]>
Commit | Line | Data |
---|---|---|
45b966db | 1 | /* CPP Library - lexical analysis. |
a5544970 | 2 | Copyright (C) 2000-2019 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, | |
1375 | "__VA_OPT__ is not available until C++2a"); | |
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" | |
1383 | " of a C++2a variadic macro"); | |
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 | ||
00a81b8b JM |
1589 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1590 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1591 | ||
1592 | static void | |
1593 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1594 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1595 | { | |
1596 | _cpp_buff *first_buff = *first_buff_p; | |
1597 | _cpp_buff *last_buff = *last_buff_p; | |
1598 | ||
1599 | if (first_buff == NULL) | |
1600 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1601 | else if (len > BUFF_ROOM (last_buff)) | |
1602 | { | |
1603 | size_t room = BUFF_ROOM (last_buff); | |
1604 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1605 | BUFF_FRONT (last_buff) += room; | |
1606 | base += room; | |
1607 | len -= room; | |
1608 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1609 | } | |
1610 | ||
1611 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1612 | BUFF_FRONT (last_buff) += len; | |
1613 | ||
1614 | *first_buff_p = first_buff; | |
1615 | *last_buff_p = last_buff; | |
1616 | } | |
1617 | ||
c865f923 ESR |
1618 | |
1619 | /* Returns true if a macro has been defined. | |
1620 | This might not work if compile with -save-temps, | |
1621 | or preprocess separately from compilation. */ | |
1622 | ||
1623 | static bool | |
1624 | is_macro(cpp_reader *pfile, const uchar *base) | |
1625 | { | |
1626 | const uchar *cur = base; | |
1627 | if (! ISIDST (*cur)) | |
1628 | return false; | |
1629 | unsigned int hash = HT_HASHSTEP (0, *cur); | |
1630 | ++cur; | |
1631 | while (ISIDNUM (*cur)) | |
1632 | { | |
1633 | hash = HT_HASHSTEP (hash, *cur); | |
1634 | ++cur; | |
1635 | } | |
1636 | hash = HT_HASHFINISH (hash, cur - base); | |
1637 | ||
1638 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1639 | base, cur - base, hash, HT_NO_INSERT)); | |
1640 | ||
3f6677f4 | 1641 | return result && cpp_macro_p (result); |
c865f923 ESR |
1642 | } |
1643 | ||
b44f8ad8 JW |
1644 | /* Returns true if a literal suffix does not have the expected form |
1645 | and is defined as a macro. */ | |
1646 | ||
1647 | static bool | |
1648 | is_macro_not_literal_suffix(cpp_reader *pfile, const uchar *base) | |
1649 | { | |
1650 | /* User-defined literals outside of namespace std must start with a single | |
1651 | underscore, so assume anything of that form really is a UDL suffix. | |
1652 | We don't need to worry about UDLs defined inside namespace std because | |
1653 | their names are reserved, so cannot be used as macro names in valid | |
1654 | programs. */ | |
1655 | if (base[0] == '_' && base[1] != '_') | |
1656 | return false; | |
1657 | return is_macro (pfile, base); | |
1658 | } | |
c865f923 | 1659 | |
2c6e3f55 | 1660 | /* Lexes a raw string. The stored string contains the spelling, including |
00a81b8b | 1661 | double quotes, delimiter string, '(' and ')', any leading |
2c6e3f55 JJ |
1662 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1663 | literal, or CPP_OTHER if it was not properly terminated. | |
1664 | ||
1665 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1666 | is the first NUL since embedded NULs are preserved. */ | |
1667 | ||
1668 | static void | |
1669 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1670 | const uchar *cur) | |
1671 | { | |
8cf88735 JJ |
1672 | uchar raw_prefix[17]; |
1673 | uchar temp_buffer[18]; | |
1674 | const uchar *orig_base; | |
1675 | unsigned int raw_prefix_len = 0, raw_suffix_len = 0; | |
1676 | enum raw_str_phase { RAW_STR_PREFIX, RAW_STR, RAW_STR_SUFFIX }; | |
1677 | raw_str_phase phase = RAW_STR_PREFIX; | |
2c6e3f55 JJ |
1678 | enum cpp_ttype type; |
1679 | size_t total_len = 0; | |
8cf88735 JJ |
1680 | /* Index into temp_buffer during phases other than RAW_STR, |
1681 | during RAW_STR phase 17 to tell BUF_APPEND that nothing should | |
1682 | be appended to temp_buffer. */ | |
1683 | size_t temp_buffer_len = 0; | |
2c6e3f55 | 1684 | _cpp_buff *first_buff = NULL, *last_buff = NULL; |
8cf88735 | 1685 | size_t raw_prefix_start; |
00a81b8b | 1686 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 JJ |
1687 | |
1688 | type = (*base == 'L' ? CPP_WSTRING : | |
1689 | *base == 'U' ? CPP_STRING32 : | |
1690 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1691 | : CPP_STRING); | |
1692 | ||
00a81b8b JM |
1693 | #define BUF_APPEND(STR,LEN) \ |
1694 | do { \ | |
1695 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1696 | &first_buff, &last_buff); \ | |
1697 | total_len += (LEN); \ | |
8cf88735 JJ |
1698 | if (__builtin_expect (temp_buffer_len < 17, 0) \ |
1699 | && (const uchar *)(STR) != base \ | |
1700 | && (LEN) <= 2) \ | |
1701 | { \ | |
1702 | memcpy (temp_buffer + temp_buffer_len, \ | |
1703 | (const uchar *)(STR), (LEN)); \ | |
1704 | temp_buffer_len += (LEN); \ | |
1705 | } \ | |
c830c7d5 | 1706 | } while (0) |
00a81b8b | 1707 | |
8cf88735 JJ |
1708 | orig_base = base; |
1709 | ++cur; | |
1710 | raw_prefix_start = cur - base; | |
1711 | for (;;) | |
1712 | { | |
00a81b8b JM |
1713 | cppchar_t c; |
1714 | ||
1715 | /* If we previously performed any trigraph or line splicing | |
8cf88735 JJ |
1716 | transformations, undo them in between the opening and closing |
1717 | double quote. */ | |
00a81b8b JM |
1718 | while (note->pos < cur) |
1719 | ++note; | |
1720 | for (; note->pos == cur; ++note) | |
1721 | { | |
1722 | switch (note->type) | |
1723 | { | |
1724 | case '\\': | |
1725 | case ' ': | |
1726 | /* Restore backslash followed by newline. */ | |
1727 | BUF_APPEND (base, cur - base); | |
1728 | base = cur; | |
1729 | BUF_APPEND ("\\", 1); | |
1730 | after_backslash: | |
1731 | if (note->type == ' ') | |
1732 | { | |
1733 | /* GNU backslash whitespace newline extension. FIXME | |
1734 | could be any sequence of non-vertical space. When we | |
1735 | can properly restore any such sequence, we should mark | |
1736 | this note as handled so _cpp_process_line_notes | |
1737 | doesn't warn. */ | |
1738 | BUF_APPEND (" ", 1); | |
1739 | } | |
1740 | ||
1741 | BUF_APPEND ("\n", 1); | |
1742 | break; | |
1743 | ||
1744 | case 0: | |
1745 | /* Already handled. */ | |
1746 | break; | |
1747 | ||
1748 | default: | |
1749 | if (_cpp_trigraph_map[note->type]) | |
1750 | { | |
1751 | /* Don't warn about this trigraph in | |
1752 | _cpp_process_line_notes, since trigraphs show up as | |
1753 | trigraphs in raw strings. */ | |
d947ada0 | 1754 | uchar type = note->type; |
00a81b8b JM |
1755 | note->type = 0; |
1756 | ||
1757 | if (!CPP_OPTION (pfile, trigraphs)) | |
1758 | /* If we didn't convert the trigraph in the first | |
1759 | place, don't do anything now either. */ | |
1760 | break; | |
1761 | ||
1762 | BUF_APPEND (base, cur - base); | |
1763 | base = cur; | |
1764 | BUF_APPEND ("??", 2); | |
1765 | ||
1766 | /* ??/ followed by newline gets two line notes, one for | |
1767 | the trigraph and one for the backslash/newline. */ | |
1768 | if (type == '/' && note[1].pos == cur) | |
1769 | { | |
1770 | if (note[1].type != '\\' | |
1771 | && note[1].type != ' ') | |
1772 | abort (); | |
1773 | BUF_APPEND ("/", 1); | |
1774 | ++note; | |
1775 | goto after_backslash; | |
1776 | } | |
00a81b8b JM |
1777 | else |
1778 | { | |
1779 | /* Skip the replacement character. */ | |
1780 | base = ++cur; | |
1781 | BUF_APPEND (&type, 1); | |
8cf88735 JJ |
1782 | c = type; |
1783 | goto check_c; | |
00a81b8b JM |
1784 | } |
1785 | } | |
1786 | else | |
1787 | abort (); | |
1788 | break; | |
1789 | } | |
1790 | } | |
1791 | c = *cur++; | |
8cf88735 JJ |
1792 | if (__builtin_expect (temp_buffer_len < 17, 0)) |
1793 | temp_buffer[temp_buffer_len++] = c; | |
2c6e3f55 | 1794 | |
8cf88735 JJ |
1795 | check_c: |
1796 | if (phase == RAW_STR_PREFIX) | |
2c6e3f55 | 1797 | { |
8cf88735 JJ |
1798 | while (raw_prefix_len < temp_buffer_len) |
1799 | { | |
1800 | raw_prefix[raw_prefix_len] = temp_buffer[raw_prefix_len]; | |
1801 | switch (raw_prefix[raw_prefix_len]) | |
1802 | { | |
1803 | case ' ': case '(': case ')': case '\\': case '\t': | |
1804 | case '\v': case '\f': case '\n': default: | |
1805 | break; | |
1806 | /* Basic source charset except the above chars. */ | |
1807 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1808 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1809 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1810 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1811 | case 'y': case 'z': | |
1812 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1813 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1814 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1815 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1816 | case 'Y': case 'Z': | |
1817 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1818 | case '6': case '7': case '8': case '9': | |
1819 | case '_': case '{': case '}': case '#': case '[': case ']': | |
1820 | case '<': case '>': case '%': case ':': case ';': case '.': | |
1821 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1822 | case '&': case '|': case '~': case '!': case '=': case ',': | |
1823 | case '"': case '\'': | |
1824 | if (raw_prefix_len < 16) | |
1825 | { | |
1826 | raw_prefix_len++; | |
1827 | continue; | |
1828 | } | |
1829 | break; | |
1830 | } | |
1831 | ||
1832 | if (raw_prefix[raw_prefix_len] != '(') | |
1833 | { | |
1834 | int col = CPP_BUF_COLUMN (pfile->buffer, cur) + 1; | |
1835 | if (raw_prefix_len == 16) | |
1836 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1837 | col, "raw string delimiter longer " | |
1838 | "than 16 characters"); | |
1839 | else if (raw_prefix[raw_prefix_len] == '\n') | |
1840 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1841 | col, "invalid new-line in raw " | |
1842 | "string delimiter"); | |
1843 | else | |
1844 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1845 | col, "invalid character '%c' in " | |
1846 | "raw string delimiter", | |
1847 | (int) raw_prefix[raw_prefix_len]); | |
1848 | pfile->buffer->cur = orig_base + raw_prefix_start - 1; | |
1849 | create_literal (pfile, token, orig_base, | |
1850 | raw_prefix_start - 1, CPP_OTHER); | |
1851 | if (first_buff) | |
1852 | _cpp_release_buff (pfile, first_buff); | |
1853 | return; | |
1854 | } | |
