqb(data, repl_bs); | |
340 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
341 | c = __builtin_ia32_pcmpeqb(data, repl_qm); | |
342 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
343 | found = __builtin_ia32_pmovmskb (t); | |
344 | found &= mask; | |
345 | } | |
346 | while (!found); | |
347 | ||
348 | __builtin_ia32_emms (); | |
349 | ||
350 | /* FOUND contains 1 in bits for which we matched a relevant | |
351 | character. Conversion to the byte index is trivial. */ | |
352 | found = __builtin_ctz(found); | |
353 | return (const uchar *)p + found; | |
354 | } | |
355 | ||
356 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ | |
357 | ||
358 | static const uchar * | |
359 | #ifndef __SSE2__ | |
360 | __attribute__((__target__("sse2"))) | |
361 | #endif | |
362 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
363 | { | |
364 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
365 | ||
366 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; | |
367 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; | |
368 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; | |
369 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; | |
370 | ||
371 | unsigned int misalign, found, mask; | |
372 | const v16qi *p; | |
373 | v16qi data, t; | |
374 | ||
375 | /* Align the source pointer. */ | |
376 | misalign = (uintptr_t)s & 15; | |
377 | p = (const v16qi *)((uintptr_t)s & -16); | |
378 | data = *p; | |
379 | ||
380 | /* Create a mask for the bytes that are valid within the first | |
381 | 16-byte block. The Idea here is that the AND with the mask | |
382 | within the loop is "free", since we need some AND or TEST | |
383 | insn in order to set the flags for the branch anyway. */ | |
384 | mask = -1u << misalign; | |
385 | ||
386 | /* Main loop processing 16 bytes at a time. */ | |
387 | goto start; | |
388 | do | |
389 | { | |
390 | data = *++p; | |
391 | mask = -1; | |
392 | ||
393 | start: | |
394 | t = __builtin_ia32_pcmpeqb128(data, repl_nl); | |
395 | t |= __builtin_ia32_pcmpeqb128(data, repl_cr); | |
396 | t |= __builtin_ia32_pcmpeqb128(data, repl_bs); | |
397 | t |= __builtin_ia32_pcmpeqb128(data, repl_qm); | |
398 | found = __builtin_ia32_pmovmskb128 (t); | |
399 | found &= mask; | |
400 | } | |
401 | while (!found); | |
402 | ||
403 | /* FOUND contains 1 in bits for which we matched a relevant | |
404 | character. Conversion to the byte index is trivial. */ | |
405 | found = __builtin_ctz(found); | |
406 | return (const uchar *)p + found; | |
407 | } | |
408 | ||
6f173e52 | 409 | #ifdef HAVE_SSE4 |
246a2fcb RH |
410 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
411 | ||
412 | static const uchar * | |
413 | #ifndef __SSE4_2__ | |
414 | __attribute__((__target__("sse4.2"))) | |
415 | #endif | |
416 | search_line_sse42 (const uchar *s, const uchar *end) | |
417 | { | |
418 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
419 | static const v16qi search = { '\n', '\r', '?', '\\' }; | |
420 | ||
421 | uintptr_t si = (uintptr_t)s; | |
422 | uintptr_t index; | |
423 | ||
424 | /* Check for unaligned input. */ | |
425 | if (si & 15) | |
426 | { | |
d35d1c0f UB |
427 | v16qi sv; |
428 | ||
246a2fcb RH |
429 | if (__builtin_expect (end - s < 16, 0) |
430 | && __builtin_expect ((si & 0xfff) > 0xff0, 0)) | |
431 | { | |
432 | /* There are less than 16 bytes left in the buffer, and less | |
433 | than 16 bytes left on the page. Reading 16 bytes at this | |
434 | point might generate a spurious page fault. Defer to the | |
435 | SSE2 implementation, which already handles alignment. */ | |
436 | return search_line_sse2 (s, end); | |
437 | } | |
438 | ||
439 | /* ??? The builtin doesn't understand that the PCMPESTRI read from | |
440 | memory need not be aligned. */ | |
d35d1c0f UB |
441 | sv = __builtin_ia32_loaddqu ((const char *) s); |
442 | index = __builtin_ia32_pcmpestri128 (search, 4, sv, 16, 0); | |
443 | ||
246a2fcb RH |
444 | if (__builtin_expect (index < 16, 0)) |
445 | goto found; | |
446 | ||
447 | /* Advance the pointer to an aligned address. We will re-scan a | |
448 | few bytes, but we no longer need care for reading past the | |
449 | end of a page, since we're guaranteed a match. */ | |
49445904 | 450 | s = (const uchar *)((si + 15) & -16); |
246a2fcb RH |
451 | } |
452 | ||
dc6bcf52 UB |
453 | /* Main loop, processing 16 bytes at a time. */ |
454 | #ifdef __GCC_ASM_FLAG_OUTPUTS__ | |
455 | while (1) | |
456 | { | |
457 | char f; | |
458 | ||
459 | /* By using inline assembly instead of the builtin, | |
460 | we can use the result, as well as the flags set. */ | |
461 | __asm ("%vpcmpestri\t$0, %2, %3" | |
462 | : "=c"(index), "=@ccc"(f) | |
463 | : "m"(*s), "x"(search), "a"(4), "d"(16)); | |
464 | if (f) | |
465 | break; | |
466 | ||
467 | s += 16; | |
468 | } | |
469 | #else | |
470 | s -= 16; | |
471 | /* By doing the whole loop in inline assembly, | |
472 | we can make proper use of the flags set. */ | |
473 | __asm ( ".balign 16\n" | |
246a2fcb | 474 | "0: add $16, %1\n" |
dc6bcf52 | 475 | " %vpcmpestri\t$0, (%1), %2\n" |
246a2fcb RH |
476 | " jnc 0b" |
477 | : "=&c"(index), "+r"(s) | |
478 | : "x"(search), "a"(4), "d"(16)); | |
dc6bcf52 | 479 | #endif |
246a2fcb RH |
480 | |
481 | found: | |
482 | return s + index; | |
483 | } | |
484 | ||
6f173e52 RH |
485 | #else |
486 | /* Work around out-dated assemblers without sse4 support. */ | |
487 | #define search_line_sse42 search_line_sse2 | |
488 | #endif | |
489 | ||
246a2fcb RH |
490 | /* Check the CPU capabilities. */ |
491 | ||
492 | #include "../gcc/config/i386/cpuid.h" | |
493 | ||
494 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); | |
495 | static search_line_fast_type search_line_fast; | |
496 | ||
b0c084b7 JJ |
497 | #define HAVE_init_vectorized_lexer 1 |
498 | static inline void | |
246a2fcb RH |
499 | init_vectorized_lexer (void) |
500 | { | |
501 | unsigned dummy, ecx = 0, edx = 0; | |
502 | search_line_fast_type impl = search_line_acc_char; | |
503 | int minimum = 0; | |
504 | ||
505 | #if defined(__SSE4_2__) | |
506 | minimum = 3; | |
507 | #elif defined(__SSE2__) | |
508 | minimum = 2; | |
ef230b38 | 509 | #elif defined(__SSE__) |
246a2fcb RH |
510 | minimum = 1; |
511 | #endif | |
512 | ||
513 | if (minimum == 3) | |
514 | impl = search_line_sse42; | |
515 | else if (__get_cpuid (1, &dummy, &dummy, &ecx, &edx) || minimum == 2) | |
516 | { | |
517 | if (minimum == 3 || (ecx & bit_SSE4_2)) | |
518 | impl = search_line_sse42; | |
519 | else if (minimum == 2 || (edx & bit_SSE2)) | |
520 | impl = search_line_sse2; | |
521 | else if (minimum == 1 || (edx & bit_SSE)) | |
522 | impl = search_line_mmx; | |
523 | } | |
524 | else if (__get_cpuid (0x80000001, &dummy, &dummy, &dummy, &edx)) | |
525 | { | |
5e70c0b5 UB |
526 | if (minimum == 1 |
527 | || (edx & (bit_MMXEXT | bit_CMOV)) == (bit_MMXEXT | bit_CMOV)) | |
246a2fcb RH |
528 | impl = search_line_mmx; |
529 | } | |
530 | ||
531 | search_line_fast = impl; | |
532 | } | |
533 | ||
0ccaaab0 | 534 | #elif defined(_ARCH_PWR8) && defined(__ALTIVEC__) |
246a2fcb | 535 | |
0ccaaab0 BS |
536 | /* A vection of the fast scanner using AltiVec vectorized byte compares |
537 | and VSX unaligned loads (when VSX is available). This is otherwise | |
538 | the same as the pre-GCC 5 version. */ | |
539 | ||
44d95244 | 540 | ATTRIBUTE_NO_SANITIZE_UNDEFINED |
0ccaaab0 BS |
541 | static const uchar * |
542 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
543 | { | |
544 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
545 | ||
546 | const vc repl_nl = { | |
547 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
548 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
549 | }; | |
550 | const vc repl_cr = { | |
551 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
552 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
553 | }; | |
554 | const vc repl_bs = { | |
555 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
556 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
557 | }; | |
558 | const vc repl_qm = { | |
559 | '?', '?', '?', '?', '?', '?', '?', '?', | |
560 | '?', '?', '?', '?', '?', '?', '?', '?', | |
561 | }; | |
562 | const vc zero = { 0 }; | |
563 | ||
564 | vc data, t; | |
565 | ||
566 | /* Main loop processing 16 bytes at a time. */ | |
567 | do | |
568 | { | |
569 | vc m_nl, m_cr, m_bs, m_qm; | |
570 | ||
571 | data = *((const vc *)s); | |
572 | s += 16; | |
573 | ||
574 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
575 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
576 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
577 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
578 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
579 | ||
580 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
581 | characters. We want to exit the loop if any byte in T is non-zero. | |
582 | Below is the expansion of vec_any_ne(t, zero). */ | |
583 | } | |
584 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
585 | ||
586 | /* Restore s to to point to the 16 bytes we just processed. */ | |
587 | s -= 16; | |
588 | ||
589 | { | |
590 | #define N (sizeof(vc) / sizeof(long)) | |
591 | ||
592 | union { | |
593 | vc v; | |
594 | /* Statically assert that N is 2 or 4. */ | |
595 | unsigned long l[(N == 2 || N == 4) ? N : -1]; | |
596 | } u; | |
597 | unsigned long l, i = 0; | |
598 | ||
599 | u.v = t; | |
600 | ||
601 | /* Find the first word of T that is non-zero. */ | |
602 | switch (N) | |
603 | { | |
604 | case 4: | |
605 | l = u.l[i++]; | |
606 | if (l != 0) | |
607 | break; | |
608 | s += sizeof(unsigned long); | |
609 | l = u.l[i++]; | |
610 | if (l != 0) | |
611 | break; | |
612 | s += sizeof(unsigned long); | |
613 | case 2: | |
614 | l = u.l[i++]; | |
615 | if (l != 0) | |
616 | break; | |
617 | s += sizeof(unsigned long); | |
618 | l = u.l[i]; | |
619 | } | |
620 | ||
621 | /* L now contains 0xff in bytes for which we matched one of the | |
622 | relevant characters. We can find the byte index by finding | |
623 | its bit index and dividing by 8. */ | |
624 | #ifdef __BIG_ENDIAN__ | |
625 | l = __builtin_clzl(l) >> 3; | |
626 | #else | |
627 | l = __builtin_ctzl(l) >> 3; | |
628 | #endif | |
629 | return s + l; | |
630 | ||
631 | #undef N | |
632 | } | |
633 | } | |
634 | ||
635 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__) | |
636 | ||
637 | /* A vection of the fast scanner using AltiVec vectorized byte compares. | |
638 | This cannot be used for little endian because vec_lvsl/lvsr are | |
639 | deprecated for little endian and the code won't work properly. */ | |
246a2fcb RH |
640 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, |
641 | so we can't compile this function without -maltivec on the command line | |
642 | (or implied by some other switch). */ | |
643 | ||
644 | static const uchar * | |
645 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
646 | { | |
647 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
648 | ||
649 | const vc repl_nl = { | |
650 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
651 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
652 | }; | |
653 | const vc repl_cr = { | |
654 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
655 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
656 | }; | |
657 | const vc repl_bs = { | |
658 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
659 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
660 | }; | |
661 | const vc repl_qm = { | |
662 | '?', '?', '?', '?', '?', '?', '?', '?', | |
663 | '?', '?', '?', '?', '?', '?', '?', '?', | |
664 | }; | |
665 | const vc ones = { | |
666 | -1, -1, -1, -1, -1, -1, -1, -1, | |
667 | -1, -1, -1, -1, -1, -1, -1, -1, | |
668 | }; | |
669 | const vc zero = { 0 }; | |
670 | ||
671 | vc data, mask, t; | |
672 | ||
673 | /* Altivec loads automatically mask addresses with -16. This lets us | |
674 | issue the first load as early as possible. */ | |
675 | data = __builtin_vec_ld(0, (const vc *)s); | |
676 | ||
677 | /* Discard bytes before the beginning of the buffer. Do this by | |
678 | beginning with all ones and shifting in zeros according to the | |
679 | mis-alignment. The LVSR instruction pulls the exact shift we | |
680 | want from the address. */ | |
681 | mask = __builtin_vec_lvsr(0, s); | |
682 | mask = __builtin_vec_perm(zero, ones, mask); | |
683 | data &= mask; | |
684 | ||
685 | /* While altivec loads mask addresses, we still need to align S so | |
686 | that the offset we compute at the end is correct. */ | |
687 | s = (const uchar *)((uintptr_t)s & -16); | |
688 | ||
689 | /* Main loop processing 16 bytes at a time. */ | |
690 | goto start; | |
691 | do | |
692 | { | |
693 | vc m_nl, m_cr, m_bs, m_qm; | |
694 | ||
695 | s += 16; | |
696 | data = __builtin_vec_ld(0, (const vc *)s); | |
697 | ||
698 | start: | |
699 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
700 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
701 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
702 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
703 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
704 | ||
705 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
706 | characters. We want to exit the loop if any byte in T is non-zero. | |
707 | Below is the expansion of vec_any_ne(t, zero). */ | |
708 | } | |
709 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
710 | ||
711 | { | |
712 | #define N (sizeof(vc) / sizeof(long)) | |
713 | ||
246a2fcb RH |
714 | union { |
715 | vc v; | |
53a103d3 DS |
716 | /* Statically assert that N is 2 or 4. */ |
717 | unsigned long l[(N == 2 || N == 4) ? N : -1]; | |
246a2fcb RH |
718 | } u; |
719 | unsigned long l, i = 0; | |
720 | ||
721 | u.v = t; | |
722 | ||
723 | /* Find the first word of T that is non-zero. */ | |
724 | switch (N) | |
725 | { | |
726 | case 4: | |
727 | l = u.l[i++]; | |
728 | if (l != 0) | |
729 | break; | |
730 | s += sizeof(unsigned long); | |
731 | l = u.l[i++]; | |
732 | if (l != 0) | |
733 | break; | |
734 | s += sizeof(unsigned long); | |
735 | case 2: | |
736 | l = u.l[i++]; | |
737 | if (l != 0) | |
738 | break; | |
739 | s += sizeof(unsigned long); | |
740 | l = u.l[i]; | |
741 | } | |
742 | ||
743 | /* L now contains 0xff in bytes for which we matched one of the | |
744 | relevant characters. We can find the byte index by finding | |
745 | its bit index and dividing by 8. */ | |
746 | l = __builtin_clzl(l) >> 3; | |
747 | return s + l; | |
