]>
Commit | Line | Data |
---|---|---|
45b966db | 1 | /* CPP Library - lexical analysis. |
8d9254fc | 2 | Copyright (C) 2000-2020 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 | ||
a3a821c9 | 571 | data = __builtin_vec_vsx_ld (0, s); |
0ccaaab0 BS |
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); | |
191816a3 | 613 | /* FALLTHRU */ |
0ccaaab0 BS |
614 | case 2: |
615 | l = u.l[i++]; | |
616 | if (l != 0) | |
617 | break; | |
618 | s += sizeof(unsigned long); | |
619 | l = u.l[i]; | |
620 | } | |
621 | ||
622 | /* L now contains 0xff in bytes for which we matched one of the | |
623 | relevant characters. We can find the byte index by finding | |
624 | its bit index and dividing by 8. */ | |
625 | #ifdef __BIG_ENDIAN__ | |
626 | l = __builtin_clzl(l) >> 3; | |
627 | #else | |
628 | l = __builtin_ctzl(l) >> 3; | |
629 | #endif | |
630 | return s + l; | |
631 | ||
632 | #undef N | |
633 | } | |
634 | } | |
635 | ||
636 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__) | |
637 | ||
638 | /* A vection of the fast scanner using AltiVec vectorized byte compares. | |
639 | This cannot be used for little endian because vec_lvsl/lvsr are | |
640 | deprecated for little endian and the code won't work properly. */ | |
246a2fcb RH |
641 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, |
642 | so we can't compile this function without -maltivec on the command line | |
643 | (or implied by some other switch). */ | |
644 | ||
645 | static const uchar * | |
646 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
647 | { | |
648 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
649 | ||
650 | const vc repl_nl = { | |
651 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
652 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
653 | }; | |
654 | const vc repl_cr = { | |
655 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
656 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
657 | }; | |
658 | const vc repl_bs = { | |
659 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
660 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
661 | }; | |
662 | const vc repl_qm = { | |
663 | '?', '?', '?', '?', '?', '?', '?', '?', | |
664 | '?', '?', '?', '?', '?', '?', '?', '?', | |
665 | }; | |
666 | const vc ones = { | |
667 | -1, -1, -1, -1, -1, -1, -1, -1, | |
668 | -1, -1, -1, -1, -1, -1, -1, -1, | |
669 | }; | |
670 | const vc zero = { 0 }; | |
671 | ||
672 | vc data, mask, t; | |
673 | ||
674 | /* Altivec loads automatically mask addresses with -16. This lets us | |
675 | issue the first load as early as possible. */ | |
676 | data = __builtin_vec_ld(0, (const vc *)s); | |
677 | ||
678 | /* Discard bytes before the beginning of the buffer. Do this by | |
679 | beginning with all ones and shifting in zeros according to the | |
680 | mis-alignment. The LVSR instruction pulls the exact shift we | |
681 | want from the address. */ | |
682 | mask = __builtin_vec_lvsr(0, s); | |
683 | mask = __builtin_vec_perm(zero, ones, mask); | |
684 | data &= mask; | |
685 | ||
686 | /* While altivec loads mask addresses, we still need to align S so | |
687 | that the offset we compute at the end is correct. */ | |
688 | s = (const uchar *)((uintptr_t)s & -16); | |
689 | ||
690 | /* Main loop processing 16 bytes at a time. */ | |
691 | goto start; | |
692 | do | |
693 | { | |
694 | vc m_nl, m_cr, m_bs, m_qm; | |
695 | ||
696 | s += 16; | |
697 | data = __builtin_vec_ld(0, (const vc *)s); | |
698 | ||
699 | start: | |
700 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
701 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
702 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
703 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
704 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
705 | ||
706 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
707 | characters. We want to exit the loop if any byte in T is non-zero. | |
708 | Below is the expansion of vec_any_ne(t, zero). */ | |
709 | } | |
710 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
711 | ||
712 | { | |
713 | #define N (sizeof(vc) / sizeof(long)) | |
714 | ||
246a2fcb RH |
715 | union { |
716 | vc v; | |
53a103d3 DS |
717 | /* Statically assert that N is 2 or 4. */ |
718 | unsigned long l[(N == 2 || N == 4) ? N : -1]; | |
246a2fcb RH |
719 | } u; |
720 | unsigned long l, i = 0; | |
721 | ||
722 | u.v = t; | |
723 | ||
724 | /* Find the first word of T that is non-zero. */ | |
725 | switch (N) | |
726 | { | |
727 | case 4: | |
728 | l = u.l[i++]; | |
729 | if (l != 0) | |
730 | break; | |
731 | s += sizeof(unsigned long); | |
732 | l = u.l[i++]; | |
733 | if (l != 0) | |
734 | break; | |
735 | s += sizeof(unsigned long); | |
67ef83c6 | 736 | /* FALLTHROUGH */ |
246a2fcb RH |
737 | case 2: |
738 | l = u.l[i++]; | |
739 | if (l != 0) | |
740 | break; | |
741 | s += sizeof(unsigned long); | |
742 | l = u.l[i]; | |
743 | } | |
744 | ||
745 | /* L now contains 0xff in bytes for which we matched one of the | |
746 | relevant characters. We can find the byte index by finding | |
747 | its bit index and dividing by 8. */ | |
748 | l = __builtin_clzl(l) >> 3; | |
749 | return s + l; | |
750 | ||
751 | #undef N | |
752 | } | |
753 | } | |
754 | ||
a6ac871c RE |
755 | #elif defined (__ARM_NEON) && defined (__ARM_64BIT_STATE) |
756 | #include "arm_neon.h" | |
757 | ||
758 | /* This doesn't have to be the exact page size, but no system may use | |
759 | a size smaller than this. ARMv8 requires a minimum page size of | |
760 | 4k. The impact of being conservative here is a small number of | |
761 | cases will take the slightly slower entry path into the main | |
762 | loop. */ | |
763 | ||
764 | #define AARCH64_MIN_PAGE_SIZE 4096 | |
765 | ||
766 | static const uchar * | |
767 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
768 | { | |
769 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); | |
770 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); | |
771 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); | |
772 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); | |
773 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); | |
774 | ||
35c4515b | 775 | #ifdef __ARM_BIG_ENDIAN |
a6ac871c RE |
776 | const int16x8_t shift = {8, 8, 8, 8, 0, 0, 0, 0}; |
777 | #else | |
778 | const int16x8_t shift = {0, 0, 0, 0, 8, 8, 8, 8}; | |
779 | #endif | |
780 | ||
781 | unsigned int found; | |
782 | const uint8_t *p; | |
783 | uint8x16_t data; | |
784 | uint8x16_t t; | |
785 | uint16x8_t m; | |
786 | uint8x16_t u, v, w; | |
787 | ||
788 | /* Align the source pointer. */ | |
789 | p = (const uint8_t *)((uintptr_t)s & -16); | |
790 | ||
791 | /* Assuming random string start positions, with a 4k page size we'll take | |
792 | the slow path about 0.37% of the time. */ | |
793 | if (__builtin_expect ((AARCH64_MIN_PAGE_SIZE | |
794 | - (((uintptr_t) s) & (AARCH64_MIN_PAGE_SIZE - 1))) | |
795 | < 16, 0)) | |
796 | { | |
797 | /* Slow path: the string starts near a possible page boundary. */ | |
798 | uint32_t misalign, mask; | |
799 | ||
800 | misalign = (uintptr_t)s & 15; | |
801 | mask = (-1u << misalign) & 0xffff; | |
802 | data = vld1q_u8 (p); | |
803 | t = vceqq_u8 (data, repl_nl); | |
804 | u = vceqq_u8 (data, repl_cr); | |
805 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
806 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
807 | t = vorrq_u8 (v, w); | |
808 | t = vandq_u8 (t, xmask); | |
809 | m = vpaddlq_u8 (t); | |
810 | m = vshlq_u16 (m, shift); | |
811 | found = vaddvq_u16 (m); | |
812 | found &= mask; | |
813 | if (found) | |
814 | return (const uchar*)p + __builtin_ctz (found); | |
815 | } | |
816 | else | |
817 | { | |
818 | data = vld1q_u8 ((const uint8_t *) s); | |
819 | t = vceqq_u8 (data, repl_nl); | |
820 | u = vceqq_u8 (data, repl_cr); | |
821 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
822 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
823 | t = vorrq_u8 (v, w); | |
8c00ae24 | 824 | if (__builtin_expect (vpaddd_u64 ((uint64x2_t)t) != 0, 0)) |
a6ac871c RE |
825 | goto done; |
826 | } | |
827 | ||
828 | do | |
829 | { | |
830 | p += 16; | |
831 | data = vld1q_u8 (p); | |
832 | t = vceqq_u8 (data, repl_nl); | |
833 | u = vceqq_u8 (data, repl_cr); | |
834 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
835 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
836 | t = vorrq_u8 (v, w); | |
837 | } while (!vpaddd_u64 ((uint64x2_t)t)); | |
838 | ||
839 | done: | |
840 | /* Now that we've found the terminating substring, work out precisely where | |
841 | we need to stop. */ | |
842 | t = vandq_u8 (t, xmask); | |
843 | m = vpaddlq_u8 (t); | |
844 | m = vshlq_u16 (m, shift); | |
845 | found = vaddvq_u16 (m); | |
846 | return (((((uintptr_t) p) < (uintptr_t) s) ? s : (const uchar *)p) | |
847 | + __builtin_ctz (found)); | |
848 | } | |
849 | ||
95d0610c | 850 | #elif defined (__ARM_NEON) |
e75b54a2 RE |
851 | #include "arm_neon.h" |
852 | ||
853 | static const uchar * | |
854 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
855 | { | |
856 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); | |
857 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); | |
858 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); | |
859 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); | |
860 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); | |
861 | ||
862 | unsigned int misalign, found, mask; | |
863 | const uint8_t *p; | |
864 | uint8x16_t data; | |
865 | ||
866 | /* Align the source pointer. */ | |
867 | misalign = (uintptr_t)s & 15; | |
868 | p = (const uint8_t *)((uintptr_t)s & -16); | |
869 | data = vld1q_u8 (p); | |
870 | ||
871 | /* Create a mask for the bytes that are valid within the first | |
872 | 16-byte block. The Idea here is that the AND with the mask | |
873 | within the loop is "free", since we need some AND or TEST | |
874 | insn in order to set the flags for the branch anyway. */ | |
875 | mask = (-1u << misalign) & 0xffff; | |
876 | ||
877 | /* Main loop, processing 16 bytes at a time. */ | |
878 | goto start; | |
879 | ||
880 | do | |
881 | { | |
882 | uint8x8_t l; | |
883 | uint16x4_t m; | |
884 | uint32x2_t n; | |
885 | uint8x16_t t, u, v, w; | |
886 | ||
887 | p += 16; | |
888 | data = vld1q_u8 (p); | |
889 | mask = 0xffff; | |
890 | ||
891 | start: | |
892 | t = vceqq_u8 (data, repl_nl); | |
893 | u = vceqq_u8 (data, repl_cr); | |
894 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
895 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
896 | t = vandq_u8 (vorrq_u8 (v, w), xmask); | |
897 | l = vpadd_u8 (vget_low_u8 (t), vget_high_u8 (t)); | |
898 | m = vpaddl_u8 (l); | |
899 | n = vpaddl_u16 (m); | |
900 | ||
901 | found = vget_lane_u32 ((uint32x2_t) vorr_u64 ((uint64x1_t) n, | |
902 | vshr_n_u64 ((uint64x1_t) n, 24)), 0); | |
903 | found &= mask; | |
904 | } | |
905 | while (!found); | |
906 | ||
907 | /* FOUND contains 1 in bits for which we matched a relevant | |
908 | character. Conversion to the byte index is trivial. */ | |
909 | found = __builtin_ctz (found); | |
910 | return (const uchar *)p + found; | |
911 | } | |
912 | ||
246a2fcb RH |
913 | #else |
914 | ||
5764ee3c | 915 | /* We only have one accelerated alternative. Use a direct call so that |
246a2fcb RH |
916 | we encourage inlining. */ |
917 | ||
918 | #define search_line_fast search_line_acc_char | |
919 | ||
920 | #endif | |
921 | ||
b0c084b7 JJ |
922 | /* Initialize the lexer if needed. */ |
923 | ||
924 | void | |
925 | _cpp_init_lexer (void) | |
926 | { | |
927 | #ifdef HAVE_init_vectorized_lexer | |
928 | init_vectorized_lexer (); | |
929 | #endif | |
930 | } | |
931 | ||
26aea073 NB |
932 | /* Returns with a logical line that contains no escaped newlines or |
933 | trigraphs. This is a time-critical inner loop. */ | |
934 | void | |
6cf87ca4 | 935 | _cpp_clean_line (cpp_reader *pfile) |
45b966db | 936 | { |
26aea073 NB |
937 | cpp_buffer *buffer; |
938 | const uchar *s; | |
939 | uchar c, *d, *p; | |
87062813 | 940 | |
26aea073 NB |
941 | buffer = pfile->buffer; |
942 | buffer->cur_note = buffer->notes_used = 0; | |
943 | buffer->cur = buffer->line_base = buffer->next_line; | |
944 | buffer->need_line = false; | |
246a2fcb | 945 | s = buffer->next_line; |
87062813 | 946 | |
26aea073 | 947 | if (!buffer->from_stage3) |
45b966db | 948 | { |
7af45bd4 ILT |
949 | const uchar *pbackslash = NULL; |
950 | ||
246a2fcb | 951 | /* Fast path. This is the common case of an un-escaped line with |
d08dcf87 ZW |
952 | no trigraphs. The primary win here is by not writing any |
953 | data back to memory until we have to. */ | |
246a2fcb | 954 | while (1) |
d08dcf87 | 955 | { |
246a2fcb RH |
956 | /* Perform an optimized search for \n, \r, \\, ?. */ |
957 | s = search_line_fast (s, buffer->rlimit); | |
d08dcf87 | 958 | |
246a2fcb RH |
959 | c = *s; |
960 | if (c == '\\') | |
961 | { | |
962 | /* Record the location of the backslash and continue. */ | |
963 | pbackslash = s++; | |
d08dcf87 | 964 | } |
246a2fcb | 965 | else if (__builtin_expect (c == '?', 0)) |
d08dcf87 | 966 | { |
246a2fcb RH |
967 | if (__builtin_expect (s[1] == '?', false) |
968 | && _cpp_trigraph_map[s[2]]) | |
d08dcf87 | 969 | { |
246a2fcb RH |
970 | /* Have a trigraph. We may or may not have to convert |
971 | it. Add a line note regardless, for -Wtrigraphs. */ | |
972 | add_line_note (buffer, s, s[2]); | |
973 | if (CPP_OPTION (pfile, trigraphs)) | |
974 | { | |
975 | /* We do, and that means we have to switch to the | |
976 | slow path. */ | |
977 | d = (uchar *) s; | |
978 | *d = _cpp_trigraph_map[s[2]]; | |
979 | s += 2; | |
980 | goto slow_path; | |
981 | } | |
d08dcf87 | 982 | } |
246a2fcb RH |
983 | /* Not a trigraph. Continue on fast-path. */ |
984 | s++; | |
d08dcf87 | 985 | } |
246a2fcb RH |
986 | else |
987 | break; | |
d08dcf87 ZW |
988 | } |
989 | ||
246a2fcb RH |
990 | /* This must be \r or \n. We're either done, or we'll be forced |
991 | to write back to the buffer and continue on the slow path. */ | |
992 | d = (uchar *) s; | |
993 | ||
994 | if (__builtin_expect (s == buffer->rlimit, false)) | |
995 | goto done; | |
996 | ||
997 | /* DOS line ending? */ | |
998 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
999 | { | |
1000 | s++; | |
1001 | if (s == buffer->rlimit) | |
1002 | goto done; | |
1003 | } | |
1004 | ||
1005 | if (__builtin_expect (pbackslash == NULL, true)) | |
1006 | goto done; | |
1007 | ||
1008 | /* Check for escaped newline. */ | |
1009 | p = d; | |
1010 | while (is_nvspace (p[-1])) | |
1011 | p--; | |
1012 | if (p - 1 != pbackslash) | |
1013 | goto done; | |
1014 | ||
1015 | /* Have an escaped newline; process it and proceed to | |
1016 | the slow path. */ | |
1017 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
1018 | d = p - 2; | |
1019 | buffer->next_line = p - 1; | |
26aea073 | 1020 | |
246a2fcb RH |
1021 | slow_path: |
1022 | while (1) | |
4a5b68a2 | 1023 | { |
26aea073 NB |
1024 | c = *++s; |
1025 | *++d = c; | |
1026 | ||
1027 | if (c == '\n' || c == '\r') | |
1028 | { | |
246a2fcb | 1029 | /* Handle DOS line endings. */ |
26aea073 NB |
1030 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
1031 | s++; | |
1032 | if (s == buffer->rlimit) | |
1033 | break; | |
1034 | ||
1035 | /* Escaped? */ | |
1036 | p = d; | |
1037 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
1038 | p--; | |
1039 | if (p == buffer->next_line || p[-1] != '\\') | |
1040 | break; | |
1041 | ||
41c32c98 | 1042 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
26aea073 NB |
1043 | d = p - 2; |
1044 | buffer->next_line = p - 1; | |
1045 | } | |
1046 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
1047 | { | |
1048 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
41c32c98 | 1049 | add_line_note (buffer, d, s[2]); |
26aea073 NB |
1050 | if (CPP_OPTION (pfile, trigraphs)) |
1051 | { | |
1052 | *d = _cpp_trigraph_map[s[2]]; | |
1053 | s += 2; | |
1054 | } | |
1055 | } | |
4a5b68a2 | 1056 | } |
45b966db | 1057 | } |
26aea073 NB |
1058 | else |
1059 | { | |
246a2fcb | 1060 | while (*s != '\n' && *s != '\r') |
26aea073 | 1061 | s++; |
26aea073 NB |
1062 | d = (uchar *) s; |
1063 | ||
1064 | /* Handle DOS line endings. */ | |
1065 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
1066 | s++; | |
1067 | } | |
0d9f234d | 1068 | |
d08dcf87 | 1069 | done: |
26aea073 | 1070 | *d = '\n'; |
41c32c98 NB |
1071 | /* A sentinel note that should never be processed. */ |
1072 | add_line_note (buffer, d + 1, '\n'); | |
26aea073 | 1073 | buffer->next_line = s + 1; |
45b966db ZW |
1074 | } |
1075 | ||
a8eb6044 NB |
1076 | /* Return true if the trigraph indicated by NOTE should be warned |
1077 | about in a comment. */ | |
1078 | static bool | |
6cf87ca4 | 1079 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
a8eb6044 NB |
1080 | { |
1081 | const uchar *p; | |
1082 | ||
1083 | /* Within comments we don't warn about trigraphs, unless the | |
1084 | trigraph forms an escaped newline, as that may change | |
6356f892 | 1085 | behavior. */ |
a8eb6044 NB |
1086 | if (note->type != '/') |
1087 | return false; | |
1088 | ||
1089 | /* If -trigraphs, then this was an escaped newline iff the next note | |
1090 | is coincident. */ | |
1091 | if (CPP_OPTION (pfile, trigraphs)) | |
1092 | return note[1].pos == note->pos; | |
1093 | ||
1094 | /* Otherwise, see if this forms an escaped newline. */ | |
1095 | p = note->pos + 3; | |