1855 | raw_prefix[raw_prefix_len] = '"'; | |
1856 | phase = RAW_STR; | |
1857 | /* Nothing should be appended to temp_buffer during | |
1858 | RAW_STR phase. */ | |
1859 | temp_buffer_len = 17; | |
1860 | break; | |
1861 | } | |
1862 | continue; | |
1863 | } | |
1864 | else if (phase == RAW_STR_SUFFIX) | |
1865 | { | |
1866 | while (raw_suffix_len <= raw_prefix_len | |
1867 | && raw_suffix_len < temp_buffer_len | |
1868 | && temp_buffer[raw_suffix_len] == raw_prefix[raw_suffix_len]) | |
1869 | raw_suffix_len++; | |
1870 | if (raw_suffix_len > raw_prefix_len) | |
1871 | break; | |
1872 | if (raw_suffix_len == temp_buffer_len) | |
1873 | continue; | |
1874 | phase = RAW_STR; | |
1875 | /* Nothing should be appended to temp_buffer during | |
1876 | RAW_STR phase. */ | |
1877 | temp_buffer_len = 17; | |
1878 | } | |
1879 | if (c == ')') | |
1880 | { | |
1881 | phase = RAW_STR_SUFFIX; | |
1882 | raw_suffix_len = 0; | |
1883 | temp_buffer_len = 0; | |
2c6e3f55 JJ |
1884 | } |
1885 | else if (c == '\n') | |
1886 | { | |
1887 | if (pfile->state.in_directive | |
d5e48350 JJ |
1888 | || (pfile->state.parsing_args |
1889 | && pfile->buffer->next_line >= pfile->buffer->rlimit)) | |
2c6e3f55 JJ |
1890 | { |
1891 | cur--; | |
1892 | type = CPP_OTHER; | |
1893 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1894 | "unterminated raw string"); | |
1895 | break; | |
1896 | } | |
1897 | ||
00a81b8b | 1898 | BUF_APPEND (base, cur - base); |
2c6e3f55 JJ |
1899 | |
1900 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1901 | CPP_INCREMENT_LINE (pfile, 0); | |
1902 | pfile->buffer->need_line = true; | |
1903 | ||
00a81b8b JM |
1904 | pfile->buffer->cur = cur-1; |
1905 | _cpp_process_line_notes (pfile, false); | |
2c6e3f55 JJ |
1906 | if (!_cpp_get_fresh_line (pfile)) |
1907 | { | |
620e594b | 1908 | location_t src_loc = token->src_loc; |
2c6e3f55 JJ |
1909 | token->type = CPP_EOF; |
1910 | /* Tell the compiler the line number of the EOF token. */ | |
1911 | token->src_loc = pfile->line_table->highest_line; | |
1912 | token->flags = BOL; | |
1913 | if (first_buff != NULL) | |
1914 | _cpp_release_buff (pfile, first_buff); | |
1915 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1916 | "unterminated raw string"); | |
1917 | return; | |
1918 | } | |
1919 | ||
1920 | cur = base = pfile->buffer->cur; | |
00a81b8b | 1921 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 1922 | } |
2c6e3f55 JJ |
1923 | } |
1924 | ||
3ce4f9e4 ESR |
1925 | if (CPP_OPTION (pfile, user_literals)) |
1926 | { | |
c865f923 ESR |
1927 | /* If a string format macro, say from inttypes.h, is placed touching |
1928 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 JW |
1929 | literal thus breaking the program. */ |
1930 | if (is_macro_not_literal_suffix (pfile, cur)) | |
7f5f5f98 | 1931 | { |
112448b4 | 1932 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 1933 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
1934 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1935 | token->src_loc, 0, | |
1936 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 1937 | "a space between literal and string macro"); |
7f5f5f98 | 1938 | } |
3ce4f9e4 | 1939 | /* Grab user defined literal suffix. */ |
561f7fc7 | 1940 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
1941 | { |
1942 | type = cpp_userdef_string_add_type (type); | |
1943 | ++cur; | |
7f5f5f98 OW |
1944 | |
1945 | while (ISIDNUM (*cur)) | |
1946 | ++cur; | |
3ce4f9e4 | 1947 | } |
3ce4f9e4 ESR |
1948 | } |
1949 | ||
2c6e3f55 JJ |
1950 | pfile->buffer->cur = cur; |
1951 | if (first_buff == NULL) | |
1952 | create_literal (pfile, token, base, cur - base, type); | |
1953 | else | |
1954 | { | |
1955 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1956 | ||
1957 | token->type = type; | |
1958 | token->val.str.len = total_len + (cur - base); | |
1959 | token->val.str.text = dest; | |
1960 | last_buff = first_buff; | |
1961 | while (last_buff != NULL) | |
1962 | { | |
1963 | memcpy (dest, last_buff->base, | |
1964 | BUFF_FRONT (last_buff) - last_buff->base); | |
1965 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1966 | last_buff = last_buff->next; | |
1967 | } | |
1968 | _cpp_release_buff (pfile, first_buff); | |
1969 | memcpy (dest, base, cur - base); | |
1970 | dest[cur - base] = '\0'; | |
1971 | } | |
1972 | } | |
1973 | ||
bced6edf | 1974 | /* Lexes a string, character constant, or angle-bracketed header file |
6338b358 | 1975 | name. The stored string contains the spelling, including opening |
2c6e3f55 JJ |
1976 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1977 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1978 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1979 | header name which must be relexed as normal tokens. | |
6338b358 NB |
1980 | |
1981 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1982 | is the first NUL since embedded NULs are preserved. */ | |
041c3194 | 1983 | static void |
6cf87ca4 | 1984 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
45b966db | 1985 | { |
6338b358 NB |
1986 | bool saw_NUL = false; |
1987 | const uchar *cur; | |
bced6edf | 1988 | cppchar_t terminator; |
6338b358 NB |
1989 | enum cpp_ttype type; |
1990 | ||
1991 | cur = base; | |
1992 | terminator = *cur++; | |
2c6e3f55 | 1993 | if (terminator == 'L' || terminator == 'U') |
6338b358 | 1994 | terminator = *cur++; |
2c6e3f55 JJ |
1995 | else if (terminator == 'u') |
1996 | { | |
1997 | terminator = *cur++; | |
1998 | if (terminator == '8') | |
1999 | terminator = *cur++; | |
2000 | } | |
2001 | if (terminator == 'R') | |
2002 | { | |
2003 | lex_raw_string (pfile, token, base, cur); | |
2004 | return; | |
2005 | } | |
2006 | if (terminator == '"') | |
b6baa67d KVH |
2007 | type = (*base == 'L' ? CPP_WSTRING : |
2008 | *base == 'U' ? CPP_STRING32 : | |
2c6e3f55 JJ |
2009 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
2010 | : CPP_STRING); | |
6338b358 | 2011 | else if (terminator == '\'') |
b6baa67d KVH |
2012 | type = (*base == 'L' ? CPP_WCHAR : |
2013 | *base == 'U' ? CPP_CHAR32 : | |
fe95b036 ESR |
2014 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8CHAR : CPP_CHAR16) |
2015 | : CPP_CHAR); | |
6338b358 NB |
2016 | else |
2017 | terminator = '>', type = CPP_HEADER_NAME; | |
93c80368 | 2018 | |
0d9f234d | 2019 | for (;;) |
45b966db | 2020 | { |
6338b358 | 2021 | cppchar_t c = *cur++; |
7868b4a2 | 2022 | |
6f572ac2 | 2023 | /* In #include-style directives, terminators are not escapable. */ |
6338b358 NB |
2024 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
2025 | cur++; | |
2026 | else if (c == terminator) | |
bced6edf | 2027 | break; |
6338b358 | 2028 | else if (c == '\n') |
0d9f234d | 2029 | { |
6338b358 | 2030 | cur--; |
4bb09c26 JM |
2031 | /* Unmatched quotes always yield undefined behavior, but |
2032 | greedy lexing means that what appears to be an unterminated | |
2033 | header name may actually be a legitimate sequence of tokens. */ | |
2034 | if (terminator == '>') | |
2035 | { | |
2036 | token->type = CPP_LESS; | |
2037 | return; | |
2038 | } | |
6338b358 NB |
2039 | type = CPP_OTHER; |
2040 | break; | |
45b966db | 2041 | } |
6338b358 NB |
2042 | else if (c == '\0') |
2043 | saw_NUL = true; | |
45b966db ZW |
2044 | } |
2045 | ||
6338b358 | 2046 | if (saw_NUL && !pfile->state.skipping) |
0527bc4e JDA |
2047 | cpp_error (pfile, CPP_DL_WARNING, |
2048 | "null character(s) preserved in literal"); | |
45b966db | 2049 | |
c663e301 JM |
2050 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
2051 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
2052 | (int) terminator); | |
2053 | ||
3ce4f9e4 ESR |
2054 | if (CPP_OPTION (pfile, user_literals)) |
2055 | { | |
c865f923 ESR |
2056 | /* If a string format macro, say from inttypes.h, is placed touching |
2057 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 JW |
2058 | literal thus breaking the program. */ |
2059 | if (is_macro_not_literal_suffix (pfile, cur)) | |
7f5f5f98 | 2060 | { |
112448b4 | 2061 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 2062 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
2063 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
2064 | token->src_loc, 0, | |
2065 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 2066 | "a space between literal and string macro"); |
7f5f5f98 | 2067 | } |
3ce4f9e4 | 2068 | /* Grab user defined literal suffix. */ |
561f7fc7 | 2069 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
2070 | { |
2071 | type = cpp_userdef_char_add_type (type); | |
2072 | type = cpp_userdef_string_add_type (type); | |
2073 | ++cur; | |
7f5f5f98 OW |
2074 | |
2075 | while (ISIDNUM (*cur)) | |
2076 | ++cur; | |
3ce4f9e4 | 2077 | } |
3ce4f9e4 | 2078 | } |
fe191308 JM |
2079 | else if (CPP_OPTION (pfile, cpp_warn_cxx11_compat) |
2080 | && is_macro (pfile, cur) | |
2081 | && !pfile->state.skipping) | |
2082 | cpp_warning_with_line (pfile, CPP_W_CXX11_COMPAT, | |
2083 | token->src_loc, 0, "C++11 requires a space " | |
2084 | "between string literal and macro"); | |
3ce4f9e4 | 2085 | |
6338b358 NB |
2086 | pfile->buffer->cur = cur; |
2087 | create_literal (pfile, token, base, cur - base, type); | |
0d9f234d | 2088 | } |
041c3194 | 2089 | |
631d0d36 MG |
2090 | /* Return the comment table. The client may not make any assumption |
2091 | about the ordering of the table. */ | |
2092 | cpp_comment_table * | |
2093 | cpp_get_comments (cpp_reader *pfile) | |
2094 | { | |
2095 | return &pfile->comments; | |
2096 | } | |
2097 | ||
2098 | /* Append a comment to the end of the comment table. */ | |
2099 | static void | |
2100 | store_comment (cpp_reader *pfile, cpp_token *token) | |
2101 | { | |
2102 | int len; | |
2103 | ||
2104 | if (pfile->comments.allocated == 0) | |
2105 | { | |
2106 | pfile->comments.allocated = 256; | |
2107 | pfile->comments.entries = (cpp_comment *) xmalloc | |
2108 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
2109 | } | |
2110 | ||
2111 | if (pfile->comments.count == pfile->comments.allocated) | |
2112 | { | |
2113 | pfile->comments.allocated *= 2; | |
2114 | pfile->comments.entries = (cpp_comment *) xrealloc | |
2115 | (pfile->comments.entries, | |
2116 | pfile->comments.allocated * sizeof (cpp_comment)); | |
2117 | } | |
2118 | ||
2119 | len = token->val.str.len; | |
2120 | ||
2121 | /* Copy comment. Note, token may not be NULL terminated. */ | |
2122 | pfile->comments.entries[pfile->comments.count].comment = | |
2123 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
2124 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