748 | ||
749 | #undef N | |
750 | } | |
751 | } | |
752 | ||
95d0610c | 753 | #elif defined (__ARM_NEON) |
e75b54a2 RE |
754 | #include "arm_neon.h" |
755 | ||
756 | static const uchar * | |
757 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
758 | { | |
759 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); | |
760 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); | |
761 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); | |
762 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); | |
763 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); | |
764 | ||
765 | unsigned int misalign, found, mask; | |
766 | const uint8_t *p; | |
767 | uint8x16_t data; | |
768 | ||
769 | /* Align the source pointer. */ | |
770 | misalign = (uintptr_t)s & 15; | |
771 | p = (const uint8_t *)((uintptr_t)s & -16); | |
772 | data = vld1q_u8 (p); | |
773 | ||
774 | /* Create a mask for the bytes that are valid within the first | |
775 | 16-byte block. The Idea here is that the AND with the mask | |
776 | within the loop is "free", since we need some AND or TEST | |
777 | insn in order to set the flags for the branch anyway. */ | |
778 | mask = (-1u << misalign) & 0xffff; | |
779 | ||
780 | /* Main loop, processing 16 bytes at a time. */ | |
781 | goto start; | |
782 | ||
783 | do | |
784 | { | |
785 | uint8x8_t l; | |
786 | uint16x4_t m; | |
787 | uint32x2_t n; | |
788 | uint8x16_t t, u, v, w; | |
789 | ||
790 | p += 16; | |
791 | data = vld1q_u8 (p); | |
792 | mask = 0xffff; | |
793 | ||
794 | start: | |
795 | t = vceqq_u8 (data, repl_nl); | |
796 | u = vceqq_u8 (data, repl_cr); | |
797 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
798 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
799 | t = vandq_u8 (vorrq_u8 (v, w), xmask); | |
800 | l = vpadd_u8 (vget_low_u8 (t), vget_high_u8 (t)); | |
801 | m = vpaddl_u8 (l); | |
802 | n = vpaddl_u16 (m); | |
803 | ||
804 | found = vget_lane_u32 ((uint32x2_t) vorr_u64 ((uint64x1_t) n, | |
805 | vshr_n_u64 ((uint64x1_t) n, 24)), 0); | |
806 | found &= mask; | |
807 | } | |
808 | while (!found); | |
809 | ||
810 | /* FOUND contains 1 in bits for which we matched a relevant | |
811 | character. Conversion to the byte index is trivial. */ | |
812 | found = __builtin_ctz (found); | |
813 | return (const uchar *)p + found; | |
814 | } | |
815 | ||
246a2fcb RH |
816 | #else |
817 | ||
818 | /* We only have one accellerated alternative. Use a direct call so that | |
819 | we encourage inlining. */ | |
820 | ||
821 | #define search_line_fast search_line_acc_char | |
822 | ||
823 | #endif | |
824 | ||
b0c084b7 JJ |
825 | /* Initialize the lexer if needed. */ |
826 | ||
827 | void | |
828 | _cpp_init_lexer (void) | |
829 | { | |
830 | #ifdef HAVE_init_vectorized_lexer | |
831 | init_vectorized_lexer (); | |
832 | #endif | |
833 | } | |
834 | ||
26aea073 NB |
835 | /* Returns with a logical line that contains no escaped newlines or |
836 | trigraphs. This is a time-critical inner loop. */ | |
837 | void | |
6cf87ca4 | 838 | _cpp_clean_line (cpp_reader *pfile) |
45b966db | 839 | { |
26aea073 NB |
840 | cpp_buffer *buffer; |
841 | const uchar *s; | |
842 | uchar c, *d, *p; | |
87062813 | 843 | |
26aea073 NB |
844 | buffer = pfile->buffer; |
845 | buffer->cur_note = buffer->notes_used = 0; | |
846 | buffer->cur = buffer->line_base = buffer->next_line; | |
847 | buffer->need_line = false; | |
246a2fcb | 848 | s = buffer->next_line; |
87062813 | 849 | |
26aea073 | 850 | if (!buffer->from_stage3) |
45b966db | 851 | { |
7af45bd4 ILT |
852 | const uchar *pbackslash = NULL; |
853 | ||
246a2fcb | 854 | /* Fast path. This is the common case of an un-escaped line with |
d08dcf87 ZW |
855 | no trigraphs. The primary win here is by not writing any |
856 | data back to memory until we have to. */ | |
246a2fcb | 857 | while (1) |
d08dcf87 | 858 | { |
246a2fcb RH |
859 | /* Perform an optimized search for \n, \r, \\, ?. */ |
860 | s = search_line_fast (s, buffer->rlimit); | |
d08dcf87 | 861 | |
246a2fcb RH |
862 | c = *s; |
863 | if (c == '\\') | |
864 | { | |
865 | /* Record the location of the backslash and continue. */ | |
866 | pbackslash = s++; | |
d08dcf87 | 867 | } |
246a2fcb | 868 | else if (__builtin_expect (c == '?', 0)) |
d08dcf87 | 869 | { |
246a2fcb RH |
870 | if (__builtin_expect (s[1] == '?', false) |
871 | && _cpp_trigraph_map[s[2]]) | |
d08dcf87 | 872 | { |
246a2fcb RH |
873 | /* Have a trigraph. We may or may not have to convert |
874 | it. Add a line note regardless, for -Wtrigraphs. */ | |
875 | add_line_note (buffer, s, s[2]); | |
876 | if (CPP_OPTION (pfile, trigraphs)) | |
877 | { | |
878 | /* We do, and that means we have to switch to the | |
879 | slow path. */ | |
880 | d = (uchar *) s; | |
881 | *d = _cpp_trigraph_map[s[2]]; | |
882 | s += 2; | |
883 | goto slow_path; | |
884 | } | |
d08dcf87 | 885 | } |
246a2fcb RH |
886 | /* Not a trigraph. Continue on fast-path. */ |
887 | s++; | |
d08dcf87 | 888 | } |
246a2fcb RH |
889 | else |
890 | break; | |
d08dcf87 ZW |
891 | } |
892 | ||
246a2fcb RH |
893 | /* This must be \r or \n. We're either done, or we'll be forced |
894 | to write back to the buffer and continue on the slow path. */ | |
895 | d = (uchar *) s; | |
896 | ||
897 | if (__builtin_expect (s == buffer->rlimit, false)) | |
898 | goto done; | |
899 | ||
900 | /* DOS line ending? */ | |
901 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
902 | { | |
903 | s++; | |
904 | if (s == buffer->rlimit) | |
905 | goto done; | |
906 | } | |
907 | ||
908 | if (__builtin_expect (pbackslash == NULL, true)) | |
909 | goto done; | |
910 | ||
911 | /* Check for escaped newline. */ | |
912 | p = d; | |
913 | while (is_nvspace (p[-1])) | |
914 | p--; | |
915 | if (p - 1 != pbackslash) | |
916 | goto done; | |
917 | ||
918 | /* Have an escaped newline; process it and proceed to | |
919 | the slow path. */ | |
920 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
921 | d = p - 2; | |
922 | buffer->next_line = p - 1; | |
26aea073 | 923 | |
246a2fcb RH |
924 | slow_path: |
925 | while (1) | |
4a5b68a2 | 926 | { |
26aea073 NB |
927 | c = *++s; |
928 | *++d = c; | |
929 | ||
930 | if (c == '\n' || c == '\r') | |
931 | { | |
246a2fcb | 932 | /* Handle DOS line endings. */ |
26aea073 NB |
933 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
934 | s++; | |
935 | if (s == buffer->rlimit) | |
936 | break; | |
937 | ||
938 | /* Escaped? */ | |
939 | p = d; | |
940 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
941 | p--; | |
942 | if (p == buffer->next_line || p[-1] != '\\') | |
943 | break; | |
944 | ||
41c32c98 | 945 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
26aea073 NB |
946 | d = p - 2; |
947 | buffer->next_line = p - 1; | |
948 | } | |
949 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
950 | { | |
951 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
41c32c98 | 952 | add_line_note (buffer, d, s[2]); |
26aea073 NB |
953 | if (CPP_OPTION (pfile, trigraphs)) |
954 | { | |
955 | *d = _cpp_trigraph_map[s[2]]; | |
956 | s += 2; | |
957 | } | |
958 | } | |
4a5b68a2 | 959 | } |
45b966db | 960 | } |
26aea073 NB |
961 | else |
962 | { | |
246a2fcb | 963 | while (*s != '\n' && *s != '\r') |
26aea073 | 964 | s++; |
26aea073 NB |
965 | d = (uchar *) s; |
966 | ||
967 | /* Handle DOS line endings. */ | |
968 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
969 | s++; | |
970 | } | |
0d9f234d | 971 | |
d08dcf87 | 972 | done: |
26aea073 | 973 | *d = '\n'; |
41c32c98 NB |
974 | /* A sentinel note that should never be processed. */ |
975 | add_line_note (buffer, d + 1, '\n'); | |
26aea073 | 976 | buffer->next_line = s + 1; |
45b966db ZW |
977 | } |
978 | ||
a8eb6044 NB |
979 | /* Return true if the trigraph indicated by NOTE should be warned |
980 | about in a comment. */ | |
981 | static bool | |
6cf87ca4 | 982 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
a8eb6044 NB |
983 | { |
984 | const uchar *p; | |
985 | ||
986 | /* Within comments we don't warn about trigraphs, unless the | |
987 | trigraph forms an escaped newline, as that may change | |
6356f892 | 988 | behavior. */ |
a8eb6044 NB |
989 | if (note->type != '/') |
990 | return false; | |
991 | ||
992 | /* If -trigraphs, then this was an escaped newline iff the next note | |
993 | is coincident. */ | |
994 | if (CPP_OPTION (pfile, trigraphs)) | |
995 | return note[1].pos == note->pos; | |
996 | ||
997 | /* Otherwise, see if this forms an escaped newline. */ | |
998 | p = note->pos + 3; | |
999 | while (is_nvspace (*p)) | |
1000 | p++; | |
1001 | ||
1002 | /* There might have been escaped newlines between the trigraph and the | |
1003 | newline we found. Hence the position test. */ | |
1004 | return (*p == '\n' && p < note[1].pos); | |
1005 | } | |
1006 | ||
26aea073 NB |
1007 | /* Process the notes created by add_line_note as far as the current |
1008 | location. */ | |
1009 | void | |
6cf87ca4 | 1010 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
45b966db | 1011 | { |
29401c30 NB |
1012 | cpp_buffer *buffer = pfile->buffer; |
1013 | ||
26aea073 | 1014 | for (;;) |
041c3194 | 1015 | { |
26aea073 NB |
1016 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
1017 | unsigned int col; | |
a5c3cccd | 1018 | |
26aea073 NB |
1019 | if (note->pos > buffer->cur) |
1020 | break; | |
a5c3cccd | 1021 | |
26aea073 NB |
1022 | buffer->cur_note++; |
1023 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
4d6baafa | 1024 | |
41c32c98 | 1025 | if (note->type == '\\' || note->type == ' ') |
26aea073 | 1026 | { |
41c32c98 | 1027 | if (note->type == ' ' && !in_comment) |
500bee0a | 1028 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
26aea073 | 1029 | "backslash and newline separated by space"); |
41c32c98 | 1030 | |
26aea073 | 1031 | if (buffer->next_line > buffer->rlimit) |
87062813 | 1032 | { |
500bee0a | 1033 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
26aea073 NB |
1034 | "backslash-newline at end of file"); |
1035 | /* Prevent "no newline at end of file" warning. */ | |
1036 | buffer->next_line = buffer->rlimit; | |
87062813 | 1037 | } |
26aea073 NB |
1038 | |
1039 | buffer->line_base = note->pos; | |
12f9df4e | 1040 | CPP_INCREMENT_LINE (pfile, 0); |
0d9f234d | 1041 | } |
41c32c98 NB |
1042 | else if (_cpp_trigraph_map[note->type]) |
1043 | { | |
a8eb6044 NB |
1044 | if (CPP_OPTION (pfile, warn_trigraphs) |
1045 | && (!in_comment || warn_in_comment (pfile, note))) | |
41c32c98 NB |
1046 | { |
1047 | if (CPP_OPTION (pfile, trigraphs)) | |
87cf0651 SB |
1048 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
1049 | pfile->line_table->highest_line, col, | |
1050 | "trigraph ??%c converted to %c", | |
1051 | note->type, | |
1052 | (int) _cpp_trigraph_map[note->type]); | |
41c32c98 | 1053 | else |
905bd7b5 | 1054 | { |
87cf0651 SB |
1055 | cpp_warning_with_line |
1056 | (pfile, CPP_W_TRIGRAPHS, | |
1057 | pfile->line_table->highest_line, col, | |
905bd7b5 GK |
1058 | "trigraph ??%c ignored, use -trigraphs to enable", |
1059 | note->type); | |
1060 | } | |
41c32c98 NB |
1061 | } |
1062 | } | |
00a81b8b JM |
1063 | else if (note->type == 0) |
1064 | /* Already processed in lex_raw_string. */; | |
41c32c98 NB |
1065 | else |
1066 | abort (); | |
041c3194 | 1067 | } |
45b966db ZW |
1068 | } |
1069 | ||
0d9f234d NB |
1070 | /* Skip a C-style block comment. We find the end of the comment by |
1071 | seeing if an asterisk is before every '/' we encounter. Returns | |
6f572ac2 NB |
1072 | nonzero if comment terminated by EOF, zero otherwise. |
1073 | ||
1074 | Buffer->cur points to the initial asterisk of the comment. */ | |
26aea073 | 1075 | bool |
6cf87ca4 | 1076 | _cpp_skip_block_comment (cpp_reader *pfile) |
45b966db | 1077 | { |
041c3194 | 1078 | cpp_buffer *buffer = pfile->buffer; |
d08dcf87 ZW |
1079 | const uchar *cur = buffer->cur; |
1080 | uchar c; | |
0d9f234d | 1081 | |
d08dcf87 ZW |
1082 | cur++; |
1083 | if (*cur == '/') | |
1084 | cur++; | |
0d9f234d | 1085 | |
26aea073 NB |
1086 | for (;;) |
1087 | { | |
0d9f234d NB |
1088 | /* People like decorating comments with '*', so check for '/' |
1089 | instead for efficiency. */ | |
d08dcf87 ZW |
1090 | c = *cur++; |
1091 | ||
041c3194 | 1092 | if (c == '/') |
45b966db | 1093 | { |
d08dcf87 | 1094 | if (cur[-2] == '*') |
0d9f234d | 1095 | break; |
041c3194 | 1096 | |
0d9f234d | 1097 | /* Warn about potential nested comments, but not if the '/' |
a1f300c0 | 1098 | comes immediately before the true comment delimiter. |
041c3194 | 1099 | Don't bother to get it right across escaped newlines. */ |
0d9f234d | 1100 | if (CPP_OPTION (pfile, warn_comments) |
d08dcf87 ZW |
1101 | && cur[0] == '*' && cur[1] != '/') |
1102 | { | |
1103 | buffer->cur = cur; | |
87cf0651 SB |
1104 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
1105 | pfile->line_table->highest_line, | |
1106 | CPP_BUF_COL (buffer), | |
1107 | "\"/*\" within comment"); | |
d08dcf87 | 1108 | } |
45b966db | 1109 | } |
26aea073 NB |
1110 | else if (c == '\n') |
1111 | { | |
12f9df4e | 1112 | unsigned int cols; |
d08dcf87 | 1113 | buffer->cur = cur - 1; |
26aea073 NB |
1114 | _cpp_process_line_notes (pfile, true); |
1115 | if (buffer->next_line >= buffer->rlimit) | |
1116 | return true; | |
1117 | _cpp_clean_line (pfile); | |
12f9df4e PB |
1118 | |
1119 | cols = buffer->next_line - buffer->line_base; | |
1120 | CPP_INCREMENT_LINE (pfile, cols); | |
1121 | ||
d08dcf87 | 1122 | cur = buffer->cur; |
26aea073 | 1123 | } |
45b966db | 1124 | } |
041c3194 | 1125 | |
d08dcf87 | 1126 | buffer->cur = cur; |
a8eb6044 | 1127 | _cpp_process_line_notes (pfile, true); |
26aea073 | 1128 | return false; |
45b966db ZW |
1129 | } |
1130 | ||
480709cc | 1131 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
da7d8304 | 1132 | terminating newline. Handles escaped newlines. Returns nonzero |
480709cc | 1133 | if a multiline comment. */ |
041c3194 | 1134 | static int |
6cf87ca4 | 1135 | skip_line_comment (cpp_reader *pfile) |
45b966db | 1136 | { |
cbcff6df | 1137 | cpp_buffer *buffer = pfile->buffer; |
1bb64668 | 1138 | source_location orig_line = pfile->line_table->highest_line; |
041c3194 | 1139 | |
26aea073 NB |
1140 | while (*buffer->cur != '\n') |
1141 | buffer->cur++; | |
480709cc | 1142 | |
26aea073 | 1143 | _cpp_process_line_notes (pfile, true); |
500bee0a | 1144 | return orig_line != pfile->line_table->highest_line; |
041c3194 | 1145 | } |
45b966db | 1146 | |
26aea073 | 1147 | /* Skips whitespace, saving the next non-whitespace character. */ |