1096 | while (is_nvspace (*p)) | |
1097 | p++; | |
1098 | ||
1099 | /* There might have been escaped newlines between the trigraph and the | |
1100 | newline we found. Hence the position test. */ | |
1101 | return (*p == '\n' && p < note[1].pos); | |
1102 | } | |
1103 | ||
26aea073 NB |
1104 | /* Process the notes created by add_line_note as far as the current |
1105 | location. */ | |
1106 | void | |
6cf87ca4 | 1107 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
45b966db | 1108 | { |
29401c30 NB |
1109 | cpp_buffer *buffer = pfile->buffer; |
1110 | ||
26aea073 | 1111 | for (;;) |
041c3194 | 1112 | { |
26aea073 NB |
1113 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
1114 | unsigned int col; | |
a5c3cccd | 1115 | |
26aea073 NB |
1116 | if (note->pos > buffer->cur) |
1117 | break; | |
a5c3cccd | 1118 | |
26aea073 NB |
1119 | buffer->cur_note++; |
1120 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
4d6baafa | 1121 | |
41c32c98 | 1122 | if (note->type == '\\' || note->type == ' ') |
26aea073 | 1123 | { |
41c32c98 | 1124 | if (note->type == ' ' && !in_comment) |
500bee0a | 1125 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
26aea073 | 1126 | "backslash and newline separated by space"); |
41c32c98 | 1127 | |
26aea073 | 1128 | if (buffer->next_line > buffer->rlimit) |
87062813 | 1129 | { |
500bee0a | 1130 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
26aea073 NB |
1131 | "backslash-newline at end of file"); |
1132 | /* Prevent "no newline at end of file" warning. */ | |
1133 | buffer->next_line = buffer->rlimit; | |
87062813 | 1134 | } |
26aea073 NB |
1135 | |
1136 | buffer->line_base = note->pos; | |
12f9df4e | 1137 | CPP_INCREMENT_LINE (pfile, 0); |
0d9f234d | 1138 | } |
41c32c98 NB |
1139 | else if (_cpp_trigraph_map[note->type]) |
1140 | { | |
a8eb6044 NB |
1141 | if (CPP_OPTION (pfile, warn_trigraphs) |
1142 | && (!in_comment || warn_in_comment (pfile, note))) | |
41c32c98 NB |
1143 | { |
1144 | if (CPP_OPTION (pfile, trigraphs)) | |
87cf0651 SB |
1145 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
1146 | pfile->line_table->highest_line, col, | |
1147 | "trigraph ??%c converted to %c", | |
1148 | note->type, | |
1149 | (int) _cpp_trigraph_map[note->type]); | |
41c32c98 | 1150 | else |
905bd7b5 | 1151 | { |
87cf0651 SB |
1152 | cpp_warning_with_line |
1153 | (pfile, CPP_W_TRIGRAPHS, | |
1154 | pfile->line_table->highest_line, col, | |
905bd7b5 GK |
1155 | "trigraph ??%c ignored, use -trigraphs to enable", |
1156 | note->type); | |
1157 | } | |
41c32c98 NB |
1158 | } |
1159 | } | |
00a81b8b JM |
1160 | else if (note->type == 0) |
1161 | /* Already processed in lex_raw_string. */; | |
41c32c98 NB |
1162 | else |
1163 | abort (); | |
041c3194 | 1164 | } |
45b966db ZW |
1165 | } |
1166 | ||
0d9f234d NB |
1167 | /* Skip a C-style block comment. We find the end of the comment by |
1168 | seeing if an asterisk is before every '/' we encounter. Returns | |
6f572ac2 NB |
1169 | nonzero if comment terminated by EOF, zero otherwise. |
1170 | ||
1171 | Buffer->cur points to the initial asterisk of the comment. */ | |
26aea073 | 1172 | bool |
6cf87ca4 | 1173 | _cpp_skip_block_comment (cpp_reader *pfile) |
45b966db | 1174 | { |
041c3194 | 1175 | cpp_buffer *buffer = pfile->buffer; |
d08dcf87 ZW |
1176 | const uchar *cur = buffer->cur; |
1177 | uchar c; | |
0d9f234d | 1178 | |
d08dcf87 ZW |
1179 | cur++; |
1180 | if (*cur == '/') | |
1181 | cur++; | |
0d9f234d | 1182 | |
26aea073 NB |
1183 | for (;;) |
1184 | { | |
0d9f234d NB |
1185 | /* People like decorating comments with '*', so check for '/' |
1186 | instead for efficiency. */ | |
d08dcf87 ZW |
1187 | c = *cur++; |
1188 | ||
041c3194 | 1189 | if (c == '/') |
45b966db | 1190 | { |
d08dcf87 | 1191 | if (cur[-2] == '*') |
0d9f234d | 1192 | break; |
041c3194 | 1193 | |
0d9f234d | 1194 | /* Warn about potential nested comments, but not if the '/' |
a1f300c0 | 1195 | comes immediately before the true comment delimiter. |
041c3194 | 1196 | Don't bother to get it right across escaped newlines. */ |
0d9f234d | 1197 | if (CPP_OPTION (pfile, warn_comments) |
d08dcf87 ZW |
1198 | && cur[0] == '*' && cur[1] != '/') |
1199 | { | |
1200 | buffer->cur = cur; | |
87cf0651 SB |
1201 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
1202 | pfile->line_table->highest_line, | |
1203 | CPP_BUF_COL (buffer), | |
1204 | "\"/*\" within comment"); | |
d08dcf87 | 1205 | } |
45b966db | 1206 | } |
26aea073 NB |
1207 | else if (c == '\n') |
1208 | { | |
12f9df4e | 1209 | unsigned int cols; |
d08dcf87 | 1210 | buffer->cur = cur - 1; |
26aea073 NB |
1211 | _cpp_process_line_notes (pfile, true); |
1212 | if (buffer->next_line >= buffer->rlimit) | |
1213 | return true; | |
1214 | _cpp_clean_line (pfile); | |
12f9df4e PB |
1215 | |
1216 | cols = buffer->next_line - buffer->line_base; | |
1217 | CPP_INCREMENT_LINE (pfile, cols); | |
1218 | ||
d08dcf87 | 1219 | cur = buffer->cur; |
26aea073 | 1220 | } |
45b966db | 1221 | } |
041c3194 | 1222 | |
d08dcf87 | 1223 | buffer->cur = cur; |
a8eb6044 | 1224 | _cpp_process_line_notes (pfile, true); |
26aea073 | 1225 | return false; |
45b966db ZW |
1226 | } |
1227 | ||
480709cc | 1228 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
da7d8304 | 1229 | terminating newline. Handles escaped newlines. Returns nonzero |
480709cc | 1230 | if a multiline comment. */ |
041c3194 | 1231 | static int |
6cf87ca4 | 1232 | skip_line_comment (cpp_reader *pfile) |
45b966db | 1233 | { |
cbcff6df | 1234 | cpp_buffer *buffer = pfile->buffer; |
620e594b | 1235 | location_t orig_line = pfile->line_table->highest_line; |
041c3194 | 1236 | |
26aea073 NB |
1237 | while (*buffer->cur != '\n') |
1238 | buffer->cur++; | |
480709cc | 1239 | |
26aea073 | 1240 | _cpp_process_line_notes (pfile, true); |
500bee0a | 1241 | return orig_line != pfile->line_table->highest_line; |
041c3194 | 1242 | } |
45b966db | 1243 | |
26aea073 | 1244 | /* Skips whitespace, saving the next non-whitespace character. */ |
52fadca8 | 1245 | static void |
6cf87ca4 | 1246 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
041c3194 ZW |
1247 | { |
1248 | cpp_buffer *buffer = pfile->buffer; | |
f7d151fb | 1249 | bool saw_NUL = false; |
45b966db | 1250 | |
0d9f234d | 1251 | do |
041c3194 | 1252 | { |
91fcd158 | 1253 | /* Horizontal space always OK. */ |
26aea073 | 1254 | if (c == ' ' || c == '\t') |
0d9f234d | 1255 | ; |
0d9f234d | 1256 | /* Just \f \v or \0 left. */ |
91fcd158 | 1257 | else if (c == '\0') |
f7d151fb | 1258 | saw_NUL = true; |
93c80368 | 1259 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
500bee0a | 1260 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
ebef4e8c NB |
1261 | CPP_BUF_COL (buffer), |
1262 | "%s in preprocessing directive", | |
1263 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d | 1264 | |
0d9f234d | 1265 | c = *buffer->cur++; |
45b966db | 1266 | } |
ec5c56db | 1267 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
1268 | while (is_nvspace (c)); |
1269 | ||
f7d151fb | 1270 | if (saw_NUL) |
0527bc4e | 1271 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
f7d151fb | 1272 | |
480709cc | 1273 | buffer->cur--; |
041c3194 | 1274 | } |
45b966db | 1275 | |
93c80368 NB |
1276 | /* See if the characters of a number token are valid in a name (no |
1277 | '.', '+' or '-'). */ | |
1278 | static int | |
6cf87ca4 | 1279 | name_p (cpp_reader *pfile, const cpp_string *string) |
93c80368 NB |
1280 | { |
1281 | unsigned int i; | |
1282 | ||
1283 | for (i = 0; i < string->len; i++) | |
1284 | if (!is_idchar (string->text[i])) | |
1285 | return 0; | |
1286 | ||
df383483 | 1287 | return 1; |
93c80368 NB |
1288 | } |
1289 | ||
50668cf6 GK |
1290 | /* After parsing an identifier or other sequence, produce a warning about |
1291 | sequences not in NFC/NFKC. */ | |
1292 | static void | |
1293 | warn_about_normalization (cpp_reader *pfile, | |
1294 | const cpp_token *token, | |
1295 | const struct normalize_state *s) | |
1296 | { | |
1297 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1298 | && !pfile->state.skipping) | |
1299 | { | |
1300 | /* Make sure that the token is printed using UCNs, even | |
1301 | if we'd otherwise happily print UTF-8. */ | |
c3f829c1 | 1302 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
50668cf6 GK |
1303 | size_t sz; |
1304 | ||
1305 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1306 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
87cf0651 SB |
1307 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1308 | "`%.*s' is not in NFKC", (int) sz, buf); | |
50668cf6 | 1309 | else |
87cf0651 SB |
1310 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1311 | "`%.*s' is not in NFC", (int) sz, buf); | |
55e7f907 | 1312 | free (buf); |
50668cf6 GK |
1313 | } |
1314 | } | |
1315 | ||
7d112d66 LH |
1316 | static const cppchar_t utf8_signifier = 0xC0; |
1317 | ||
1318 | /* Returns TRUE if the sequence starting at buffer->cur is valid in | |
1613e52b | 1319 | an identifier. FIRST is TRUE if this starts an identifier. */ |
bced6edf | 1320 | static bool |
50668cf6 GK |
1321 | forms_identifier_p (cpp_reader *pfile, int first, |
1322 | struct normalize_state *state) | |
bced6edf | 1323 | { |
1613e52b NB |
1324 | cpp_buffer *buffer = pfile->buffer; |
1325 | ||
1326 | if (*buffer->cur == '$') | |
1327 | { | |
1328 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1329 | return false; | |
1330 | ||
1331 | buffer->cur++; | |
78b8811a | 1332 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
1613e52b | 1333 | { |
78b8811a | 1334 | CPP_OPTION (pfile, warn_dollars) = 0; |
0527bc4e | 1335 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
1613e52b NB |
1336 | } |
1337 | ||
1338 | return true; | |
1339 | } | |
bced6edf | 1340 | |
7d112d66 LH |
1341 | /* Is this a syntactically valid UCN or a valid UTF-8 char? */ |
1342 | if (CPP_OPTION (pfile, extended_identifiers)) | |
bced6edf | 1343 | { |
fbb22910 | 1344 | cppchar_t s; |
7d112d66 LH |
1345 | if (*buffer->cur >= utf8_signifier) |
1346 | { | |
1347 | if (_cpp_valid_utf8 (pfile, &buffer->cur, buffer->rlimit, 1 + !first, | |
1348 | state, &s)) | |
1349 | return true; | |
1350 | } | |
1351 | else if (*buffer->cur == '\\' | |
1352 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) | |
1353 | { | |
1354 | buffer->cur += 2; | |
1355 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, | |
1356 | state, &s, NULL, NULL)) | |
1357 | return true; | |
1358 | buffer->cur -= 2; | |
1359 | } | |
bced6edf | 1360 | } |
bced6edf | 1361 | |
1613e52b | 1362 | return false; |
bced6edf NB |
1363 | } |
1364 | ||
fb771b9d TT |
1365 | /* Helper function to issue error about improper __VA_OPT__ use. */ |
1366 | static void | |
1367 | maybe_va_opt_error (cpp_reader *pfile) | |
1368 | { | |
1369 | if (CPP_PEDANTIC (pfile) && !CPP_OPTION (pfile, va_opt)) | |
1370 | { | |
1371 | /* __VA_OPT__ should not be accepted at all, but allow it in | |
1372 | system headers. */ | |
1373 | if (!cpp_in_system_header (pfile)) | |
1374 | cpp_error (pfile, CPP_DL_PEDWARN, | |
b04445d4 | 1375 | "__VA_OPT__ is not available until C++20"); |
fb771b9d TT |
1376 | } |
1377 | else if (!pfile->state.va_args_ok) | |
1378 | { | |
1379 | /* __VA_OPT__ should only appear in the replacement list of a | |
1380 | variadic macro. */ | |
1381 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1382 | "__VA_OPT__ can only appear in the expansion" | |
b04445d4 | 1383 | " of a C++20 variadic macro"); |
fb771b9d TT |
1384 | } |
1385 | } | |
1386 | ||
17e7cb85 KT |
1387 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1388 | static cpp_hashnode * | |
1389 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1390 | { | |
1391 | cpp_hashnode *result; | |
1392 | const uchar *cur; | |
1393 | unsigned int len; | |
1394 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1395 | ||
1396 | cur = base + 1; | |
1397 | while (ISIDNUM (*cur)) | |
1398 | { | |
1399 | hash = HT_HASHSTEP (hash, *cur); | |
1400 | cur++; | |
1401 | } | |
1402 | len = cur - base; | |
1403 | hash = HT_HASHFINISH (hash, len); | |
1404 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1405 | base, len, hash, HT_ALLOC)); | |
1406 | ||
1407 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1408 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1409 | && !pfile->state.skipping, 0)) | |
1410 | { | |
1411 | /* It is allowed to poison the same identifier twice. */ | |
1412 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1413 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1414 | NODE_NAME (result)); | |
1415 | ||
1416 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1417 | replacement list of a variadic macro. */ | |
1418 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1419 | && !pfile->state.va_args_ok) | |
3976796b ESR |
1420 | { |
1421 | if (CPP_OPTION (pfile, cplusplus)) | |
1422 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1423 | "__VA_ARGS__ can only appear in the expansion" | |
1424 | " of a C++11 variadic macro"); | |
1425 | else | |
1426 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1427 | "__VA_ARGS__ can only appear in the expansion" | |
1428 | " of a C99 variadic macro"); | |
1429 | } | |
17e7cb85 | 1430 | |
fb771b9d TT |
1431 | if (result == pfile->spec_nodes.n__VA_OPT__) |
1432 | maybe_va_opt_error (pfile); | |
1433 | ||
17e7cb85 KT |
1434 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
1435 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1436 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1437 | "identifier \"%s\" is a special operator name in C++", | |
1438 | NODE_NAME (result)); | |
17e7cb85 KT |
1439 | } |
1440 | ||
1441 | return result; | |
1442 | } | |
1443 | ||
1444 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1445 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1446 | cpp_hashnode * | |
1447 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1448 | { | |
1449 | cpp_hashnode *result; | |
1450 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1451 | return result; | |
1452 | } | |
1453 | ||
bced6edf | 1454 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
0d9f234d | 1455 | static cpp_hashnode * |
50668cf6 | 1456 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
be5ffc59 | 1457 | struct normalize_state *nst, cpp_hashnode **spelling) |
45b966db | 1458 | { |
93c80368 | 1459 | cpp_hashnode *result; |
47e20491 | 1460 | const uchar *cur; |
c6e83800 ZW |
1461 | unsigned int len; |
1462 | unsigned int hash = HT_HASHSTEP (0, *base); | |
2c3fcba6 | 1463 | |
c6e83800 | 1464 | cur = pfile->buffer->cur; |
47e20491 | 1465 | if (! starts_ucn) |
d3f4ff8b JM |
1466 | { |
1467 | while (ISIDNUM (*cur)) | |
1468 | { | |
1469 | hash = HT_HASHSTEP (hash, *cur); | |
1470 | cur++; | |
1471 | } | |
1472 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *(cur - 1)); | |
1473 | } | |
47e20491 | 1474 | pfile->buffer->cur = cur; |
50668cf6 | 1475 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
10cf9bde | 1476 | { |
7d112d66 LH |
1477 | /* Slower version for identifiers containing UCNs |
1478 | or extended chars (including $). */ | |
47e20491 GK |
1479 | do { |
1480 | while (ISIDNUM (*pfile->buffer->cur)) | |
50668cf6 | 1481 | { |
d3f4ff8b | 1482 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *pfile->buffer->cur); |
50668cf6 | 1483 | pfile->buffer->cur++; |
50668cf6 GK |
1484 | } |
1485 | } while (forms_identifier_p (pfile, false, nst)); | |
47e20491 GK |
1486 | result = _cpp_interpret_identifier (pfile, base, |
1487 | pfile->buffer->cur - base); | |
be5ffc59 | 1488 | *spelling = cpp_lookup (pfile, base, pfile->buffer->cur - base); |
2c3fcba6 | 1489 | } |
47e20491 GK |
1490 | else |
1491 | { | |
1492 | len = cur - base; | |
1493 | hash = HT_HASHFINISH (hash, len); | |
bced6edf | 1494 | |
2bf41bf0 TT |
1495 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1496 | base, len, hash, HT_ALLOC)); | |
be5ffc59 | 1497 | *spelling = result; |
47e20491 | 1498 | } |
2c3fcba6 | 1499 | |
bced6edf | 1500 | /* Rarely, identifiers require diagnostics when lexed. */ |
2c3fcba6 ZW |
1501 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1502 | && !pfile->state.skipping, 0)) | |
1503 | { | |
1504 | /* It is allowed to poison the same identifier twice. */ | |
1505 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
0527bc4e | 1506 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
2c3fcba6 ZW |
1507 | NODE_NAME (result)); |
1508 | ||
1509 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1510 | replacement list of a variadic macro. */ | |
1511 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1512 | && !pfile->state.va_args_ok) | |
3976796b ESR |
1513 | { |
1514 | if (CPP_OPTION (pfile, cplusplus)) | |
1515 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1516 | "__VA_ARGS__ can only appear in the expansion" | |
1517 | " of a C++11 variadic macro"); | |
1518 | else | |
1519 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1520 | "__VA_ARGS__ can only appear in the expansion" | |
1521 | " of a C99 variadic macro"); | |
1522 | } | |
3d8b2a98 | 1523 | |
fb771b9d TT |
1524 | /* __VA_OPT__ should only appear in the replacement list of a |
1525 | variadic macro. */ | |
1526 | if (result == pfile->spec_nodes.n__VA_OPT__) | |
1527 | maybe_va_opt_error (pfile); | |
1528 | ||
3d8b2a98 ILT |
1529 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