2125 | token->val.str.text, len); | |
2126 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
2127 | ||
2128 | /* Set source location. */ | |
2129 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
2130 | ||
2131 | /* Increment the count of entries in the comment table. */ | |
2132 | pfile->comments.count++; | |
2133 | } | |
2134 | ||
93c80368 | 2135 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 2136 | static void |
6cf87ca4 ZW |
2137 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
2138 | cppchar_t type) | |
9e62c811 | 2139 | { |
041c3194 | 2140 | unsigned char *buffer; |
651a20b5 | 2141 | unsigned int len, clen, i; |
df383483 | 2142 | |
1c6d33ef | 2143 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
480709cc | 2144 | |
3542203b NB |
2145 | /* C++ comments probably (not definitely) have moved past a new |
2146 | line, which we don't want to save in the comment. */ | |
480709cc | 2147 | if (is_vspace (pfile->buffer->cur[-1])) |
3542203b | 2148 | len--; |
477cdac7 | 2149 | |
651a20b5 KT |
2150 | /* If we are currently in a directive or in argument parsing, then |
2151 | we need to store all C++ comments as C comments internally, and | |
2152 | so we need to allocate a little extra space in that case. | |
477cdac7 JT |
2153 | |
2154 | Note that the only time we encounter a directive here is | |
2155 | when we are saving comments in a "#define". */ | |
651a20b5 KT |
2156 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
2157 | && type == '/') ? len + 2 : len; | |
477cdac7 JT |
2158 | |
2159 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
df383483 | 2160 | |
041c3194 | 2161 | token->type = CPP_COMMENT; |
477cdac7 | 2162 | token->val.str.len = clen; |
0d9f234d | 2163 | token->val.str.text = buffer; |
45b966db | 2164 | |
1c6d33ef NB |
2165 | buffer[0] = '/'; |
2166 | memcpy (buffer + 1, from, len - 1); | |
477cdac7 | 2167 | |
1eeeb6a4 | 2168 | /* Finish conversion to a C comment, if necessary. */ |
651a20b5 | 2169 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
477cdac7 JT |
2170 | { |
2171 | buffer[1] = '*'; | |
2172 | buffer[clen - 2] = '*'; | |
2173 | buffer[clen - 1] = '/'; | |
651a20b5 KT |
2174 | /* As there can be in a C++ comments illegal sequences for C comments |
2175 | we need to filter them out. */ | |
2176 | for (i = 2; i < (clen - 2); i++) | |
2177 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
2178 | buffer[i] = '|'; | |
477cdac7 | 2179 | } |
631d0d36 MG |
2180 | |
2181 | /* Finally store this comment for use by clients of libcpp. */ | |
2182 | store_comment (pfile, token); | |
0d9f234d | 2183 | } |
45b966db | 2184 | |
81fea426 MP |
2185 | /* Returns true if comment at COMMENT_START is a recognized FALLTHROUGH |
2186 | comment. */ | |
2187 | ||
2188 | static bool | |
2189 | fallthrough_comment_p (cpp_reader *pfile, const unsigned char *comment_start) | |
2190 | { | |
2191 | const unsigned char *from = comment_start + 1; | |
70f6d5e1 JJ |
2192 | |
2193 | switch (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough)) | |
2194 | { | |
2195 | /* For both -Wimplicit-fallthrough=0 and -Wimplicit-fallthrough=5 we | |
2196 | don't recognize any comments. The latter only checks attributes, | |
2197 | the former doesn't warn. */ | |
2198 | case 0: | |
2199 | default: | |
2200 | return false; | |
2201 | /* -Wimplicit-fallthrough=1 considers any comment, no matter what | |
2202 | content it has. */ | |
2203 | case 1: | |
2204 | return true; | |
2205 | case 2: | |
2206 | /* -Wimplicit-fallthrough=2 looks for (case insensitive) | |
2207 | .*falls?[ \t-]*thr(u|ough).* regex. */ | |
2208 | for (; (size_t) (pfile->buffer->cur - from) >= sizeof "fallthru" - 1; | |
2209 | from++) | |
2210 | { | |
2211 | /* Is there anything like strpbrk with upper boundary, or | |
2212 | memchr looking for 2 characters rather than just one? */ | |
2213 | if (from[0] != 'f' && from[0] != 'F') | |
2214 | continue; | |
2215 | if (from[1] != 'a' && from[1] != 'A') | |
2216 | continue; | |
2217 | if (from[2] != 'l' && from[2] != 'L') | |
2218 | continue; | |
2219 | if (from[3] != 'l' && from[3] != 'L') | |
2220 | continue; | |
2221 | from += sizeof "fall" - 1; | |
2222 | if (from[0] == 's' || from[0] == 'S') | |
2223 | from++; | |
2224 | while (*from == ' ' || *from == '\t' || *from == '-') | |
2225 | from++; | |
2226 | if (from[0] != 't' && from[0] != 'T') | |
2227 | continue; | |
2228 | if (from[1] != 'h' && from[1] != 'H') | |
2229 | continue; | |
2230 | if (from[2] != 'r' && from[2] != 'R') | |
2231 | continue; | |
2232 | if (from[3] == 'u' || from[3] == 'U') | |
2233 | return true; | |
2234 | if (from[3] != 'o' && from[3] != 'O') | |
2235 | continue; | |
2236 | if (from[4] != 'u' && from[4] != 'U') | |
2237 | continue; | |
2238 | if (from[5] != 'g' && from[5] != 'G') | |
2239 | continue; | |
2240 | if (from[6] != 'h' && from[6] != 'H') | |
2241 | continue; | |
2242 | return true; | |
2243 | } | |
2244 | return false; | |
2245 | case 3: | |
2246 | case 4: | |
2247 | break; | |
2248 | } | |
2249 | ||
81fea426 MP |
2250 | /* Whole comment contents: |
2251 | -fallthrough | |
2252 | @fallthrough@ | |
2253 | */ | |
2254 | if (*from == '-' || *from == '@') | |
2255 | { | |
2256 | size_t len = sizeof "fallthrough" - 1; | |
2257 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2258 | return false; | |
2259 | if (memcmp (from + 1, "fallthrough", len)) | |
2260 | return false; | |
2261 | if (*from == '@') | |
2262 | { | |
2263 | if (from[len + 1] != '@') | |
2264 | return false; | |
2265 | len++; | |
2266 | } | |
2267 | from += 1 + len; | |
2268 | } | |
2269 | /* Whole comment contents (regex): | |
70f6d5e1 | 2270 | lint -fallthrough[ \t]* |
81b02905 JJ |
2271 | */ |
2272 | else if (*from == 'l') | |
2273 | { | |
2274 | size_t len = sizeof "int -fallthrough" - 1; | |
2275 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2276 | return false; | |
2277 | if (memcmp (from + 1, "int -fallthrough", len)) | |
70f6d5e1 | 2278 | return false; |
81b02905 | 2279 | from += 1 + len; |
70f6d5e1 JJ |
2280 | while (*from == ' ' || *from == '\t') |
2281 | from++; | |
2282 | } | |
2283 | /* Whole comment contents (regex): | |
2284 | [ \t]*FALLTHR(U|OUGH)[ \t]* | |
2285 | */ | |
2286 | else if (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough) == 4) | |
2287 | { | |
2288 | while (*from == ' ' || *from == '\t') | |
2289 | from++; | |
2290 | if ((size_t) (pfile->buffer->cur - from) < sizeof "FALLTHRU" - 1) | |
2291 | return false; | |
2292 | if (memcmp (from, "FALLTHR", sizeof "FALLTHR" - 1)) | |
2293 | return false; | |
2294 | from += sizeof "FALLTHR" - 1; | |
2295 | if (*from == 'U') | |
2296 | from++; | |
2297 | else if ((size_t) (pfile->buffer->cur - from) < sizeof "OUGH" - 1) | |
2298 | return false; | |
2299 | else if (memcmp (from, "OUGH", sizeof "OUGH" - 1)) | |
2300 | return false; | |
2301 | else | |
2302 | from += sizeof "OUGH" - 1; | |
2303 | while (*from == ' ' || *from == '\t') | |
2304 | from++; | |
81b02905 JJ |
2305 | } |
2306 | /* Whole comment contents (regex): | |
ee19ef45 JJ |
2307 | [ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)? |
2308 | [ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
2309 | [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
81fea426 MP |
2310 | */ |
2311 | else | |
2312 | { | |
81b02905 | 2313 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 MP |
2314 | from++; |
2315 | unsigned char f = *from; | |
81b02905 JJ |
2316 | bool all_upper = false; |
2317 | if (f == 'E' || f == 'e') | |
70f6d5e1 JJ |
2318 | { |
2319 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2320 | < sizeof "else fallthru" - 1) |
2321 | return false; | |
ee19ef45 | 2322 | if (f == 'E' && memcmp (from + 1, "LSE", sizeof "LSE" - 1) == 0) |
81b02905 | 2323 | all_upper = true; |
ee19ef45 JJ |
2324 | else if (memcmp (from + 1, "lse", sizeof "lse" - 1)) |
2325 | return false; | |
2326 | from += sizeof "else" - 1; | |
2327 | if (*from == ',') | |
2328 | from++; | |
70f6d5e1 | 2329 | if (*from != ' ') |
ee19ef45 JJ |
2330 | return false; |
2331 | from++; | |
2332 | if (all_upper && *from == 'f') | |
81b02905 | 2333 | return false; |
81b02905 JJ |
2334 | if (f == 'e' && *from == 'F') |
2335 | return false; | |
2336 | f = *from; | |
70f6d5e1 | 2337 | } |
81b02905 | 2338 | else if (f == 'I' || f == 'i') |
70f6d5e1 JJ |
2339 | { |
2340 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2341 | < sizeof "intentional fallthru" - 1) |
2342 | return false; | |
2343 | if (f == 'I' && memcmp (from + 1, "NTENTIONAL", | |
2344 | sizeof "NTENTIONAL" - 1) == 0) | |
2345 | all_upper = true; | |
2346 | else if (memcmp (from + 1, "ntentional", | |
2347 | sizeof "ntentional" - 1)) | |
2348 | return false; | |
2349 | from += sizeof "intentional" - 1; | |
2350 | if (*from == ' ') | |
2351 | { | |
2352 | from++; | |
2353 | if (all_upper && *from == 'f') | |
2354 | return false; | |
2355 | } | |
2356 | else if (all_upper) | |
2357 | { | |
2358 | if (memcmp (from, "LY F", sizeof "LY F" - 1)) | |
2359 | return false; | |
2360 | from += sizeof "LY " - 1; | |
2361 | } | |
2362 | else | |
2363 | { | |
2364 | if (memcmp (from, "ly ", sizeof "ly " - 1)) | |
2365 | return false; | |
2366 | from += sizeof "ly " - 1; | |
2367 | } | |
2368 | if (f == 'i' && *from == 'F') | |
2369 | return false; | |
2370 | f = *from; | |
70f6d5e1 | 2371 | } |
81fea426 MP |
2372 | if (f != 'F' && f != 'f') |
2373 | return false; | |
7bad794a | 2374 | if ((size_t) (pfile->buffer->cur - from) < sizeof "fallthru" - 1) |
81fea426 | 2375 | return false; |
81fea426 MP |
2376 | if (f == 'F' && memcmp (from + 1, "ALL", sizeof "ALL" - 1) == 0) |
2377 | all_upper = true; | |
81b02905 JJ |
2378 | else if (all_upper) |
2379 | return false; | |
81fea426 MP |
2380 | else if (memcmp (from + 1, "all", sizeof "all" - 1)) |
2381 | return false; | |
7bad794a JJ |
2382 | from += sizeof "fall" - 1; |
2383 | if (*from == (all_upper ? 'S' : 's') && from[1] == ' ') | |
2384 | from += 2; | |
2385 | else if (*from == ' ' || *from == '-') | |
2386 | from++; | |
2387 | else if (*from != (all_upper ? 'T' : 't')) | |
81fea426 | 2388 | return false; |
81fea426 MP |
2389 | if ((f == 'f' || *from != 'T') && (all_upper || *from != 't')) |
2390 | return false; | |
7bad794a | 2391 | if ((size_t) (pfile->buffer->cur - from) < sizeof "thru" - 1) |
81fea426 MP |
2392 | return false; |
2393 | if (memcmp (from + 1, all_upper ? "HRU" : "hru", sizeof "hru" - 1)) | |
2394 | { | |
7bad794a | 2395 | if ((size_t) (pfile->buffer->cur - from) < sizeof "through" - 1) |
81fea426 MP |