52fadca8 | 1148 | static void |
6cf87ca4 | 1149 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
041c3194 ZW |
1150 | { |
1151 | cpp_buffer *buffer = pfile->buffer; | |
f7d151fb | 1152 | bool saw_NUL = false; |
45b966db | 1153 | |
0d9f234d | 1154 | do |
041c3194 | 1155 | { |
91fcd158 | 1156 | /* Horizontal space always OK. */ |
26aea073 | 1157 | if (c == ' ' || c == '\t') |
0d9f234d | 1158 | ; |
0d9f234d | 1159 | /* Just \f \v or \0 left. */ |
91fcd158 | 1160 | else if (c == '\0') |
f7d151fb | 1161 | saw_NUL = true; |
93c80368 | 1162 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
500bee0a | 1163 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
ebef4e8c NB |
1164 | CPP_BUF_COL (buffer), |
1165 | "%s in preprocessing directive", | |
1166 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d | 1167 | |
0d9f234d | 1168 | c = *buffer->cur++; |
45b966db | 1169 | } |
ec5c56db | 1170 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
1171 | while (is_nvspace (c)); |
1172 | ||
f7d151fb | 1173 | if (saw_NUL) |
0527bc4e | 1174 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
f7d151fb | 1175 | |
480709cc | 1176 | buffer->cur--; |
041c3194 | 1177 | } |
45b966db | 1178 | |
93c80368 NB |
1179 | /* See if the characters of a number token are valid in a name (no |
1180 | '.', '+' or '-'). */ | |
1181 | static int | |
6cf87ca4 | 1182 | name_p (cpp_reader *pfile, const cpp_string *string) |
93c80368 NB |
1183 | { |
1184 | unsigned int i; | |
1185 | ||
1186 | for (i = 0; i < string->len; i++) | |
1187 | if (!is_idchar (string->text[i])) | |
1188 | return 0; | |
1189 | ||
df383483 | 1190 | return 1; |
93c80368 NB |
1191 | } |
1192 | ||
50668cf6 GK |
1193 | /* After parsing an identifier or other sequence, produce a warning about |
1194 | sequences not in NFC/NFKC. */ | |
1195 | static void | |
1196 | warn_about_normalization (cpp_reader *pfile, | |
1197 | const cpp_token *token, | |
1198 | const struct normalize_state *s) | |
1199 | { | |
1200 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1201 | && !pfile->state.skipping) | |
1202 | { | |
1203 | /* Make sure that the token is printed using UCNs, even | |
1204 | if we'd otherwise happily print UTF-8. */ | |
c3f829c1 | 1205 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
50668cf6 GK |
1206 | size_t sz; |
1207 | ||
1208 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1209 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
87cf0651 SB |
1210 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1211 | "`%.*s' is not in NFKC", (int) sz, buf); | |
50668cf6 | 1212 | else |
87cf0651 SB |
1213 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1214 | "`%.*s' is not in NFC", (int) sz, buf); | |
55e7f907 | 1215 | free (buf); |
50668cf6 GK |
1216 | } |
1217 | } | |
1218 | ||
bced6edf | 1219 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in |
1613e52b | 1220 | an identifier. FIRST is TRUE if this starts an identifier. */ |
bced6edf | 1221 | static bool |
50668cf6 GK |
1222 | forms_identifier_p (cpp_reader *pfile, int first, |
1223 | struct normalize_state *state) | |
bced6edf | 1224 | { |
1613e52b NB |
1225 | cpp_buffer *buffer = pfile->buffer; |
1226 | ||
1227 | if (*buffer->cur == '$') | |
1228 | { | |
1229 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1230 | return false; | |
1231 | ||
1232 | buffer->cur++; | |
78b8811a | 1233 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
1613e52b | 1234 | { |
78b8811a | 1235 | CPP_OPTION (pfile, warn_dollars) = 0; |
0527bc4e | 1236 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
1613e52b NB |
1237 | } |
1238 | ||
1239 | return true; | |
1240 | } | |
bced6edf | 1241 | |
1613e52b | 1242 | /* Is this a syntactically valid UCN? */ |
af15a2fe | 1243 | if (CPP_OPTION (pfile, extended_identifiers) |
6baba9bb | 1244 | && *buffer->cur == '\\' |
1613e52b | 1245 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) |
bced6edf | 1246 | { |
fbb22910 | 1247 | cppchar_t s; |
1613e52b | 1248 | buffer->cur += 2; |
50668cf6 | 1249 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
fbb22910 | 1250 | state, &s)) |
1613e52b NB |
1251 | return true; |
1252 | buffer->cur -= 2; | |
bced6edf | 1253 | } |
bced6edf | 1254 | |
1613e52b | 1255 | return false; |
bced6edf NB |
1256 | } |
1257 | ||
17e7cb85 KT |
1258 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1259 | static cpp_hashnode * | |
1260 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1261 | { | |
1262 | cpp_hashnode *result; | |
1263 | const uchar *cur; | |
1264 | unsigned int len; | |
1265 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1266 | ||
1267 | cur = base + 1; | |
1268 | while (ISIDNUM (*cur)) | |
1269 | { | |
1270 | hash = HT_HASHSTEP (hash, *cur); | |
1271 | cur++; | |
1272 | } | |
1273 | len = cur - base; | |
1274 | hash = HT_HASHFINISH (hash, len); | |
1275 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1276 | base, len, hash, HT_ALLOC)); | |
1277 | ||
1278 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1279 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1280 | && !pfile->state.skipping, 0)) | |
1281 | { | |
1282 | /* It is allowed to poison the same identifier twice. */ | |
1283 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1284 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1285 | NODE_NAME (result)); | |
1286 | ||
1287 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1288 | replacement list of a variadic macro. */ | |
1289 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1290 | && !pfile->state.va_args_ok) | |
3976796b ESR |
1291 | { |
1292 | if (CPP_OPTION (pfile, cplusplus)) | |
1293 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1294 | "__VA_ARGS__ can only appear in the expansion" | |
1295 | " of a C++11 variadic macro"); | |
1296 | else | |
1297 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1298 | "__VA_ARGS__ can only appear in the expansion" | |
1299 | " of a C99 variadic macro"); | |
1300 | } | |
17e7cb85 KT |
1301 | |
1302 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1303 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1304 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1305 | "identifier \"%s\" is a special operator name in C++", | |
1306 | NODE_NAME (result)); | |
17e7cb85 KT |
1307 | } |
1308 | ||
1309 | return result; | |
1310 | } | |
1311 | ||
1312 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1313 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1314 | cpp_hashnode * | |
1315 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1316 | { | |
1317 | cpp_hashnode *result; | |
1318 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1319 | return result; | |
1320 | } | |
1321 | ||
bced6edf | 1322 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
0d9f234d | 1323 | static cpp_hashnode * |
50668cf6 | 1324 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
be5ffc59 | 1325 | struct normalize_state *nst, cpp_hashnode **spelling) |
45b966db | 1326 | { |
93c80368 | 1327 | cpp_hashnode *result; |
47e20491 | 1328 | const uchar *cur; |
c6e83800 ZW |
1329 | unsigned int len; |
1330 | unsigned int hash = HT_HASHSTEP (0, *base); | |
2c3fcba6 | 1331 | |
c6e83800 | 1332 | cur = pfile->buffer->cur; |
47e20491 | 1333 | if (! starts_ucn) |
d3f4ff8b JM |
1334 | { |
1335 | while (ISIDNUM (*cur)) | |
1336 | { | |
1337 | hash = HT_HASHSTEP (hash, *cur); | |
1338 | cur++; | |
1339 | } | |
1340 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *(cur - 1)); | |
1341 | } | |
47e20491 | 1342 | pfile->buffer->cur = cur; |
50668cf6 | 1343 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
10cf9bde | 1344 | { |
47e20491 GK |
1345 | /* Slower version for identifiers containing UCNs (or $). */ |
1346 | do { | |
1347 | while (ISIDNUM (*pfile->buffer->cur)) | |
50668cf6 | 1348 | { |
d3f4ff8b | 1349 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *pfile->buffer->cur); |
50668cf6 | 1350 | pfile->buffer->cur++; |
50668cf6 GK |
1351 | } |
1352 | } while (forms_identifier_p (pfile, false, nst)); | |
47e20491 GK |
1353 | result = _cpp_interpret_identifier (pfile, base, |
1354 | pfile->buffer->cur - base); | |
be5ffc59 | 1355 | *spelling = cpp_lookup (pfile, base, pfile->buffer->cur - base); |
2c3fcba6 | 1356 | } |
47e20491 GK |
1357 | else |
1358 | { | |
1359 | len = cur - base; | |
1360 | hash = HT_HASHFINISH (hash, len); | |
bced6edf | 1361 | |
2bf41bf0 TT |
1362 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1363 | base, len, hash, HT_ALLOC)); | |
be5ffc59 | 1364 | *spelling = result; |
47e20491 | 1365 | } |
2c3fcba6 | 1366 | |
bced6edf | 1367 | /* Rarely, identifiers require diagnostics when lexed. */ |
2c3fcba6 ZW |
1368 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1369 | && !pfile->state.skipping, 0)) | |
1370 | { | |
1371 | /* It is allowed to poison the same identifier twice. */ | |
1372 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
0527bc4e | 1373 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
2c3fcba6 ZW |
1374 | NODE_NAME (result)); |
1375 | ||
1376 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1377 | replacement list of a variadic macro. */ | |
1378 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1379 | && !pfile->state.va_args_ok) | |
3976796b ESR |
1380 | { |
1381 | if (CPP_OPTION (pfile, cplusplus)) | |
1382 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1383 | "__VA_ARGS__ can only appear in the expansion" | |
1384 | " of a C++11 variadic macro"); | |
1385 | else | |
1386 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1387 | "__VA_ARGS__ can only appear in the expansion" | |
1388 | " of a C99 variadic macro"); | |
1389 | } | |
3d8b2a98 ILT |
1390 | |
1391 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1392 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1393 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1394 | "identifier \"%s\" is a special operator name in C++", | |
1395 | NODE_NAME (result)); | |
2c3fcba6 ZW |
1396 | } |
1397 | ||
1398 | return result; | |
1399 | } | |
1400 | ||
bced6edf | 1401 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
45b966db | 1402 | static void |
50668cf6 GK |
1403 | lex_number (cpp_reader *pfile, cpp_string *number, |
1404 | struct normalize_state *nst) | |
45b966db | 1405 | { |
562a5c27 | 1406 | const uchar *cur; |
bced6edf NB |
1407 | const uchar *base; |
1408 | uchar *dest; | |
45b966db | 1409 | |
bced6edf NB |
1410 | base = pfile->buffer->cur - 1; |
1411 | do | |
041c3194 | 1412 | { |
bced6edf | 1413 | cur = pfile->buffer->cur; |
0d9f234d | 1414 | |
bced6edf | 1415 | /* N.B. ISIDNUM does not include $. */ |
7057e645 ESR |
1416 | while (ISIDNUM (*cur) || *cur == '.' || DIGIT_SEP (*cur) |
1417 | || VALID_SIGN (*cur, cur[-1])) | |
50668cf6 | 1418 | { |
d3f4ff8b | 1419 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *cur); |
50668cf6 | 1420 | cur++; |
50668cf6 | 1421 | } |
a5858a3d ESR |
1422 | /* A number can't end with a digit separator. */ |
1423 | while (cur > pfile->buffer->cur && DIGIT_SEP (cur[-1])) | |
1424 | --cur; | |
45b966db | 1425 | |
10cf9bde | 1426 | pfile->buffer->cur = cur; |
45b966db | 1427 | } |
50668cf6 | 1428 | while (forms_identifier_p (pfile, false, nst)); |
93c80368 | 1429 | |
bced6edf NB |
1430 | number->len = cur - base; |
1431 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1432 | memcpy (dest, base, number->len); | |
1433 | dest[number->len] = '\0'; | |
1434 | number->text = dest; | |
93c80368 NB |
1435 | } |
1436 | ||
6338b358 NB |
1437 | /* Create a token of type TYPE with a literal spelling. */ |
1438 | static void | |
6cf87ca4 ZW |
1439 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1440 | unsigned int len, enum cpp_ttype type) | |
6338b358 NB |
1441 | { |
1442 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1443 | ||
1444 | memcpy (dest, base, len); | |
1445 | dest[len] = '\0'; | |
1446 | token->type = type; | |
1447 | token->val.str.len = len; | |
1448 | token->val.str.text = dest; | |
1449 | } | |
1450 | ||
00a81b8b JM |
1451 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1452 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1453 | ||
1454 | static void | |
1455 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1456 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1457 | { | |
1458 | _cpp_buff *first_buff = *first_buff_p; | |
1459 | _cpp_buff *last_buff = *last_buff_p; | |
1460 | ||
1461 | if (first_buff == NULL) | |
1462 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1463 | else if (len > BUFF_ROOM (last_buff)) | |
1464 | { | |
1465 | size_t room = BUFF_ROOM (last_buff); | |
1466 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1467 | BUFF_FRONT (last_buff) += room; | |
1468 | base += room; | |
1469 | len -= room; | |
1470 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1471 | } | |
1472 | ||
1473 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1474 | BUFF_FRONT (last_buff) += len; | |
1475 | ||
1476 | *first_buff_p = first_buff; | |
1477 | *last_buff_p = last_buff; | |
1478 | } | |
1479 | ||
c865f923 ESR |
1480 | |
1481 | /* Returns true if a macro has been defined. | |
1482 | This might not work if compile with -save-temps, | |
1483 | or preprocess separately from compilation. */ | |
1484 | ||
1485 | static bool | |
1486 | is_macro(cpp_reader *pfile, const uchar *base) | |
1487 | { | |
1488 | const uchar *cur = base; | |
1489 | if (! ISIDST (*cur)) | |
1490 | return false; | |
1491 | unsigned int hash = HT_HASHSTEP (0, *cur); | |
1492 | ++cur; | |
1493 | while (ISIDNUM (*cur)) | |
1494 | { | |
1495 | hash = HT_HASHSTEP (hash, *cur); | |
1496 | ++cur; | |
1497 | } | |
1498 | hash = HT_HASHFINISH (hash, cur - base); | |
1499 | ||
1500 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1501 | base, cur - base, hash, HT_NO_INSERT)); | |
1502 | ||
1503 | return !result ? false : (result->type == NT_MACRO); | |
1504 | } | |
1505 | ||
1506 | ||
2c6e3f55 | 1507 | /* Lexes a raw string. The stored string contains the spelling, including |
00a81b8b | 1508 | double quotes, delimiter string, '(' and ')', any leading |
2c6e3f55 JJ |
1509 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1510 | literal, or CPP_OTHER if it was not properly terminated. | |
1511 | ||
1512 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1513 | is the first NUL since embedded NULs are preserved. */ | |
1514 | ||
1515 | static void | |