1530 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1531 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1532 | "identifier \"%s\" is a special operator name in C++", | |
1533 | NODE_NAME (result)); | |
2c3fcba6 ZW |
1534 | } |
1535 | ||
1536 | return result; | |
1537 | } | |
1538 | ||
bced6edf | 1539 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
45b966db | 1540 | static void |
50668cf6 GK |
1541 | lex_number (cpp_reader *pfile, cpp_string *number, |
1542 | struct normalize_state *nst) | |
45b966db | 1543 | { |
562a5c27 | 1544 | const uchar *cur; |
bced6edf NB |
1545 | const uchar *base; |
1546 | uchar *dest; | |
45b966db | 1547 | |
bced6edf NB |
1548 | base = pfile->buffer->cur - 1; |
1549 | do | |
041c3194 | 1550 | { |
bced6edf | 1551 | cur = pfile->buffer->cur; |
0d9f234d | 1552 | |
bced6edf | 1553 | /* N.B. ISIDNUM does not include $. */ |
7057e645 ESR |
1554 | while (ISIDNUM (*cur) || *cur == '.' || DIGIT_SEP (*cur) |
1555 | || VALID_SIGN (*cur, cur[-1])) | |
50668cf6 | 1556 | { |
d3f4ff8b | 1557 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *cur); |
50668cf6 | 1558 | cur++; |
50668cf6 | 1559 | } |
a5858a3d ESR |
1560 | /* A number can't end with a digit separator. */ |
1561 | while (cur > pfile->buffer->cur && DIGIT_SEP (cur[-1])) | |
1562 | --cur; | |
45b966db | 1563 | |
10cf9bde | 1564 | pfile->buffer->cur = cur; |
45b966db | 1565 | } |
50668cf6 | 1566 | while (forms_identifier_p (pfile, false, nst)); |
93c80368 | 1567 | |
bced6edf NB |
1568 | number->len = cur - base; |
1569 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1570 | memcpy (dest, base, number->len); | |
1571 | dest[number->len] = '\0'; | |
1572 | number->text = dest; | |
93c80368 NB |
1573 | } |
1574 | ||
6338b358 NB |
1575 | /* Create a token of type TYPE with a literal spelling. */ |
1576 | static void | |
6cf87ca4 ZW |
1577 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1578 | unsigned int len, enum cpp_ttype type) | |
6338b358 NB |
1579 | { |
1580 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1581 | ||
1582 | memcpy (dest, base, len); | |
1583 | dest[len] = '\0'; | |
1584 | token->type = type; | |
1585 | token->val.str.len = len; | |
1586 | token->val.str.text = dest; | |
1587 | } | |
1588 | ||
ed63c387 NS |
1589 | /* A pair of raw buffer pointers. The currently open one is [1], the |
1590 | first one is [0]. Used for string literal lexing. */ | |
1591 | struct lit_accum { | |
1592 | _cpp_buff *first; | |
1593 | _cpp_buff *last; | |
1594 | const uchar *rpos; | |
1595 | size_t accum; | |
1596 | ||
1597 | lit_accum () | |
1598 | : first (NULL), last (NULL), rpos (0), accum (0) | |
1599 | { | |
1600 | } | |
1601 | ||
1602 | void append (cpp_reader *, const uchar *, size_t); | |
1603 | ||
1604 | void read_begin (cpp_reader *); | |
1605 | bool reading_p () const | |
1606 | { | |
1607 | return rpos != NULL; | |
1608 | } | |
1609 | char read_char () | |
1610 | { | |
1611 | char c = *rpos++; | |
1612 | if (rpos == BUFF_FRONT (last)) | |
1613 | rpos = NULL; | |
1614 | return c; | |
1615 | } | |
1616 | }; | |
1617 | ||
00a81b8b JM |
1618 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1619 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1620 | ||
ed63c387 NS |
1621 | void |
1622 | lit_accum::append (cpp_reader *pfile, const uchar *base, size_t len) | |
00a81b8b | 1623 | { |
ed63c387 NS |
1624 | if (!last) |
1625 | /* Starting. */ | |
1626 | first = last = _cpp_get_buff (pfile, len); | |
1627 | else if (len > BUFF_ROOM (last)) | |
00a81b8b | 1628 | { |
ed63c387 NS |
1629 | /* There is insufficient room in the buffer. Copy what we can, |
1630 | and then either extend or create a new one. */ | |
1631 | size_t room = BUFF_ROOM (last); | |
1632 | memcpy (BUFF_FRONT (last), base, room); | |
1633 | BUFF_FRONT (last) += room; | |
00a81b8b JM |
1634 | base += room; |
1635 | len -= room; | |
ed63c387 | 1636 | accum += room; |
00a81b8b | 1637 | |
ed63c387 NS |
1638 | gcc_checking_assert (!rpos); |
1639 | ||
1640 | last = _cpp_append_extend_buff (pfile, last, len); | |
1641 | } | |
00a81b8b | 1642 | |
ed63c387 NS |
1643 | memcpy (BUFF_FRONT (last), base, len); |
1644 | BUFF_FRONT (last) += len; | |
1645 | accum += len; | |
00a81b8b JM |
1646 | } |
1647 | ||
ed63c387 NS |
1648 | void |
1649 | lit_accum::read_begin (cpp_reader *pfile) | |
1650 | { | |
1651 | /* We never accumulate more than 4 chars to read. */ | |
1652 | if (BUFF_ROOM (last) < 4) | |
1653 | ||
1654 | last = _cpp_append_extend_buff (pfile, last, 4); | |
1655 | rpos = BUFF_FRONT (last); | |
1656 | } | |
c865f923 ESR |
1657 | |
1658 | /* Returns true if a macro has been defined. | |
1659 | This might not work if compile with -save-temps, | |
1660 | or preprocess separately from compilation. */ | |
1661 | ||
1662 | static bool | |
1663 | is_macro(cpp_reader *pfile, const uchar *base) | |
1664 | { | |
1665 | const uchar *cur = base; | |
1666 | if (! ISIDST (*cur)) | |
1667 | return false; | |
1668 | unsigned int hash = HT_HASHSTEP (0, *cur); | |
1669 | ++cur; | |
1670 | while (ISIDNUM (*cur)) | |
1671 | { | |
1672 | hash = HT_HASHSTEP (hash, *cur); | |
1673 | ++cur; | |
1674 | } | |
1675 | hash = HT_HASHFINISH (hash, cur - base); | |
1676 | ||
1677 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1678 | base, cur - base, hash, HT_NO_INSERT)); | |
1679 | ||
3f6677f4 | 1680 | return result && cpp_macro_p (result); |
c865f923 ESR |
1681 | } |
1682 | ||
b44f8ad8 JW |
1683 | /* Returns true if a literal suffix does not have the expected form |
1684 | and is defined as a macro. */ | |
1685 | ||
1686 | static bool | |
1687 | is_macro_not_literal_suffix(cpp_reader *pfile, const uchar *base) | |
1688 | { | |
1689 | /* User-defined literals outside of namespace std must start with a single | |
1690 | underscore, so assume anything of that form really is a UDL suffix. | |
1691 | We don't need to worry about UDLs defined inside namespace std because | |
1692 | their names are reserved, so cannot be used as macro names in valid | |
1693 | programs. */ | |
1694 | if (base[0] == '_' && base[1] != '_') | |
1695 | return false; | |
1696 | return is_macro (pfile, base); | |
1697 | } | |
c865f923 | 1698 | |
ed63c387 NS |
1699 | /* Lexes a raw string. The stored string contains the spelling, |
1700 | including double quotes, delimiter string, '(' and ')', any leading | |
1701 | 'L', 'u', 'U' or 'u8' and 'R' modifier. The created token contains | |
1702 | the type of the literal, or CPP_OTHER if it was not properly | |
1703 | terminated. | |
1704 | ||
1705 | BASE is the start of the token. Updates pfile->buffer->cur to just | |
1706 | after the lexed string. | |
2c6e3f55 JJ |
1707 | |
1708 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1709 | is the first NUL since embedded NULs are preserved. */ | |
1710 | ||
1711 | static void | |
ed63c387 | 1712 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
2c6e3f55 | 1713 | { |
ed63c387 NS |
1714 | const uchar *pos = base; |
1715 | ||
1716 | /* 'tis a pity this information isn't passed down from the lexer's | |
1717 | initial categorization of the token. */ | |
1718 | enum cpp_ttype type = CPP_STRING; | |
1719 | ||
1720 | if (*pos == 'L') | |
1721 | { | |
1722 | type = CPP_WSTRING; | |
1723 | pos++; | |
1724 | } | |
1725 | else if (*pos == 'U') | |
1726 | { | |
1727 | type = CPP_STRING32; | |
1728 | pos++; | |
1729 | } | |
1730 | else if (*pos == 'u') | |
1731 | { | |
1732 | if (pos[1] == '8') | |
1733 | { | |
1734 | type = CPP_UTF8STRING; | |
1735 | pos++; | |
1736 | } | |
1737 | else | |
1738 | type = CPP_STRING16; | |
1739 | pos++; | |
1740 | } | |
1741 | ||
1742 | gcc_checking_assert (pos[0] == 'R' && pos[1] == '"'); | |
1743 | pos += 2; | |
1744 | ||
00a81b8b | 1745 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 1746 | |
ed63c387 NS |
1747 | /* Skip notes before the ". */ |
1748 | while (note->pos < pos) | |
1749 | ++note; | |
1750 | ||
1751 | lit_accum accum; | |
1752 | ||
1753 | uchar prefix[17]; | |
1754 | unsigned prefix_len = 0; | |
1755 | enum Phase | |
1756 | { | |
1757 | PHASE_PREFIX = -2, | |
1758 | PHASE_NONE = -1, | |
1759 | PHASE_SUFFIX = 0 | |
1760 | } phase = PHASE_PREFIX; | |
1761 | ||
8cf88735 JJ |
1762 | for (;;) |
1763 | { | |
ed63c387 | 1764 | gcc_checking_assert (note->pos >= pos); |
00a81b8b | 1765 | |
ed63c387 NS |
1766 | /* Undo any escaped newlines and trigraphs. */ |
1767 | if (!accum.reading_p () && note->pos == pos) | |
1768 | switch (note->type) | |
1769 | { | |
1770 | case '\\': | |
1771 | case ' ': | |
1772 | /* Restore backslash followed by newline. */ | |
1773 | accum.append (pfile, base, pos - base); | |
1774 | base = pos; | |
1775 | accum.read_begin (pfile); | |
1776 | accum.append (pfile, UC"\\", 1); | |
1777 | ||
1778 | after_backslash: | |
1779 | if (note->type == ' ') | |
1780 | /* GNU backslash whitespace newline extension. FIXME | |
1781 | could be any sequence of non-vertical space. When we | |
1782 | can properly restore any such sequence, we should | |
1783 | mark this note as handled so _cpp_process_line_notes | |
1784 | doesn't warn. */ | |
1785 | accum.append (pfile, UC" ", 1); | |
1786 | ||
1787 | accum.append (pfile, UC"\n", 1); | |
1788 | note++; | |
1789 | break; | |
00a81b8b | 1790 | |
ed63c387 NS |
1791 | case '\n': |
1792 | /* This can happen for ??/<NEWLINE> when trigraphs are not | |
1793 | being interpretted. */ | |
1794 | gcc_checking_assert (!CPP_OPTION (pfile, trigraphs)); | |
1795 | note->type = 0; | |
1796 | note++; | |
1797 | break; | |
00a81b8b | 1798 | |
ed63c387 NS |
1799 | default: |
1800 | gcc_checking_assert (_cpp_trigraph_map[note->type]); | |
1801 | ||
1802 | /* Don't warn about this trigraph in | |
1803 | _cpp_process_line_notes, since trigraphs show up as | |
1804 | trigraphs in raw strings. */ | |
1805 | uchar type = note->type; | |
1806 | note->type = 0; | |
1807 | ||
1808 | if (CPP_OPTION (pfile, trigraphs)) | |
1809 | { | |
1810 | accum.append (pfile, base, pos - base); | |
1811 | base = pos; | |
1812 | accum.read_begin (pfile); | |
1813 | accum.append (pfile, UC"??", 2); | |
1814 | accum.append (pfile, &type, 1); | |
1815 | ||
1816 | /* ??/ followed by newline gets two line notes, one for | |
1817 | the trigraph and one for the backslash/newline. */ | |
1818 | if (type == '/' && note[1].pos == pos) | |
1819 | { | |
1820 | note++; | |
1821 | gcc_assert (note->type == '\\' || note->type == ' '); | |
1822 | goto after_backslash; | |
1823 | } | |
1824 | /* Skip the replacement character. */ | |
1825 | base = ++pos; | |
1826 | } | |
1827 | ||
1828 | note++; | |
1829 | break; | |
1830 | } | |
1831 | ||
1832 | /* Now get a char to process. Either from an expanded note, or | |
1833 | from the line buffer. */ | |
1834 | bool read_note = accum.reading_p (); | |
1835 | char c = read_note ? accum.read_char () : *pos++; | |
2c6e3f55 | 1836 | |
ed63c387 | 1837 | if (phase == PHASE_PREFIX) |
2c6e3f55 | 1838 | { |
ed63c387 | 1839 | if (c == '(') |
8cf88735 | 1840 | { |
ed63c387 NS |
1841 | /* Done. */ |
1842 | phase = PHASE_NONE; | |
1843 | prefix[prefix_len++] = '"'; | |
1844 | } | |
1845 | else if (prefix_len < 16 | |
1846 | /* Prefix chars are any of the basic character set, | |
1847 | [lex.charset] except for ' | |
1848 | ()\\\t\v\f\n'. Optimized for a contiguous | |
1849 | alphabet. */ | |
1850 | /* Unlike a switch, this collapses down to one or | |
1851 | two shift and bitmask operations on an ASCII | |
1852 | system, with an outlier or two. */ | |
1853 | && (('Z' - 'A' == 25 | |
1854 | ? ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) | |
1855 | : ISIDST (c)) | |
1856 | || (c >= '0' && c <= '9') | |
1857 | || c == '_' || c == '{' || c == '}' | |
1858 | || c == '[' || c == ']' || c == '#' | |
1859 | || c == '<' || c == '>' || c == '%' | |
1860 | || c == ':' || c == ';' || c == '.' || c == '?' | |
1861 | || c == '*' || c == '+' || c == '-' || c == '/' | |
1862 | || c == '^' || c == '&' || c == '|' || c == '~' | |
1863 | || c == '!' || c == '=' || c == ',' | |
1864 | || c == '"' || c == '\'')) | |
1865 | prefix[prefix_len++] = c; | |
1866 | else | |
1867 | { | |
1868 | /* Something is wrong. */ | |
1869 | int col = CPP_BUF_COLUMN (pfile->buffer, pos) + read_note; | |
1870 | if (prefix_len == 16) | |
1871 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1872 | col, "raw string delimiter longer " | |
1873 | "than 16 characters"); | |
1874 | else if (c == '\n') | |
1875 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1876 | col, "invalid new-line in raw " | |
1877 | "string delimiter"); | |
1878 | else | |
1879 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1880 | col, "invalid character '%c' in " | |
1881 | "raw string delimiter", c); | |
1882 | type = CPP_OTHER; | |
1883 | phase = PHASE_NONE; | |
1884 | /* Continue until we get a close quote, that's probably | |
1885 | the best failure mode. */ | |
1886 | prefix_len = 0; | |
8cf88735 JJ |
1887 | } |
1888 | continue; | |
1889 | } | |
ed63c387 NS |
1890 | |
1891 | if (phase != PHASE_NONE) | |
8cf88735 | 1892 | { |
ed63c387 NS |
1893 | if (prefix[phase] != c) |
1894 | phase = PHASE_NONE; | |
1895 | else if (unsigned (phase + 1) == prefix_len) | |
8cf88735 | 1896 | break; |
ed63c387 NS |
1897 | else |
1898 | { | |
1899 | phase = Phase (phase + 1); | |
1900 | continue; | |
1901 | } | |
2c6e3f55 | 1902 | } |
ed63c387 NS |
1903 | |
1904 | if (!prefix_len && c == '"') | |
1905 | /* Failure mode lexing. */ | |
1906 | goto out; | |
1907 | else if (prefix_len && c == ')') | |
1908 | phase = PHASE_SUFFIX; | |
1909 | else if (!read_note && c == '\n') | |
2c6e3f55 | 1910 | { |
ed63c387 NS |
1911 | pos--; |
1912 | pfile->buffer->cur = pos; | |
2c6e3f55 | 1913 | if (pfile->state.in_directive |
d5e48350 JJ |
1914 | || (pfile->state.parsing_args |
1915 | && pfile->buffer->next_line >= pfile->buffer->rlimit)) | |
2c6e3f55 | 1916 | { |
2c6e3f55 JJ |
1917 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, |
1918 | "unterminated raw string"); | |
ed63c387 NS |
1919 | type = CPP_OTHER; |
1920 | goto out; | |
2c6e3f55 JJ |
1921 | } |
1922 | ||
ed63c387 | 1923 | accum.append (pfile, base, pos - base + 1); |
2a0225e4 NS |
1924 | _cpp_process_line_notes (pfile, false); |
1925 | ||
1926 | if (pfile->buffer->next_line < pfile->buffer->rlimit) | |
2c6e3f55 JJ |
1927 | CPP_INCREMENT_LINE (pfile, 0); |
1928 | pfile->buffer->need_line = true; | |
1929 | ||
1930 | if (!_cpp_get_fresh_line (pfile)) | |
1931 | { | |
ed63c387 | 1932 | /* We ran out of file and failed to get a line. */ |
620e594b | 1933 | location_t src_loc = token->src_loc; |
2c6e3f55 JJ |
1934 | token->type = CPP_EOF; |
1935 | /* Tell the compiler the line number of the EOF token. */ | |
1936 | token->src_loc = pfile->line_table->highest_line; | |
1937 | token->flags = BOL; | |
ed63c387 NS |
1938 | if (accum.first) |
1939 | _cpp_release_buff (pfile, accum.first); | |
2c6e3f55 JJ |
1940 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, |
1941 | "unterminated raw string"); | |
2a0225e4 NS |
1942 | /* Now pop the buffer that _cpp_get_fresh_line did not. */ |
1943 | _cpp_pop_buffer (pfile); | |
2c6e3f55 JJ |
1944 | return; |
1945 | } | |
1946 | ||
ed63c387 | 1947 | pos = base = pfile->buffer->cur; |
00a81b8b | 1948 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 1949 | } |
2c6e3f55 JJ |
1950 | } |
1951 | ||
3ce4f9e4 ESR |
1952 | if (CPP_OPTION (pfile, user_literals)) |
1953 | { | |
c865f923 ESR |
1954 | /* If a string format macro, say from inttypes.h, is placed touching |
1955 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 | 1956 | literal thus breaking the program. */ |
ed63c387 | 1957 | if (is_macro_not_literal_suffix (pfile, pos)) |
7f5f5f98 | 1958 | { |
112448b4 | 1959 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 1960 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
1961 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1962 | token->src_loc, 0, | |
1963 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 1964 | "a space between literal and string macro"); |
7f5f5f98 | 1965 | } |
3ce4f9e4 | 1966 | /* Grab user defined literal suffix. */ |
ed63c387 | 1967 | else if (ISIDST (*pos)) |
3ce4f9e4 ESR |
1968 | { |
1969 | type = cpp_userdef_string_add_type (type); | |
ed63c387 | 1970 | ++pos; |
7f5f5f98 | 1971 | |
ed63c387 NS |
1972 | while (ISIDNUM (*pos)) |
1973 | ++pos; | |
3ce4f9e4 | 1974 | } |
3ce4f9e4 ESR |
1975 | } |
1976 | ||
ed63c387 NS |
1977 | out: |
1978 | pfile->buffer->cur = pos; | |
1979 | if (!accum.accum) | |
1980 | create_literal (pfile, token, base, pos - base, type); | |
2c6e3f55 JJ |
1981 | else |
1982 | { | |
ed63c387 NS |
1983 | size_t extra_len = pos - base; |
1984 | uchar *dest = _cpp_unaligned_alloc (pfile, accum.accum + extra_len + 1); | |
2c6e3f55 JJ |
1985 | |
1986 | token->type = type; | |
ed63c387 | 1987 | token->val.str.len = accum.accum + extra_len; |
2c6e3f55 | 1988 | token->val.str.text = dest; |
ed63c387 | 1989 | for (_cpp_buff *buf = accum.first; buf; buf = buf->next) |
2c6e3f55 | 1990 | { |
ed63c387 NS |
1991 | size_t len = BUFF_FRONT (buf) - buf->base; |
1992 | memcpy (dest, buf->base, len); | |
1993 | dest += len; | |
2c6e3f55 | 1994 | } |
ed63c387 NS |
1995 | _cpp_release_buff (pfile, accum.first); |
1996 | memcpy (dest, base, extra_len); | |
1997 | dest[extra_len] = '\0'; | |
2c6e3f55 JJ |
1998 | } |
1999 | } | |
2000 | ||
bced6edf | 2001 | /* Lexes a string, character constant, or angle-bracketed header file |
6338b358 | 2002 | name. The stored string contains the spelling, including opening |