2396 | return false; |
2397 | if (memcmp (from + 1, all_upper ? "HROUGH" : "hrough", | |
2398 | sizeof "hrough" - 1)) | |
2399 | return false; | |
2400 | from += sizeof "through" - 1; | |
2401 | } | |
2402 | else | |
2403 | from += sizeof "thru" - 1; | |
81b02905 | 2404 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 | 2405 | from++; |
81b02905 JJ |
2406 | if (*from == '-') |
2407 | { | |
2408 | from++; | |
2409 | if (*comment_start == '*') | |
2410 | { | |
2411 | do | |
2412 | { | |
2413 | while (*from && *from != '*' | |
2414 | && *from != '\n' && *from != '\r') | |
2415 | from++; | |
2416 | if (*from != '*' || from[1] == '/') | |
2417 | break; | |
2418 | from++; | |
2419 | } | |
2420 | while (1); | |
2421 | } | |
2422 | else | |
2423 | while (*from && *from != '\n' && *from != '\r') | |
2424 | from++; | |
2425 | } | |
81fea426 MP |
2426 | } |
2427 | /* C block comment. */ | |
2428 | if (*comment_start == '*') | |
2429 | { | |
2430 | if (*from != '*' || from[1] != '/') | |
2431 | return false; | |
2432 | } | |
2433 | /* C++ line comment. */ | |
2434 | else if (*from != '\n') | |
2435 | return false; | |
2436 | ||
2437 | return true; | |
2438 | } | |
2439 | ||
5fddcffc NB |
2440 | /* Allocate COUNT tokens for RUN. */ |
2441 | void | |
6cf87ca4 | 2442 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
5fddcffc | 2443 | { |
72bb2c39 | 2444 | run->base = XNEWVEC (cpp_token, count); |
5fddcffc NB |
2445 | run->limit = run->base + count; |
2446 | run->next = NULL; | |
2447 | } | |
2448 | ||
2449 | /* Returns the next tokenrun, or creates one if there is none. */ | |
2450 | static tokenrun * | |
6cf87ca4 | 2451 | next_tokenrun (tokenrun *run) |
5fddcffc NB |
2452 | { |
2453 | if (run->next == NULL) | |
2454 | { | |
72bb2c39 | 2455 | run->next = XNEW (tokenrun); |
bdcbe496 | 2456 | run->next->prev = run; |
5fddcffc NB |
2457 | _cpp_init_tokenrun (run->next, 250); |
2458 | } | |
2459 | ||
2460 | return run->next; | |
2461 | } | |
2462 | ||
ad2305ad | 2463 | /* Return the number of not yet processed token in a given |
92582b75 TT |
2464 | context. */ |
2465 | int | |
ad2305ad | 2466 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
92582b75 | 2467 | { |
92582b75 | 2468 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
cbbcf655 | 2469 | return (LAST (context).token - FIRST (context).token); |
92582b75 TT |
2470 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
2471 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
cbbcf655 | 2472 | return (LAST (context).ptoken - FIRST (context).ptoken); |
92582b75 TT |
2473 | else |
2474 | abort (); | |
2475 | } | |
2476 | ||
ad2305ad DS |
2477 | /* Returns the token present at index INDEX in a given context. If |
2478 | INDEX is zero, the next token to be processed is returned. */ | |
92582b75 | 2479 | static const cpp_token* |
ad2305ad | 2480 | _cpp_token_from_context_at (cpp_context *context, int index) |
92582b75 | 2481 | { |
92582b75 TT |
2482 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
2483 | return &(FIRST (context).token[index]); | |
2484 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
2485 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
2486 | return FIRST (context).ptoken[index]; | |
2487 | else | |
2488 | abort (); | |
2489 | } | |
2490 | ||
5950c3c9 BE |
2491 | /* Look ahead in the input stream. */ |
2492 | const cpp_token * | |
2493 | cpp_peek_token (cpp_reader *pfile, int index) | |
2494 | { | |
2495 | cpp_context *context = pfile->context; | |
2496 | const cpp_token *peektok; | |
2497 | int count; | |
2498 | ||
2499 | /* First, scan through any pending cpp_context objects. */ | |
2500 | while (context->prev) | |
2501 | { | |
ad2305ad | 2502 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
5950c3c9 BE |
2503 | |
2504 | if (index < (int) sz) | |
ad2305ad | 2505 | return _cpp_token_from_context_at (context, index); |
5950c3c9 BE |
2506 | index -= (int) sz; |
2507 | context = context->prev; | |
2508 | } | |
2509 | ||
2510 | /* We will have to read some new tokens after all (and do so | |
2511 | without invalidating preceding tokens). */ | |
2512 | count = index; | |
2513 | pfile->keep_tokens++; | |
2514 | ||
b8cd77f4 JJ |
2515 | /* For peeked tokens temporarily disable line_change reporting, |
2516 | until the tokens are parsed for real. */ | |
2517 | void (*line_change) (cpp_reader *, const cpp_token *, int) | |
2518 | = pfile->cb.line_change; | |
2519 | pfile->cb.line_change = NULL; | |
2520 | ||
5950c3c9 BE |
2521 | do |
2522 | { | |
2523 | peektok = _cpp_lex_token (pfile); | |
2524 | if (peektok->type == CPP_EOF) | |
e4b33ee5 JJ |
2525 | { |
2526 | index--; | |
2527 | break; | |
2528 | } | |
5950c3c9 BE |
2529 | } |
2530 | while (index--); | |
2531 | ||
e4b33ee5 | 2532 | _cpp_backup_tokens_direct (pfile, count - index); |
5950c3c9 | 2533 | pfile->keep_tokens--; |
b8cd77f4 | 2534 | pfile->cb.line_change = line_change; |
5950c3c9 BE |
2535 | |
2536 | return peektok; | |
2537 | } | |
2538 | ||
4ed5bcfb NB |
2539 | /* Allocate a single token that is invalidated at the same time as the |
2540 | rest of the tokens on the line. Has its line and col set to the | |
2541 | same as the last lexed token, so that diagnostics appear in the | |
2542 | right place. */ | |
2543 | cpp_token * | |
6cf87ca4 | 2544 | _cpp_temp_token (cpp_reader *pfile) |
4ed5bcfb NB |
2545 | { |
2546 | cpp_token *old, *result; | |
5950c3c9 BE |
2547 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
2548 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
4ed5bcfb NB |
2549 | |
2550 | old = pfile->cur_token - 1; | |
5950c3c9 BE |
2551 | /* Any pre-existing lookaheads must not be clobbered. */ |
2552 | if (la) | |
2553 | { | |
2554 | if (sz <= la) | |
2555 | { | |
2556 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
2557 | ||
2558 | if (sz < la) | |
2559 | memmove (next->base + 1, next->base, | |
2560 | (la - sz) * sizeof (cpp_token)); | |
2561 | ||
2562 | next->base[0] = pfile->cur_run->limit[-1]; | |
2563 | } | |
2564 | ||
2565 | if (sz > 1) | |
2566 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
2567 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
2568 | } | |
2569 | ||
2570 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
4ed5bcfb NB |
2571 | { |
2572 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
2573 | pfile->cur_token = pfile->cur_run->base; | |
2574 | } | |
2575 | ||
2576 | result = pfile->cur_token++; | |
12f9df4e | 2577 | result->src_loc = old->src_loc; |
4ed5bcfb NB |
2578 | return result; |
2579 | } | |
2580 | ||
14baae01 NB |
2581 | /* Lex a token into RESULT (external interface). Takes care of issues |
2582 | like directive handling, token lookahead, multiple include | |
a1f300c0 | 2583 | optimization and skipping. */ |
345894b4 | 2584 | const cpp_token * |
6cf87ca4 | 2585 | _cpp_lex_token (cpp_reader *pfile) |
5fddcffc | 2586 | { |
bdcbe496 | 2587 | cpp_token *result; |
5fddcffc | 2588 | |
bdcbe496 | 2589 | for (;;) |
5fddcffc | 2590 | { |
bdcbe496 | 2591 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 2592 | { |
bdcbe496 NB |
2593 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
2594 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc | 2595 | } |
ee380365 TT |
2596 | /* We assume that the current token is somewhere in the current |
2597 | run. */ | |
2598 | if (pfile->cur_token < pfile->cur_run->base | |
2599 | || pfile->cur_token >= pfile->cur_run->limit) | |
2600 | abort (); | |
5fddcffc | 2601 | |
bdcbe496 | 2602 | if (pfile->lookaheads) |
14baae01 NB |
2603 | { |
2604 | pfile->lookaheads--; | |
2605 | result = pfile->cur_token++; | |
2606 | } | |
bdcbe496 | 2607 | else |
14baae01 | 2608 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
2609 | |
2610 | if (result->flags & BOL) | |
5fddcffc | 2611 | { |
bdcbe496 NB |
2612 | /* Is this a directive. If _cpp_handle_directive returns |
2613 | false, it is an assembler #. */ | |
2614 | if (result->type == CPP_HASH | |
e808ec9c NB |
2615 | /* 6.10.3 p 11: Directives in a list of macro arguments |
2616 | gives undefined behavior. This implementation | |
2617 | handles the directive as normal. */ | |
bc4071dd | 2618 | && pfile->state.parsing_args != 1) |
21b11495 | 2619 | { |
bc4071dd | 2620 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
21b11495 | 2621 | { |
bc4071dd RH |
2622 | if (pfile->directive_result.type == CPP_PADDING) |
2623 | continue; | |
21b11495 | 2624 | result = &pfile->directive_result; |
21b11495 ZW |
2625 | } |
2626 | } | |
bc4071dd RH |
2627 | else if (pfile->state.in_deferred_pragma) |
2628 | result = &pfile->directive_result; | |
21b11495 | 2629 | |
97293897 | 2630 | if (pfile->cb.line_change && !pfile->state.skipping) |
6cf87ca4 | 2631 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
5fddcffc | 2632 | } |
5fddcffc | 2633 | |
bdcbe496 | 2634 | /* We don't skip tokens in directives. */ |
bc4071dd | 2635 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
bdcbe496 | 2636 | break; |
5fddcffc | 2637 | |
bdcbe496 | 2638 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 2639 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
6356f892 | 2640 | get here and MI optimization works. */ |
5fddcffc | 2641 | pfile->mi_valid = false; |
bdcbe496 NB |
2642 | |
2643 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
2644 | break; | |
5fddcffc NB |
2645 | } |
2646 | ||
345894b4 | 2647 | return result; |
5fddcffc NB |
2648 | } |
2649 | ||
26aea073 NB |
2650 | /* Returns true if a fresh line has been loaded. */ |
2651 | bool | |
6cf87ca4 | 2652 | _cpp_get_fresh_line (cpp_reader *pfile) |
004cb263 | 2653 | { |
22234f56 PB |
2654 | int return_at_eof; |
2655 | ||
26aea073 NB |
2656 | /* We can't get a new line until we leave the current directive. */ |
2657 | if (pfile->state.in_directive) | |
2658 | return false; | |
df383483 | 2659 | |
26aea073 | 2660 | for (;;) |
1a76916c | 2661 | { |
26aea073 | 2662 | cpp_buffer *buffer = pfile->buffer; |
1a76916c | 2663 | |
26aea073 NB |
2664 | if (!buffer->need_line) |
2665 | return true; | |
2666 | ||
2667 | if (buffer->next_line < buffer->rlimit) | |
004cb263 | 2668 | { |
26aea073 NB |
2669 | _cpp_clean_line (pfile); |
2670 | return true; | |
2671 | } | |
004cb263 | 2672 | |
26aea073 NB |
2673 | /* First, get out of parsing arguments state. */ |
2674 | if (pfile->state.parsing_args) | |
2675 | return false; | |
2676 | ||
2677 | /* End of buffer. Non-empty files should end in a newline. */ | |
2678 | if (buffer->buf != buffer->rlimit | |
2679 | && buffer->next_line > buffer->rlimit | |