1516 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1517 | const uchar *cur) | |
1518 | { | |
8cf88735 JJ |
1519 | uchar raw_prefix[17]; |
1520 | uchar temp_buffer[18]; | |
1521 | const uchar *orig_base; | |
1522 | unsigned int raw_prefix_len = 0, raw_suffix_len = 0; | |
1523 | enum raw_str_phase { RAW_STR_PREFIX, RAW_STR, RAW_STR_SUFFIX }; | |
1524 | raw_str_phase phase = RAW_STR_PREFIX; | |
2c6e3f55 JJ |
1525 | enum cpp_ttype type; |
1526 | size_t total_len = 0; | |
8cf88735 JJ |
1527 | /* Index into temp_buffer during phases other than RAW_STR, |
1528 | during RAW_STR phase 17 to tell BUF_APPEND that nothing should | |
1529 | be appended to temp_buffer. */ | |
1530 | size_t temp_buffer_len = 0; | |
2c6e3f55 | 1531 | _cpp_buff *first_buff = NULL, *last_buff = NULL; |
8cf88735 | 1532 | size_t raw_prefix_start; |
00a81b8b | 1533 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 JJ |
1534 | |
1535 | type = (*base == 'L' ? CPP_WSTRING : | |
1536 | *base == 'U' ? CPP_STRING32 : | |
1537 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1538 | : CPP_STRING); | |
1539 | ||
00a81b8b JM |
1540 | #define BUF_APPEND(STR,LEN) \ |
1541 | do { \ | |
1542 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1543 | &first_buff, &last_buff); \ | |
1544 | total_len += (LEN); \ | |
8cf88735 JJ |
1545 | if (__builtin_expect (temp_buffer_len < 17, 0) \ |
1546 | && (const uchar *)(STR) != base \ | |
1547 | && (LEN) <= 2) \ | |
1548 | { \ | |
1549 | memcpy (temp_buffer + temp_buffer_len, \ | |
1550 | (const uchar *)(STR), (LEN)); \ | |
1551 | temp_buffer_len += (LEN); \ | |
1552 | } \ | |
00a81b8b JM |
1553 | } while (0); |
1554 | ||
8cf88735 JJ |
1555 | orig_base = base; |
1556 | ++cur; | |
1557 | raw_prefix_start = cur - base; | |
1558 | for (;;) | |
1559 | { | |
00a81b8b JM |
1560 | cppchar_t c; |
1561 | ||
1562 | /* If we previously performed any trigraph or line splicing | |
8cf88735 JJ |
1563 | transformations, undo them in between the opening and closing |
1564 | double quote. */ | |
00a81b8b JM |
1565 | while (note->pos < cur) |
1566 | ++note; | |
1567 | for (; note->pos == cur; ++note) | |
1568 | { | |
1569 | switch (note->type) | |
1570 | { | |
1571 | case '\\': | |
1572 | case ' ': | |
1573 | /* Restore backslash followed by newline. */ | |
1574 | BUF_APPEND (base, cur - base); | |
1575 | base = cur; | |
1576 | BUF_APPEND ("\\", 1); | |
1577 | after_backslash: | |
1578 | if (note->type == ' ') | |
1579 | { | |
1580 | /* GNU backslash whitespace newline extension. FIXME | |
1581 | could be any sequence of non-vertical space. When we | |
1582 | can properly restore any such sequence, we should mark | |
1583 | this note as handled so _cpp_process_line_notes | |
1584 | doesn't warn. */ | |
1585 | BUF_APPEND (" ", 1); | |
1586 | } | |
1587 | ||
1588 | BUF_APPEND ("\n", 1); | |
1589 | break; | |
1590 | ||
1591 | case 0: | |
1592 | /* Already handled. */ | |
1593 | break; | |
1594 | ||
1595 | default: | |
1596 | if (_cpp_trigraph_map[note->type]) | |
1597 | { | |
1598 | /* Don't warn about this trigraph in | |
1599 | _cpp_process_line_notes, since trigraphs show up as | |
1600 | trigraphs in raw strings. */ | |
d947ada0 | 1601 | uchar type = note->type; |
00a81b8b JM |
1602 | note->type = 0; |
1603 | ||
1604 | if (!CPP_OPTION (pfile, trigraphs)) | |
1605 | /* If we didn't convert the trigraph in the first | |
1606 | place, don't do anything now either. */ | |
1607 | break; | |
1608 | ||
1609 | BUF_APPEND (base, cur - base); | |
1610 | base = cur; | |
1611 | BUF_APPEND ("??", 2); | |
1612 | ||
1613 | /* ??/ followed by newline gets two line notes, one for | |
1614 | the trigraph and one for the backslash/newline. */ | |
1615 | if (type == '/' && note[1].pos == cur) | |
1616 | { | |
1617 | if (note[1].type != '\\' | |
1618 | && note[1].type != ' ') | |
1619 | abort (); | |
1620 | BUF_APPEND ("/", 1); | |
1621 | ++note; | |
1622 | goto after_backslash; | |
1623 | } | |
00a81b8b JM |
1624 | else |
1625 | { | |
1626 | /* Skip the replacement character. */ | |
1627 | base = ++cur; | |
1628 | BUF_APPEND (&type, 1); | |
8cf88735 JJ |
1629 | c = type; |
1630 | goto check_c; | |
00a81b8b JM |
1631 | } |
1632 | } | |
1633 | else | |
1634 | abort (); | |
1635 | break; | |
1636 | } | |
1637 | } | |
1638 | c = *cur++; | |
8cf88735 JJ |
1639 | if (__builtin_expect (temp_buffer_len < 17, 0)) |
1640 | temp_buffer[temp_buffer_len++] = c; | |
2c6e3f55 | 1641 | |
8cf88735 JJ |
1642 | check_c: |
1643 | if (phase == RAW_STR_PREFIX) | |
2c6e3f55 | 1644 | { |
8cf88735 JJ |
1645 | while (raw_prefix_len < temp_buffer_len) |
1646 | { | |
1647 | raw_prefix[raw_prefix_len] = temp_buffer[raw_prefix_len]; | |
1648 | switch (raw_prefix[raw_prefix_len]) | |
1649 | { | |
1650 | case ' ': case '(': case ')': case '\\': case '\t': | |
1651 | case '\v': case '\f': case '\n': default: | |
1652 | break; | |
1653 | /* Basic source charset except the above chars. */ | |
1654 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1655 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1656 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1657 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1658 | case 'y': case 'z': | |
1659 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1660 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1661 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1662 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1663 | case 'Y': case 'Z': | |
1664 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1665 | case '6': case '7': case '8': case '9': | |
1666 | case '_': case '{': case '}': case '#': case '[': case ']': | |
1667 | case '<': case '>': case '%': case ':': case ';': case '.': | |
1668 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1669 | case '&': case '|': case '~': case '!': case '=': case ',': | |
1670 | case '"': case '\'': | |
1671 | if (raw_prefix_len < 16) | |
1672 | { | |
1673 | raw_prefix_len++; | |
1674 | continue; | |
1675 | } | |
1676 | break; | |
1677 | } | |
1678 | ||
1679 | if (raw_prefix[raw_prefix_len] != '(') | |
1680 | { | |
1681 | int col = CPP_BUF_COLUMN (pfile->buffer, cur) + 1; | |
1682 | if (raw_prefix_len == 16) | |
1683 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1684 | col, "raw string delimiter longer " | |
1685 | "than 16 characters"); | |
1686 | else if (raw_prefix[raw_prefix_len] == '\n') | |
1687 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1688 | col, "invalid new-line in raw " | |
1689 | "string delimiter"); | |
1690 | else | |
1691 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1692 | col, "invalid character '%c' in " | |
1693 | "raw string delimiter", | |
1694 | (int) raw_prefix[raw_prefix_len]); | |
1695 | pfile->buffer->cur = orig_base + raw_prefix_start - 1; | |
1696 | create_literal (pfile, token, orig_base, | |
1697 | raw_prefix_start - 1, CPP_OTHER); | |
1698 | if (first_buff) | |
1699 | _cpp_release_buff (pfile, first_buff); | |
1700 | return; | |
1701 | } | |
1702 | raw_prefix[raw_prefix_len] = '"'; | |
1703 | phase = RAW_STR; | |
1704 | /* Nothing should be appended to temp_buffer during | |
1705 | RAW_STR phase. */ | |
1706 | temp_buffer_len = 17; | |
1707 | break; | |
1708 | } | |
1709 | continue; | |
1710 | } | |
1711 | else if (phase == RAW_STR_SUFFIX) | |
1712 | { | |
1713 | while (raw_suffix_len <= raw_prefix_len | |
1714 | && raw_suffix_len < temp_buffer_len | |
1715 | && temp_buffer[raw_suffix_len] == raw_prefix[raw_suffix_len]) | |
1716 | raw_suffix_len++; | |
1717 | if (raw_suffix_len > raw_prefix_len) | |
1718 | break; | |
1719 | if (raw_suffix_len == temp_buffer_len) | |
1720 | continue; | |
1721 | phase = RAW_STR; | |
1722 | /* Nothing should be appended to temp_buffer during | |
1723 | RAW_STR phase. */ | |
1724 | temp_buffer_len = 17; | |
1725 | } | |
1726 | if (c == ')') | |
1727 | { | |
1728 | phase = RAW_STR_SUFFIX; | |
1729 | raw_suffix_len = 0; | |
1730 | temp_buffer_len = 0; | |
2c6e3f55 JJ |
1731 | } |
1732 | else if (c == '\n') | |
1733 | { | |
1734 | if (pfile->state.in_directive | |
d5e48350 JJ |
1735 | || (pfile->state.parsing_args |
1736 | && pfile->buffer->next_line >= pfile->buffer->rlimit)) | |
2c6e3f55 JJ |
1737 | { |
1738 | cur--; | |
1739 | type = CPP_OTHER; | |
1740 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1741 | "unterminated raw string"); | |
1742 | break; | |
1743 | } | |
1744 | ||
00a81b8b | 1745 | BUF_APPEND (base, cur - base); |
2c6e3f55 JJ |
1746 | |
1747 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1748 | CPP_INCREMENT_LINE (pfile, 0); | |
1749 | pfile->buffer->need_line = true; | |
1750 | ||
00a81b8b JM |
1751 | pfile->buffer->cur = cur-1; |
1752 | _cpp_process_line_notes (pfile, false); | |
2c6e3f55 JJ |
1753 | if (!_cpp_get_fresh_line (pfile)) |
1754 | { | |
1755 | source_location src_loc = token->src_loc; | |
1756 | token->type = CPP_EOF; | |
1757 | /* Tell the compiler the line number of the EOF token. */ | |
1758 | token->src_loc = pfile->line_table->highest_line; | |
1759 | token->flags = BOL; | |
1760 | if (first_buff != NULL) | |
1761 | _cpp_release_buff (pfile, first_buff); | |
1762 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1763 | "unterminated raw string"); | |
1764 | return; | |
1765 | } | |
1766 | ||
1767 | cur = base = pfile->buffer->cur; | |
00a81b8b | 1768 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 1769 | } |
2c6e3f55 JJ |
1770 | } |
1771 | ||
3ce4f9e4 ESR |
1772 | if (CPP_OPTION (pfile, user_literals)) |
1773 | { | |
c865f923 ESR |
1774 | /* If a string format macro, say from inttypes.h, is placed touching |
1775 | a string literal it could be parsed as a C++11 user-defined string | |
1776 | literal thus breaking the program. | |
1777 | Try to identify macros with is_macro. A warning is issued. */ | |
1778 | if (is_macro (pfile, cur)) | |
7f5f5f98 | 1779 | { |
112448b4 | 1780 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 1781 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
1782 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1783 | token->src_loc, 0, | |
1784 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 1785 | "a space between literal and string macro"); |
7f5f5f98 | 1786 | } |
3ce4f9e4 | 1787 | /* Grab user defined literal suffix. */ |
561f7fc7 | 1788 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
1789 | { |
1790 | type = cpp_userdef_string_add_type (type); | |
1791 | ++cur; | |
7f5f5f98 OW |
1792 | |
1793 | while (ISIDNUM (*cur)) | |
1794 | ++cur; | |
3ce4f9e4 | 1795 | } |
3ce4f9e4 ESR |
1796 | } |
1797 | ||
2c6e3f55 JJ |
1798 | pfile->buffer->cur = cur; |
1799 | if (first_buff == NULL) | |
1800 | create_literal (pfile, token, base, cur - base, type); | |
1801 | else | |
1802 | { | |
1803 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1804 | ||
1805 | token->type = type; | |
1806 | token->val.str.len = total_len + (cur - base); | |
1807 | token->val.str.text = dest; | |
1808 | last_buff = first_buff; | |
1809 | while (last_buff != NULL) | |
1810 | { | |
1811 | memcpy (dest, last_buff->base, | |
1812 | BUFF_FRONT (last_buff) - last_buff->base); | |
1813 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1814 | last_buff = last_buff->next; | |
1815 | } | |
1816 | _cpp_release_buff (pfile, first_buff); | |
1817 | memcpy (dest, base, cur - base); | |
1818 | dest[cur - base] = '\0'; | |
1819 | } | |
1820 | } | |
1821 | ||
bced6edf | 1822 | /* Lexes a string, character constant, or angle-bracketed header file |
6338b358 | 1823 | name. The stored string contains the spelling, including opening |
2c6e3f55 JJ |
1824 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1825 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1826 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1827 | header name which must be relexed as normal tokens. | |
6338b358 NB |
1828 | |
1829 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1830 | is the first NUL since embedded NULs are preserved. */ | |
041c3194 | 1831 | static void |
6cf87ca4 | 1832 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
45b966db | 1833 | { |
6338b358 NB |
1834 | bool saw_NUL = false; |
1835 | const uchar *cur; | |
bced6edf | 1836 | cppchar_t terminator; |
6338b358 NB |
1837 | enum cpp_ttype type; |
1838 | ||
1839 | cur = base; | |
1840 | terminator = *cur++; | |
2c6e3f55 | 1841 | if (terminator == 'L' || terminator == 'U') |
6338b358 | 1842 | terminator = *cur++; |
2c6e3f55 JJ |
1843 | else if (terminator == 'u') |
1844 | { | |
1845 | terminator = *cur++; | |
1846 | if (terminator == '8') | |
1847 | terminator = *cur++; | |
1848 | } | |
1849 | if (terminator == 'R') | |
1850 | { | |
1851 | lex_raw_string (pfile, token, base, cur); | |
1852 | return; | |
1853 | } | |
1854 | if (terminator == '"') | |
b6baa67d KVH |
1855 | type = (*base == 'L' ? CPP_WSTRING : |
1856 | *base == 'U' ? CPP_STRING32 : | |
2c6e3f55 JJ |
1857 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
1858 | : CPP_STRING); | |
6338b358 | 1859 | else if (terminator == '\'') |
b6baa67d KVH |
1860 | type = (*base == 'L' ? CPP_WCHAR : |
1861 | *base == 'U' ? CPP_CHAR32 : | |
fe95b036 ESR |
1862 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8CHAR : CPP_CHAR16) |
1863 | : CPP_CHAR); | |
6338b358 NB |
1864 | else |
1865 | terminator = '>', type = CPP_HEADER_NAME; | |
93c80368 | 1866 | |
0d9f234d | 1867 | for (;;) |
45b966db | 1868 | { |
6338b358 | 1869 | cppchar_t c = *cur++; |
7868b4a2 | 1870 | |
6f572ac2 | 1871 | /* In #include-style directives, terminators are not escapable. */ |
6338b358 NB |
1872 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
1873 | cur++; | |
1874 | else if (c == terminator) | |
bced6edf | 1875 | break; |
6338b358 | 1876 | else if (c == '\n') |
0d9f234d | 1877 | { |
6338b358 | 1878 | cur--; |
4bb09c26 JM |
1879 | /* Unmatched quotes always yield undefined behavior, but |
1880 | greedy lexing means that what appears to be an unterminated | |
1881 | header name may actually be a legitimate sequence of tokens. */ | |
1882 | if (terminator == '>') | |
1883 | { | |
1884 | token->type = CPP_LESS; | |
1885 | return; | |
1886 | } | |
6338b358 NB |
1887 | type = CPP_OTHER; |
1888 | break; | |
45b966db | 1889 | } |
6338b358 NB |
1890 | else if (c == '\0') |
1891 | saw_NUL = true; | |
45b966db ZW |
1892 | } |
1893 | ||
6338b358 | 1894 | if (saw_NUL && !pfile->state.skipping) |
0527bc4e JDA |
1895 | cpp_error (pfile, CPP_DL_WARNING, |
1896 | "null character(s) preserved in literal"); | |
45b966db | 1897 | |
c663e301 JM |