2c6e3f55 JJ |
2003 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
2004 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
2005 | if it was not properly terminated, or CPP_LESS for an unterminated | |
2006 | header name which must be relexed as normal tokens. | |
6338b358 NB |
2007 | |
2008 | The spelling is NUL-terminated, but it is not guaranteed that this | |
2009 | is the first NUL since embedded NULs are preserved. */ | |
041c3194 | 2010 | static void |
6cf87ca4 | 2011 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
45b966db | 2012 | { |
6338b358 NB |
2013 | bool saw_NUL = false; |
2014 | const uchar *cur; | |
bced6edf | 2015 | cppchar_t terminator; |
6338b358 NB |
2016 | enum cpp_ttype type; |
2017 | ||
2018 | cur = base; | |
2019 | terminator = *cur++; | |
2c6e3f55 | 2020 | if (terminator == 'L' || terminator == 'U') |
6338b358 | 2021 | terminator = *cur++; |
2c6e3f55 JJ |
2022 | else if (terminator == 'u') |
2023 | { | |
2024 | terminator = *cur++; | |
2025 | if (terminator == '8') | |
2026 | terminator = *cur++; | |
2027 | } | |
2028 | if (terminator == 'R') | |
2029 | { | |
ed63c387 | 2030 | lex_raw_string (pfile, token, base); |
2c6e3f55 JJ |
2031 | return; |
2032 | } | |
2033 | if (terminator == '"') | |
b6baa67d KVH |
2034 | type = (*base == 'L' ? CPP_WSTRING : |
2035 | *base == 'U' ? CPP_STRING32 : | |
2c6e3f55 JJ |
2036 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
2037 | : CPP_STRING); | |
6338b358 | 2038 | else if (terminator == '\'') |
b6baa67d KVH |
2039 | type = (*base == 'L' ? CPP_WCHAR : |
2040 | *base == 'U' ? CPP_CHAR32 : | |
fe95b036 ESR |
2041 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8CHAR : CPP_CHAR16) |
2042 | : CPP_CHAR); | |
6338b358 NB |
2043 | else |
2044 | terminator = '>', type = CPP_HEADER_NAME; | |
93c80368 | 2045 | |
0d9f234d | 2046 | for (;;) |
45b966db | 2047 | { |
6338b358 | 2048 | cppchar_t c = *cur++; |
7868b4a2 | 2049 | |
6f572ac2 | 2050 | /* In #include-style directives, terminators are not escapable. */ |
6338b358 NB |
2051 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
2052 | cur++; | |
2053 | else if (c == terminator) | |
bced6edf | 2054 | break; |
6338b358 | 2055 | else if (c == '\n') |
0d9f234d | 2056 | { |
6338b358 | 2057 | cur--; |
4bb09c26 JM |
2058 | /* Unmatched quotes always yield undefined behavior, but |
2059 | greedy lexing means that what appears to be an unterminated | |
2060 | header name may actually be a legitimate sequence of tokens. */ | |
2061 | if (terminator == '>') | |
2062 | { | |
2063 | token->type = CPP_LESS; | |
2064 | return; | |
2065 | } | |
6338b358 NB |
2066 | type = CPP_OTHER; |
2067 | break; | |
45b966db | 2068 | } |
6338b358 NB |
2069 | else if (c == '\0') |
2070 | saw_NUL = true; | |
45b966db ZW |
2071 | } |
2072 | ||
6338b358 | 2073 | if (saw_NUL && !pfile->state.skipping) |
0527bc4e JDA |
2074 | cpp_error (pfile, CPP_DL_WARNING, |
2075 | "null character(s) preserved in literal"); | |
45b966db | 2076 | |
c663e301 JM |
2077 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
2078 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
2079 | (int) terminator); | |
2080 | ||
3ce4f9e4 ESR |
2081 | if (CPP_OPTION (pfile, user_literals)) |
2082 | { | |
c865f923 ESR |
2083 | /* If a string format macro, say from inttypes.h, is placed touching |
2084 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 JW |
2085 | literal thus breaking the program. */ |
2086 | if (is_macro_not_literal_suffix (pfile, cur)) | |
7f5f5f98 | 2087 | { |
112448b4 | 2088 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 2089 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
2090 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
2091 | token->src_loc, 0, | |
2092 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 2093 | "a space between literal and string macro"); |
7f5f5f98 | 2094 | } |
3ce4f9e4 | 2095 | /* Grab user defined literal suffix. */ |
561f7fc7 | 2096 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
2097 | { |
2098 | type = cpp_userdef_char_add_type (type); | |
2099 | type = cpp_userdef_string_add_type (type); | |
2100 | ++cur; | |
7f5f5f98 OW |
2101 | |
2102 | while (ISIDNUM (*cur)) | |
2103 | ++cur; | |
3ce4f9e4 | 2104 | } |
3ce4f9e4 | 2105 | } |
fe191308 JM |
2106 | else if (CPP_OPTION (pfile, cpp_warn_cxx11_compat) |
2107 | && is_macro (pfile, cur) | |
2108 | && !pfile->state.skipping) | |
2109 | cpp_warning_with_line (pfile, CPP_W_CXX11_COMPAT, | |
2110 | token->src_loc, 0, "C++11 requires a space " | |
2111 | "between string literal and macro"); | |
3ce4f9e4 | 2112 | |
6338b358 NB |
2113 | pfile->buffer->cur = cur; |
2114 | create_literal (pfile, token, base, cur - base, type); | |
0d9f234d | 2115 | } |
041c3194 | 2116 | |
631d0d36 MG |
2117 | /* Return the comment table. The client may not make any assumption |
2118 | about the ordering of the table. */ | |
2119 | cpp_comment_table * | |
2120 | cpp_get_comments (cpp_reader *pfile) | |
2121 | { | |
2122 | return &pfile->comments; | |
2123 | } | |
2124 | ||
2125 | /* Append a comment to the end of the comment table. */ | |
2126 | static void | |
2127 | store_comment (cpp_reader *pfile, cpp_token *token) | |
2128 | { | |
2129 | int len; | |
2130 | ||
2131 | if (pfile->comments.allocated == 0) | |
2132 | { | |
2133 | pfile->comments.allocated = 256; | |
2134 | pfile->comments.entries = (cpp_comment *) xmalloc | |
2135 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
2136 | } | |
2137 | ||
2138 | if (pfile->comments.count == pfile->comments.allocated) | |
2139 | { | |
2140 | pfile->comments.allocated *= 2; | |
2141 | pfile->comments.entries = (cpp_comment *) xrealloc | |
2142 | (pfile->comments.entries, | |
2143 | pfile->comments.allocated * sizeof (cpp_comment)); | |
2144 | } | |
2145 | ||
2146 | len = token->val.str.len; | |
2147 | ||
2148 | /* Copy comment. Note, token may not be NULL terminated. */ | |
2149 | pfile->comments.entries[pfile->comments.count].comment = | |
2150 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
2151 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
2152 | token->val.str.text, len); | |
2153 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
2154 | ||
2155 | /* Set source location. */ | |
2156 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
2157 | ||
2158 | /* Increment the count of entries in the comment table. */ | |
2159 | pfile->comments.count++; | |
2160 | } | |
2161 | ||
93c80368 | 2162 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 2163 | static void |
6cf87ca4 ZW |
2164 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
2165 | cppchar_t type) | |
9e62c811 | 2166 | { |
041c3194 | 2167 | unsigned char *buffer; |
651a20b5 | 2168 | unsigned int len, clen, i; |
df383483 | 2169 | |
1c6d33ef | 2170 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
480709cc | 2171 | |
3542203b NB |
2172 | /* C++ comments probably (not definitely) have moved past a new |
2173 | line, which we don't want to save in the comment. */ | |
480709cc | 2174 | if (is_vspace (pfile->buffer->cur[-1])) |
3542203b | 2175 | len--; |
477cdac7 | 2176 | |
651a20b5 KT |
2177 | /* If we are currently in a directive or in argument parsing, then |
2178 | we need to store all C++ comments as C comments internally, and | |
2179 | so we need to allocate a little extra space in that case. | |
477cdac7 JT |
2180 | |
2181 | Note that the only time we encounter a directive here is | |
2182 | when we are saving comments in a "#define". */ | |
651a20b5 KT |
2183 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
2184 | && type == '/') ? len + 2 : len; | |
477cdac7 JT |
2185 | |
2186 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
df383483 | 2187 | |
041c3194 | 2188 | token->type = CPP_COMMENT; |
477cdac7 | 2189 | token->val.str.len = clen; |
0d9f234d | 2190 | token->val.str.text = buffer; |
45b966db | 2191 | |
1c6d33ef NB |
2192 | buffer[0] = '/'; |
2193 | memcpy (buffer + 1, from, len - 1); | |
477cdac7 | 2194 | |
1eeeb6a4 | 2195 | /* Finish conversion to a C comment, if necessary. */ |
651a20b5 | 2196 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
477cdac7 JT |
2197 | { |
2198 | buffer[1] = '*'; | |
2199 | buffer[clen - 2] = '*'; | |
2200 | buffer[clen - 1] = '/'; | |
651a20b5 KT |
2201 | /* As there can be in a C++ comments illegal sequences for C comments |
2202 | we need to filter them out. */ | |
2203 | for (i = 2; i < (clen - 2); i++) | |
2204 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
2205 | buffer[i] = '|'; | |
477cdac7 | 2206 | } |
631d0d36 MG |
2207 | |
2208 | /* Finally store this comment for use by clients of libcpp. */ | |
2209 | store_comment (pfile, token); | |
0d9f234d | 2210 | } |
45b966db | 2211 | |
81fea426 MP |
2212 | /* Returns true if comment at COMMENT_START is a recognized FALLTHROUGH |
2213 | comment. */ | |
2214 | ||
2215 | static bool | |
2216 | fallthrough_comment_p (cpp_reader *pfile, const unsigned char *comment_start) | |
2217 | { | |
2218 | const unsigned char *from = comment_start + 1; | |
70f6d5e1 JJ |
2219 | |
2220 | switch (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough)) | |
2221 | { | |
2222 | /* For both -Wimplicit-fallthrough=0 and -Wimplicit-fallthrough=5 we | |
2223 | don't recognize any comments. The latter only checks attributes, | |
2224 | the former doesn't warn. */ | |
2225 | case 0: | |
2226 | default: | |
2227 | return false; | |
2228 | /* -Wimplicit-fallthrough=1 considers any comment, no matter what | |
2229 | content it has. */ | |
2230 | case 1: | |
2231 | return true; | |
2232 | case 2: | |
2233 | /* -Wimplicit-fallthrough=2 looks for (case insensitive) | |
2234 | .*falls?[ \t-]*thr(u|ough).* regex. */ | |
2235 | for (; (size_t) (pfile->buffer->cur - from) >= sizeof "fallthru" - 1; | |
2236 | from++) | |
2237 | { | |
2238 | /* Is there anything like strpbrk with upper boundary, or | |
2239 | memchr looking for 2 characters rather than just one? */ | |
2240 | if (from[0] != 'f' && from[0] != 'F') | |
2241 | continue; | |
2242 | if (from[1] != 'a' && from[1] != 'A') | |
2243 | continue; | |
2244 | if (from[2] != 'l' && from[2] != 'L') | |
2245 | continue; | |
2246 | if (from[3] != 'l' && from[3] != 'L') | |
2247 | continue; | |
2248 | from += sizeof "fall" - 1; | |
2249 | if (from[0] == 's' || from[0] == 'S') | |
2250 | from++; | |
2251 | while (*from == ' ' || *from == '\t' || *from == '-') | |
2252 | from++; | |
2253 | if (from[0] != 't' && from[0] != 'T') | |
2254 | continue; | |
2255 | if (from[1] != 'h' && from[1] != 'H') | |
2256 | continue; | |
2257 | if (from[2] != 'r' && from[2] != 'R') | |
2258 | continue; | |
2259 | if (from[3] == 'u' || from[3] == 'U') | |
2260 | return true; | |
2261 | if (from[3] != 'o' && from[3] != 'O') | |
2262 | continue; | |
2263 | if (from[4] != 'u' && from[4] != 'U') | |
2264 | continue; | |
2265 | if (from[5] != 'g' && from[5] != 'G') | |
2266 | continue; | |
2267 | if (from[6] != 'h' && from[6] != 'H') | |
2268 | continue; | |
2269 | return true; | |
2270 | } | |
2271 | return false; | |
2272 | case 3: | |
2273 | case 4: | |
2274 | break; | |
2275 | } | |
2276 | ||
81fea426 MP |
2277 | /* Whole comment contents: |
2278 | -fallthrough | |
2279 | @fallthrough@ | |
2280 | */ | |
2281 | if (*from == '-' || *from == '@') | |
2282 | { | |
2283 | size_t len = sizeof "fallthrough" - 1; | |
2284 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2285 | return false; | |
2286 | if (memcmp (from + 1, "fallthrough", len)) | |
2287 | return false; | |
2288 | if (*from == '@') | |
2289 | { | |
2290 | if (from[len + 1] != '@') | |
2291 | return false; | |
2292 | len++; | |
2293 | } | |
2294 | from += 1 + len; | |
2295 | } | |
2296 | /* Whole comment contents (regex): | |
70f6d5e1 | 2297 | lint -fallthrough[ \t]* |
81b02905 JJ |
2298 | */ |
2299 | else if (*from == 'l') | |
2300 | { | |
2301 | size_t len = sizeof "int -fallthrough" - 1; | |
2302 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2303 | return false; | |
2304 | if (memcmp (from + 1, "int -fallthrough", len)) | |
70f6d5e1 | 2305 | return false; |
81b02905 | 2306 | from += 1 + len; |
70f6d5e1 JJ |
2307 | while (*from == ' ' || *from == '\t') |
2308 | from++; | |
2309 | } | |
2310 | /* Whole comment contents (regex): | |
2311 | [ \t]*FALLTHR(U|OUGH)[ \t]* | |
2312 | */ | |
2313 | else if (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough) == 4) | |
2314 | { | |
2315 | while (*from == ' ' || *from == '\t') | |
2316 | from++; | |
2317 | if ((size_t) (pfile->buffer->cur - from) < sizeof "FALLTHRU" - 1) | |
2318 | return false; | |
2319 | if (memcmp (from, "FALLTHR", sizeof "FALLTHR" - 1)) | |
2320 | return false; | |
2321 | from += sizeof "FALLTHR" - 1; | |
2322 | if (*from == 'U') | |
2323 | from++; | |
2324 | else if ((size_t) (pfile->buffer->cur - from) < sizeof "OUGH" - 1) | |
2325 | return false; | |
2326 | else if (memcmp (from, "OUGH", sizeof "OUGH" - 1)) | |
2327 | return false; | |
2328 | else | |
2329 | from += sizeof "OUGH" - 1; | |
2330 | while (*from == ' ' || *from == '\t') | |
2331 | from++; | |
81b02905 JJ |
2332 | } |
2333 | /* Whole comment contents (regex): | |
ee19ef45 JJ |
2334 | [ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)? |
2335 | [ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
2336 | [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
81fea426 MP |
2337 | */ |
2338 | else | |
2339 | { | |
81b02905 | 2340 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 MP |
2341 | from++; |
2342 | unsigned char f = *from; | |
81b02905 JJ |
2343 | bool all_upper = false; |
2344 | if (f == 'E' || f == 'e') | |
70f6d5e1 JJ |
2345 | { |
2346 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2347 | < sizeof "else fallthru" - 1) |
2348 | return false; | |
ee19ef45 | 2349 | if (f == 'E' && memcmp (from + 1, "LSE", sizeof "LSE" - 1) == 0) |
81b02905 | 2350 | all_upper = true; |
ee19ef45 JJ |
2351 | else if (memcmp (from + 1, "lse", sizeof "lse" - 1)) |
2352 | return false; | |
2353 | from += sizeof "else" - 1; | |
2354 | if (*from == ',') | |
2355 | from++; | |
70f6d5e1 | 2356 | if (*from != ' ') |
ee19ef45 JJ |
2357 | return false; |
2358 | from++; | |
2359 | if (all_upper && *from == 'f') | |
81b02905 | 2360 | return false; |
81b02905 JJ |
2361 | if (f == 'e' && *from == 'F') |
2362 | return false; | |
2363 | f = *from; | |
70f6d5e1 | 2364 | } |
81b02905 | 2365 | else if (f == 'I' || f == 'i') |
70f6d5e1 JJ |
2366 | { |
2367 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2368 | < sizeof "intentional fallthru" - 1) |
2369 | return false; | |
2370 | if (f == 'I' && memcmp (from + 1, "NTENTIONAL", | |
2371 | sizeof "NTENTIONAL" - 1) == 0) | |
2372 | all_upper = true; | |
2373 | else if (memcmp (from + 1, "ntentional", | |
2374 | sizeof "ntentional" - 1)) | |
2375 | return false; | |
2376 | from += sizeof "intentional" - 1; | |
2377 | if (*from == ' ') | |
2378 | { | |
2379 | from++; | |
2380 | if (all_upper && *from == 'f') | |
2381 | return false; | |
2382 | } | |
2383 | else if (all_upper) | |
2384 | { | |
2385 | if (memcmp (from, "LY F", sizeof "LY F" - 1)) | |
2386 | return false; | |
2387 | from += sizeof "LY " - 1; | |
2388 | } | |
2389 | else | |
2390 | { | |
2391 | if (memcmp (from, "ly ", sizeof "ly " - 1)) | |
2392 | return false; | |
2393 | from += sizeof "ly " - 1; | |
2394 | } | |
2395 | if (f == 'i' && *from == 'F') | |
2396 | return false; | |
2397 | f = *from; | |
70f6d5e1 | 2398 | } |
81fea426 MP |
2399 | if (f != 'F' && f != 'f') |
2400 | return false; | |
7bad794a | 2401 | if ((size_t) (pfile->buffer->cur - from) < sizeof "fallthru" - 1) |
81fea426 | 2402 | return false; |
81fea426 MP |
2403 | if (f == 'F' && memcmp (from + 1, "ALL", sizeof "ALL" - 1) == 0) |
2404 | all_upper = true; | |
81b02905 JJ |
2405 | else if (all_upper) |
2406 | return false; | |
81fea426 MP |
2407 | else if (memcmp (from + 1, "all", sizeof "all" - 1)) |
2408 | return false; | |
7bad794a JJ |
2409 | from += sizeof "fall" - 1; |
2410 | if (*from == (all_upper ? 'S' : 's') && from[1] == ' ') | |
2411 | from += 2; | |
2412 | else if (*from == ' ' || *from == '-') | |
2413 | from++; | |
2414 | else if (*from != (all_upper ? 'T' : 't')) | |
81fea426 | 2415 | return false; |
81fea426 MP |
2416 | if ((f == 'f' || *from != 'T') && (all_upper || *from != 't')) |
2417 | return false; | |
7bad794a | 2418 | if ((size_t) (pfile->buffer->cur - from) < sizeof "thru" - 1) |
81fea426 MP |
2419 | return false; |
2420 | if (memcmp (from + 1, all_upper ? "HRU" : "hru", sizeof "hru" - 1)) | |
2421 | { | |
7bad794a | 2422 | if ((size_t) (pfile->buffer->cur - from) < sizeof "through" - 1) |
81fea426 MP |
2423 | return false; |
2424 | if (memcmp (from + 1, all_upper ? "HROUGH" : "hrough", | |
2425 | sizeof "hrough" - 1)) | |
2426 | return false; | |
2427 | from += sizeof "through" - 1; | |
2428 | } | |
2429 | else | |
2430 | from += sizeof "thru" - 1; | |
81b02905 | 2431 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 | 2432 | from++; |
81b02905 JJ |
2433 | if (*from == '-') |
2434 | { | |
2435 | from++; | |
2436 | if (*comment_start == '*') | |
2437 | { | |
2438 | do | |
2439 | { | |
2440 | while (*from && *from != '*' | |
2441 | && *from != '\n' && *from != '\r') | |
2442 | from++; | |
2443 | if (*from != '*' || from[1] == '/') | |
2444 | break; | |
2445 | from++; | |
2446 | } | |
2447 | while (1); | |
2448 | } | |
2449 | else | |