2680 | && !buffer->from_stage3) | |
2681 | { | |
ed0e74e0 | 2682 | /* Clip to buffer size. */ |
26aea073 | 2683 | buffer->next_line = buffer->rlimit; |
26aea073 | 2684 | } |
22234f56 PB |
2685 | |
2686 | return_at_eof = buffer->return_at_eof; | |
26aea073 | 2687 | _cpp_pop_buffer (pfile); |
22234f56 | 2688 | if (pfile->buffer == NULL || return_at_eof) |
a506c55c | 2689 | return false; |
26aea073 | 2690 | } |
004cb263 NB |
2691 | } |
2692 | ||
6f572ac2 NB |
2693 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
2694 | do \ | |
2695 | { \ | |
2696 | result->type = ELSE_TYPE; \ | |
2697 | if (*buffer->cur == CHAR) \ | |
2698 | buffer->cur++, result->type = THEN_TYPE; \ | |
2699 | } \ | |
2700 | while (0) | |
480709cc | 2701 | |
14baae01 NB |
2702 | /* Lex a token into pfile->cur_token, which is also incremented, to |
2703 | get diagnostics pointing to the correct location. | |
2704 | ||
2705 | Does not handle issues such as token lookahead, multiple-include | |
f1ba665b | 2706 | optimization, directives, skipping etc. This function is only |
14baae01 NB |
2707 | suitable for use by _cpp_lex_token, and in special cases like |
2708 | lex_expansion_token which doesn't care for any of these issues. | |
2709 | ||
2710 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
2711 | otherwise returns to the start of the token buffer if permissible. | |
2712 | Returns the location of the lexed token. */ | |
2713 | cpp_token * | |
6cf87ca4 | 2714 | _cpp_lex_direct (cpp_reader *pfile) |
45b966db | 2715 | { |
0d9f234d | 2716 | cppchar_t c; |
adb84b42 | 2717 | cpp_buffer *buffer; |
7bad794a JJ |
2718 | const unsigned char *comment_start; |
2719 | bool fallthrough_comment = false; | |
14baae01 | 2720 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 2721 | |
5fddcffc | 2722 | fresh_line: |
26aea073 | 2723 | result->flags = 0; |
2be570f9 | 2724 | buffer = pfile->buffer; |
a506c55c | 2725 | if (buffer->need_line) |
26aea073 | 2726 | { |
bc4071dd RH |
2727 | if (pfile->state.in_deferred_pragma) |
2728 | { | |
2729 | result->type = CPP_PRAGMA_EOL; | |
2730 | pfile->state.in_deferred_pragma = false; | |
2731 | if (!pfile->state.pragma_allow_expansion) | |
2732 | pfile->state.prevent_expansion--; | |
2733 | return result; | |
2734 | } | |
26aea073 NB |
2735 | if (!_cpp_get_fresh_line (pfile)) |
2736 | { | |
2737 | result->type = CPP_EOF; | |
9ff7868d NB |
2738 | if (!pfile->state.in_directive) |
2739 | { | |
2740 | /* Tell the compiler the line number of the EOF token. */ | |
500bee0a | 2741 | result->src_loc = pfile->line_table->highest_line; |
9ff7868d NB |
2742 | result->flags = BOL; |
2743 | } | |
26aea073 NB |
2744 | return result; |
2745 | } | |
81fea426 | 2746 | if (buffer != pfile->buffer) |
7bad794a | 2747 | fallthrough_comment = false; |
26aea073 NB |
2748 | if (!pfile->keep_tokens) |
2749 | { | |
2750 | pfile->cur_run = &pfile->base_run; | |
2751 | result = pfile->base_run.base; | |
2752 | pfile->cur_token = result + 1; | |
2753 | } | |
2754 | result->flags = BOL; | |
2755 | if (pfile->state.parsing_args == 2) | |
2756 | result->flags |= PREV_WHITE; | |
2757 | } | |
a506c55c | 2758 | buffer = pfile->buffer; |
5fddcffc | 2759 | update_tokens_line: |
500bee0a | 2760 | result->src_loc = pfile->line_table->highest_line; |
041c3194 | 2761 | |
5fddcffc | 2762 | skipped_white: |
26aea073 NB |
2763 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
2764 | && !pfile->overlaid_buffer) | |
2765 | { | |
2766 | _cpp_process_line_notes (pfile, false); | |
500bee0a | 2767 | result->src_loc = pfile->line_table->highest_line; |
26aea073 | 2768 | } |
480709cc | 2769 | c = *buffer->cur++; |
12f9df4e | 2770 | |
f3f6029d NS |
2771 | if (pfile->forced_token_location) |
2772 | result->src_loc = pfile->forced_token_location; | |
e3dfef44 GC |
2773 | else |
2774 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
2775 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
5fddcffc | 2776 | |
0d9f234d | 2777 | switch (c) |
45b966db | 2778 | { |
4d6baafa NB |
2779 | case ' ': case '\t': case '\f': case '\v': case '\0': |
2780 | result->flags |= PREV_WHITE; | |
26aea073 NB |
2781 | skip_whitespace (pfile, c); |
2782 | goto skipped_white; | |
0d9f234d | 2783 | |
26aea073 | 2784 | case '\n': |
056f95ec NS |
2785 | /* Increment the line, unless this is the last line ... */ |
2786 | if (buffer->cur < buffer->rlimit | |
2787 | /* ... or this is a #include, (where _cpp_stack_file needs to | |
2788 | unwind by one line) ... */ | |
2789 | || (pfile->state.in_directive > 1 | |
2790 | /* ... except traditional-cpp increments this elsewhere. */ | |
2791 | && !CPP_OPTION (pfile, traditional))) | |
12f9df4e | 2792 | CPP_INCREMENT_LINE (pfile, 0); |
26aea073 NB |
2793 | buffer->need_line = true; |
2794 | goto fresh_line; | |
46d07497 | 2795 | |
0d9f234d NB |
2796 | case '0': case '1': case '2': case '3': case '4': |
2797 | case '5': case '6': case '7': case '8': case '9': | |
50668cf6 GK |
2798 | { |
2799 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2800 | result->type = CPP_NUMBER; | |
2801 | lex_number (pfile, &result->val.str, &nst); | |
2802 | warn_about_normalization (pfile, result, &nst); | |
2803 | break; | |
2804 | } | |
46d07497 | 2805 | |
0abc6a6a | 2806 | case 'L': |
b6baa67d KVH |
2807 | case 'u': |
2808 | case 'U': | |
2c6e3f55 JJ |
2809 | case 'R': |
2810 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2811 | wide strings or raw strings. */ | |
a48e3dd1 JM |
2812 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
2813 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
bced6edf | 2814 | { |
2c6e3f55 JJ |
2815 | if ((*buffer->cur == '\'' && c != 'R') |
2816 | || *buffer->cur == '"' | |
2817 | || (*buffer->cur == 'R' | |
2818 | && c != 'R' | |
2819 | && buffer->cur[1] == '"' | |
a48e3dd1 | 2820 | && CPP_OPTION (pfile, rliterals)) |
2c6e3f55 JJ |
2821 | || (*buffer->cur == '8' |
2822 | && c == 'u' | |
fe95b036 ESR |
2823 | && ((buffer->cur[1] == '"' || (buffer->cur[1] == '\'' |
2824 | && CPP_OPTION (pfile, utf8_char_literals))) | |
a48e3dd1 JM |
2825 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
2826 | && CPP_OPTION (pfile, rliterals))))) | |
b6baa67d KVH |
2827 | { |
2828 | lex_string (pfile, result, buffer->cur - 1); | |
2829 | break; | |
2830 | } | |
bced6edf | 2831 | } |
df383483 | 2832 | /* Fall through. */ |
0abc6a6a | 2833 | |
0d9f234d NB |
2834 | case '_': |
2835 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2836 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2837 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
b6baa67d | 2838 | case 's': case 't': case 'v': case 'w': case 'x': |
0d9f234d NB |
2839 | case 'y': case 'z': |
2840 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
0abc6a6a | 2841 | case 'G': case 'H': case 'I': case 'J': case 'K': |
2c6e3f55 | 2842 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
b6baa67d | 2843 | case 'S': case 'T': case 'V': case 'W': case 'X': |
0d9f234d NB |
2844 | case 'Y': case 'Z': |
2845 | result->type = CPP_NAME; | |
50668cf6 GK |
2846 | { |
2847 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
9a0c6187 | 2848 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
be5ffc59 JM |
2849 | &nst, |
2850 | &result->val.node.spelling); | |
50668cf6 GK |
2851 | warn_about_normalization (pfile, result, &nst); |
2852 | } | |
0d9f234d | 2853 | |
0d9f234d | 2854 | /* Convert named operators to their proper types. */ |
9a0c6187 | 2855 | if (result->val.node.node->flags & NODE_OPERATOR) |
0d9f234d NB |
2856 | { |
2857 | result->flags |= NAMED_OP; | |
9a0c6187 | 2858 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
0d9f234d | 2859 | } |
81fea426 MP |
2860 | |
2861 | /* Signal FALLTHROUGH comment followed by another token. */ | |
7bad794a | 2862 | if (fallthrough_comment) |
81fea426 | 2863 | result->flags |= PREV_FALLTHROUGH; |
0d9f234d NB |
2864 | break; |
2865 | ||
2866 | case '\'': | |
2867 | case '"': | |
6338b358 | 2868 | lex_string (pfile, result, buffer->cur - 1); |
0d9f234d | 2869 | break; |
041c3194 | 2870 | |
0d9f234d | 2871 | case '/': |
1c6d33ef NB |
2872 | /* A potential block or line comment. */ |
2873 | comment_start = buffer->cur; | |
6f572ac2 NB |
2874 | c = *buffer->cur; |
2875 | ||
1c6d33ef NB |
2876 | if (c == '*') |
2877 | { | |
26aea073 | 2878 | if (_cpp_skip_block_comment (pfile)) |
0527bc4e | 2879 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
0d9f234d | 2880 | } |
909eb89c | 2881 | else if (c == '/' && ! CPP_OPTION (pfile, traditional)) |
0d9f234d | 2882 | { |
909eb89c MP |
2883 | /* Don't warn for system headers. */ |
2884 | if (cpp_in_system_header (pfile)) | |
2885 | ; | |
f3bede71 | 2886 | /* Warn about comments if pedantically GNUC89, and not |
bdb05a7b | 2887 | in system headers. */ |
909eb89c MP |
2888 | else if (CPP_OPTION (pfile, lang) == CLK_GNUC89 |
2889 | && CPP_PEDANTIC (pfile) | |
2890 | && ! buffer->warned_cplusplus_comments) | |
041c3194 | 2891 | { |
0c86a39d JJ |
2892 | if (cpp_error (pfile, CPP_DL_PEDWARN, |
2893 | "C++ style comments are not allowed in ISO C90")) | |
2894 | cpp_error (pfile, CPP_DL_NOTE, | |
2895 | "(this will be reported only once per input file)"); | |
1c6d33ef NB |
2896 | buffer->warned_cplusplus_comments = 1; |
2897 | } | |
f3bede71 | 2898 | /* Or if specifically desired via -Wc90-c99-compat. */ |
177cce46 | 2899 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
dd3ff077 | 2900 | && ! CPP_OPTION (pfile, cplusplus) |
f3bede71 MP |
2901 | && ! buffer->warned_cplusplus_comments) |
2902 | { | |
0c86a39d JJ |
2903 | if (cpp_error (pfile, CPP_DL_WARNING, |
2904 | "C++ style comments are incompatible with C90")) | |
2905 | cpp_error (pfile, CPP_DL_NOTE, | |
2906 | "(this will be reported only once per input file)"); | |
f3bede71 MP |
2907 | buffer->warned_cplusplus_comments = 1; |
2908 | } | |
909eb89c MP |
2909 | /* In C89/C94, C++ style comments are forbidden. */ |
2910 | else if ((CPP_OPTION (pfile, lang) == CLK_STDC89 | |
2911 | || CPP_OPTION (pfile, lang) == CLK_STDC94)) | |
2912 | { | |
2913 | /* But don't be confused about valid code such as | |
2914 | - // immediately followed by *, | |
2915 | - // in a preprocessing directive, | |
2916 | - // in an #if 0 block. */ | |
2917 | if (buffer->cur[1] == '*' | |
2918 | || pfile->state.in_directive | |
2919 | || pfile->state.skipping) | |
2920 | { | |
2921 | result->type = CPP_DIV; | |
2922 | break; | |
2923 | } | |
2924 | else if (! buffer->warned_cplusplus_comments) | |
2925 | { | |
0c86a39d JJ |
2926 | if (cpp_error (pfile, CPP_DL_ERROR, |