1898 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
1899 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
1900 | (int) terminator); | |
1901 | ||
3ce4f9e4 ESR |
1902 | if (CPP_OPTION (pfile, user_literals)) |
1903 | { | |
c865f923 ESR |
1904 | /* If a string format macro, say from inttypes.h, is placed touching |
1905 | a string literal it could be parsed as a C++11 user-defined string | |
1906 | literal thus breaking the program. | |
1907 | Try to identify macros with is_macro. A warning is issued. */ | |
1908 | if (is_macro (pfile, cur)) | |
7f5f5f98 | 1909 | { |
112448b4 | 1910 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 1911 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
1912 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1913 | token->src_loc, 0, | |
1914 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 1915 | "a space between literal and string macro"); |
7f5f5f98 | 1916 | } |
3ce4f9e4 | 1917 | /* Grab user defined literal suffix. */ |
561f7fc7 | 1918 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
1919 | { |
1920 | type = cpp_userdef_char_add_type (type); | |
1921 | type = cpp_userdef_string_add_type (type); | |
1922 | ++cur; | |
7f5f5f98 OW |
1923 | |
1924 | while (ISIDNUM (*cur)) | |
1925 | ++cur; | |
3ce4f9e4 | 1926 | } |
3ce4f9e4 | 1927 | } |
fe191308 JM |
1928 | else if (CPP_OPTION (pfile, cpp_warn_cxx11_compat) |
1929 | && is_macro (pfile, cur) | |
1930 | && !pfile->state.skipping) | |
1931 | cpp_warning_with_line (pfile, CPP_W_CXX11_COMPAT, | |
1932 | token->src_loc, 0, "C++11 requires a space " | |
1933 | "between string literal and macro"); | |
3ce4f9e4 | 1934 | |
6338b358 NB |
1935 | pfile->buffer->cur = cur; |
1936 | create_literal (pfile, token, base, cur - base, type); | |
0d9f234d | 1937 | } |
041c3194 | 1938 | |
631d0d36 MG |
1939 | /* Return the comment table. The client may not make any assumption |
1940 | about the ordering of the table. */ | |
1941 | cpp_comment_table * | |
1942 | cpp_get_comments (cpp_reader *pfile) | |
1943 | { | |
1944 | return &pfile->comments; | |
1945 | } | |
1946 | ||
1947 | /* Append a comment to the end of the comment table. */ | |
1948 | static void | |
1949 | store_comment (cpp_reader *pfile, cpp_token *token) | |
1950 | { | |
1951 | int len; | |
1952 | ||
1953 | if (pfile->comments.allocated == 0) | |
1954 | { | |
1955 | pfile->comments.allocated = 256; | |
1956 | pfile->comments.entries = (cpp_comment *) xmalloc | |
1957 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
1958 | } | |
1959 | ||
1960 | if (pfile->comments.count == pfile->comments.allocated) | |
1961 | { | |
1962 | pfile->comments.allocated *= 2; | |
1963 | pfile->comments.entries = (cpp_comment *) xrealloc | |
1964 | (pfile->comments.entries, | |
1965 | pfile->comments.allocated * sizeof (cpp_comment)); | |
1966 | } | |
1967 | ||
1968 | len = token->val.str.len; | |
1969 | ||
1970 | /* Copy comment. Note, token may not be NULL terminated. */ | |
1971 | pfile->comments.entries[pfile->comments.count].comment = | |
1972 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
1973 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
1974 | token->val.str.text, len); | |
1975 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
1976 | ||
1977 | /* Set source location. */ | |
1978 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
1979 | ||
1980 | /* Increment the count of entries in the comment table. */ | |
1981 | pfile->comments.count++; | |
1982 | } | |
1983 | ||
93c80368 | 1984 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 1985 | static void |
6cf87ca4 ZW |
1986 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
1987 | cppchar_t type) | |
9e62c811 | 1988 | { |
041c3194 | 1989 | unsigned char *buffer; |
651a20b5 | 1990 | unsigned int len, clen, i; |
df383483 | 1991 | |
1c6d33ef | 1992 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
480709cc | 1993 | |
3542203b NB |
1994 | /* C++ comments probably (not definitely) have moved past a new |
1995 | line, which we don't want to save in the comment. */ | |
480709cc | 1996 | if (is_vspace (pfile->buffer->cur[-1])) |
3542203b | 1997 | len--; |
477cdac7 | 1998 | |
651a20b5 KT |
1999 | /* If we are currently in a directive or in argument parsing, then |
2000 | we need to store all C++ comments as C comments internally, and | |
2001 | so we need to allocate a little extra space in that case. | |
477cdac7 JT |
2002 | |
2003 | Note that the only time we encounter a directive here is | |
2004 | when we are saving comments in a "#define". */ | |
651a20b5 KT |
2005 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
2006 | && type == '/') ? len + 2 : len; | |
477cdac7 JT |
2007 | |
2008 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
df383483 | 2009 | |
041c3194 | 2010 | token->type = CPP_COMMENT; |
477cdac7 | 2011 | token->val.str.len = clen; |
0d9f234d | 2012 | token->val.str.text = buffer; |
45b966db | 2013 | |
1c6d33ef NB |
2014 | buffer[0] = '/'; |
2015 | memcpy (buffer + 1, from, len - 1); | |
477cdac7 | 2016 | |
1eeeb6a4 | 2017 | /* Finish conversion to a C comment, if necessary. */ |
651a20b5 | 2018 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
477cdac7 JT |
2019 | { |
2020 | buffer[1] = '*'; | |
2021 | buffer[clen - 2] = '*'; | |
2022 | buffer[clen - 1] = '/'; | |
651a20b5 KT |
2023 | /* As there can be in a C++ comments illegal sequences for C comments |
2024 | we need to filter them out. */ | |
2025 | for (i = 2; i < (clen - 2); i++) | |
2026 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
2027 | buffer[i] = '|'; | |
477cdac7 | 2028 | } |
631d0d36 MG |
2029 | |
2030 | /* Finally store this comment for use by clients of libcpp. */ | |
2031 | store_comment (pfile, token); | |
0d9f234d | 2032 | } |
45b966db | 2033 | |
5fddcffc NB |
2034 | /* Allocate COUNT tokens for RUN. */ |
2035 | void | |
6cf87ca4 | 2036 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
5fddcffc | 2037 | { |
72bb2c39 | 2038 | run->base = XNEWVEC (cpp_token, count); |
5fddcffc NB |
2039 | run->limit = run->base + count; |
2040 | run->next = NULL; | |
2041 | } | |
2042 | ||
2043 | /* Returns the next tokenrun, or creates one if there is none. */ | |
2044 | static tokenrun * | |
6cf87ca4 | 2045 | next_tokenrun (tokenrun *run) |
5fddcffc NB |
2046 | { |
2047 | if (run->next == NULL) | |
2048 | { | |
72bb2c39 | 2049 | run->next = XNEW (tokenrun); |
bdcbe496 | 2050 | run->next->prev = run; |
5fddcffc NB |
2051 | _cpp_init_tokenrun (run->next, 250); |
2052 | } | |
2053 | ||
2054 | return run->next; | |
2055 | } | |
2056 | ||
ad2305ad | 2057 | /* Return the number of not yet processed token in a given |
92582b75 TT |
2058 | context. */ |
2059 | int | |
ad2305ad | 2060 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
92582b75 | 2061 | { |
92582b75 | 2062 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
cbbcf655 | 2063 | return (LAST (context).token - FIRST (context).token); |
92582b75 TT |
2064 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
2065 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
cbbcf655 | 2066 | return (LAST (context).ptoken - FIRST (context).ptoken); |
92582b75 TT |
2067 | else |
2068 | abort (); | |
2069 | } | |
2070 | ||
ad2305ad DS |
2071 | /* Returns the token present at index INDEX in a given context. If |
2072 | INDEX is zero, the next token to be processed is returned. */ | |
92582b75 | 2073 | static const cpp_token* |
ad2305ad | 2074 | _cpp_token_from_context_at (cpp_context *context, int index) |
92582b75 | 2075 | { |
92582b75 TT |
2076 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
2077 | return &(FIRST (context).token[index]); | |
2078 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
2079 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
2080 | return FIRST (context).ptoken[index]; | |
2081 | else | |
2082 | abort (); | |
2083 | } | |
2084 | ||
5950c3c9 BE |
2085 | /* Look ahead in the input stream. */ |
2086 | const cpp_token * | |
2087 | cpp_peek_token (cpp_reader *pfile, int index) | |
2088 | { | |
2089 | cpp_context *context = pfile->context; | |
2090 | const cpp_token *peektok; | |
2091 | int count; | |
2092 | ||
2093 | /* First, scan through any pending cpp_context objects. */ | |
2094 | while (context->prev) | |
2095 | { | |
ad2305ad | 2096 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
5950c3c9 BE |
2097 | |
2098 | if (index < (int) sz) | |
ad2305ad | 2099 | return _cpp_token_from_context_at (context, index); |
5950c3c9 BE |
2100 | index -= (int) sz; |
2101 | context = context->prev; | |
2102 | } | |
2103 | ||
2104 | /* We will have to read some new tokens after all (and do so | |
2105 | without invalidating preceding tokens). */ | |
2106 | count = index; | |
2107 | pfile->keep_tokens++; | |
2108 | ||
b8cd77f4 JJ |
2109 | /* For peeked tokens temporarily disable line_change reporting, |
2110 | until the tokens are parsed for real. */ | |
2111 | void (*line_change) (cpp_reader *, const cpp_token *, int) | |
2112 | = pfile->cb.line_change; | |
2113 | pfile->cb.line_change = NULL; | |
2114 | ||
5950c3c9 BE |
2115 | do |
2116 | { | |
2117 | peektok = _cpp_lex_token (pfile); | |
2118 | if (peektok->type == CPP_EOF) | |
e4b33ee5 JJ |
2119 | { |
2120 | index--; | |
2121 | break; | |
2122 | } | |
5950c3c9 BE |
2123 | } |
2124 | while (index--); | |
2125 | ||
e4b33ee5 | 2126 | _cpp_backup_tokens_direct (pfile, count - index); |
5950c3c9 | 2127 | pfile->keep_tokens--; |
b8cd77f4 | 2128 | pfile->cb.line_change = line_change; |
5950c3c9 BE |
2129 | |
2130 | return peektok; | |
2131 | } | |
2132 | ||
4ed5bcfb NB |
2133 | /* Allocate a single token that is invalidated at the same time as the |
2134 | rest of the tokens on the line. Has its line and col set to the | |
2135 | same as the last lexed token, so that diagnostics appear in the | |
2136 | right place. */ | |
2137 | cpp_token * | |
6cf87ca4 | 2138 | _cpp_temp_token (cpp_reader *pfile) |
4ed5bcfb NB |
2139 | { |
2140 | cpp_token *old, *result; | |
5950c3c9 BE |
2141 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
2142 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
4ed5bcfb NB |
2143 | |
2144 | old = pfile->cur_token - 1; | |
5950c3c9 BE |
2145 | /* Any pre-existing lookaheads must not be clobbered. */ |
2146 | if (la) | |
2147 | { | |
2148 | if (sz <= la) | |
2149 | { | |
2150 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
2151 | ||
2152 | if (sz < la) | |
2153 | memmove (next->base + 1, next->base, | |
2154 | (la - sz) * sizeof (cpp_token)); | |
2155 | ||
2156 | next->base[0] = pfile->cur_run->limit[-1]; | |
2157 | } | |
2158 | ||
2159 | if (sz > 1) | |
2160 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
2161 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
2162 | } | |
2163 | ||
2164 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
4ed5bcfb NB |
2165 | { |
2166 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
2167 | pfile->cur_token = pfile->cur_run->base; | |
2168 | } | |
2169 | ||
2170 | result = pfile->cur_token++; | |
12f9df4e | 2171 | result->src_loc = old->src_loc; |
4ed5bcfb NB |
2172 | return result; |
2173 | } | |
2174 | ||
14baae01 NB |
2175 | /* Lex a token into RESULT (external interface). Takes care of issues |
2176 | like directive handling, token lookahead, multiple include | |
a1f300c0 | 2177 | optimization and skipping. */ |
345894b4 | 2178 | const cpp_token * |
6cf87ca4 | 2179 | _cpp_lex_token (cpp_reader *pfile) |
5fddcffc | 2180 | { |
bdcbe496 | 2181 | cpp_token *result; |
5fddcffc | 2182 | |
bdcbe496 | 2183 | for (;;) |
5fddcffc | 2184 | { |
bdcbe496 | 2185 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 2186 | { |
bdcbe496 NB |
2187 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
2188 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc | 2189 | } |
ee380365 TT |
2190 | /* We assume that the current token is somewhere in the current |
2191 | run. */ | |
2192 | if (pfile->cur_token < pfile->cur_run->base | |
2193 | || pfile->cur_token >= pfile->cur_run->limit) | |
2194 | abort (); | |
5fddcffc | 2195 | |
bdcbe496 | 2196 | if (pfile->lookaheads) |
14baae01 NB |
2197 | { |
2198 | pfile->lookaheads--; | |
2199 | result = pfile->cur_token++; | |
2200 | } | |
bdcbe496 | 2201 | else |
14baae01 | 2202 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
2203 | |
2204 | if (result->flags & BOL) | |
5fddcffc | 2205 | { |
bdcbe496 NB |
2206 | /* Is this a directive. If _cpp_handle_directive returns |
2207 | false, it is an assembler #. */ | |
2208 | if (result->type == CPP_HASH | |
e808ec9c NB |
2209 | /* 6.10.3 p 11: Directives in a list of macro arguments |
2210 | gives undefined behavior. This implementation | |
2211 | handles the directive as normal. */ | |
bc4071dd | 2212 | && pfile->state.parsing_args != 1) |
21b11495 | 2213 | { |
bc4071dd | 2214 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
21b11495 | 2215 | { |
bc4071dd RH |
2216 | if (pfile->directive_result.type == CPP_PADDING) |
2217 | continue; | |
21b11495 | 2218 | result = &pfile->directive_result; |
21b11495 ZW |
2219 | } |
2220 | } | |
bc4071dd RH |
2221 | else if (pfile->state.in_deferred_pragma) |
2222 | result = &pfile->directive_result; | |
21b11495 | 2223 | |
97293897 | 2224 | if (pfile->cb.line_change && !pfile->state.skipping) |
6cf87ca4 | 2225 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
5fddcffc | 2226 | } |
5fddcffc | 2227 | |
bdcbe496 | 2228 | /* We don't skip tokens in directives. */ |
bc4071dd | 2229 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
bdcbe496 | 2230 | break; |
5fddcffc | 2231 | |
bdcbe496 | 2232 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 2233 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
6356f892 | 2234 | get here and MI optimization works. */ |
5fddcffc | 2235 | pfile->mi_valid = false; |
bdcbe496 NB |
2236 | |
2237 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
2238 | break; | |
5fddcffc NB |
2239 | } |
2240 | ||
345894b4 | 2241 | return result; |
5fddcffc NB |
2242 | } |
2243 | ||
26aea073 NB |
2244 | /* Returns true if a fresh line has been loaded. */ |
2245 | bool | |
6cf87ca4 | 2246 | _cpp_get_fresh_line (cpp_reader *pfile) |
004cb263 | 2247 | { |
22234f56 PB |
2248 | int return_at_eof; |
2249 | ||
26aea073 NB |
2250 | /* We can't get a new line until we leave the current directive. */ |
2251 | if (pfile->state.in_directive) | |
2252 | return false; | |
df383483 | 2253 | |
26aea073 | 2254 | for (;;) |