2450 | while (*from && *from != '\n' && *from != '\r') | |
2451 | from++; | |
2452 | } | |
81fea426 MP |
2453 | } |
2454 | /* C block comment. */ | |
2455 | if (*comment_start == '*') | |
2456 | { | |
2457 | if (*from != '*' || from[1] != '/') | |
2458 | return false; | |
2459 | } | |
2460 | /* C++ line comment. */ | |
2461 | else if (*from != '\n') | |
2462 | return false; | |
2463 | ||
2464 | return true; | |
2465 | } | |
2466 | ||
5fddcffc NB |
2467 | /* Allocate COUNT tokens for RUN. */ |
2468 | void | |
6cf87ca4 | 2469 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
5fddcffc | 2470 | { |
72bb2c39 | 2471 | run->base = XNEWVEC (cpp_token, count); |
5fddcffc NB |
2472 | run->limit = run->base + count; |
2473 | run->next = NULL; | |
2474 | } | |
2475 | ||
2476 | /* Returns the next tokenrun, or creates one if there is none. */ | |
2477 | static tokenrun * | |
6cf87ca4 | 2478 | next_tokenrun (tokenrun *run) |
5fddcffc NB |
2479 | { |
2480 | if (run->next == NULL) | |
2481 | { | |
72bb2c39 | 2482 | run->next = XNEW (tokenrun); |
bdcbe496 | 2483 | run->next->prev = run; |
5fddcffc NB |
2484 | _cpp_init_tokenrun (run->next, 250); |
2485 | } | |
2486 | ||
2487 | return run->next; | |
2488 | } | |
2489 | ||
ad2305ad | 2490 | /* Return the number of not yet processed token in a given |
92582b75 TT |
2491 | context. */ |
2492 | int | |
ad2305ad | 2493 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
92582b75 | 2494 | { |
92582b75 | 2495 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
cbbcf655 | 2496 | return (LAST (context).token - FIRST (context).token); |
92582b75 TT |
2497 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
2498 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
cbbcf655 | 2499 | return (LAST (context).ptoken - FIRST (context).ptoken); |
92582b75 TT |
2500 | else |
2501 | abort (); | |
2502 | } | |
2503 | ||
ad2305ad DS |
2504 | /* Returns the token present at index INDEX in a given context. If |
2505 | INDEX is zero, the next token to be processed is returned. */ | |
92582b75 | 2506 | static const cpp_token* |
ad2305ad | 2507 | _cpp_token_from_context_at (cpp_context *context, int index) |
92582b75 | 2508 | { |
92582b75 TT |
2509 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
2510 | return &(FIRST (context).token[index]); | |
2511 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
2512 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
2513 | return FIRST (context).ptoken[index]; | |
2514 | else | |
2515 | abort (); | |
2516 | } | |
2517 | ||
5950c3c9 BE |
2518 | /* Look ahead in the input stream. */ |
2519 | const cpp_token * | |
2520 | cpp_peek_token (cpp_reader *pfile, int index) | |
2521 | { | |
2522 | cpp_context *context = pfile->context; | |
2523 | const cpp_token *peektok; | |
2524 | int count; | |
2525 | ||
2526 | /* First, scan through any pending cpp_context objects. */ | |
2527 | while (context->prev) | |
2528 | { | |
ad2305ad | 2529 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
5950c3c9 BE |
2530 | |
2531 | if (index < (int) sz) | |
ad2305ad | 2532 | return _cpp_token_from_context_at (context, index); |
5950c3c9 BE |
2533 | index -= (int) sz; |
2534 | context = context->prev; | |
2535 | } | |
2536 | ||
2537 | /* We will have to read some new tokens after all (and do so | |
2538 | without invalidating preceding tokens). */ | |
2539 | count = index; | |
2540 | pfile->keep_tokens++; | |
2541 | ||
b8cd77f4 JJ |
2542 | /* For peeked tokens temporarily disable line_change reporting, |
2543 | until the tokens are parsed for real. */ | |
2544 | void (*line_change) (cpp_reader *, const cpp_token *, int) | |
2545 | = pfile->cb.line_change; | |
2546 | pfile->cb.line_change = NULL; | |
2547 | ||
5950c3c9 BE |
2548 | do |
2549 | { | |
2550 | peektok = _cpp_lex_token (pfile); | |
2551 | if (peektok->type == CPP_EOF) | |
e4b33ee5 JJ |
2552 | { |
2553 | index--; | |
2554 | break; | |
2555 | } | |
5950c3c9 BE |
2556 | } |
2557 | while (index--); | |
2558 | ||
e4b33ee5 | 2559 | _cpp_backup_tokens_direct (pfile, count - index); |
5950c3c9 | 2560 | pfile->keep_tokens--; |
b8cd77f4 | 2561 | pfile->cb.line_change = line_change; |
5950c3c9 BE |
2562 | |
2563 | return peektok; | |
2564 | } | |
2565 | ||
4ed5bcfb NB |
2566 | /* Allocate a single token that is invalidated at the same time as the |
2567 | rest of the tokens on the line. Has its line and col set to the | |
2568 | same as the last lexed token, so that diagnostics appear in the | |
2569 | right place. */ | |
2570 | cpp_token * | |
6cf87ca4 | 2571 | _cpp_temp_token (cpp_reader *pfile) |
4ed5bcfb NB |
2572 | { |
2573 | cpp_token *old, *result; | |
5950c3c9 BE |
2574 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
2575 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
4ed5bcfb NB |
2576 | |
2577 | old = pfile->cur_token - 1; | |
5950c3c9 BE |
2578 | /* Any pre-existing lookaheads must not be clobbered. */ |
2579 | if (la) | |
2580 | { | |
2581 | if (sz <= la) | |
2582 | { | |
2583 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
2584 | ||
2585 | if (sz < la) | |
2586 | memmove (next->base + 1, next->base, | |
2587 | (la - sz) * sizeof (cpp_token)); | |
2588 | ||
2589 | next->base[0] = pfile->cur_run->limit[-1]; | |
2590 | } | |
2591 | ||
2592 | if (sz > 1) | |
2593 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
2594 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
2595 | } | |
2596 | ||
2597 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
4ed5bcfb NB |
2598 | { |
2599 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
2600 | pfile->cur_token = pfile->cur_run->base; | |
2601 | } | |
2602 | ||
2603 | result = pfile->cur_token++; | |
12f9df4e | 2604 | result->src_loc = old->src_loc; |
4ed5bcfb NB |
2605 | return result; |
2606 | } | |
2607 | ||
14baae01 NB |
2608 | /* Lex a token into RESULT (external interface). Takes care of issues |
2609 | like directive handling, token lookahead, multiple include | |
a1f300c0 | 2610 | optimization and skipping. */ |
345894b4 | 2611 | const cpp_token * |
6cf87ca4 | 2612 | _cpp_lex_token (cpp_reader *pfile) |
5fddcffc | 2613 | { |
bdcbe496 | 2614 | cpp_token *result; |
5fddcffc | 2615 | |
bdcbe496 | 2616 | for (;;) |
5fddcffc | 2617 | { |
bdcbe496 | 2618 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 2619 | { |
bdcbe496 NB |
2620 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
2621 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc | 2622 | } |
ee380365 TT |
2623 | /* We assume that the current token is somewhere in the current |
2624 | run. */ | |
2625 | if (pfile->cur_token < pfile->cur_run->base | |
2626 | || pfile->cur_token >= pfile->cur_run->limit) | |
2627 | abort (); | |
5fddcffc | 2628 | |
bdcbe496 | 2629 | if (pfile->lookaheads) |
14baae01 NB |
2630 | { |
2631 | pfile->lookaheads--; | |
2632 | result = pfile->cur_token++; | |
2633 | } | |
bdcbe496 | 2634 | else |
14baae01 | 2635 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
2636 | |
2637 | if (result->flags & BOL) | |
5fddcffc | 2638 | { |
bdcbe496 NB |
2639 | /* Is this a directive. If _cpp_handle_directive returns |
2640 | false, it is an assembler #. */ | |
2641 | if (result->type == CPP_HASH | |
e808ec9c NB |
2642 | /* 6.10.3 p 11: Directives in a list of macro arguments |
2643 | gives undefined behavior. This implementation | |
2644 | handles the directive as normal. */ | |
bc4071dd | 2645 | && pfile->state.parsing_args != 1) |
21b11495 | 2646 | { |
bc4071dd | 2647 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
21b11495 | 2648 | { |
bc4071dd RH |
2649 | if (pfile->directive_result.type == CPP_PADDING) |
2650 | continue; | |
21b11495 | 2651 | result = &pfile->directive_result; |
21b11495 ZW |
2652 | } |
2653 | } | |
bc4071dd RH |
2654 | else if (pfile->state.in_deferred_pragma) |
2655 | result = &pfile->directive_result; | |
21b11495 | 2656 | |
97293897 | 2657 | if (pfile->cb.line_change && !pfile->state.skipping) |
6cf87ca4 | 2658 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
5fddcffc | 2659 | } |
5fddcffc | 2660 | |
bdcbe496 | 2661 | /* We don't skip tokens in directives. */ |
bc4071dd | 2662 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
bdcbe496 | 2663 | break; |
5fddcffc | 2664 | |
bdcbe496 | 2665 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 2666 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
6356f892 | 2667 | get here and MI optimization works. */ |
5fddcffc | 2668 | pfile->mi_valid = false; |
bdcbe496 NB |
2669 | |
2670 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
2671 | break; | |
5fddcffc NB |
2672 | } |
2673 | ||
345894b4 | 2674 | return result; |
5fddcffc NB |
2675 | } |
2676 | ||
26aea073 NB |
2677 | /* Returns true if a fresh line has been loaded. */ |
2678 | bool | |
6cf87ca4 | 2679 | _cpp_get_fresh_line (cpp_reader *pfile) |
004cb263 | 2680 | { |
26aea073 NB |
2681 | /* We can't get a new line until we leave the current directive. */ |
2682 | if (pfile->state.in_directive) | |
2683 | return false; | |
df383483 | 2684 | |
26aea073 | 2685 | for (;;) |
1a76916c | 2686 | { |
26aea073 | 2687 | cpp_buffer *buffer = pfile->buffer; |
1a76916c | 2688 | |
26aea073 NB |
2689 | if (!buffer->need_line) |
2690 | return true; | |
2691 | ||
2692 | if (buffer->next_line < buffer->rlimit) | |
004cb263 | 2693 | { |
26aea073 NB |
2694 | _cpp_clean_line (pfile); |
2695 | return true; | |
2696 | } | |
004cb263 | 2697 | |
26aea073 NB |
2698 | /* First, get out of parsing arguments state. */ |
2699 | if (pfile->state.parsing_args) | |
2700 | return false; | |
2701 | ||
2702 | /* End of buffer. Non-empty files should end in a newline. */ | |
2703 | if (buffer->buf != buffer->rlimit | |
2704 | && buffer->next_line > buffer->rlimit | |
2705 | && !buffer->from_stage3) | |
2706 | { | |
ed0e74e0 | 2707 | /* Clip to buffer size. */ |
26aea073 | 2708 | buffer->next_line = buffer->rlimit; |
26aea073 | 2709 | } |
22234f56 | 2710 | |
2a0225e4 NS |
2711 | if (buffer->prev && !buffer->return_at_eof) |
2712 | _cpp_pop_buffer (pfile); | |
2713 | else | |
2714 | { | |
2715 | /* End of translation. Do not pop the buffer yet. Increment | |
2716 | line number so that the EOF token is on a line of its own | |
2717 | (_cpp_lex_direct doesn't increment in that case, because | |
2718 | it's hard for it to distinguish this special case). */ | |
2719 | CPP_INCREMENT_LINE (pfile, 0); | |
2720 | return false; | |
2721 | } | |
26aea073 | 2722 | } |
004cb263 NB |
2723 | } |
2724 | ||
6f572ac2 NB |
2725 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
2726 | do \ | |
2727 | { \ | |
2728 | result->type = ELSE_TYPE; \ | |
2729 | if (*buffer->cur == CHAR) \ | |
2730 | buffer->cur++, result->type = THEN_TYPE; \ | |
2731 | } \ | |
2732 | while (0) | |
480709cc | 2733 | |
14baae01 NB |
2734 | /* Lex a token into pfile->cur_token, which is also incremented, to |
2735 | get diagnostics pointing to the correct location. | |
2736 | ||
2737 | Does not handle issues such as token lookahead, multiple-include | |
f1ba665b | 2738 | optimization, directives, skipping etc. This function is only |
14baae01 NB |
2739 | suitable for use by _cpp_lex_token, and in special cases like |
2740 | lex_expansion_token which doesn't care for any of these issues. | |
2741 | ||
2742 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
2743 | otherwise returns to the start of the token buffer if permissible. | |
2744 | Returns the location of the lexed token. */ | |
2745 | cpp_token * | |
6cf87ca4 | 2746 | _cpp_lex_direct (cpp_reader *pfile) |
45b966db | 2747 | { |
0d9f234d | 2748 | cppchar_t c; |
adb84b42 | 2749 | cpp_buffer *buffer; |
7bad794a JJ |
2750 | const unsigned char *comment_start; |
2751 | bool fallthrough_comment = false; | |
14baae01 | 2752 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 2753 | |
5fddcffc | 2754 | fresh_line: |
26aea073 | 2755 | result->flags = 0; |
2be570f9 | 2756 | buffer = pfile->buffer; |
a506c55c | 2757 | if (buffer->need_line) |
26aea073 | 2758 | { |
bc4071dd RH |
2759 | if (pfile->state.in_deferred_pragma) |
2760 | { | |
2761 | result->type = CPP_PRAGMA_EOL; | |
2762 | pfile->state.in_deferred_pragma = false; | |
2763 | if (!pfile->state.pragma_allow_expansion) | |
2764 | pfile->state.prevent_expansion--; | |
2765 | return result; | |
2766 | } | |
26aea073 NB |
2767 | if (!_cpp_get_fresh_line (pfile)) |
2768 | { | |
2769 | result->type = CPP_EOF; | |
9ff7868d NB |
2770 | if (!pfile->state.in_directive) |
2771 | { | |
2772 | /* Tell the compiler the line number of the EOF token. */ | |
500bee0a | 2773 | result->src_loc = pfile->line_table->highest_line; |
9ff7868d | 2774 | result->flags = BOL; |
2a0225e4 NS |
2775 | /* Now pop the buffer that _cpp_get_fresh_line did not. */ |
2776 | _cpp_pop_buffer (pfile); | |
9ff7868d | 2777 | } |
26aea073 NB |
2778 | return result; |
2779 | } | |
81fea426 | 2780 | if (buffer != pfile->buffer) |
7bad794a | 2781 | fallthrough_comment = false; |
26aea073 NB |
2782 | if (!pfile->keep_tokens) |
2783 | { | |
2784 | pfile->cur_run = &pfile->base_run; | |
2785 | result = pfile->base_run.base; | |
2786 | pfile->cur_token = result + 1; | |
2787 | } | |
2788 | result->flags = BOL; | |
2789 | if (pfile->state.parsing_args == 2) | |
2790 | result->flags |= PREV_WHITE; | |
2791 | } | |
a506c55c | 2792 | buffer = pfile->buffer; |
5fddcffc | 2793 | update_tokens_line: |
500bee0a | 2794 | result->src_loc = pfile->line_table->highest_line; |
041c3194 | 2795 | |
5fddcffc | 2796 | skipped_white: |
26aea073 NB |
2797 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
2798 | && !pfile->overlaid_buffer) | |
2799 | { | |
2800 | _cpp_process_line_notes (pfile, false); | |
500bee0a | 2801 | result->src_loc = pfile->line_table->highest_line; |
26aea073 | 2802 | } |
480709cc | 2803 | c = *buffer->cur++; |
12f9df4e | 2804 | |
f3f6029d NS |
2805 | if (pfile->forced_token_location) |
2806 | result->src_loc = pfile->forced_token_location; | |
e3dfef44 GC |
2807 | else |
2808 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
2809 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
5fddcffc | 2810 | |
0d9f234d | 2811 | switch (c) |
45b966db | 2812 | { |
4d6baafa NB |
2813 | case ' ': case '\t': case '\f': case '\v': case '\0': |
2814 | result->flags |= PREV_WHITE; | |
26aea073 NB |
2815 | skip_whitespace (pfile, c); |
2816 | goto skipped_white; | |
0d9f234d | 2817 | |
26aea073 | 2818 | case '\n': |
056f95ec NS |
2819 | /* Increment the line, unless this is the last line ... */ |
2820 | if (buffer->cur < buffer->rlimit | |
2821 | /* ... or this is a #include, (where _cpp_stack_file needs to | |
2822 | unwind by one line) ... */ | |
2823 | || (pfile->state.in_directive > 1 | |
2824 | /* ... except traditional-cpp increments this elsewhere. */ | |
2825 | && !CPP_OPTION (pfile, traditional))) | |
12f9df4e | 2826 | CPP_INCREMENT_LINE (pfile, 0); |
26aea073 NB |
2827 | buffer->need_line = true; |
2828 | goto fresh_line; | |
46d07497 | 2829 | |
0d9f234d NB |
2830 | case '0': case '1': case '2': case '3': case '4': |
2831 | case '5': case '6': case '7': case '8': case '9': | |
50668cf6 GK |
2832 | { |
2833 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2834 | result->type = CPP_NUMBER; | |
2835 | lex_number (pfile, &result->val.str, &nst); | |
2836 | warn_about_normalization (pfile, result, &nst); | |
2837 | break; | |
2838 | } | |
46d07497 | 2839 | |
0abc6a6a | 2840 | case 'L': |
b6baa67d KVH |
2841 | case 'u': |
2842 | case 'U': | |
2c6e3f55 JJ |
2843 | case 'R': |
2844 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2845 | wide strings or raw strings. */ | |
a48e3dd1 JM |
2846 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
2847 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
bced6edf | 2848 | { |
2c6e3f55 JJ |
2849 | if ((*buffer->cur == '\'' && c != 'R') |
2850 | || *buffer->cur == '"' | |
2851 | || (*buffer->cur == 'R' | |
2852 | && c != 'R' | |
2853 | && buffer->cur[1] == '"' | |
a48e3dd1 | 2854 | && CPP_OPTION (pfile, rliterals)) |
2c6e3f55 JJ |
2855 | || (*buffer->cur == '8' |
2856 | && c == 'u' | |
fe95b036 ESR |
2857 | && ((buffer->cur[1] == '"' || (buffer->cur[1] == '\'' |
2858 | && CPP_OPTION (pfile, utf8_char_literals))) | |
a48e3dd1 JM |
2859 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
2860 | && CPP_OPTION (pfile, rliterals))))) | |
b6baa67d KVH |
2861 | { |
2862 | lex_string (pfile, result, buffer->cur - 1); | |
2863 | break; | |
2864 | } | |
bced6edf | 2865 | } |
df383483 | 2866 | /* Fall through. */ |
0abc6a6a | 2867 | |
0d9f234d NB |
2868 | case '_': |
2869 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2870 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2871 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
b6baa67d | 2872 | case 's': case 't': case 'v': case 'w': case 'x': |
0d9f234d NB |
2873 | case 'y': case 'z': |
2874 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
0abc6a6a | 2875 | case 'G': case 'H': case 'I': case 'J': case 'K': |
2c6e3f55 | 2876 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
b6baa67d | 2877 | case 'S': case 'T': case 'V': case 'W': case 'X': |
0d9f234d NB |
2878 | case 'Y': case 'Z': |
2879 | result->type = CPP_NAME; | |
50668cf6 GK |
2880 | { |
2881 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
9a0c6187 | 2882 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
be5ffc59 JM |
2883 | &nst, |
2884 | &result->val.node.spelling); | |
50668cf6 GK |