2927 | "C++ style comments are not allowed in " | |
2928 | "ISO C90")) | |
2929 | cpp_error (pfile, CPP_DL_NOTE, | |
2930 | "(this will be reported only once per input " | |
2931 | "file)"); | |
909eb89c MP |
2932 | buffer->warned_cplusplus_comments = 1; |
2933 | } | |
2934 | } | |
01ef6563 | 2935 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
87cf0651 | 2936 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
1c6d33ef | 2937 | } |
480709cc NB |
2938 | else if (c == '=') |
2939 | { | |
6f572ac2 | 2940 | buffer->cur++; |
480709cc NB |
2941 | result->type = CPP_DIV_EQ; |
2942 | break; | |
2943 | } | |
2944 | else | |
2945 | { | |
480709cc NB |
2946 | result->type = CPP_DIV; |
2947 | break; | |
2948 | } | |
0d9f234d | 2949 | |
7bad794a JJ |
2950 | if (fallthrough_comment_p (pfile, comment_start)) |
2951 | fallthrough_comment = true; | |
2952 | ||
05945a1b DM |
2953 | if (pfile->cb.comment) |
2954 | { | |
2955 | size_t len = pfile->buffer->cur - comment_start; | |
2956 | pfile->cb.comment (pfile, result->src_loc, comment_start - 1, | |
2957 | len + 1); | |
2958 | } | |
2959 | ||
1c6d33ef NB |
2960 | if (!pfile->state.save_comments) |
2961 | { | |
2962 | result->flags |= PREV_WHITE; | |
5fddcffc | 2963 | goto update_tokens_line; |
0d9f234d | 2964 | } |
1c6d33ef | 2965 | |
7bad794a | 2966 | if (fallthrough_comment) |
81fea426 MP |
2967 | result->flags |= PREV_FALLTHROUGH; |
2968 | ||
1c6d33ef | 2969 | /* Save the comment as a token in its own right. */ |
477cdac7 | 2970 | save_comment (pfile, result, comment_start, c); |
bdcbe496 | 2971 | break; |
0d9f234d NB |
2972 | |
2973 | case '<': | |
2974 | if (pfile->state.angled_headers) | |
2975 | { | |
6338b358 | 2976 | lex_string (pfile, result, buffer->cur - 1); |
4bb09c26 JM |
2977 | if (result->type != CPP_LESS) |
2978 | break; | |
0d9f234d | 2979 | } |
45b966db | 2980 | |
6f572ac2 NB |
2981 | result->type = CPP_LESS; |
2982 | if (*buffer->cur == '=') | |
2983 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2984 | else if (*buffer->cur == '<') | |
0d9f234d | 2985 | { |
6f572ac2 NB |
2986 | buffer->cur++; |
2987 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
0d9f234d | 2988 | } |
6f572ac2 | 2989 | else if (CPP_OPTION (pfile, digraphs)) |
480709cc | 2990 | { |
6f572ac2 NB |
2991 | if (*buffer->cur == ':') |
2992 | { | |
1582c677 PC |
2993 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
2994 | three characters are <:: and the subsequent character | |
2995 | is neither : nor >, the < is treated as a preprocessor | |
2996 | token by itself". */ | |
2997 | if (CPP_OPTION (pfile, cplusplus) | |
61949153 PC |
2998 | && CPP_OPTION (pfile, lang) != CLK_CXX98 |
2999 | && CPP_OPTION (pfile, lang) != CLK_GNUCXX | |
1582c677 PC |
3000 | && buffer->cur[1] == ':' |
3001 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
3002 | break; | |
3003 | ||
6f572ac2 NB |
3004 | buffer->cur++; |
3005 | result->flags |= DIGRAPH; | |
3006 | result->type = CPP_OPEN_SQUARE; | |
3007 | } | |
3008 | else if (*buffer->cur == '%') | |
3009 | { | |
3010 | buffer->cur++; | |
3011 | result->flags |= DIGRAPH; | |
3012 | result->type = CPP_OPEN_BRACE; | |
3013 | } | |
480709cc | 3014 | } |
0d9f234d NB |
3015 | break; |
3016 | ||
3017 | case '>': | |
6f572ac2 NB |
3018 | result->type = CPP_GREATER; |
3019 | if (*buffer->cur == '=') | |
3020 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
3021 | else if (*buffer->cur == '>') | |
0d9f234d | 3022 | { |
6f572ac2 NB |
3023 | buffer->cur++; |
3024 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
3025 | } | |
0d9f234d NB |
3026 | break; |
3027 | ||
cbcff6df | 3028 | case '%': |
6f572ac2 NB |
3029 | result->type = CPP_MOD; |
3030 | if (*buffer->cur == '=') | |
3031 | buffer->cur++, result->type = CPP_MOD_EQ; | |
3032 | else if (CPP_OPTION (pfile, digraphs)) | |
480709cc | 3033 | { |
6f572ac2 | 3034 | if (*buffer->cur == ':') |
480709cc | 3035 | { |
6f572ac2 NB |
3036 | buffer->cur++; |
3037 | result->flags |= DIGRAPH; | |
3038 | result->type = CPP_HASH; | |
3039 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
9a0c6187 | 3040 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
6f572ac2 NB |
3041 | } |
3042 | else if (*buffer->cur == '>') | |
3043 | { | |
3044 | buffer->cur++; | |
3045 | result->flags |= DIGRAPH; | |
3046 | result->type = CPP_CLOSE_BRACE; | |
480709cc | 3047 | } |
480709cc | 3048 | } |
0d9f234d NB |
3049 | break; |
3050 | ||
cbcff6df | 3051 | case '.': |
480709cc | 3052 | result->type = CPP_DOT; |
6f572ac2 | 3053 | if (ISDIGIT (*buffer->cur)) |
480709cc | 3054 | { |
50668cf6 | 3055 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
480709cc | 3056 | result->type = CPP_NUMBER; |
50668cf6 GK |
3057 | lex_number (pfile, &result->val.str, &nst); |
3058 | warn_about_normalization (pfile, result, &nst); | |
480709cc | 3059 | } |
6f572ac2 NB |
3060 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
3061 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
3062 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
3063 | buffer->cur++, result->type = CPP_DOT_STAR; | |
0d9f234d | 3064 | break; |
45b966db | 3065 | |
0d9f234d | 3066 | case '+': |
6f572ac2 NB |
3067 | result->type = CPP_PLUS; |
3068 | if (*buffer->cur == '+') | |
3069 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
3070 | else if (*buffer->cur == '=') | |
3071 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
0d9f234d | 3072 | break; |
04e3ec78 | 3073 | |
0d9f234d | 3074 | case '-': |
6f572ac2 NB |
3075 | result->type = CPP_MINUS; |
3076 | if (*buffer->cur == '>') | |
0d9f234d | 3077 | { |
6f572ac2 | 3078 | buffer->cur++; |
480709cc | 3079 | result->type = CPP_DEREF; |
6f572ac2 NB |
3080 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
3081 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
480709cc | 3082 | } |
6f572ac2 NB |
3083 | else if (*buffer->cur == '-') |
3084 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
3085 | else if (*buffer->cur == '=') | |
3086 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
0d9f234d | 3087 | break; |
45b966db | 3088 | |
0d9f234d | 3089 | case '&': |
6f572ac2 NB |
3090 | result->type = CPP_AND; |
3091 | if (*buffer->cur == '&') | |
3092 | buffer->cur++, result->type = CPP_AND_AND; | |
3093 | else if (*buffer->cur == '=') | |
3094 | buffer->cur++, result->type = CPP_AND_EQ; | |
0d9f234d | 3095 | break; |
df383483 | 3096 | |
0d9f234d | 3097 | case '|': |
6f572ac2 NB |
3098 | result->type = CPP_OR; |
3099 | if (*buffer->cur == '|') | |
3100 | buffer->cur++, result->type = CPP_OR_OR; | |
3101 | else if (*buffer->cur == '=') | |
3102 | buffer->cur++, result->type = CPP_OR_EQ; | |
0d9f234d | 3103 | break; |
45b966db | 3104 | |
0d9f234d | 3105 | case ':': |
6f572ac2 | 3106 | result->type = CPP_COLON; |
93313b94 | 3107 | if (*buffer->cur == ':' && CPP_OPTION (pfile, scope)) |
6f572ac2 NB |
3108 | buffer->cur++, result->type = CPP_SCOPE; |
3109 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
0d9f234d | 3110 | { |
6f572ac2 | 3111 | buffer->cur++; |
0d9f234d | 3112 | result->flags |= DIGRAPH; |
480709cc NB |
3113 | result->type = CPP_CLOSE_SQUARE; |
3114 | } | |
0d9f234d | 3115 | break; |
45b966db | 3116 | |
480709cc NB |
3117 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
3118 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
3119 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
3120 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
9a0c6187 | 3121 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
480709cc | 3122 | |
26aea073 | 3123 | case '?': result->type = CPP_QUERY; break; |
0d9f234d NB |
3124 | case '~': result->type = CPP_COMPL; break; |
3125 | case ',': result->type = CPP_COMMA; break; | |
3126 | case '(': result->type = CPP_OPEN_PAREN; break; | |
3127 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
3128 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
3129 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
3130 | case '{': result->type = CPP_OPEN_BRACE; break; | |
3131 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
3132 | case ';': result->type = CPP_SEMICOLON; break; | |
3133 | ||
40f03658 | 3134 | /* @ is a punctuator in Objective-C. */ |
cc937581 | 3135 | case '@': result->type = CPP_ATSIGN; break; |
0d9f234d | 3136 | |
7d112d66 | 3137 | default: |
1613e52b NB |
3138 | { |
3139 | const uchar *base = --buffer->cur; | |
0abc6a6a | 3140 | |
7d112d66 LH |
3141 | /* Check for an extended identifier ($ or UCN or UTF-8). */ |
3142 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
50668cf6 | 3143 | if (forms_identifier_p (pfile, true, &nst)) |
1613e52b NB |
3144 | { |
3145 | result->type = CPP_NAME; | |
be5ffc59 JM |
3146 | result->val.node.node = lex_identifier (pfile, base, true, &nst, |
3147 | &result->val.node.spelling); | |
50668cf6 | 3148 | warn_about_normalization (pfile, result, &nst); |
1613e52b NB |
3149 | break; |
3150 | } | |
7d112d66 LH |
3151 | |
3152 | /* Otherwise this will form a CPP_OTHER token. Parse valid UTF-8 as a | |
3153 | single token. */ | |
1613e52b | 3154 | buffer->cur++; |
7d112d66 LH |
3155 | if (c >= utf8_signifier) |
3156 | { | |
3157 | const uchar *pstr = base; | |
3158 | cppchar_t s; | |
3159 | if (_cpp_valid_utf8 (pfile, &pstr, buffer->rlimit, 0, NULL, &s)) | |
3160 | buffer->cur = pstr; | |
3161 | } | |
3162 | create_literal (pfile, result, base, buffer->cur - base, CPP_OTHER); | |
3163 | break; | |
1067694a | 3164 | } |
1613e52b | 3165 | |
0d9f234d | 3166 | } |
bdcbe496 | 3167 | |
a3998c2f DM |
3168 | /* Potentially convert the location of the token to a range. */ |
3169 | if (result->src_loc >= RESERVED_LOCATION_COUNT | |
3170 | && result->type != CPP_EOF) | |
3171 | { | |
3172 | /* Ensure that any line notes are processed, so that we have the | |
3173 | correct physical line/column for the end-point of the token even | |
3174 | when a logical line is split via one or more backslashes. */ | |
3175 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos | |
3176 | && !pfile->overlaid_buffer) | |
3177 | _cpp_process_line_notes (pfile, false); | |
3178 | ||
3179 | source_range tok_range; | |
3180 | tok_range.m_start = result->src_loc; | |
3181 | tok_range.m_finish | |
3182 | = linemap_position_for_column (pfile->line_table, | |
3183 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
3184 | ||
3185 | result->src_loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
3186 | result->src_loc, | |
3187 | tok_range, NULL); | |
3188 | } | |
ebedc9a3 | 3189 | |
bdcbe496 | 3190 | return result; |
0d9f234d NB |
3191 | } |
3192 | ||