1a76916c | 2255 | { |
26aea073 | 2256 | cpp_buffer *buffer = pfile->buffer; |
1a76916c | 2257 | |
26aea073 NB |
2258 | if (!buffer->need_line) |
2259 | return true; | |
2260 | ||
2261 | if (buffer->next_line < buffer->rlimit) | |
004cb263 | 2262 | { |
26aea073 NB |
2263 | _cpp_clean_line (pfile); |
2264 | return true; | |
2265 | } | |
004cb263 | 2266 | |
26aea073 NB |
2267 | /* First, get out of parsing arguments state. */ |
2268 | if (pfile->state.parsing_args) | |
2269 | return false; | |
2270 | ||
2271 | /* End of buffer. Non-empty files should end in a newline. */ | |
2272 | if (buffer->buf != buffer->rlimit | |
2273 | && buffer->next_line > buffer->rlimit | |
2274 | && !buffer->from_stage3) | |
2275 | { | |
ed0e74e0 | 2276 | /* Clip to buffer size. */ |
26aea073 | 2277 | buffer->next_line = buffer->rlimit; |
26aea073 | 2278 | } |
22234f56 PB |
2279 | |
2280 | return_at_eof = buffer->return_at_eof; | |
26aea073 | 2281 | _cpp_pop_buffer (pfile); |
22234f56 | 2282 | if (pfile->buffer == NULL || return_at_eof) |
a506c55c | 2283 | return false; |
26aea073 | 2284 | } |
004cb263 NB |
2285 | } |
2286 | ||
6f572ac2 NB |
2287 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
2288 | do \ | |
2289 | { \ | |
2290 | result->type = ELSE_TYPE; \ | |
2291 | if (*buffer->cur == CHAR) \ | |
2292 | buffer->cur++, result->type = THEN_TYPE; \ | |
2293 | } \ | |
2294 | while (0) | |
480709cc | 2295 | |
14baae01 NB |
2296 | /* Lex a token into pfile->cur_token, which is also incremented, to |
2297 | get diagnostics pointing to the correct location. | |
2298 | ||
2299 | Does not handle issues such as token lookahead, multiple-include | |
f1ba665b | 2300 | optimization, directives, skipping etc. This function is only |
14baae01 NB |
2301 | suitable for use by _cpp_lex_token, and in special cases like |
2302 | lex_expansion_token which doesn't care for any of these issues. | |
2303 | ||
2304 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
2305 | otherwise returns to the start of the token buffer if permissible. | |
2306 | Returns the location of the lexed token. */ | |
2307 | cpp_token * | |
6cf87ca4 | 2308 | _cpp_lex_direct (cpp_reader *pfile) |
45b966db | 2309 | { |
0d9f234d | 2310 | cppchar_t c; |
adb84b42 | 2311 | cpp_buffer *buffer; |
0d9f234d | 2312 | const unsigned char *comment_start; |
14baae01 | 2313 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 2314 | |
5fddcffc | 2315 | fresh_line: |
26aea073 | 2316 | result->flags = 0; |
2be570f9 | 2317 | buffer = pfile->buffer; |
a506c55c | 2318 | if (buffer->need_line) |
26aea073 | 2319 | { |
bc4071dd RH |
2320 | if (pfile->state.in_deferred_pragma) |
2321 | { | |
2322 | result->type = CPP_PRAGMA_EOL; | |
2323 | pfile->state.in_deferred_pragma = false; | |
2324 | if (!pfile->state.pragma_allow_expansion) | |
2325 | pfile->state.prevent_expansion--; | |
2326 | return result; | |
2327 | } | |
26aea073 NB |
2328 | if (!_cpp_get_fresh_line (pfile)) |
2329 | { | |
2330 | result->type = CPP_EOF; | |
9ff7868d NB |
2331 | if (!pfile->state.in_directive) |
2332 | { | |
2333 | /* Tell the compiler the line number of the EOF token. */ | |
500bee0a | 2334 | result->src_loc = pfile->line_table->highest_line; |
9ff7868d NB |
2335 | result->flags = BOL; |
2336 | } | |
26aea073 NB |
2337 | return result; |
2338 | } | |
2339 | if (!pfile->keep_tokens) | |
2340 | { | |
2341 | pfile->cur_run = &pfile->base_run; | |
2342 | result = pfile->base_run.base; | |
2343 | pfile->cur_token = result + 1; | |
2344 | } | |
2345 | result->flags = BOL; | |
2346 | if (pfile->state.parsing_args == 2) | |
2347 | result->flags |= PREV_WHITE; | |
2348 | } | |
a506c55c | 2349 | buffer = pfile->buffer; |
5fddcffc | 2350 | update_tokens_line: |
500bee0a | 2351 | result->src_loc = pfile->line_table->highest_line; |
041c3194 | 2352 | |
5fddcffc | 2353 | skipped_white: |
26aea073 NB |
2354 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
2355 | && !pfile->overlaid_buffer) | |
2356 | { | |
2357 | _cpp_process_line_notes (pfile, false); | |
500bee0a | 2358 | result->src_loc = pfile->line_table->highest_line; |
26aea073 | 2359 | } |
480709cc | 2360 | c = *buffer->cur++; |
12f9df4e | 2361 | |
e3dfef44 GC |
2362 | if (pfile->forced_token_location_p) |
2363 | result->src_loc = *pfile->forced_token_location_p; | |
2364 | else | |
2365 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
2366 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
5fddcffc | 2367 | |
0d9f234d | 2368 | switch (c) |
45b966db | 2369 | { |
4d6baafa NB |
2370 | case ' ': case '\t': case '\f': case '\v': case '\0': |
2371 | result->flags |= PREV_WHITE; | |
26aea073 NB |
2372 | skip_whitespace (pfile, c); |
2373 | goto skipped_white; | |
0d9f234d | 2374 | |
26aea073 | 2375 | case '\n': |
12f9df4e PB |
2376 | if (buffer->cur < buffer->rlimit) |
2377 | CPP_INCREMENT_LINE (pfile, 0); | |
26aea073 NB |
2378 | buffer->need_line = true; |
2379 | goto fresh_line; | |
46d07497 | 2380 | |
0d9f234d NB |
2381 | case '0': case '1': case '2': case '3': case '4': |
2382 | case '5': case '6': case '7': case '8': case '9': | |
50668cf6 GK |
2383 | { |
2384 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2385 | result->type = CPP_NUMBER; | |
2386 | lex_number (pfile, &result->val.str, &nst); | |
2387 | warn_about_normalization (pfile, result, &nst); | |
2388 | break; | |
2389 | } | |
46d07497 | 2390 | |
0abc6a6a | 2391 | case 'L': |
b6baa67d KVH |
2392 | case 'u': |
2393 | case 'U': | |
2c6e3f55 JJ |
2394 | case 'R': |
2395 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2396 | wide strings or raw strings. */ | |
a48e3dd1 JM |
2397 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
2398 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
bced6edf | 2399 | { |
2c6e3f55 JJ |
2400 | if ((*buffer->cur == '\'' && c != 'R') |
2401 | || *buffer->cur == '"' | |
2402 | || (*buffer->cur == 'R' | |
2403 | && c != 'R' | |
2404 | && buffer->cur[1] == '"' | |
a48e3dd1 | 2405 | && CPP_OPTION (pfile, rliterals)) |
2c6e3f55 JJ |
2406 | || (*buffer->cur == '8' |
2407 | && c == 'u' | |
fe95b036 ESR |
2408 | && ((buffer->cur[1] == '"' || (buffer->cur[1] == '\'' |
2409 | && CPP_OPTION (pfile, utf8_char_literals))) | |
a48e3dd1 JM |
2410 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
2411 | && CPP_OPTION (pfile, rliterals))))) | |
b6baa67d KVH |
2412 | { |
2413 | lex_string (pfile, result, buffer->cur - 1); | |
2414 | break; | |
2415 | } | |
bced6edf | 2416 | } |
df383483 | 2417 | /* Fall through. */ |
0abc6a6a | 2418 | |
0d9f234d NB |
2419 | case '_': |
2420 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2421 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2422 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
b6baa67d | 2423 | case 's': case 't': case 'v': case 'w': case 'x': |
0d9f234d NB |
2424 | case 'y': case 'z': |
2425 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
0abc6a6a | 2426 | case 'G': case 'H': case 'I': case 'J': case 'K': |
2c6e3f55 | 2427 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
b6baa67d | 2428 | case 'S': case 'T': case 'V': case 'W': case 'X': |
0d9f234d NB |
2429 | case 'Y': case 'Z': |
2430 | result->type = CPP_NAME; | |
50668cf6 GK |
2431 | { |
2432 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
9a0c6187 | 2433 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
be5ffc59 JM |
2434 | &nst, |
2435 | &result->val.node.spelling); | |
50668cf6 GK |
2436 | warn_about_normalization (pfile, result, &nst); |
2437 | } | |
0d9f234d | 2438 | |
0d9f234d | 2439 | /* Convert named operators to their proper types. */ |
9a0c6187 | 2440 | if (result->val.node.node->flags & NODE_OPERATOR) |
0d9f234d NB |
2441 | { |
2442 | result->flags |= NAMED_OP; | |
9a0c6187 | 2443 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
0d9f234d NB |
2444 | } |
2445 | break; | |
2446 | ||
2447 | case '\'': | |
2448 | case '"': | |
6338b358 | 2449 | lex_string (pfile, result, buffer->cur - 1); |
0d9f234d | 2450 | break; |
041c3194 | 2451 | |
0d9f234d | 2452 | case '/': |
1c6d33ef NB |
2453 | /* A potential block or line comment. */ |
2454 | comment_start = buffer->cur; | |
6f572ac2 NB |
2455 | c = *buffer->cur; |
2456 | ||
1c6d33ef NB |
2457 | if (c == '*') |
2458 | { | |
26aea073 | 2459 | if (_cpp_skip_block_comment (pfile)) |
0527bc4e | 2460 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
0d9f234d | 2461 | } |
909eb89c | 2462 | else if (c == '/' && ! CPP_OPTION (pfile, traditional)) |
0d9f234d | 2463 | { |
909eb89c MP |
2464 | /* Don't warn for system headers. */ |
2465 | if (cpp_in_system_header (pfile)) | |
2466 | ; | |
f3bede71 | 2467 | /* Warn about comments if pedantically GNUC89, and not |
bdb05a7b | 2468 | in system headers. */ |
909eb89c MP |
2469 | else if (CPP_OPTION (pfile, lang) == CLK_GNUC89 |
2470 | && CPP_PEDANTIC (pfile) | |
2471 | && ! buffer->warned_cplusplus_comments) | |
041c3194 | 2472 | { |
0527bc4e | 2473 | cpp_error (pfile, CPP_DL_PEDWARN, |
56508306 | 2474 | "C++ style comments are not allowed in ISO C90"); |
0527bc4e | 2475 | cpp_error (pfile, CPP_DL_PEDWARN, |
ebef4e8c | 2476 | "(this will be reported only once per input file)"); |
1c6d33ef NB |
2477 | buffer->warned_cplusplus_comments = 1; |
2478 | } | |
f3bede71 | 2479 | /* Or if specifically desired via -Wc90-c99-compat. */ |
177cce46 | 2480 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
dd3ff077 | 2481 | && ! CPP_OPTION (pfile, cplusplus) |
f3bede71 MP |
2482 | && ! buffer->warned_cplusplus_comments) |
2483 | { | |
2484 | cpp_error (pfile, CPP_DL_WARNING, | |
3f4f5c9a | 2485 | "C++ style comments are incompatible with C90"); |
f3bede71 MP |
2486 | cpp_error (pfile, CPP_DL_WARNING, |
2487 | "(this will be reported only once per input file)"); | |
2488 | buffer->warned_cplusplus_comments = 1; | |
2489 | } | |
909eb89c MP |
2490 | /* In C89/C94, C++ style comments are forbidden. */ |
2491 | else if ((CPP_OPTION (pfile, lang) == CLK_STDC89 | |
2492 | || CPP_OPTION (pfile, lang) == CLK_STDC94)) | |
2493 | { | |
2494 | /* But don't be confused about valid code such as | |
2495 | - // immediately followed by *, | |
2496 | - // in a preprocessing directive, | |
2497 | - // in an #if 0 block. */ | |
2498 | if (buffer->cur[1] == '*' | |
2499 | || pfile->state.in_directive | |
2500 | || pfile->state.skipping) | |
2501 | { | |
2502 | result->type = CPP_DIV; | |
2503 | break; | |
2504 | } | |
2505 | else if (! buffer->warned_cplusplus_comments) | |
2506 | { | |
2507 | cpp_error (pfile, CPP_DL_ERROR, | |
2508 | "C++ style comments are not allowed in ISO C90"); | |
2509 | cpp_error (pfile, CPP_DL_ERROR, | |
2510 | "(this will be reported only once per input " | |
2511 | "file)"); | |
2512 | buffer->warned_cplusplus_comments = 1; | |
2513 | } | |
2514 | } | |
01ef6563 | 2515 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
87cf0651 | 2516 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
1c6d33ef | 2517 | } |
480709cc NB |
2518 | else if (c == '=') |
2519 | { | |
6f572ac2 | 2520 | buffer->cur++; |
480709cc NB |
2521 | result->type = CPP_DIV_EQ; |
2522 | break; | |
2523 | } | |
2524 | else | |
2525 | { | |
480709cc NB |
2526 | result->type = CPP_DIV; |
2527 | break; | |
2528 | } | |
0d9f234d | 2529 | |
1c6d33ef NB |
2530 | if (!pfile->state.save_comments) |
2531 | { | |
2532 | result->flags |= PREV_WHITE; | |
5fddcffc | 2533 | goto update_tokens_line; |
0d9f234d | 2534 | } |
1c6d33ef NB |
2535 | |
2536 | /* Save the comment as a token in its own right. */ | |
477cdac7 | 2537 | save_comment (pfile, result, comment_start, c); |
bdcbe496 | 2538 | break; |
0d9f234d NB |
2539 | |
2540 | case '<': | |
2541 | if (pfile->state.angled_headers) | |
2542 | { | |
6338b358 | 2543 | lex_string (pfile, result, buffer->cur - 1); |
4bb09c26 JM |
2544 | if (result->type != CPP_LESS) |
2545 | break; | |
0d9f234d | 2546 | } |
45b966db | 2547 | |
6f572ac2 NB |
2548 | result->type = CPP_LESS; |
2549 | if (*buffer->cur == '=') | |
2550 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2551 | else if (*buffer->cur == '<') | |
0d9f234d | 2552 | { |
6f572ac2 NB |
2553 | buffer->cur++; |
2554 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
0d9f234d | 2555 | } |
6f572ac2 | 2556 | else if (CPP_OPTION (pfile, digraphs)) |
480709cc | 2557 | { |
6f572ac2 NB |
2558 | if (*buffer->cur == ':') |
2559 | { | |
1582c677 PC |
2560 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
2561 | three characters are <:: and the subsequent character | |
2562 | is neither : nor >, the < is treated as a preprocessor | |
2563 | token by itself". */ | |
2564 | if (CPP_OPTION (pfile, cplusplus) | |
61949153 PC |
2565 | && CPP_OPTION (pfile, lang) != CLK_CXX98 |
2566 | && CPP_OPTION (pfile, lang) != CLK_GNUCXX | |
1582c677 PC |
2567 | && buffer->cur[1] == ':' |
2568 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
2569 | break; | |
2570 | ||
6f572ac2 NB |
2571 | buffer->cur++; |
2572 | result->flags |= DIGRAPH; | |
2573 | result->type = CPP_OPEN_SQUARE; | |
2574 | } | |
2575 | else if (*buffer->cur == '%') | |
2576 | { | |
2577 | buffer->cur++; | |
2578 | result->flags |= DIGRAPH; | |
2579 | result->type = CPP_OPEN_BRACE; | |
2580 | } | |
480709cc | 2581 | } |
0d9f234d NB |
2582 | break; |
2583 | ||
2584 | case '>': | |
6f572ac2 NB |
2585 | result->type = CPP_GREATER; |
2586 | if (*buffer->cur == '=') | |
2587 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
2588 | else if (*buffer->cur == '>') | |
0d9f234d | 2589 | { |
6f572ac2 NB |
2590 | buffer->cur++; |
2591 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
2592 | } | |
0d9f234d NB |
2593 | break; |
2594 | ||
cbcff6df | 2595 | case '%': |
6f572ac2 NB |
2596 | result->type = CPP_MOD; |
2597 | if (*buffer->cur == '=') | |
2598 | buffer->cur++, result->type = CPP_MOD_EQ; | |
2599 | else if (CPP_OPTION (pfile, digraphs)) | |
480709cc | 2600 | { |
6f572ac2 | 2601 | if (*buffer->cur == ':') |
480709cc | 2602 | { |
6f572ac2 NB |
2603 | buffer->cur++; |
2604 | result->flags |= DIGRAPH; | |
2605 | result->type = CPP_HASH; | |
2606 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
9a0c6187 | 2607 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
6f572ac2 NB |
2608 | } |
2609 | else if (*buffer->cur == '>') | |
2610 | { | |
2611 | buffer->cur++; | |
2612 | result->flags |= DIGRAPH; | |
2613 | result->type = CPP_CLOSE_BRACE; | |
480709cc | 2614 | } |
480709cc | 2615 | } |
0d9f234d NB |
2616 | break; |
2617 | ||
cbcff6df | 2618 | case '.': |