2885 | warn_about_normalization (pfile, result, &nst); |
2886 | } | |
0d9f234d | 2887 | |
0d9f234d | 2888 | /* Convert named operators to their proper types. */ |
9a0c6187 | 2889 | if (result->val.node.node->flags & NODE_OPERATOR) |
0d9f234d NB |
2890 | { |
2891 | result->flags |= NAMED_OP; | |
9a0c6187 | 2892 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
0d9f234d | 2893 | } |
81fea426 MP |
2894 | |
2895 | /* Signal FALLTHROUGH comment followed by another token. */ | |
7bad794a | 2896 | if (fallthrough_comment) |
81fea426 | 2897 | result->flags |= PREV_FALLTHROUGH; |
0d9f234d NB |
2898 | break; |
2899 | ||
2900 | case '\'': | |
2901 | case '"': | |
6338b358 | 2902 | lex_string (pfile, result, buffer->cur - 1); |
0d9f234d | 2903 | break; |
041c3194 | 2904 | |
0d9f234d | 2905 | case '/': |
1c6d33ef NB |
2906 | /* A potential block or line comment. */ |
2907 | comment_start = buffer->cur; | |
6f572ac2 NB |
2908 | c = *buffer->cur; |
2909 | ||
1c6d33ef NB |
2910 | if (c == '*') |
2911 | { | |
26aea073 | 2912 | if (_cpp_skip_block_comment (pfile)) |
0527bc4e | 2913 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
0d9f234d | 2914 | } |
909eb89c | 2915 | else if (c == '/' && ! CPP_OPTION (pfile, traditional)) |
0d9f234d | 2916 | { |
909eb89c MP |
2917 | /* Don't warn for system headers. */ |
2918 | if (cpp_in_system_header (pfile)) | |
2919 | ; | |
f3bede71 | 2920 | /* Warn about comments if pedantically GNUC89, and not |
bdb05a7b | 2921 | in system headers. */ |
909eb89c MP |
2922 | else if (CPP_OPTION (pfile, lang) == CLK_GNUC89 |
2923 | && CPP_PEDANTIC (pfile) | |
2924 | && ! buffer->warned_cplusplus_comments) | |
041c3194 | 2925 | { |
0c86a39d JJ |
2926 | if (cpp_error (pfile, CPP_DL_PEDWARN, |
2927 | "C++ style comments are not allowed in ISO C90")) | |
2928 | cpp_error (pfile, CPP_DL_NOTE, | |
2929 | "(this will be reported only once per input file)"); | |
1c6d33ef NB |
2930 | buffer->warned_cplusplus_comments = 1; |
2931 | } | |
f3bede71 | 2932 | /* Or if specifically desired via -Wc90-c99-compat. */ |
177cce46 | 2933 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
dd3ff077 | 2934 | && ! CPP_OPTION (pfile, cplusplus) |
f3bede71 MP |
2935 | && ! buffer->warned_cplusplus_comments) |
2936 | { | |
0c86a39d JJ |
2937 | if (cpp_error (pfile, CPP_DL_WARNING, |
2938 | "C++ style comments are incompatible with C90")) | |
2939 | cpp_error (pfile, CPP_DL_NOTE, | |
2940 | "(this will be reported only once per input file)"); | |
f3bede71 MP |
2941 | buffer->warned_cplusplus_comments = 1; |
2942 | } | |
909eb89c MP |
2943 | /* In C89/C94, C++ style comments are forbidden. */ |
2944 | else if ((CPP_OPTION (pfile, lang) == CLK_STDC89 | |
2945 | || CPP_OPTION (pfile, lang) == CLK_STDC94)) | |
2946 | { | |
2947 | /* But don't be confused about valid code such as | |
2948 | - // immediately followed by *, | |
2949 | - // in a preprocessing directive, | |
2950 | - // in an #if 0 block. */ | |
2951 | if (buffer->cur[1] == '*' | |
2952 | || pfile->state.in_directive | |
2953 | || pfile->state.skipping) | |
2954 | { | |
2955 | result->type = CPP_DIV; | |
2956 | break; | |
2957 | } | |
2958 | else if (! buffer->warned_cplusplus_comments) | |
2959 | { | |
0c86a39d JJ |
2960 | if (cpp_error (pfile, CPP_DL_ERROR, |
2961 | "C++ style comments are not allowed in " | |
2962 | "ISO C90")) | |
2963 | cpp_error (pfile, CPP_DL_NOTE, | |
2964 | "(this will be reported only once per input " | |
2965 | "file)"); | |
909eb89c MP |
2966 | buffer->warned_cplusplus_comments = 1; |
2967 | } | |
2968 | } | |
01ef6563 | 2969 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
87cf0651 | 2970 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
1c6d33ef | 2971 | } |
480709cc NB |
2972 | else if (c == '=') |
2973 | { | |
6f572ac2 | 2974 | buffer->cur++; |
480709cc NB |
2975 | result->type = CPP_DIV_EQ; |
2976 | break; | |
2977 | } | |
2978 | else | |
2979 | { | |
480709cc NB |
2980 | result->type = CPP_DIV; |
2981 | break; | |
2982 | } | |
0d9f234d | 2983 | |
7bad794a JJ |
2984 | if (fallthrough_comment_p (pfile, comment_start)) |
2985 | fallthrough_comment = true; | |
2986 | ||
05945a1b DM |
2987 | if (pfile->cb.comment) |
2988 | { | |
2989 | size_t len = pfile->buffer->cur - comment_start; | |
2990 | pfile->cb.comment (pfile, result->src_loc, comment_start - 1, | |
2991 | len + 1); | |
2992 | } | |
2993 | ||
1c6d33ef NB |
2994 | if (!pfile->state.save_comments) |
2995 | { | |
2996 | result->flags |= PREV_WHITE; | |
5fddcffc | 2997 | goto update_tokens_line; |
0d9f234d | 2998 | } |
1c6d33ef | 2999 | |
7bad794a | 3000 | if (fallthrough_comment) |
81fea426 MP |
3001 | result->flags |= PREV_FALLTHROUGH; |
3002 | ||
1c6d33ef | 3003 | /* Save the comment as a token in its own right. */ |
477cdac7 | 3004 | save_comment (pfile, result, comment_start, c); |
bdcbe496 | 3005 | break; |
0d9f234d NB |
3006 | |
3007 | case '<': | |
3008 | if (pfile->state.angled_headers) | |
3009 | { | |
6338b358 | 3010 | lex_string (pfile, result, buffer->cur - 1); |
4bb09c26 JM |
3011 | if (result->type != CPP_LESS) |
3012 | break; | |
0d9f234d | 3013 | } |
45b966db | 3014 | |
6f572ac2 NB |
3015 | result->type = CPP_LESS; |
3016 | if (*buffer->cur == '=') | |
b7689b96 JM |
3017 | { |
3018 | buffer->cur++, result->type = CPP_LESS_EQ; | |
3019 | if (*buffer->cur == '>' | |
3020 | && CPP_OPTION (pfile, cplusplus) | |
b04445d4 | 3021 | && CPP_OPTION (pfile, lang) >= CLK_GNUCXX20) |
b7689b96 JM |
3022 | buffer->cur++, result->type = CPP_SPACESHIP; |
3023 | } | |
6f572ac2 | 3024 | else if (*buffer->cur == '<') |
0d9f234d | 3025 | { |
6f572ac2 NB |
3026 | buffer->cur++; |
3027 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
0d9f234d | 3028 | } |
6f572ac2 | 3029 | else if (CPP_OPTION (pfile, digraphs)) |
480709cc | 3030 | { |
6f572ac2 NB |
3031 | if (*buffer->cur == ':') |
3032 | { | |
1582c677 PC |
3033 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
3034 | three characters are <:: and the subsequent character | |
3035 | is neither : nor >, the < is treated as a preprocessor | |
3036 | token by itself". */ | |
3037 | if (CPP_OPTION (pfile, cplusplus) | |
61949153 PC |
3038 | && CPP_OPTION (pfile, lang) != CLK_CXX98 |
3039 | && CPP_OPTION (pfile, lang) != CLK_GNUCXX | |
1582c677 PC |
3040 | && buffer->cur[1] == ':' |
3041 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
3042 | break; | |
3043 | ||
6f572ac2 NB |
3044 | buffer->cur++; |
3045 | result->flags |= DIGRAPH; | |
3046 | result->type = CPP_OPEN_SQUARE; | |
3047 | } | |
3048 | else if (*buffer->cur == '%') | |
3049 | { | |
3050 | buffer->cur++; | |
3051 | result->flags |= DIGRAPH; | |
3052 | result->type = CPP_OPEN_BRACE; | |
3053 | } | |
480709cc | 3054 | } |
0d9f234d NB |
3055 | break; |
3056 | ||
3057 | case '>': | |
6f572ac2 NB |
3058 | result->type = CPP_GREATER; |
3059 | if (*buffer->cur == '=') | |
3060 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
3061 | else if (*buffer->cur == '>') | |
0d9f234d | 3062 | { |
6f572ac2 NB |
3063 | buffer->cur++; |
3064 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
3065 | } | |
0d9f234d NB |
3066 | break; |
3067 | ||
cbcff6df | 3068 | case '%': |
6f572ac2 NB |
3069 | result->type = CPP_MOD; |
3070 | if (*buffer->cur == '=') | |
3071 | buffer->cur++, result->type = CPP_MOD_EQ; | |
3072 | else if (CPP_OPTION (pfile, digraphs)) | |
480709cc | 3073 | { |
6f572ac2 | 3074 | if (*buffer->cur == ':') |
480709cc | 3075 | { |
6f572ac2 NB |
3076 | buffer->cur++; |
3077 | result->flags |= DIGRAPH; | |
3078 | result->type = CPP_HASH; | |
3079 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
9a0c6187 | 3080 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
6f572ac2 NB |
3081 | } |
3082 | else if (*buffer->cur == '>') | |
3083 | { | |
3084 | buffer->cur++; | |
3085 | result->flags |= DIGRAPH; | |
3086 | result->type = CPP_CLOSE_BRACE; | |
480709cc | 3087 | } |
480709cc | 3088 | } |
0d9f234d NB |
3089 | break; |
3090 | ||
cbcff6df | 3091 | case '.': |
480709cc | 3092 | result->type = CPP_DOT; |
6f572ac2 | 3093 | if (ISDIGIT (*buffer->cur)) |
480709cc | 3094 | { |
50668cf6 | 3095 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
480709cc | 3096 | result->type = CPP_NUMBER; |
50668cf6 GK |
3097 | lex_number (pfile, &result->val.str, &nst); |
3098 | warn_about_normalization (pfile, result, &nst); | |
480709cc | 3099 | } |
6f572ac2 NB |
3100 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
3101 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
3102 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
3103 | buffer->cur++, result->type = CPP_DOT_STAR; | |
0d9f234d | 3104 | break; |
45b966db | 3105 | |
0d9f234d | 3106 | case '+': |
6f572ac2 NB |
3107 | result->type = CPP_PLUS; |
3108 | if (*buffer->cur == '+') | |
3109 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
3110 | else if (*buffer->cur == '=') | |
3111 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
0d9f234d | 3112 | break; |
04e3ec78 | 3113 | |
0d9f234d | 3114 | case '-': |
6f572ac2 NB |
3115 | result->type = CPP_MINUS; |
3116 | if (*buffer->cur == '>') | |
0d9f234d | 3117 | { |
6f572ac2 | 3118 | buffer->cur++; |
480709cc | 3119 | result->type = CPP_DEREF; |
6f572ac2 NB |
3120 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
3121 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
480709cc | 3122 | } |
6f572ac2 NB |
3123 | else if (*buffer->cur == '-') |
3124 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
3125 | else if (*buffer->cur == '=') | |
3126 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
0d9f234d | 3127 | break; |
45b966db | 3128 | |
0d9f234d | 3129 | case '&': |
6f572ac2 NB |
3130 | result->type = CPP_AND; |
3131 | if (*buffer->cur == '&') | |
3132 | buffer->cur++, result->type = CPP_AND_AND; | |
3133 | else if (*buffer->cur == '=') | |
3134 | buffer->cur++, result->type = CPP_AND_EQ; | |
0d9f234d | 3135 | break; |
df383483 | 3136 | |
0d9f234d | 3137 | case '|': |
6f572ac2 NB |
3138 | result->type = CPP_OR; |
3139 | if (*buffer->cur == '|') | |
3140 | buffer->cur++, result->type = CPP_OR_OR; | |
3141 | else if (*buffer->cur == '=') | |
3142 | buffer->cur++, result->type = CPP_OR_EQ; | |
0d9f234d | 3143 | break; |
45b966db | 3144 | |
0d9f234d | 3145 | case ':': |
6f572ac2 | 3146 | result->type = CPP_COLON; |
93313b94 | 3147 | if (*buffer->cur == ':' && CPP_OPTION (pfile, scope)) |
6f572ac2 NB |
3148 | buffer->cur++, result->type = CPP_SCOPE; |
3149 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
0d9f234d | 3150 | { |
6f572ac2 | 3151 | buffer->cur++; |
0d9f234d | 3152 | result->flags |= DIGRAPH; |
480709cc NB |
3153 | result->type = CPP_CLOSE_SQUARE; |
3154 | } | |
0d9f234d | 3155 | break; |
45b966db | 3156 | |
480709cc NB |
3157 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
3158 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
3159 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
3160 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
9a0c6187 | 3161 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
480709cc | 3162 | |
26aea073 | 3163 | case '?': result->type = CPP_QUERY; break; |
0d9f234d NB |
3164 | case '~': result->type = CPP_COMPL; break; |
3165 | case ',': result->type = CPP_COMMA; break; | |
3166 | case '(': result->type = CPP_OPEN_PAREN; break; | |
3167 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
3168 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
3169 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
3170 | case '{': result->type = CPP_OPEN_BRACE; break; | |
3171 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
3172 | case ';': result->type = CPP_SEMICOLON; break; | |
3173 | ||
40f03658 | 3174 | /* @ is a punctuator in Objective-C. */ |
cc937581 | 3175 | case '@': result->type = CPP_ATSIGN; break; |
0d9f234d | 3176 | |
7d112d66 | 3177 | default: |
1613e52b NB |
3178 | { |
3179 | const uchar *base = --buffer->cur; | |
0abc6a6a | 3180 | |
7d112d66 LH |
3181 | /* Check for an extended identifier ($ or UCN or UTF-8). */ |
3182 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
50668cf6 | 3183 | if (forms_identifier_p (pfile, true, &nst)) |
1613e52b NB |
3184 | { |
3185 | result->type = CPP_NAME; | |
be5ffc59 JM |
3186 | result->val.node.node = lex_identifier (pfile, base, true, &nst, |
3187 | &result->val.node.spelling); | |
50668cf6 | 3188 | warn_about_normalization (pfile, result, &nst); |
1613e52b NB |
3189 | break; |
3190 | } | |
7d112d66 LH |
3191 | |
3192 | /* Otherwise this will form a CPP_OTHER token. Parse valid UTF-8 as a | |
3193 | single token. */ | |
1613e52b | 3194 | buffer->cur++; |
7d112d66 LH |
3195 | if (c >= utf8_signifier) |
3196 | { | |
3197 | const uchar *pstr = base; | |
3198 | cppchar_t s; | |
3199 | if (_cpp_valid_utf8 (pfile, &pstr, buffer->rlimit, 0, NULL, &s)) | |
3200 | buffer->cur = pstr; | |
3201 | } | |
3202 | create_literal (pfile, result, base, buffer->cur - base, CPP_OTHER); | |
3203 | break; | |
1067694a | 3204 | } |
1613e52b | 3205 | |
0d9f234d | 3206 | } |
bdcbe496 | 3207 | |
a3998c2f DM |
3208 | /* Potentially convert the location of the token to a range. */ |
3209 | if (result->src_loc >= RESERVED_LOCATION_COUNT | |
3210 | && result->type != CPP_EOF) | |
3211 | { | |
3212 | /* Ensure that any line notes are processed, so that we have the | |
3213 | correct physical line/column for the end-point of the token even | |
3214 | when a logical line is split via one or more backslashes. */ | |
3215 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos | |
3216 | && !pfile->overlaid_buffer) | |
3217 | _cpp_process_line_notes (pfile, false); | |
3218 | ||
3219 | source_range tok_range; | |
3220 | tok_range.m_start = result->src_loc; | |
3221 | tok_range.m_finish | |
3222 | = linemap_position_for_column (pfile->line_table, | |
3223 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
3224 | ||
3225 | result->src_loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
3226 | result->src_loc, | |
3227 | tok_range, NULL); | |
3228 | } | |
ebedc9a3 | 3229 | |
bdcbe496 | 3230 | return result; |
0d9f234d NB |
3231 | } |
3232 | ||
59325650 NB |
3233 | /* An upper bound on the number of bytes needed to spell TOKEN. |
3234 | Does not include preceding whitespace. */ | |
93c80368 | 3235 | unsigned int |
6cf87ca4 | 3236 | cpp_token_len (const cpp_token *token) |
0d9f234d | 3237 | { |
93c80368 | 3238 | unsigned int len; |
6d2c2047 | 3239 | |
93c80368 | 3240 | switch (TOKEN_SPELL (token)) |
041c3194 | 3241 | { |
cc955282 | 3242 | default: len = 6; break; |
6338b358 | 3243 | case SPELL_LITERAL: len = token->val.str.len; break; |
9a0c6187 | 3244 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
041c3194 | 3245 | } |
59325650 NB |
3246 | |
3247 | return len; | |
6d2c2047 ZW |
3248 | } |
3249 | ||
47e20491 GK |
3250 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
3251 | Return the number of bytes read out of NAME. (There are always | |
3252 | 10 bytes written to BUFFER.) */ | |
3253 | ||
3254 | static size_t | |
3255 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
3256 | { | |
3257 | int j; | |
3258 | int ucn_len = 0; | |
3259 | int ucn_len_c; | |
3260 | unsigned t; | |
3261 | unsigned long utf32; | |
3262 | ||
3263 | /* Compute the length of the UTF-8 sequence. */ | |
3264 | for (t = *name; t & 0x80; t <<= 1) | |
3265 | ucn_len++; | |
3266 | ||
3267 | utf32 = *name & (0x7F >> ucn_len); | |
3268 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
3269 | { | |
3270 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
3271 | ||
3272 | /* Ill-formed UTF-8. */ | |
3273 | if ((*name & ~0x3F) != 0x80) | |
3274 | abort (); | |
3275 | } | |
3276 | ||
3277 | *buffer++ = '\\'; | |
3278 | *buffer++ = 'U'; | |
3279 | for (j = 7; j >= 0; j--) | |
3280 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
3281 | return ucn_len; | |
3282 | } | |
3283 | ||
cfc93532 MLI |
3284 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
3285 | the spelling of the digraph. */ | |
3286 | static const unsigned char * | |
3287 | cpp_digraph2name (enum cpp_ttype type) | |
3288 | { | |
3289 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
3290 | } | |
47e20491 | 3291 | |
be5ffc59 JM |
3292 | /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER. |
3293 | The buffer must already contain the enough space to hold the | |
3294 | token's spelling. Returns a pointer to the character after the | |
3295 | last character written. */ | |
3296 | unsigned char * | |
3297 | _cpp_spell_ident_ucns (unsigned char *buffer, cpp_hashnode *ident) | |
3298 | { | |
3299 | size_t i; | |
3300 | const unsigned char *name = NODE_NAME (ident); | |
3301 | ||
3302 | for (i = 0; i < NODE_LEN (ident); i++) | |
3303 | if (name[i] & ~0x7F) | |
3304 | { | |
3305 | i += utf8_to_ucn (buffer, name + i) - 1; | |
3306 | buffer += 10; | |
3307 | } | |
3308 | else | |
3309 | *buffer++ = name[i]; | |
3310 | ||
3311 | return buffer; | |
3312 | } | |
3313 | ||
041c3194 | 3314 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 | 3315 | already contain the enough space to hold the token's spelling. |
6cf87ca4 | 3316 | Returns a pointer to the character after the last character written. |
47e20491 | 3317 | FORSTRING is true if this is to be the spelling after translation |
be5ffc59 JM |
3318 | phase 1 (with the original spelling of extended identifiers), false |
3319 | if extended identifiers should always be written using UCNs (there is | |