59325650 NB |
3193 | /* An upper bound on the number of bytes needed to spell TOKEN. |
3194 | Does not include preceding whitespace. */ | |
93c80368 | 3195 | unsigned int |
6cf87ca4 | 3196 | cpp_token_len (const cpp_token *token) |
0d9f234d | 3197 | { |
93c80368 | 3198 | unsigned int len; |
6d2c2047 | 3199 | |
93c80368 | 3200 | switch (TOKEN_SPELL (token)) |
041c3194 | 3201 | { |
cc955282 | 3202 | default: len = 6; break; |
6338b358 | 3203 | case SPELL_LITERAL: len = token->val.str.len; break; |
9a0c6187 | 3204 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
041c3194 | 3205 | } |
59325650 NB |
3206 | |
3207 | return len; | |
6d2c2047 ZW |
3208 | } |
3209 | ||
47e20491 GK |
3210 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
3211 | Return the number of bytes read out of NAME. (There are always | |
3212 | 10 bytes written to BUFFER.) */ | |
3213 | ||
3214 | static size_t | |
3215 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
3216 | { | |
3217 | int j; | |
3218 | int ucn_len = 0; | |
3219 | int ucn_len_c; | |
3220 | unsigned t; | |
3221 | unsigned long utf32; | |
3222 | ||
3223 | /* Compute the length of the UTF-8 sequence. */ | |
3224 | for (t = *name; t & 0x80; t <<= 1) | |
3225 | ucn_len++; | |
3226 | ||
3227 | utf32 = *name & (0x7F >> ucn_len); | |
3228 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
3229 | { | |
3230 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
3231 | ||
3232 | /* Ill-formed UTF-8. */ | |
3233 | if ((*name & ~0x3F) != 0x80) | |
3234 | abort (); | |
3235 | } | |
3236 | ||
3237 | *buffer++ = '\\'; | |
3238 | *buffer++ = 'U'; | |
3239 | for (j = 7; j >= 0; j--) | |
3240 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
3241 | return ucn_len; | |
3242 | } | |
3243 | ||
cfc93532 MLI |
3244 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
3245 | the spelling of the digraph. */ | |
3246 | static const unsigned char * | |
3247 | cpp_digraph2name (enum cpp_ttype type) | |
3248 | { | |
3249 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
3250 | } | |
47e20491 | 3251 | |
be5ffc59 JM |
3252 | /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER. |
3253 | The buffer must already contain the enough space to hold the | |
3254 | token's spelling. Returns a pointer to the character after the | |
3255 | last character written. */ | |
3256 | unsigned char * | |
3257 | _cpp_spell_ident_ucns (unsigned char *buffer, cpp_hashnode *ident) | |
3258 | { | |
3259 | size_t i; | |
3260 | const unsigned char *name = NODE_NAME (ident); | |
3261 | ||
3262 | for (i = 0; i < NODE_LEN (ident); i++) | |
3263 | if (name[i] & ~0x7F) | |
3264 | { | |
3265 | i += utf8_to_ucn (buffer, name + i) - 1; | |
3266 | buffer += 10; | |
3267 | } | |
3268 | else | |
3269 | *buffer++ = name[i]; | |
3270 | ||
3271 | return buffer; | |
3272 | } | |
3273 | ||
041c3194 | 3274 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 | 3275 | already contain the enough space to hold the token's spelling. |
6cf87ca4 | 3276 | Returns a pointer to the character after the last character written. |
47e20491 | 3277 | FORSTRING is true if this is to be the spelling after translation |
be5ffc59 JM |
3278 | phase 1 (with the original spelling of extended identifiers), false |
3279 | if extended identifiers should always be written using UCNs (there is | |
3280 | no option for always writing them in the internal UTF-8 form). | |
6cf87ca4 | 3281 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
93c80368 | 3282 | unsigned char * |
6cf87ca4 | 3283 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
47e20491 | 3284 | unsigned char *buffer, bool forstring) |
041c3194 | 3285 | { |
96be6998 | 3286 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
3287 | { |
3288 | case SPELL_OPERATOR: | |
3289 | { | |
3290 | const unsigned char *spelling; | |
3291 | unsigned char c; | |
d6d5f795 | 3292 | |
041c3194 | 3293 | if (token->flags & DIGRAPH) |
cfc93532 | 3294 | spelling = cpp_digraph2name (token->type); |
92936ecf ZW |
3295 | else if (token->flags & NAMED_OP) |
3296 | goto spell_ident; | |
041c3194 | 3297 | else |
96be6998 | 3298 | spelling = TOKEN_NAME (token); |
df383483 | 3299 | |
041c3194 ZW |
3300 | while ((c = *spelling++) != '\0') |
3301 | *buffer++ = c; | |
3302 | } | |
3303 | break; | |
d6d5f795 | 3304 | |
47ad4138 | 3305 | spell_ident: |
041c3194 | 3306 | case SPELL_IDENT: |
47e20491 GK |
3307 | if (forstring) |
3308 | { | |
be5ffc59 JM |
3309 | memcpy (buffer, NODE_NAME (token->val.node.spelling), |
3310 | NODE_LEN (token->val.node.spelling)); | |
3311 | buffer += NODE_LEN (token->val.node.spelling); | |
47e20491 GK |
3312 | } |
3313 | else | |
be5ffc59 | 3314 | buffer = _cpp_spell_ident_ucns (buffer, token->val.node.node); |
041c3194 | 3315 | break; |
d6d5f795 | 3316 | |
6338b358 | 3317 | case SPELL_LITERAL: |
47ad4138 ZW |
3318 | memcpy (buffer, token->val.str.text, token->val.str.len); |
3319 | buffer += token->val.str.len; | |
3320 | break; | |
3321 | ||
041c3194 | 3322 | case SPELL_NONE: |
0527bc4e JDA |
3323 | cpp_error (pfile, CPP_DL_ICE, |
3324 | "unspellable token %s", TOKEN_NAME (token)); | |
041c3194 ZW |
3325 | break; |
3326 | } | |
d6d5f795 | 3327 | |
041c3194 ZW |
3328 | return buffer; |
3329 | } | |
d6d5f795 | 3330 | |
5d8ebbd8 NB |
3331 | /* Returns TOKEN spelt as a null-terminated string. The string is |
3332 | freed when the reader is destroyed. Useful for diagnostics. */ | |
93c80368 | 3333 | unsigned char * |
6cf87ca4 | 3334 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
59325650 NB |
3335 | { |
3336 | unsigned int len = cpp_token_len (token) + 1; | |
ece54d54 | 3337 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 3338 | |
47e20491 | 3339 | end = cpp_spell_token (pfile, token, start, false); |
93c80368 | 3340 | end[0] = '\0'; |
c5a04734 | 3341 | |
93c80368 NB |
3342 | return start; |
3343 | } | |
c5a04734 | 3344 | |
cfc93532 MLI |
3345 | /* Returns a pointer to a string which spells the token defined by |
3346 | TYPE and FLAGS. Used by C front ends, which really should move to | |
3347 | using cpp_token_as_text. */ | |
93c80368 | 3348 | const char * |
cfc93532 | 3349 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
93c80368 | 3350 | { |
cfc93532 MLI |
3351 | if (flags & DIGRAPH) |
3352 | return (const char *) cpp_digraph2name (type); | |
3353 | else if (flags & NAMED_OP) | |
3354 | return cpp_named_operator2name (type); | |
3355 | ||
93c80368 NB |
3356 | return (const char *) token_spellings[type].name; |
3357 | } | |
c5a04734 | 3358 | |
4ed5bcfb NB |
3359 | /* Writes the spelling of token to FP, without any preceding space. |
3360 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
3361 | double-buffering. */ | |
93c80368 | 3362 | void |
6cf87ca4 | 3363 | cpp_output_token (const cpp_token *token, FILE *fp) |
93c80368 | 3364 | { |
93c80368 | 3365 | switch (TOKEN_SPELL (token)) |
c5a04734 | 3366 | { |
93c80368 NB |
3367 | case SPELL_OPERATOR: |
3368 | { | |
3369 | const unsigned char *spelling; | |
3b681e9d | 3370 | int c; |
c5a04734 | 3371 | |
93c80368 | 3372 | if (token->flags & DIGRAPH) |
cfc93532 | 3373 | spelling = cpp_digraph2name (token->type); |
93c80368 NB |
3374 | else if (token->flags & NAMED_OP) |
3375 | goto spell_ident; | |
3376 | else | |
3377 | spelling = TOKEN_NAME (token); | |
041c3194 | 3378 | |
3b681e9d ZW |
3379 | c = *spelling; |
3380 | do | |
3381 | putc (c, fp); | |
3382 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
3383 | } |
3384 | break; | |
041c3194 | 3385 | |
93c80368 NB |
3386 | spell_ident: |
3387 | case SPELL_IDENT: | |
47e20491 GK |
3388 | { |
3389 | size_t i; | |
9a0c6187 | 3390 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 3391 | |
9a0c6187 | 3392 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
3393 | if (name[i] & ~0x7F) |
3394 | { | |
3395 | unsigned char buffer[10]; | |
3396 | i += utf8_to_ucn (buffer, name + i) - 1; | |
3397 | fwrite (buffer, 1, 10, fp); | |
3398 | } | |
3399 | else | |
9a0c6187 | 3400 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
47e20491 GK |
3401 | } |
3402 | break; | |
041c3194 | 3403 | |
6338b358 | 3404 | case SPELL_LITERAL: |
47ad4138 ZW |
3405 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
3406 | break; | |
3407 | ||
93c80368 NB |
3408 | case SPELL_NONE: |
3409 | /* An error, most probably. */ | |
3410 | break; | |
041c3194 | 3411 | } |
c5a04734 ZW |
3412 | } |
3413 | ||
93c80368 NB |
3414 | /* Compare two tokens. */ |
3415 | int | |
6cf87ca4 | 3416 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
c5a04734 | 3417 | { |
93c80368 NB |
3418 | if (a->type == b->type && a->flags == b->flags) |
3419 | switch (TOKEN_SPELL (a)) | |
3420 | { | |
3421 | default: /* Keep compiler happy. */ | |
3422 | case SPELL_OPERATOR: | |
9a0c6187 | 3423 | /* token_no is used to track where multiple consecutive ## |
aa508502 | 3424 | tokens were originally located. */ |
9a0c6187 | 3425 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
93c80368 | 3426 | case SPELL_NONE: |
9a0c6187 | 3427 | return (a->type != CPP_MACRO_ARG |
be5ffc59 JM |
3428 | || (a->val.macro_arg.arg_no == b->val.macro_arg.arg_no |
3429 | && a->val.macro_arg.spelling == b->val.macro_arg.spelling)); | |
93c80368 | 3430 | case SPELL_IDENT: |
be5ffc59 JM |
3431 | return (a->val.node.node == b->val.node.node |
3432 | && a->val.node.spelling == b->val.node.spelling); | |
6338b358 | 3433 | case SPELL_LITERAL: |
93c80368 NB |
3434 | return (a->val.str.len == b->val.str.len |
3435 | && !memcmp (a->val.str.text, b->val.str.text, | |
3436 | a->val.str.len)); | |
3437 | } | |
c5a04734 | 3438 | |
041c3194 ZW |
3439 | return 0; |
3440 | } | |
3441 | ||
93c80368 NB |
3442 | /* Returns nonzero if a space should be inserted to avoid an |
3443 | accidental token paste for output. For simplicity, it is | |
3444 | conservative, and occasionally advises a space where one is not | |
3445 | needed, e.g. "." and ".2". */ | |
93c80368 | 3446 | int |
6cf87ca4 ZW |
3447 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
3448 | const cpp_token *token2) | |
c5a04734 | 3449 | { |
93c80368 NB |
3450 | enum cpp_ttype a = token1->type, b = token2->type; |
3451 | cppchar_t c; | |
c5a04734 | 3452 | |
93c80368 NB |
3453 | if (token1->flags & NAMED_OP) |
3454 | a = CPP_NAME; | |
3455 | if (token2->flags & NAMED_OP) | |
3456 | b = CPP_NAME; | |
c5a04734 | 3457 | |
93c80368 NB |
3458 | c = EOF; |
3459 | if (token2->flags & DIGRAPH) | |
37b8524c | 3460 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
3461 | else if (token_spellings[b].category == SPELL_OPERATOR) |