480709cc | 2619 | result->type = CPP_DOT; |
6f572ac2 | 2620 | if (ISDIGIT (*buffer->cur)) |
480709cc | 2621 | { |
50668cf6 | 2622 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
480709cc | 2623 | result->type = CPP_NUMBER; |
50668cf6 GK |
2624 | lex_number (pfile, &result->val.str, &nst); |
2625 | warn_about_normalization (pfile, result, &nst); | |
480709cc | 2626 | } |
6f572ac2 NB |
2627 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
2628 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
2629 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
2630 | buffer->cur++, result->type = CPP_DOT_STAR; | |
0d9f234d | 2631 | break; |
45b966db | 2632 | |
0d9f234d | 2633 | case '+': |
6f572ac2 NB |
2634 | result->type = CPP_PLUS; |
2635 | if (*buffer->cur == '+') | |
2636 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
2637 | else if (*buffer->cur == '=') | |
2638 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
0d9f234d | 2639 | break; |
04e3ec78 | 2640 | |
0d9f234d | 2641 | case '-': |
6f572ac2 NB |
2642 | result->type = CPP_MINUS; |
2643 | if (*buffer->cur == '>') | |
0d9f234d | 2644 | { |
6f572ac2 | 2645 | buffer->cur++; |
480709cc | 2646 | result->type = CPP_DEREF; |
6f572ac2 NB |
2647 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
2648 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
480709cc | 2649 | } |
6f572ac2 NB |
2650 | else if (*buffer->cur == '-') |
2651 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
2652 | else if (*buffer->cur == '=') | |
2653 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
0d9f234d | 2654 | break; |
45b966db | 2655 | |
0d9f234d | 2656 | case '&': |
6f572ac2 NB |
2657 | result->type = CPP_AND; |
2658 | if (*buffer->cur == '&') | |
2659 | buffer->cur++, result->type = CPP_AND_AND; | |
2660 | else if (*buffer->cur == '=') | |
2661 | buffer->cur++, result->type = CPP_AND_EQ; | |
0d9f234d | 2662 | break; |
df383483 | 2663 | |
0d9f234d | 2664 | case '|': |
6f572ac2 NB |
2665 | result->type = CPP_OR; |
2666 | if (*buffer->cur == '|') | |
2667 | buffer->cur++, result->type = CPP_OR_OR; | |
2668 | else if (*buffer->cur == '=') | |
2669 | buffer->cur++, result->type = CPP_OR_EQ; | |
0d9f234d | 2670 | break; |
45b966db | 2671 | |
0d9f234d | 2672 | case ':': |
6f572ac2 NB |
2673 | result->type = CPP_COLON; |
2674 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
2675 | buffer->cur++, result->type = CPP_SCOPE; | |
2676 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
0d9f234d | 2677 | { |
6f572ac2 | 2678 | buffer->cur++; |
0d9f234d | 2679 | result->flags |= DIGRAPH; |
480709cc NB |
2680 | result->type = CPP_CLOSE_SQUARE; |
2681 | } | |
0d9f234d | 2682 | break; |
45b966db | 2683 | |
480709cc NB |
2684 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
2685 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
2686 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
2687 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
9a0c6187 | 2688 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
480709cc | 2689 | |
26aea073 | 2690 | case '?': result->type = CPP_QUERY; break; |
0d9f234d NB |
2691 | case '~': result->type = CPP_COMPL; break; |
2692 | case ',': result->type = CPP_COMMA; break; | |
2693 | case '(': result->type = CPP_OPEN_PAREN; break; | |
2694 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
2695 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
2696 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
2697 | case '{': result->type = CPP_OPEN_BRACE; break; | |
2698 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
2699 | case ';': result->type = CPP_SEMICOLON; break; | |
2700 | ||
40f03658 | 2701 | /* @ is a punctuator in Objective-C. */ |
cc937581 | 2702 | case '@': result->type = CPP_ATSIGN; break; |
0d9f234d | 2703 | |
0abc6a6a | 2704 | case '$': |
1613e52b NB |
2705 | case '\\': |
2706 | { | |
2707 | const uchar *base = --buffer->cur; | |
50668cf6 | 2708 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
0abc6a6a | 2709 | |
50668cf6 | 2710 | if (forms_identifier_p (pfile, true, &nst)) |
1613e52b NB |
2711 | { |
2712 | result->type = CPP_NAME; | |
be5ffc59 JM |
2713 | result->val.node.node = lex_identifier (pfile, base, true, &nst, |
2714 | &result->val.node.spelling); | |
50668cf6 | 2715 | warn_about_normalization (pfile, result, &nst); |
1613e52b NB |
2716 | break; |
2717 | } | |
2718 | buffer->cur++; | |
1067694a | 2719 | } |
1613e52b | 2720 | |
1067694a | 2721 | default: |
6338b358 NB |
2722 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); |
2723 | break; | |
0d9f234d | 2724 | } |
bdcbe496 | 2725 | |
ebedc9a3 DM |
2726 | source_range tok_range; |
2727 | tok_range.m_start = result->src_loc; | |
2728 | if (result->src_loc >= RESERVED_LOCATION_COUNT) | |
2729 | tok_range.m_finish | |
2730 | = linemap_position_for_column (pfile->line_table, | |
2731 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
2732 | else | |
2733 | tok_range.m_finish = tok_range.m_start; | |
2734 | ||
2735 | result->src_loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
2736 | result->src_loc, | |
2737 | tok_range, NULL); | |
2738 | ||
bdcbe496 | 2739 | return result; |
0d9f234d NB |
2740 | } |
2741 | ||
59325650 NB |
2742 | /* An upper bound on the number of bytes needed to spell TOKEN. |
2743 | Does not include preceding whitespace. */ | |
93c80368 | 2744 | unsigned int |
6cf87ca4 | 2745 | cpp_token_len (const cpp_token *token) |
0d9f234d | 2746 | { |
93c80368 | 2747 | unsigned int len; |
6d2c2047 | 2748 | |
93c80368 | 2749 | switch (TOKEN_SPELL (token)) |
041c3194 | 2750 | { |
cc955282 | 2751 | default: len = 6; break; |
6338b358 | 2752 | case SPELL_LITERAL: len = token->val.str.len; break; |
9a0c6187 | 2753 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
041c3194 | 2754 | } |
59325650 NB |
2755 | |
2756 | return len; | |
6d2c2047 ZW |
2757 | } |
2758 | ||
47e20491 GK |
2759 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
2760 | Return the number of bytes read out of NAME. (There are always | |
2761 | 10 bytes written to BUFFER.) */ | |
2762 | ||
2763 | static size_t | |
2764 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
2765 | { | |
2766 | int j; | |
2767 | int ucn_len = 0; | |
2768 | int ucn_len_c; | |
2769 | unsigned t; | |
2770 | unsigned long utf32; | |
2771 | ||
2772 | /* Compute the length of the UTF-8 sequence. */ | |
2773 | for (t = *name; t & 0x80; t <<= 1) | |
2774 | ucn_len++; | |
2775 | ||
2776 | utf32 = *name & (0x7F >> ucn_len); | |
2777 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
2778 | { | |
2779 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
2780 | ||
2781 | /* Ill-formed UTF-8. */ | |
2782 | if ((*name & ~0x3F) != 0x80) | |
2783 | abort (); | |
2784 | } | |
2785 | ||
2786 | *buffer++ = '\\'; | |
2787 | *buffer++ = 'U'; | |
2788 | for (j = 7; j >= 0; j--) | |
2789 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
2790 | return ucn_len; | |
2791 | } | |
2792 | ||
cfc93532 MLI |
2793 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
2794 | the spelling of the digraph. */ | |
2795 | static const unsigned char * | |
2796 | cpp_digraph2name (enum cpp_ttype type) | |
2797 | { | |
2798 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
2799 | } | |
47e20491 | 2800 | |
be5ffc59 JM |
2801 | /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER. |
2802 | The buffer must already contain the enough space to hold the | |
2803 | token's spelling. Returns a pointer to the character after the | |
2804 | last character written. */ | |
2805 | unsigned char * | |
2806 | _cpp_spell_ident_ucns (unsigned char *buffer, cpp_hashnode *ident) | |
2807 | { | |
2808 | size_t i; | |
2809 | const unsigned char *name = NODE_NAME (ident); | |
2810 | ||
2811 | for (i = 0; i < NODE_LEN (ident); i++) | |
2812 | if (name[i] & ~0x7F) | |
2813 | { | |
2814 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2815 | buffer += 10; | |
2816 | } | |
2817 | else | |
2818 | *buffer++ = name[i]; | |
2819 | ||
2820 | return buffer; | |
2821 | } | |
2822 | ||
041c3194 | 2823 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 | 2824 | already contain the enough space to hold the token's spelling. |
6cf87ca4 | 2825 | Returns a pointer to the character after the last character written. |
47e20491 | 2826 | FORSTRING is true if this is to be the spelling after translation |
be5ffc59 JM |
2827 | phase 1 (with the original spelling of extended identifiers), false |
2828 | if extended identifiers should always be written using UCNs (there is | |
2829 | no option for always writing them in the internal UTF-8 form). | |
6cf87ca4 | 2830 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
93c80368 | 2831 | unsigned char * |
6cf87ca4 | 2832 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
47e20491 | 2833 | unsigned char *buffer, bool forstring) |
041c3194 | 2834 | { |
96be6998 | 2835 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
2836 | { |
2837 | case SPELL_OPERATOR: | |
2838 | { | |
2839 | const unsigned char *spelling; | |
2840 | unsigned char c; | |
d6d5f795 | 2841 | |
041c3194 | 2842 | if (token->flags & DIGRAPH) |
cfc93532 | 2843 | spelling = cpp_digraph2name (token->type); |
92936ecf ZW |
2844 | else if (token->flags & NAMED_OP) |
2845 | goto spell_ident; | |
041c3194 | 2846 | else |
96be6998 | 2847 | spelling = TOKEN_NAME (token); |
df383483 | 2848 | |
041c3194 ZW |
2849 | while ((c = *spelling++) != '\0') |
2850 | *buffer++ = c; | |
2851 | } | |
2852 | break; | |
d6d5f795 | 2853 | |
47ad4138 | 2854 | spell_ident: |
041c3194 | 2855 | case SPELL_IDENT: |
47e20491 GK |
2856 | if (forstring) |
2857 | { | |
be5ffc59 JM |
2858 | memcpy (buffer, NODE_NAME (token->val.node.spelling), |
2859 | NODE_LEN (token->val.node.spelling)); | |
2860 | buffer += NODE_LEN (token->val.node.spelling); | |
47e20491 GK |
2861 | } |
2862 | else | |
be5ffc59 | 2863 | buffer = _cpp_spell_ident_ucns (buffer, token->val.node.node); |
041c3194 | 2864 | break; |
d6d5f795 | 2865 | |
6338b358 | 2866 | case SPELL_LITERAL: |
47ad4138 ZW |
2867 | memcpy (buffer, token->val.str.text, token->val.str.len); |
2868 | buffer += token->val.str.len; | |
2869 | break; | |
2870 | ||
041c3194 | 2871 | case SPELL_NONE: |
0527bc4e JDA |
2872 | cpp_error (pfile, CPP_DL_ICE, |
2873 | "unspellable token %s", TOKEN_NAME (token)); | |
041c3194 ZW |
2874 | break; |
2875 | } | |
d6d5f795 | 2876 | |
041c3194 ZW |
2877 | return buffer; |
2878 | } | |
d6d5f795 | 2879 | |
5d8ebbd8 NB |
2880 | /* Returns TOKEN spelt as a null-terminated string. The string is |
2881 | freed when the reader is destroyed. Useful for diagnostics. */ | |
93c80368 | 2882 | unsigned char * |
6cf87ca4 | 2883 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
59325650 NB |
2884 | { |
2885 | unsigned int len = cpp_token_len (token) + 1; | |
ece54d54 | 2886 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 2887 | |
47e20491 | 2888 | end = cpp_spell_token (pfile, token, start, false); |
93c80368 | 2889 | end[0] = '\0'; |
c5a04734 | 2890 | |
93c80368 NB |
2891 | return start; |
2892 | } | |
c5a04734 | 2893 | |
cfc93532 MLI |
2894 | /* Returns a pointer to a string which spells the token defined by |
2895 | TYPE and FLAGS. Used by C front ends, which really should move to | |
2896 | using cpp_token_as_text. */ | |
93c80368 | 2897 | const char * |
cfc93532 | 2898 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
93c80368 | 2899 | { |
cfc93532 MLI |
2900 | if (flags & DIGRAPH) |
2901 | return (const char *) cpp_digraph2name (type); | |
2902 | else if (flags & NAMED_OP) | |
2903 | return cpp_named_operator2name (type); | |
2904 | ||
93c80368 NB |
2905 | return (const char *) token_spellings[type].name; |
2906 | } | |
c5a04734 | 2907 | |
4ed5bcfb NB |
2908 | /* Writes the spelling of token to FP, without any preceding space. |
2909 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
2910 | double-buffering. */ | |
93c80368 | 2911 | void |
6cf87ca4 | 2912 | cpp_output_token (const cpp_token *token, FILE *fp) |
93c80368 | 2913 | { |
93c80368 | 2914 | switch (TOKEN_SPELL (token)) |
c5a04734 | 2915 | { |
93c80368 NB |
2916 | case SPELL_OPERATOR: |
2917 | { | |
2918 | const unsigned char *spelling; | |
3b681e9d | 2919 | int c; |
c5a04734 | 2920 | |
93c80368 | 2921 | if (token->flags & DIGRAPH) |
cfc93532 | 2922 | spelling = cpp_digraph2name (token->type); |
93c80368 NB |
2923 | else if (token->flags & NAMED_OP) |
2924 | goto spell_ident; | |
2925 | else | |
2926 | spelling = TOKEN_NAME (token); | |
041c3194 | 2927 | |
3b681e9d ZW |
2928 | c = *spelling; |
2929 | do | |
2930 | putc (c, fp); | |
2931 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
2932 | } |
2933 | break; | |
041c3194 | 2934 | |
93c80368 NB |
2935 | spell_ident: |
2936 | case SPELL_IDENT: | |
47e20491 GK |
2937 | { |
2938 | size_t i; | |
9a0c6187 | 2939 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 2940 | |
9a0c6187 | 2941 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
2942 | if (name[i] & ~0x7F) |
2943 | { | |
2944 | unsigned char buffer[10]; | |
2945 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2946 | fwrite (buffer, 1, 10, fp); | |
2947 | } | |
2948 | else | |
9a0c6187 | 2949 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
47e20491 GK |
2950 | } |
2951 | break; | |
041c3194 | 2952 | |
6338b358 | 2953 | case SPELL_LITERAL: |
47ad4138 ZW |
2954 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
2955 | break; | |
2956 | ||
93c80368 NB |
2957 | case SPELL_NONE: |
2958 | /* An error, most probably. */ | |
2959 | break; | |
041c3194 | 2960 | } |
c5a04734 ZW |
2961 | } |
2962 | ||
93c80368 NB |
2963 | /* Compare two tokens. */ |
2964 | int | |
6cf87ca4 | 2965 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
c5a04734 | 2966 | { |
93c80368 NB |
2967 | if (a->type == b->type && a->flags == b->flags) |
2968 | switch (TOKEN_SPELL (a)) | |
2969 | { | |
2970 | default: /* Keep compiler happy. */ | |
2971 | case SPELL_OPERATOR: | |
9a0c6187 | 2972 | /* token_no is used to track where multiple consecutive ## |
aa508502 | 2973 | tokens were originally located. */ |
9a0c6187 | 2974 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
93c80368 | 2975 | case SPELL_NONE: |
9a0c6187 | 2976 | return (a->type != CPP_MACRO_ARG |
be5ffc59 JM |
2977 | || (a->val.macro_arg.arg_no == b->val.macro_arg.arg_no |
2978 | && a->val.macro_arg.spelling == b->val.macro_arg.spelling)); | |
93c80368 | 2979 | case SPELL_IDENT: |
be5ffc59 JM |