3320 | no option for always writing them in the internal UTF-8 form). | |
6cf87ca4 | 3321 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
93c80368 | 3322 | unsigned char * |
6cf87ca4 | 3323 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
47e20491 | 3324 | unsigned char *buffer, bool forstring) |
041c3194 | 3325 | { |
96be6998 | 3326 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
3327 | { |
3328 | case SPELL_OPERATOR: | |
3329 | { | |
3330 | const unsigned char *spelling; | |
3331 | unsigned char c; | |
d6d5f795 | 3332 | |
041c3194 | 3333 | if (token->flags & DIGRAPH) |
cfc93532 | 3334 | spelling = cpp_digraph2name (token->type); |
92936ecf ZW |
3335 | else if (token->flags & NAMED_OP) |
3336 | goto spell_ident; | |
041c3194 | 3337 | else |
96be6998 | 3338 | spelling = TOKEN_NAME (token); |
df383483 | 3339 | |
041c3194 ZW |
3340 | while ((c = *spelling++) != '\0') |
3341 | *buffer++ = c; | |
3342 | } | |
3343 | break; | |
d6d5f795 | 3344 | |
47ad4138 | 3345 | spell_ident: |
041c3194 | 3346 | case SPELL_IDENT: |
47e20491 GK |
3347 | if (forstring) |
3348 | { | |
be5ffc59 JM |
3349 | memcpy (buffer, NODE_NAME (token->val.node.spelling), |
3350 | NODE_LEN (token->val.node.spelling)); | |
3351 | buffer += NODE_LEN (token->val.node.spelling); | |
47e20491 GK |
3352 | } |
3353 | else | |
be5ffc59 | 3354 | buffer = _cpp_spell_ident_ucns (buffer, token->val.node.node); |
041c3194 | 3355 | break; |
d6d5f795 | 3356 | |
6338b358 | 3357 | case SPELL_LITERAL: |
47ad4138 ZW |
3358 | memcpy (buffer, token->val.str.text, token->val.str.len); |
3359 | buffer += token->val.str.len; | |
3360 | break; | |
3361 | ||
041c3194 | 3362 | case SPELL_NONE: |
0527bc4e JDA |
3363 | cpp_error (pfile, CPP_DL_ICE, |
3364 | "unspellable token %s", TOKEN_NAME (token)); | |
041c3194 ZW |
3365 | break; |
3366 | } | |
d6d5f795 | 3367 | |
041c3194 ZW |
3368 | return buffer; |
3369 | } | |
d6d5f795 | 3370 | |
5d8ebbd8 NB |
3371 | /* Returns TOKEN spelt as a null-terminated string. The string is |
3372 | freed when the reader is destroyed. Useful for diagnostics. */ | |
93c80368 | 3373 | unsigned char * |
6cf87ca4 | 3374 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
59325650 NB |
3375 | { |
3376 | unsigned int len = cpp_token_len (token) + 1; | |
ece54d54 | 3377 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 3378 | |
47e20491 | 3379 | end = cpp_spell_token (pfile, token, start, false); |
93c80368 | 3380 | end[0] = '\0'; |
c5a04734 | 3381 | |
93c80368 NB |
3382 | return start; |
3383 | } | |
c5a04734 | 3384 | |
cfc93532 MLI |
3385 | /* Returns a pointer to a string which spells the token defined by |
3386 | TYPE and FLAGS. Used by C front ends, which really should move to | |
3387 | using cpp_token_as_text. */ | |
93c80368 | 3388 | const char * |
cfc93532 | 3389 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
93c80368 | 3390 | { |
cfc93532 MLI |
3391 | if (flags & DIGRAPH) |
3392 | return (const char *) cpp_digraph2name (type); | |
3393 | else if (flags & NAMED_OP) | |
3394 | return cpp_named_operator2name (type); | |
3395 | ||
93c80368 NB |
3396 | return (const char *) token_spellings[type].name; |
3397 | } | |
c5a04734 | 3398 | |
4ed5bcfb NB |
3399 | /* Writes the spelling of token to FP, without any preceding space. |
3400 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
3401 | double-buffering. */ | |
93c80368 | 3402 | void |
6cf87ca4 | 3403 | cpp_output_token (const cpp_token *token, FILE *fp) |
93c80368 | 3404 | { |
93c80368 | 3405 | switch (TOKEN_SPELL (token)) |
c5a04734 | 3406 | { |
93c80368 NB |
3407 | case SPELL_OPERATOR: |
3408 | { | |
3409 | const unsigned char *spelling; | |
3b681e9d | 3410 | int c; |
c5a04734 | 3411 | |
93c80368 | 3412 | if (token->flags & DIGRAPH) |
cfc93532 | 3413 | spelling = cpp_digraph2name (token->type); |
93c80368 NB |
3414 | else if (token->flags & NAMED_OP) |
3415 | goto spell_ident; | |
3416 | else | |
3417 | spelling = TOKEN_NAME (token); | |
041c3194 | 3418 | |
3b681e9d ZW |
3419 | c = *spelling; |
3420 | do | |
3421 | putc (c, fp); | |
3422 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
3423 | } |
3424 | break; | |
041c3194 | 3425 | |
93c80368 NB |
3426 | spell_ident: |
3427 | case SPELL_IDENT: | |
47e20491 GK |
3428 | { |
3429 | size_t i; | |
9a0c6187 | 3430 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 3431 | |
9a0c6187 | 3432 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
3433 | if (name[i] & ~0x7F) |
3434 | { | |
3435 | unsigned char buffer[10]; | |
3436 | i += utf8_to_ucn (buffer, name + i) - 1; | |
3437 | fwrite (buffer, 1, 10, fp); | |
3438 | } | |
3439 | else | |
9a0c6187 | 3440 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
47e20491 GK |
3441 | } |
3442 | break; | |
041c3194 | 3443 | |
6338b358 | 3444 | case SPELL_LITERAL: |
47ad4138 ZW |
3445 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
3446 | break; | |
3447 | ||
93c80368 NB |
3448 | case SPELL_NONE: |
3449 | /* An error, most probably. */ | |
3450 | break; | |
041c3194 | 3451 | } |
c5a04734 ZW |
3452 | } |
3453 | ||
93c80368 NB |
3454 | /* Compare two tokens. */ |
3455 | int | |
6cf87ca4 | 3456 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
c5a04734 | 3457 | { |
93c80368 NB |
3458 | if (a->type == b->type && a->flags == b->flags) |
3459 | switch (TOKEN_SPELL (a)) | |
3460 | { | |
3461 | default: /* Keep compiler happy. */ | |
3462 | case SPELL_OPERATOR: | |
9a0c6187 | 3463 | /* token_no is used to track where multiple consecutive ## |
aa508502 | 3464 | tokens were originally located. */ |
9a0c6187 | 3465 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
93c80368 | 3466 | case SPELL_NONE: |
9a0c6187 | 3467 | return (a->type != CPP_MACRO_ARG |
be5ffc59 JM |
3468 | || (a->val.macro_arg.arg_no == b->val.macro_arg.arg_no |
3469 | && a->val.macro_arg.spelling == b->val.macro_arg.spelling)); | |
93c80368 | 3470 | case SPELL_IDENT: |
be5ffc59 JM |
3471 | return (a->val.node.node == b->val.node.node |
3472 | && a->val.node.spelling == b->val.node.spelling); | |
6338b358 | 3473 | case SPELL_LITERAL: |
93c80368 NB |
3474 | return (a->val.str.len == b->val.str.len |
3475 | && !memcmp (a->val.str.text, b->val.str.text, | |
3476 | a->val.str.len)); | |
3477 | } | |
c5a04734 | 3478 | |
041c3194 ZW |
3479 | return 0; |
3480 | } | |
3481 | ||
93c80368 NB |
3482 | /* Returns nonzero if a space should be inserted to avoid an |
3483 | accidental token paste for output. For simplicity, it is | |
3484 | conservative, and occasionally advises a space where one is not | |
3485 | needed, e.g. "." and ".2". */ | |
93c80368 | 3486 | int |
6cf87ca4 ZW |
3487 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
3488 | const cpp_token *token2) | |
c5a04734 | 3489 | { |
93c80368 NB |
3490 | enum cpp_ttype a = token1->type, b = token2->type; |
3491 | cppchar_t c; | |
c5a04734 | 3492 | |
93c80368 NB |
3493 | if (token1->flags & NAMED_OP) |
3494 | a = CPP_NAME; | |
3495 | if (token2->flags & NAMED_OP) | |
3496 | b = CPP_NAME; | |
c5a04734 | 3497 | |
93c80368 NB |
3498 | c = EOF; |
3499 | if (token2->flags & DIGRAPH) | |
37b8524c | 3500 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
3501 | else if (token_spellings[b].category == SPELL_OPERATOR) |
3502 | c = token_spellings[b].name[0]; | |
c5a04734 | 3503 | |
93c80368 | 3504 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 3505 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 3506 | return 1; |
c5a04734 | 3507 | |
93c80368 | 3508 | switch (a) |
c5a04734 | 3509 | { |
b52dbbf8 SE |
3510 | case CPP_GREATER: return c == '>'; |
3511 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
93c80368 NB |
3512 | case CPP_PLUS: return c == '+'; |
3513 | case CPP_MINUS: return c == '-' || c == '>'; | |
3514 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
3515 | case CPP_MOD: return c == ':' || c == '>'; | |
3516 | case CPP_AND: return c == '&'; | |
3517 | case CPP_OR: return c == '|'; | |
3518 | case CPP_COLON: return c == ':' || c == '>'; | |
3519 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 3520 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
3521 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
3522 | case CPP_NAME: return ((b == CPP_NUMBER | |
3523 | && name_p (pfile, &token2->val.str)) | |
3524 | || b == CPP_NAME | |
3525 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
3526 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
3527 | || c == '.' || c == '+' || c == '-'); | |
1613e52b | 3528 | /* UCNs */ |
1067694a NB |
3529 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
3530 | && b == CPP_NAME) | |
1613e52b | 3531 | || (CPP_OPTION (pfile, objc) |
1067694a | 3532 | && token1->val.str.text[0] == '@' |
1613e52b | 3533 | && (b == CPP_NAME || b == CPP_STRING))); |
b7689b96 | 3534 | case CPP_LESS_EQ: return c == '>'; |
87e356ba JJ |
3535 | case CPP_STRING: |
3536 | case CPP_WSTRING: | |
3537 | case CPP_UTF8STRING: | |
3538 | case CPP_STRING16: | |
3539 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals) | |
3540 | && (b == CPP_NAME | |
3541 | || (TOKEN_SPELL (token2) == SPELL_LITERAL | |
3542 | && ISIDST (token2->val.str.text[0])))); | |
3543 | ||
93c80368 | 3544 | default: break; |
c5a04734 | 3545 | } |
c5a04734 | 3546 | |
417f3e3a | 3547 | return 0; |
c5a04734 ZW |
3548 | } |
3549 | ||
93c80368 | 3550 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
3551 | character, to FP. Leading whitespace is removed. If there are |
3552 | macros, special token padding is not performed. */ | |
c5a04734 | 3553 | void |
6cf87ca4 | 3554 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
c5a04734 | 3555 | { |
4ed5bcfb | 3556 | const cpp_token *token; |
96be6998 | 3557 | |
4ed5bcfb NB |
3558 | token = cpp_get_token (pfile); |
3559 | while (token->type != CPP_EOF) | |
96be6998 | 3560 | { |
4ed5bcfb NB |
3561 | cpp_output_token (token, fp); |
3562 | token = cpp_get_token (pfile); | |
3563 | if (token->flags & PREV_WHITE) | |
3564 | putc (' ', fp); | |
96be6998 ZW |
3565 | } |
3566 | ||
93c80368 | 3567 | putc ('\n', fp); |
041c3194 | 3568 | } |
c5a04734 | 3569 | |
5d6342eb TT |
3570 | /* Return a string representation of all the remaining tokens on the |
3571 | current line. The result is allocated using xmalloc and must be | |
3572 | freed by the caller. */ | |
3573 | unsigned char * | |
3574 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
3575 | { | |
3576 | const cpp_token *token; | |
3577 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
3578 | unsigned int alloced = 120 + out; | |
3579 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
3580 | ||
3581 | /* If DIR_NAME is empty, there are no initial contents. */ | |
3582 | if (dir_name) | |
3583 | { | |
3584 | sprintf ((char *) result, "#%s ", dir_name); | |
3585 | out += 2; | |
3586 | } | |
3587 | ||
3588 | token = cpp_get_token (pfile); | |
3589 | while (token->type != CPP_EOF) | |
3590 | { | |
3591 | unsigned char *last; | |
3592 | /* Include room for a possible space and the terminating nul. */ | |
3593 | unsigned int len = cpp_token_len (token) + 2; | |
3594 | ||
3595 | if (out + len > alloced) | |
3596 | { | |
3597 | alloced *= 2; | |
3598 | if (out + len > alloced) | |
3599 | alloced = out + len; | |
3600 | result = (unsigned char *) xrealloc (result, alloced); | |
3601 | } | |
3602 | ||
3603 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
3604 | out = last - result; | |
3605 | ||
3606 | token = cpp_get_token (pfile); | |
3607 | if (token->flags & PREV_WHITE) | |
3608 | result[out++] = ' '; | |
3609 | } | |
3610 | ||
3611 | result[out] = '\0'; | |
3612 | return result; | |
3613 | } | |
3614 | ||
1e013d2e NB |
3615 | /* Memory buffers. Changing these three constants can have a dramatic |
3616 | effect on performance. The values here are reasonable defaults, | |
3617 | but might be tuned. If you adjust them, be sure to test across a | |
3618 | range of uses of cpplib, including heavy nested function-like macro | |
3619 | expansion. Also check the change in peak memory usage (NJAMD is a | |
3620 | good tool for this). */ | |
3621 | #define MIN_BUFF_SIZE 8000 | |
87062813 | 3622 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
1e013d2e NB |
3623 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
3624 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 3625 | |
87062813 NB |
3626 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
3627 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
3628 | #endif | |
3629 | ||
c9e7a609 NB |
3630 | /* Create a new allocation buffer. Place the control block at the end |
3631 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 | 3632 | static _cpp_buff * |
6cf87ca4 | 3633 | new_buff (size_t len) |
b8af0ca5 NB |
3634 | { |
3635 | _cpp_buff *result; | |
ece54d54 | 3636 | unsigned char *base; |
b8af0ca5 | 3637 | |
1e013d2e NB |
3638 | if (len < MIN_BUFF_SIZE) |
3639 | len = MIN_BUFF_SIZE; | |
c70f6ed3 | 3640 | len = CPP_ALIGN (len); |
b8af0ca5 | 3641 | |
ceb17928 | 3642 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
3643 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff |
3644 | struct first. */ | |
3645 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
3646 | base = XNEWVEC (unsigned char, len + slen); | |
3647 | result = (_cpp_buff *) base; | |
3648 | base += slen; | |
3649 | #else | |
c3f829c1 | 3650 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
b8af0ca5 | 3651 | result = (_cpp_buff *) (base + len); |
1a80db97 | 3652 | #endif |
b8af0ca5 NB |
3653 | result->base = base; |
3654 | result->cur = base; | |
3655 | result->limit = base + len; | |
3656 | result->next = NULL; | |
3657 | return result; | |
3658 | } | |
3659 | ||
3660 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
3661 | void | |
6cf87ca4 | 3662 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
b8af0ca5 NB |
3663 | { |
3664 | _cpp_buff *end = buff; | |
3665 | ||
3666 | while (end->next) | |
3667 | end = end->next; | |
3668 | end->next = pfile->free_buffs; | |
3669 | pfile->free_buffs = buff; | |
3670 | } | |
3671 | ||
3672 | /* Return a free buffer of size at least MIN_SIZE. */ | |
3673 | _cpp_buff * | |
6cf87ca4 | 3674 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
b8af0ca5 NB |
3675 | { |
3676 | _cpp_buff *result, **p; | |
3677 | ||
3678 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
3679 | { | |
6142088c | 3680 | size_t size; |
1e013d2e NB |
3681 | |
3682 | if (*p == NULL) | |
b8af0ca5 | 3683 | return new_buff (min_size); |
1e013d2e NB |
3684 | result = *p; |
3685 | size = result->limit - result->base; | |
3686 | /* Return a buffer that's big enough, but don't waste one that's | |
3687 | way too big. */ | |
34f5271d | 3688 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
b8af0ca5 NB |
3689 | break; |
3690 | } | |
3691 | ||
3692 | *p = result->next; | |
3693 | result->next = NULL; | |
3694 | result->cur = result->base; | |
3695 | return result; | |
3696 | } | |
3697 | ||
4fe9b91c | 3698 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3699 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
3700 | the excess bytes to the new buffer. Chains the new buffer after | |
3701 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 3702 | _cpp_buff * |
6cf87ca4 | 3703 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
b8af0ca5 | 3704 | { |
6142088c | 3705 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 3706 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 3707 | |
8c3b2693 NB |
3708 | buff->next = new_buff; |
3709 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
3710 | return new_buff; | |
3711 | } | |
3712 | ||
4fe9b91c | 3713 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3714 | remaining bytes of the buffer pointed to by BUFF, and at least |
3715 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
3716 | Chains the new buffer before the buffer pointed to by BUFF, and | |
3717 | updates the pointer to point to the new buffer. */ | |
3718 | void | |
6cf87ca4 | 3719 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
8c3b2693 NB |
3720 | { |
3721 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
3722 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
3723 | ||
3724 | new_buff = _cpp_get_buff (pfile, size); | |
3725 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
3726 | new_buff->next = old_buff; | |
3727 | *pbuff = new_buff; | |
b8af0ca5 NB |
3728 | } |
3729 | ||
3730 | /* Free a chain of buffers starting at BUFF. */ | |
3731 | void | |
5671bf27 | 3732 | _cpp_free_buff (_cpp_buff *buff) |
b8af0ca5 NB |
3733 | { |
3734 | _cpp_buff *next; | |
3735 | ||
3736 | for (; buff; buff = next) | |
3737 | { | |
3738 | next = buff->next; | |
ceb17928 | 3739 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
3740 | free (buff); |
3741 | #else | |
b8af0ca5 | 3742 | free (buff->base); |
1a80db97 | 3743 | #endif |
b8af0ca5 NB |
3744 | } |
3745 | } | |
417f3e3a | 3746 | |
ece54d54 NB |
3747 | /* Allocate permanent, unaligned storage of length LEN. */ |
3748 | unsigned char * | |
6cf87ca4 | 3749 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
ece54d54 NB |
3750 | { |
3751 | _cpp_buff *buff = pfile->u_buff; | |
3752 | unsigned char *result = buff->cur; | |
3753 | ||
3754 | if (len > (size_t) (buff->limit - result)) | |
3755 | { | |
3756 | buff = _cpp_get_buff (pfile, len); | |
3757 | buff->next = pfile->u_buff; | |
3758 | pfile->u_buff = buff; | |
3759 | result = buff->cur; | |
3760 | } | |
3761 | ||
3762 | buff->cur = result + len; | |