2980 | return (a->val.node.node == b->val.node.node |
2981 | && a->val.node.spelling == b->val.node.spelling); | |
6338b358 | 2982 | case SPELL_LITERAL: |
93c80368 NB |
2983 | return (a->val.str.len == b->val.str.len |
2984 | && !memcmp (a->val.str.text, b->val.str.text, | |
2985 | a->val.str.len)); | |
2986 | } | |
c5a04734 | 2987 | |
041c3194 ZW |
2988 | return 0; |
2989 | } | |
2990 | ||
93c80368 NB |
2991 | /* Returns nonzero if a space should be inserted to avoid an |
2992 | accidental token paste for output. For simplicity, it is | |
2993 | conservative, and occasionally advises a space where one is not | |
2994 | needed, e.g. "." and ".2". */ | |
93c80368 | 2995 | int |
6cf87ca4 ZW |
2996 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
2997 | const cpp_token *token2) | |
c5a04734 | 2998 | { |
93c80368 NB |
2999 | enum cpp_ttype a = token1->type, b = token2->type; |
3000 | cppchar_t c; | |
c5a04734 | 3001 | |
93c80368 NB |
3002 | if (token1->flags & NAMED_OP) |
3003 | a = CPP_NAME; | |
3004 | if (token2->flags & NAMED_OP) | |
3005 | b = CPP_NAME; | |
c5a04734 | 3006 | |
93c80368 NB |
3007 | c = EOF; |
3008 | if (token2->flags & DIGRAPH) | |
37b8524c | 3009 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
3010 | else if (token_spellings[b].category == SPELL_OPERATOR) |
3011 | c = token_spellings[b].name[0]; | |
c5a04734 | 3012 | |
93c80368 | 3013 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 3014 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 3015 | return 1; |
c5a04734 | 3016 | |
93c80368 | 3017 | switch (a) |
c5a04734 | 3018 | { |
b52dbbf8 SE |
3019 | case CPP_GREATER: return c == '>'; |
3020 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
93c80368 NB |
3021 | case CPP_PLUS: return c == '+'; |
3022 | case CPP_MINUS: return c == '-' || c == '>'; | |
3023 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
3024 | case CPP_MOD: return c == ':' || c == '>'; | |
3025 | case CPP_AND: return c == '&'; | |
3026 | case CPP_OR: return c == '|'; | |
3027 | case CPP_COLON: return c == ':' || c == '>'; | |
3028 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 3029 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
3030 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
3031 | case CPP_NAME: return ((b == CPP_NUMBER | |
3032 | && name_p (pfile, &token2->val.str)) | |
3033 | || b == CPP_NAME | |
3034 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
3035 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
3036 | || c == '.' || c == '+' || c == '-'); | |
1613e52b | 3037 | /* UCNs */ |
1067694a NB |
3038 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
3039 | && b == CPP_NAME) | |
1613e52b | 3040 | || (CPP_OPTION (pfile, objc) |
1067694a | 3041 | && token1->val.str.text[0] == '@' |
1613e52b | 3042 | && (b == CPP_NAME || b == CPP_STRING))); |
87e356ba JJ |
3043 | case CPP_STRING: |
3044 | case CPP_WSTRING: | |
3045 | case CPP_UTF8STRING: | |
3046 | case CPP_STRING16: | |
3047 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals) | |
3048 | && (b == CPP_NAME | |
3049 | || (TOKEN_SPELL (token2) == SPELL_LITERAL | |
3050 | && ISIDST (token2->val.str.text[0])))); | |
3051 | ||
93c80368 | 3052 | default: break; |
c5a04734 | 3053 | } |
c5a04734 | 3054 | |
417f3e3a | 3055 | return 0; |
c5a04734 ZW |
3056 | } |
3057 | ||
93c80368 | 3058 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
3059 | character, to FP. Leading whitespace is removed. If there are |
3060 | macros, special token padding is not performed. */ | |
c5a04734 | 3061 | void |
6cf87ca4 | 3062 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
c5a04734 | 3063 | { |
4ed5bcfb | 3064 | const cpp_token *token; |
96be6998 | 3065 | |
4ed5bcfb NB |
3066 | token = cpp_get_token (pfile); |
3067 | while (token->type != CPP_EOF) | |
96be6998 | 3068 | { |
4ed5bcfb NB |
3069 | cpp_output_token (token, fp); |
3070 | token = cpp_get_token (pfile); | |
3071 | if (token->flags & PREV_WHITE) | |
3072 | putc (' ', fp); | |
96be6998 ZW |
3073 | } |
3074 | ||
93c80368 | 3075 | putc ('\n', fp); |
041c3194 | 3076 | } |
c5a04734 | 3077 | |
5d6342eb TT |
3078 | /* Return a string representation of all the remaining tokens on the |
3079 | current line. The result is allocated using xmalloc and must be | |
3080 | freed by the caller. */ | |
3081 | unsigned char * | |
3082 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
3083 | { | |
3084 | const cpp_token *token; | |
3085 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
3086 | unsigned int alloced = 120 + out; | |
3087 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
3088 | ||
3089 | /* If DIR_NAME is empty, there are no initial contents. */ | |
3090 | if (dir_name) | |
3091 | { | |
3092 | sprintf ((char *) result, "#%s ", dir_name); | |
3093 | out += 2; | |
3094 | } | |
3095 | ||
3096 | token = cpp_get_token (pfile); | |
3097 | while (token->type != CPP_EOF) | |
3098 | { | |
3099 | unsigned char *last; | |
3100 | /* Include room for a possible space and the terminating nul. */ | |
3101 | unsigned int len = cpp_token_len (token) + 2; | |
3102 | ||
3103 | if (out + len > alloced) | |
3104 | { | |
3105 | alloced *= 2; | |
3106 | if (out + len > alloced) | |
3107 | alloced = out + len; | |
3108 | result = (unsigned char *) xrealloc (result, alloced); | |
3109 | } | |
3110 | ||
3111 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
3112 | out = last - result; | |
3113 | ||
3114 | token = cpp_get_token (pfile); | |
3115 | if (token->flags & PREV_WHITE) | |
3116 | result[out++] = ' '; | |
3117 | } | |
3118 | ||
3119 | result[out] = '\0'; | |
3120 | return result; | |
3121 | } | |
3122 | ||
1e013d2e NB |
3123 | /* Memory buffers. Changing these three constants can have a dramatic |
3124 | effect on performance. The values here are reasonable defaults, | |
3125 | but might be tuned. If you adjust them, be sure to test across a | |
3126 | range of uses of cpplib, including heavy nested function-like macro | |
3127 | expansion. Also check the change in peak memory usage (NJAMD is a | |
3128 | good tool for this). */ | |
3129 | #define MIN_BUFF_SIZE 8000 | |
87062813 | 3130 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
1e013d2e NB |
3131 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
3132 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 3133 | |
87062813 NB |
3134 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
3135 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
3136 | #endif | |
3137 | ||
c9e7a609 NB |
3138 | /* Create a new allocation buffer. Place the control block at the end |
3139 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 | 3140 | static _cpp_buff * |
6cf87ca4 | 3141 | new_buff (size_t len) |
b8af0ca5 NB |
3142 | { |
3143 | _cpp_buff *result; | |
ece54d54 | 3144 | unsigned char *base; |
b8af0ca5 | 3145 | |
1e013d2e NB |
3146 | if (len < MIN_BUFF_SIZE) |
3147 | len = MIN_BUFF_SIZE; | |
c70f6ed3 | 3148 | len = CPP_ALIGN (len); |
b8af0ca5 | 3149 | |
1a80db97 JJ |
3150 | #ifdef ENABLE_VALGRIND_CHECKING |
3151 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff | |
3152 | struct first. */ | |
3153 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
3154 | base = XNEWVEC (unsigned char, len + slen); | |
3155 | result = (_cpp_buff *) base; | |
3156 | base += slen; | |
3157 | #else | |
c3f829c1 | 3158 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
b8af0ca5 | 3159 | result = (_cpp_buff *) (base + len); |
1a80db97 | 3160 | #endif |
b8af0ca5 NB |
3161 | result->base = base; |
3162 | result->cur = base; | |
3163 | result->limit = base + len; | |
3164 | result->next = NULL; | |
3165 | return result; | |
3166 | } | |
3167 | ||
3168 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
3169 | void | |
6cf87ca4 | 3170 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
b8af0ca5 NB |
3171 | { |
3172 | _cpp_buff *end = buff; | |
3173 | ||
3174 | while (end->next) | |
3175 | end = end->next; | |
3176 | end->next = pfile->free_buffs; | |
3177 | pfile->free_buffs = buff; | |
3178 | } | |
3179 | ||
3180 | /* Return a free buffer of size at least MIN_SIZE. */ | |
3181 | _cpp_buff * | |
6cf87ca4 | 3182 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
b8af0ca5 NB |
3183 | { |
3184 | _cpp_buff *result, **p; | |
3185 | ||
3186 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
3187 | { | |
6142088c | 3188 | size_t size; |
1e013d2e NB |
3189 | |
3190 | if (*p == NULL) | |
b8af0ca5 | 3191 | return new_buff (min_size); |
1e013d2e NB |
3192 | result = *p; |
3193 | size = result->limit - result->base; | |
3194 | /* Return a buffer that's big enough, but don't waste one that's | |
3195 | way too big. */ | |
34f5271d | 3196 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
b8af0ca5 NB |
3197 | break; |
3198 | } | |
3199 | ||
3200 | *p = result->next; | |
3201 | result->next = NULL; | |
3202 | result->cur = result->base; | |
3203 | return result; | |
3204 | } | |
3205 | ||
4fe9b91c | 3206 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3207 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
3208 | the excess bytes to the new buffer. Chains the new buffer after | |
3209 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 3210 | _cpp_buff * |
6cf87ca4 | 3211 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
b8af0ca5 | 3212 | { |
6142088c | 3213 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 3214 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 3215 | |
8c3b2693 NB |
3216 | buff->next = new_buff; |
3217 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
3218 | return new_buff; | |
3219 | } | |
3220 | ||
4fe9b91c | 3221 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3222 | remaining bytes of the buffer pointed to by BUFF, and at least |
3223 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
3224 | Chains the new buffer before the buffer pointed to by BUFF, and | |
3225 | updates the pointer to point to the new buffer. */ | |
3226 | void | |
6cf87ca4 | 3227 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
8c3b2693 NB |
3228 | { |
3229 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
3230 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
3231 | ||
3232 | new_buff = _cpp_get_buff (pfile, size); | |
3233 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
3234 | new_buff->next = old_buff; | |
3235 | *pbuff = new_buff; | |
b8af0ca5 NB |
3236 | } |
3237 | ||
3238 | /* Free a chain of buffers starting at BUFF. */ | |
3239 | void | |
5671bf27 | 3240 | _cpp_free_buff (_cpp_buff *buff) |
b8af0ca5 NB |
3241 | { |
3242 | _cpp_buff *next; | |
3243 | ||
3244 | for (; buff; buff = next) | |
3245 | { | |
3246 | next = buff->next; | |
1a80db97 JJ |
3247 | #ifdef ENABLE_VALGRIND_CHECKING |
3248 | free (buff); | |
3249 | #else | |
b8af0ca5 | 3250 | free (buff->base); |
1a80db97 | 3251 | #endif |
b8af0ca5 NB |
3252 | } |
3253 | } | |
417f3e3a | 3254 | |
ece54d54 NB |
3255 | /* Allocate permanent, unaligned storage of length LEN. */ |
3256 | unsigned char * | |
6cf87ca4 | 3257 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
ece54d54 NB |
3258 | { |
3259 | _cpp_buff *buff = pfile->u_buff; | |
3260 | unsigned char *result = buff->cur; | |
3261 | ||
3262 | if (len > (size_t) (buff->limit - result)) | |
3263 | { | |
3264 | buff = _cpp_get_buff (pfile, len); | |
3265 | buff->next = pfile->u_buff; | |
3266 | pfile->u_buff = buff; | |
3267 | result = buff->cur; | |
3268 | } | |
3269 | ||
3270 | buff->cur = result + len; | |
3271 | return result; | |
3272 | } | |
3273 | ||
87062813 NB |
3274 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
3275 | That buffer is used for growing allocations when saving macro | |
3276 | replacement lists in a #define, and when parsing an answer to an | |
3277 | assertion in #assert, #unassert or #if (and therefore possibly | |
3278 | whilst expanding macros). It therefore must not be used by any | |
3279 | code that they might call: specifically the lexer and the guts of | |
3280 | the macro expander. | |
3281 | ||
3282 | All existing other uses clearly fit this restriction: storing | |
3283 | registered pragmas during initialization. */ | |
93c80368 | 3284 | unsigned char * |
6cf87ca4 | 3285 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
3fef5b2b | 3286 | { |
8c3b2693 NB |
3287 | _cpp_buff *buff = pfile->a_buff; |
3288 | unsigned char *result = buff->cur; | |
3fef5b2b | 3289 | |
8c3b2693 | 3290 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 3291 | { |
8c3b2693 NB |
3292 | buff = _cpp_get_buff (pfile, len); |
3293 | buff->next = pfile->a_buff; | |
3294 | pfile->a_buff = buff; | |
3295 | result = buff->cur; | |
3fef5b2b | 3296 | } |
041c3194 | 3297 | |
8c3b2693 | 3298 | buff->cur = result + len; |
93c80368 | 3299 | return result; |
041c3194 | 3300 | } |
d8044160 GK |
3301 | |
3302 | /* Say which field of TOK is in use. */ | |
3303 | ||
3304 | enum cpp_token_fld_kind | |
c26302d5 | 3305 | cpp_token_val_index (const cpp_token *tok) |
d8044160 GK |
3306 | { |
3307 | switch (TOKEN_SPELL (tok)) | |
3308 | { | |
3309 | case SPELL_IDENT: | |
3310 | return CPP_TOKEN_FLD_NODE; | |
3311 | case SPELL_LITERAL: | |
3312 | return CPP_TOKEN_FLD_STR; | |
aa508502 JM |
3313 | case SPELL_OPERATOR: |
3314 | if (tok->type == CPP_PASTE) | |
9a0c6187 | 3315 | return CPP_TOKEN_FLD_TOKEN_NO; |
aa508502 JM |
3316 | else |
3317 | return CPP_TOKEN_FLD_NONE; | |
d8044160 GK |
3318 | case SPELL_NONE: |
3319 | if (tok->type == CPP_MACRO_ARG) | |
3320 | return CPP_TOKEN_FLD_ARG_NO; | |
3321 | else if (tok->type == CPP_PADDING) | |
3322 | return CPP_TOKEN_FLD_SOURCE; | |
21b11495 | 3323 | else if (tok->type == CPP_PRAGMA) |
bc4071dd | 3324 | return CPP_TOKEN_FLD_PRAGMA; |
d8044160 GK |
3325 | /* else fall through */ |
3326 | default: | |
3327 | return CPP_TOKEN_FLD_NONE; | |
3328 | } | |
3329 | } | |
e3dfef44 GC |
3330 | |
3331 | /* All tokens lexed in R after calling this function will be forced to have | |
3332 | their source_location the same as the location referenced by P, until | |
3333 | cpp_stop_forcing_token_locations is called for R. */ | |
3334 | ||
3335 | void | |
3336 | cpp_force_token_locations (cpp_reader *r, source_location *p) | |
3337 | { | |
3338 | r->forced_token_location_p = p; | |
3339 | } | |
3340 | ||
3341 | /* Go back to assigning locations naturally for lexed tokens. */ | |
3342 | ||
3343 | void | |
3344 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
3345 | { | |
3346 | r->forced_token_location_p = NULL; | |
3347 | } |