3763 | return result; | |
3764 | } | |
3765 | ||
87062813 NB |
3766 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
3767 | That buffer is used for growing allocations when saving macro | |
3768 | replacement lists in a #define, and when parsing an answer to an | |
3769 | assertion in #assert, #unassert or #if (and therefore possibly | |
3770 | whilst expanding macros). It therefore must not be used by any | |
3771 | code that they might call: specifically the lexer and the guts of | |
3772 | the macro expander. | |
3773 | ||
3774 | All existing other uses clearly fit this restriction: storing | |
3775 | registered pragmas during initialization. */ | |
93c80368 | 3776 | unsigned char * |
6cf87ca4 | 3777 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
3fef5b2b | 3778 | { |
8c3b2693 NB |
3779 | _cpp_buff *buff = pfile->a_buff; |
3780 | unsigned char *result = buff->cur; | |
3fef5b2b | 3781 | |
8c3b2693 | 3782 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 3783 | { |
8c3b2693 NB |
3784 | buff = _cpp_get_buff (pfile, len); |
3785 | buff->next = pfile->a_buff; | |
3786 | pfile->a_buff = buff; | |
3787 | result = buff->cur; | |
3fef5b2b | 3788 | } |
041c3194 | 3789 | |
8c3b2693 | 3790 | buff->cur = result + len; |
93c80368 | 3791 | return result; |
041c3194 | 3792 | } |
d8044160 | 3793 | |
10f04917 NS |
3794 | /* Commit or allocate storage from a buffer. */ |
3795 | ||
3796 | void * | |
3797 | _cpp_commit_buff (cpp_reader *pfile, size_t size) | |
3798 | { | |
3799 | void *ptr = BUFF_FRONT (pfile->a_buff); | |
3800 | ||
3801 | if (pfile->hash_table->alloc_subobject) | |
3802 | { | |
3803 | void *copy = pfile->hash_table->alloc_subobject (size); | |
3804 | memcpy (copy, ptr, size); | |
3805 | ptr = copy; | |
3806 | } | |
3807 | else | |
3808 | BUFF_FRONT (pfile->a_buff) += size; | |
3809 | ||
3810 | return ptr; | |
3811 | } | |
3812 | ||
d8044160 GK |
3813 | /* Say which field of TOK is in use. */ |
3814 | ||
3815 | enum cpp_token_fld_kind | |
c26302d5 | 3816 | cpp_token_val_index (const cpp_token *tok) |
d8044160 GK |
3817 | { |
3818 | switch (TOKEN_SPELL (tok)) | |
3819 | { | |
3820 | case SPELL_IDENT: | |
3821 | return CPP_TOKEN_FLD_NODE; | |
3822 | case SPELL_LITERAL: | |
3823 | return CPP_TOKEN_FLD_STR; | |
aa508502 | 3824 | case SPELL_OPERATOR: |
3f23e487 AP |
3825 | /* Operands which were originally spelled as ident keep around |
3826 | the node for the exact spelling. */ | |
3827 | if (tok->flags & NAMED_OP) | |
3828 | return CPP_TOKEN_FLD_NODE; | |
3829 | else if (tok->type == CPP_PASTE) | |
9a0c6187 | 3830 | return CPP_TOKEN_FLD_TOKEN_NO; |
aa508502 JM |
3831 | else |
3832 | return CPP_TOKEN_FLD_NONE; | |
d8044160 GK |
3833 | case SPELL_NONE: |
3834 | if (tok->type == CPP_MACRO_ARG) | |
3835 | return CPP_TOKEN_FLD_ARG_NO; | |
3836 | else if (tok->type == CPP_PADDING) | |
3837 | return CPP_TOKEN_FLD_SOURCE; | |
21b11495 | 3838 | else if (tok->type == CPP_PRAGMA) |
bc4071dd | 3839 | return CPP_TOKEN_FLD_PRAGMA; |
191816a3 | 3840 | /* fall through */ |
d8044160 GK |
3841 | default: |
3842 | return CPP_TOKEN_FLD_NONE; | |
3843 | } | |
3844 | } | |
e3dfef44 | 3845 | |
f3f6029d | 3846 | /* All tokens lexed in R after calling this function will be forced to |
620e594b | 3847 | have their location_t to be P, until |
e3dfef44 GC |
3848 | cpp_stop_forcing_token_locations is called for R. */ |
3849 | ||
3850 | void | |
620e594b | 3851 | cpp_force_token_locations (cpp_reader *r, location_t loc) |
e3dfef44 | 3852 | { |
f3f6029d | 3853 | r->forced_token_location = loc; |
e3dfef44 GC |
3854 | } |
3855 | ||
3856 | /* Go back to assigning locations naturally for lexed tokens. */ | |
3857 | ||
3858 | void | |
3859 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
3860 | { | |
f3f6029d | 3861 | r->forced_token_location = 0; |
e3dfef44 | 3862 | } |
b224c376 NS |
3863 | |
3864 | /* We're looking at \, if it's escaping EOL, look past it. If at | |
3865 | LIMIT, don't advance. */ | |
3866 | ||
3867 | static const unsigned char * | |
3868 | do_peek_backslash (const unsigned char *peek, const unsigned char *limit) | |
3869 | { | |
3870 | const unsigned char *probe = peek; | |
3871 | ||
3872 | if (__builtin_expect (peek[1] == '\n', true)) | |
3873 | { | |
3874 | eol: | |
3875 | probe += 2; | |
3876 | if (__builtin_expect (probe < limit, true)) | |
3877 | { | |
3878 | peek = probe; | |
3879 | if (*peek == '\\') | |
3880 | /* The user might be perverse. */ | |
3881 | return do_peek_backslash (peek, limit); | |
3882 | } | |
3883 | } | |
3884 | else if (__builtin_expect (peek[1] == '\r', false)) | |
3885 | { | |
3886 | if (probe[2] == '\n') | |
3887 | probe++; | |
3888 | goto eol; | |
3889 | } | |
3890 | ||
3891 | return peek; | |
3892 | } | |
3893 | ||
3894 | static const unsigned char * | |
3895 | do_peek_next (const unsigned char *peek, const unsigned char *limit) | |
3896 | { | |
3897 | if (__builtin_expect (*peek == '\\', false)) | |
3898 | peek = do_peek_backslash (peek, limit); | |
3899 | return peek; | |
3900 | } | |
3901 | ||
3902 | static const unsigned char * | |
3903 | do_peek_prev (const unsigned char *peek, const unsigned char *bound) | |
3904 | { | |
3905 | if (peek == bound) | |
3906 | return NULL; | |
3907 | ||
3908 | unsigned char c = *--peek; | |
3909 | if (__builtin_expect (c == '\n', false) | |
3910 | || __builtin_expect (c == 'r', false)) | |
3911 | { | |
3912 | if (peek == bound) | |
3913 | return peek; | |
3914 | int ix = -1; | |
3915 | if (c == '\n' && peek[ix] == '\r') | |
3916 | { | |
3917 | if (peek + ix == bound) | |
3918 | return peek; | |
3919 | ix--; | |
3920 | } | |
3921 | ||
3922 | if (peek[ix] == '\\') | |
3923 | return do_peek_prev (peek + ix, bound); | |
3924 | ||
3925 | return peek; | |
3926 | } | |
3927 | else | |
3928 | return peek; | |
3929 | } | |
3930 | ||
3931 | /* Directives-only scanning. Somewhat more relaxed than correct | |
3932 | parsing -- some ill-formed programs will not be rejected. */ | |
3933 | ||
3934 | void | |
3935 | cpp_directive_only_process (cpp_reader *pfile, | |
3936 | void *data, | |
3937 | void (*cb) (cpp_reader *, CPP_DO_task, void *, ...)) | |
3938 | { | |
3939 | do | |
3940 | { | |
3941 | restart: | |
3942 | /* Buffer initialization, but no line cleaning. */ | |
3943 | cpp_buffer *buffer = pfile->buffer; | |
3944 | buffer->cur_note = buffer->notes_used = 0; | |
3945 | buffer->cur = buffer->line_base = buffer->next_line; | |
3946 | buffer->need_line = false; | |
3947 | /* Files always end in a newline. We rely on this for | |
3948 | character peeking safety. */ | |
3949 | gcc_assert (buffer->rlimit[-1] == '\n'); | |
3950 | ||
3951 | const unsigned char *base = buffer->cur; | |
3952 | unsigned line_count = 0; | |
3953 | const unsigned char *line_start = base; | |
3954 | ||
3955 | bool bol = true; | |
3956 | bool raw = false; | |
3957 | ||
3958 | const unsigned char *lwm = base; | |
3959 | for (const unsigned char *pos = base, *limit = buffer->rlimit; | |
3960 | pos < limit;) | |
3961 | { | |
3962 | unsigned char c = *pos++; | |
3963 | /* This matches the switch in _cpp_lex_direct. */ | |
3964 | switch (c) | |
3965 | { | |
3966 | case ' ': case '\t': case '\f': case '\v': | |
3967 | /* Whitespace, do nothing. */ | |
3968 | break; | |
3969 | ||
3970 | case '\r': /* MAC line ending, or Windows \r\n */ | |
3971 | if (*pos == '\n') | |
3972 | pos++; | |
3973 | /* FALLTHROUGH */ | |
3974 | ||
3975 | case '\n': | |
3976 | bol = true; | |
3977 | ||
3978 | next_line: | |
3979 | CPP_INCREMENT_LINE (pfile, 0); | |
3980 | line_count++; | |
3981 | line_start = pos; | |
3982 | break; | |
3983 | ||
3984 | case '\\': | |
3985 | /* <backslash><newline> is removed, and doesn't undo any | |
3986 | preceeding escape or whatnot. */ | |
3987 | if (*pos == '\n') | |
3988 | { | |
3989 | pos++; | |
3990 | goto next_line; | |
3991 | } | |
3992 | else if (*pos == '\r') | |
3993 | { | |
3994 | if (pos[1] == '\n') | |
3995 | pos++; | |
3996 | pos++; | |
3997 | goto next_line; | |
3998 | } | |
3999 | goto dflt; | |
4000 | ||
4001 | case '#': | |
4002 | if (bol) | |
4003 | { | |
4004 | /* Line directive. */ | |
4005 | if (pos - 1 > base && !pfile->state.skipping) | |
4006 | cb (pfile, CPP_DO_print, data, | |
4007 | line_count, base, pos - 1 - base); | |
4008 | ||
4009 | /* Prep things for directive handling. */ | |
4010 | buffer->next_line = pos; | |
4011 | buffer->need_line = true; | |
2a0225e4 NS |
4012 | bool ok = _cpp_get_fresh_line (pfile); |
4013 | gcc_checking_assert (ok); | |
b224c376 NS |
4014 | |
4015 | /* Ensure proper column numbering for generated | |
4016 | error messages. */ | |
4017 | buffer->line_base -= pos - line_start; | |
4018 | ||
4019 | _cpp_handle_directive (pfile, line_start + 1 != pos); | |
4020 | ||
4021 | /* Sanitize the line settings. Duplicate #include's can | |
4022 | mess things up. */ | |
4023 | // FIXME: Necessary? | |
4024 | pfile->line_table->highest_location | |
4025 | = pfile->line_table->highest_line; | |
4026 | ||
4027 | if (!pfile->state.skipping | |
4028 | && pfile->buffer->next_line < pfile->buffer->rlimit) | |
4029 | cb (pfile, CPP_DO_location, data, | |
4030 | pfile->line_table->highest_line); | |
4031 | ||
4032 | goto restart; | |
4033 | } | |
4034 | goto dflt; | |
4035 | ||
4036 | case '/': | |
4037 | { | |
4038 | const unsigned char *peek = do_peek_next (pos, limit); | |
4039 | if (!(*peek == '/' || *peek == '*')) | |
4040 | goto dflt; | |
4041 | ||
4042 | /* Line or block comment */ | |
4043 | bool is_block = *peek == '*'; | |
4044 | bool star = false; | |
4045 | bool esc = false; | |
4046 | location_t sloc | |
4047 | = linemap_position_for_column (pfile->line_table, | |
4048 | pos - line_start); | |
4049 | ||
4050 | while (pos < limit) | |
4051 | { | |
4052 | char c = *pos++; | |
4053 | switch (c) | |
4054 | { | |
4055 | case '\\': | |
4056 | esc = true; | |
4057 | break; | |
4058 | ||
4059 | case '\r': | |
4060 | if (*pos == '\n') | |
4061 | pos++; | |
4062 | /* FALLTHROUGH */ | |
4063 | ||
4064 | case '\n': | |
4065 | { | |
4066 | CPP_INCREMENT_LINE (pfile, 0); | |
4067 | line_count++; | |
4068 | line_start = pos; | |
4069 | if (!esc && !is_block) | |
4070 | { | |
4071 | bol = true; | |
4072 | goto done_comment; | |
4073 | } | |
4074 | } | |
4075 | if (!esc) | |
4076 | star = false; | |
4077 | esc = false; | |
4078 | break; | |
4079 | ||
4080 | case '*': | |
4081 | if (pos > peek && !esc) | |
4082 | star = is_block; | |
4083 | esc = false; | |
4084 | break; | |
4085 | ||
4086 | case '/': | |
4087 | if (star) | |
4088 | goto done_comment; | |
4089 | /* FALLTHROUGH */ | |
4090 | ||
4091 | default: | |
4092 | star = false; | |
4093 | esc = false; | |
4094 | break; | |
4095 | } | |
4096 | } | |
4097 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, | |
4098 | "unterminated comment"); | |
4099 | done_comment: | |
4100 | lwm = pos; | |
4101 | break; | |
4102 | } | |
4103 | ||
4104 | case '\'': | |
4105 | if (!CPP_OPTION (pfile, digit_separators)) | |
4106 | goto delimited_string; | |
4107 | ||
4108 | /* Possibly a number punctuator. */ | |
4109 | if (!ISIDNUM (*do_peek_next (pos, limit))) | |
4110 | goto delimited_string; | |
4111 | ||
4112 | goto quote_peek; | |
4113 | ||
4114 | case '\"': | |
4115 | if (!CPP_OPTION (pfile, rliterals)) | |
4116 | goto delimited_string; | |
4117 | ||
4118 | quote_peek: | |
4119 | { | |
4120 | /* For ' see if it's a number punctuator | |
4121 | \.?<digit>(<digit>|<identifier-nondigit> | |
4122 | |'<digit>|'<nondigit>|[eEpP]<sign>|\.)* */ | |
4123 | /* For " see if it's a raw string | |
4124 | {U,L,u,u8}R. This includes CPP_NUMBER detection, | |
4125 | because that could be 0e+R. */ | |
4126 | const unsigned char *peek = pos - 1; | |
4127 | bool quote_first = c == '"'; | |
4128 | bool quote_eight = false; | |
4129 | bool maybe_number_start = false; | |
4130 | bool want_number = false; | |
4131 | ||
4132 | while ((peek = do_peek_prev (peek, lwm))) | |
4133 | { | |
4134 | unsigned char p = *peek; | |
4135 | if (quote_first) | |
4136 | { | |
4137 | if (!raw) | |
4138 | { | |
4139 | if (p != 'R') | |
4140 | break; | |
4141 | raw = true; | |
4142 | continue; | |
4143 | } | |
4144 | ||
4145 | quote_first = false; | |
4146 | if (p == 'L' || p == 'U' || p == 'u') | |
4147 | ; | |
4148 | else if (p == '8') | |
4149 | quote_eight = true; | |
4150 | else | |
4151 | goto second_raw; | |
4152 | } | |
4153 | else if (quote_eight) | |
4154 | { | |
4155 | if (p != 'u') | |
4156 | { | |
4157 | raw = false; | |
4158 | break; | |
4159 | } | |
4160 | quote_eight = false; | |
4161 | } | |
4162 | else if (c == '"') | |
4163 | { | |
4164 | second_raw:; | |
4165 | if (!want_number && ISIDNUM (p)) | |
4166 | { | |
4167 | raw = false; | |
4168 | break; | |
4169 | } | |
4170 | } | |
4171 | ||
4172 | if (ISDIGIT (p)) | |
4173 | maybe_number_start = true; | |
4174 | else if (p == '.') | |
4175 | want_number = true; | |
4176 | else if (ISIDNUM (p)) | |
4177 | maybe_number_start = false; | |
4178 | else if (p == '+' || p == '-') | |
4179 | { | |
4180 | if (const unsigned char *peek_prev | |
4181 | = do_peek_prev (peek, lwm)) | |
4182 | { | |
4183 | p = *peek_prev; | |
4184 | if (p == 'e' || p == 'E' | |
4185 | || p == 'p' || p == 'P') | |
4186 | { | |
4187 | want_number = true; | |
4188 | maybe_number_start = false; | |
4189 | } | |
4190 | else | |
4191 | break; | |
4192 | } | |
4193 | else | |
4194 | break; | |
4195 | } | |
4196 | else if (p == '\'' || p == '\"') | |
4197 | { | |
4198 | /* If this is lwm, this must be the end of a | |
4199 | previous string. So this is a trailing | |
4200 | literal type, (a) if those are allowed, | |
4201 | and (b) maybe_start is false. Otherwise | |
4202 | this must be a CPP_NUMBER because we've | |
4203 | met another ', and we'd have checked that | |
4204 | in its own right. */ | |
4205 | if (peek == lwm && CPP_OPTION (pfile, uliterals)) | |
4206 | { | |
4207 | if (!maybe_number_start && !want_number) | |
4208 | /* Must be a literal type. */ | |
4209 | raw = false; | |
4210 | } | |
4211 | else if (p == '\'' | |
4212 | && CPP_OPTION (pfile, digit_separators)) | |
4213 | maybe_number_start = true; | |
4214 | break; | |
4215 | } | |
4216 | else if (c == '\'') | |
4217 | break; | |
4218 | else if (!quote_first && !quote_eight) | |
4219 | break; | |
4220 | } | |
4221 | ||
4222 | if (maybe_number_start) | |
4223 | { | |
4224 | if (c == '\'') | |
4225 | /* A CPP NUMBER. */ | |
4226 | goto dflt; | |
4227 | raw = false; | |
4228 | } | |
4229 | ||
4230 | goto delimited_string; | |
4231 | } | |
4232 | ||
4233 | delimited_string: | |
4234 | { | |
4235 | /* (Possibly raw) string or char literal. */ | |
4236 | unsigned char end = c; | |
4237 | int delim_len = -1; | |
4238 | const unsigned char *delim = NULL; | |
4239 | location_t sloc = linemap_position_for_column (pfile->line_table, | |
4240 | pos - line_start); | |
4241 | int esc = 0; | |
4242 | ||
4243 | if (raw) | |
4244 | { | |
4245 | /* There can be no line breaks in the delimiter. */ | |
4246 | delim = pos; | |
4247 | for (delim_len = 0; (c = *pos++) != '('; delim_len++) | |
4248 | { | |
4249 | if (delim_len == 16) | |
4250 | { | |
4251 | cpp_error_with_line (pfile, CPP_DL_ERROR, | |
4252 | sloc, 0, | |
4253 | "raw string delimiter" | |
4254 | " longer than %d" | |
4255 | " characters", | |
4256 | delim_len); | |
4257 | raw = false; | |
4258 | pos = delim; | |
4259 | break; | |
4260 | } | |
4261 | if (strchr (") \\\t\v\f\n", c)) | |
4262 | { | |
4263 | cpp_error_with_line (pfile, CPP_DL_ERROR, | |
4264 | sloc, 0, | |
4265 | "invalid character '%c'" | |
4266 | " in raw string" | |
4267 | " delimiter", c); | |
4268 | raw = false; | |
4269 | pos = delim; | |
4270 | break; | |
4271 | } | |
4272 | if (pos >= limit) | |
4273 | goto bad_string; | |
4274 | } | |
4275 | } | |
4276 | ||
4277 | while (pos < limit) | |
4278 | { | |
4279 | char c = *pos++; | |
4280 | switch (c) | |
4281 | { | |
4282 | case '\\': | |
4283 | if (!raw) | |
4284 | esc++; | |
4285 | break; | |
4286 | ||
4287 | case '\r': | |
4288 | if (*pos == '\n') | |
4289 | pos++; | |
4290 | /* FALLTHROUGH */ | |
4291 | ||
4292 | case '\n': | |
4293 | { | |
4294 | CPP_INCREMENT_LINE (pfile, 0); | |
4295 | line_count++; | |
4296 | line_start = pos; | |
4297 | } | |
4298 | if (esc) | |
4299 | esc--; | |
4300 | break; | |
4301 | ||
4302 | case ')': | |
4303 | if (raw | |
4304 | && pos + delim_len + 1 < limit | |
4305 | && pos[delim_len] == end | |
4306 | && !memcmp (delim, pos, delim_len)) | |
4307 | { | |
4308 | pos += delim_len + 1; | |
4309 | raw = false; | |
4310 | goto done_string; | |
4311 | } | |
4312 | break; | |
4313 | ||
4314 | default: | |
4315 | if (!raw && !(esc & 1) && c == end) | |
4316 | goto done_string; | |
4317 | esc = 0; | |
4318 | break; | |
4319 | } | |
4320 | } | |
4321 | bad_string: | |
4322 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, | |
4323 | "unterminated literal"); | |
4324 | ||
4325 | done_string: | |
4326 | raw = false; | |
4327 | lwm = pos - 1; | |
4328 | } | |
4329 | goto dflt; | |
4330 | ||
4331 | default: | |
4332 | dflt: | |
4333 | bol = false; | |
4334 | pfile->mi_valid = false; | |
4335 | break; | |
4336 | } | |
4337 | } | |
4338 | ||
4339 | if (buffer->rlimit > base && !pfile->state.skipping) | |
4340 | cb (pfile, CPP_DO_print, data, line_count, base, buffer->rlimit - base); | |
4341 | ||
4342 | _cpp_pop_buffer (pfile); | |
4343 | } | |
4344 | while (pfile->buffer); | |
4345 | } |