3462 | c = token_spellings[b].name[0]; | |
c5a04734 | 3463 | |
93c80368 | 3464 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 3465 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 3466 | return 1; |
c5a04734 | 3467 | |
93c80368 | 3468 | switch (a) |
c5a04734 | 3469 | { |
b52dbbf8 SE |
3470 | case CPP_GREATER: return c == '>'; |
3471 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
93c80368 NB |
3472 | case CPP_PLUS: return c == '+'; |
3473 | case CPP_MINUS: return c == '-' || c == '>'; | |
3474 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
3475 | case CPP_MOD: return c == ':' || c == '>'; | |
3476 | case CPP_AND: return c == '&'; | |
3477 | case CPP_OR: return c == '|'; | |
3478 | case CPP_COLON: return c == ':' || c == '>'; | |
3479 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 3480 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
3481 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
3482 | case CPP_NAME: return ((b == CPP_NUMBER | |
3483 | && name_p (pfile, &token2->val.str)) | |
3484 | || b == CPP_NAME | |
3485 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
3486 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
3487 | || c == '.' || c == '+' || c == '-'); | |
1613e52b | 3488 | /* UCNs */ |
1067694a NB |
3489 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
3490 | && b == CPP_NAME) | |
1613e52b | 3491 | || (CPP_OPTION (pfile, objc) |
1067694a | 3492 | && token1->val.str.text[0] == '@' |
1613e52b | 3493 | && (b == CPP_NAME || b == CPP_STRING))); |
87e356ba JJ |
3494 | case CPP_STRING: |
3495 | case CPP_WSTRING: | |
3496 | case CPP_UTF8STRING: | |
3497 | case CPP_STRING16: | |
3498 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals) | |
3499 | && (b == CPP_NAME | |
3500 | || (TOKEN_SPELL (token2) == SPELL_LITERAL | |
3501 | && ISIDST (token2->val.str.text[0])))); | |
3502 | ||
93c80368 | 3503 | default: break; |
c5a04734 | 3504 | } |
c5a04734 | 3505 | |
417f3e3a | 3506 | return 0; |
c5a04734 ZW |
3507 | } |
3508 | ||
93c80368 | 3509 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
3510 | character, to FP. Leading whitespace is removed. If there are |
3511 | macros, special token padding is not performed. */ | |
c5a04734 | 3512 | void |
6cf87ca4 | 3513 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
c5a04734 | 3514 | { |
4ed5bcfb | 3515 | const cpp_token *token; |
96be6998 | 3516 | |
4ed5bcfb NB |
3517 | token = cpp_get_token (pfile); |
3518 | while (token->type != CPP_EOF) | |
96be6998 | 3519 | { |
4ed5bcfb NB |
3520 | cpp_output_token (token, fp); |
3521 | token = cpp_get_token (pfile); | |
3522 | if (token->flags & PREV_WHITE) | |
3523 | putc (' ', fp); | |
96be6998 ZW |
3524 | } |
3525 | ||
93c80368 | 3526 | putc ('\n', fp); |
041c3194 | 3527 | } |
c5a04734 | 3528 | |
5d6342eb TT |
3529 | /* Return a string representation of all the remaining tokens on the |
3530 | current line. The result is allocated using xmalloc and must be | |
3531 | freed by the caller. */ | |
3532 | unsigned char * | |
3533 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
3534 | { | |
3535 | const cpp_token *token; | |
3536 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
3537 | unsigned int alloced = 120 + out; | |
3538 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
3539 | ||
3540 | /* If DIR_NAME is empty, there are no initial contents. */ | |
3541 | if (dir_name) | |
3542 | { | |
3543 | sprintf ((char *) result, "#%s ", dir_name); | |
3544 | out += 2; | |
3545 | } | |
3546 | ||
3547 | token = cpp_get_token (pfile); | |
3548 | while (token->type != CPP_EOF) | |
3549 | { | |
3550 | unsigned char *last; | |
3551 | /* Include room for a possible space and the terminating nul. */ | |
3552 | unsigned int len = cpp_token_len (token) + 2; | |
3553 | ||
3554 | if (out + len > alloced) | |
3555 | { | |
3556 | alloced *= 2; | |
3557 | if (out + len > alloced) | |
3558 | alloced = out + len; | |
3559 | result = (unsigned char *) xrealloc (result, alloced); | |
3560 | } | |
3561 | ||
3562 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
3563 | out = last - result; | |
3564 | ||
3565 | token = cpp_get_token (pfile); | |
3566 | if (token->flags & PREV_WHITE) | |
3567 | result[out++] = ' '; | |
3568 | } | |
3569 | ||
3570 | result[out] = '\0'; | |
3571 | return result; | |
3572 | } | |
3573 | ||
1e013d2e NB |
3574 | /* Memory buffers. Changing these three constants can have a dramatic |
3575 | effect on performance. The values here are reasonable defaults, | |
3576 | but might be tuned. If you adjust them, be sure to test across a | |
3577 | range of uses of cpplib, including heavy nested function-like macro | |
3578 | expansion. Also check the change in peak memory usage (NJAMD is a | |
3579 | good tool for this). */ | |
3580 | #define MIN_BUFF_SIZE 8000 | |
87062813 | 3581 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
1e013d2e NB |
3582 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
3583 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 3584 | |
87062813 NB |
3585 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
3586 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
3587 | #endif | |
3588 | ||
c9e7a609 NB |
3589 | /* Create a new allocation buffer. Place the control block at the end |
3590 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 | 3591 | static _cpp_buff * |
6cf87ca4 | 3592 | new_buff (size_t len) |
b8af0ca5 NB |
3593 | { |
3594 | _cpp_buff *result; | |
ece54d54 | 3595 | unsigned char *base; |
b8af0ca5 | 3596 | |
1e013d2e NB |
3597 | if (len < MIN_BUFF_SIZE) |
3598 | len = MIN_BUFF_SIZE; | |
c70f6ed3 | 3599 | len = CPP_ALIGN (len); |
b8af0ca5 | 3600 | |
ceb17928 | 3601 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
3602 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff |
3603 | struct first. */ | |
3604 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
3605 | base = XNEWVEC (unsigned char, len + slen); | |
3606 | result = (_cpp_buff *) base; | |
3607 | base += slen; | |
3608 | #else | |
c3f829c1 | 3609 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
b8af0ca5 | 3610 | result = (_cpp_buff *) (base + len); |
1a80db97 | 3611 | #endif |
b8af0ca5 NB |
3612 | result->base = base; |
3613 | result->cur = base; | |
3614 | result->limit = base + len; | |
3615 | result->next = NULL; | |
3616 | return result; | |
3617 | } | |
3618 | ||
3619 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
3620 | void | |
6cf87ca4 | 3621 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
b8af0ca5 NB |
3622 | { |
3623 | _cpp_buff *end = buff; | |
3624 | ||
3625 | while (end->next) | |
3626 | end = end->next; | |
3627 | end->next = pfile->free_buffs; | |
3628 | pfile->free_buffs = buff; | |
3629 | } | |
3630 | ||
3631 | /* Return a free buffer of size at least MIN_SIZE. */ | |
3632 | _cpp_buff * | |
6cf87ca4 | 3633 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
b8af0ca5 NB |
3634 | { |
3635 | _cpp_buff *result, **p; | |
3636 | ||
3637 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
3638 | { | |
6142088c | 3639 | size_t size; |
1e013d2e NB |
3640 | |
3641 | if (*p == NULL) | |
b8af0ca5 | 3642 | return new_buff (min_size); |
1e013d2e NB |
3643 | result = *p; |
3644 | size = result->limit - result->base; | |
3645 | /* Return a buffer that's big enough, but don't waste one that's | |
3646 | way too big. */ | |
34f5271d | 3647 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
b8af0ca5 NB |
3648 | break; |
3649 | } | |
3650 | ||
3651 | *p = result->next; | |
3652 | result->next = NULL; | |
3653 | result->cur = result->base; | |
3654 | return result; | |
3655 | } | |
3656 | ||
4fe9b91c | 3657 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3658 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
3659 | the excess bytes to the new buffer. Chains the new buffer after | |
3660 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 3661 | _cpp_buff * |
6cf87ca4 | 3662 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
b8af0ca5 | 3663 | { |
6142088c | 3664 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 3665 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 3666 | |
8c3b2693 NB |
3667 | buff->next = new_buff; |
3668 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
3669 | return new_buff; | |
3670 | } | |
3671 | ||
4fe9b91c | 3672 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
3673 | remaining bytes of the buffer pointed to by BUFF, and at least |
3674 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
3675 | Chains the new buffer before the buffer pointed to by BUFF, and | |
3676 | updates the pointer to point to the new buffer. */ | |
3677 | void | |
6cf87ca4 | 3678 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
8c3b2693 NB |
3679 | { |
3680 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
3681 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
3682 | ||
3683 | new_buff = _cpp_get_buff (pfile, size); | |
3684 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
3685 | new_buff->next = old_buff; | |
3686 | *pbuff = new_buff; | |
b8af0ca5 NB |
3687 | } |
3688 | ||
3689 | /* Free a chain of buffers starting at BUFF. */ | |
3690 | void | |
5671bf27 | 3691 | _cpp_free_buff (_cpp_buff *buff) |
b8af0ca5 NB |
3692 | { |
3693 | _cpp_buff *next; | |
3694 | ||
3695 | for (; buff; buff = next) | |
3696 | { | |
3697 | next = buff->next; | |
ceb17928 | 3698 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
3699 | free (buff); |
3700 | #else | |
b8af0ca5 | 3701 | free (buff->base); |
1a80db97 | 3702 | #endif |
b8af0ca5 NB |
3703 | } |
3704 | } | |
417f3e3a | 3705 | |
ece54d54 NB |
3706 | /* Allocate permanent, unaligned storage of length LEN. */ |
3707 | unsigned char * | |
6cf87ca4 | 3708 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
ece54d54 NB |
3709 | { |
3710 | _cpp_buff *buff = pfile->u_buff; | |
3711 | unsigned char *result = buff->cur; | |
3712 | ||
3713 | if (len > (size_t) (buff->limit - result)) | |
3714 | { | |
3715 | buff = _cpp_get_buff (pfile, len); | |
3716 | buff->next = pfile->u_buff; | |
3717 | pfile->u_buff = buff; | |
3718 | result = buff->cur; | |
3719 | } | |
3720 | ||
3721 | buff->cur = result + len; | |
3722 | return result; | |
3723 | } | |
3724 | ||
87062813 NB |
3725 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
3726 | That buffer is used for growing allocations when saving macro | |
3727 | replacement lists in a #define, and when parsing an answer to an | |
3728 | assertion in #assert, #unassert or #if (and therefore possibly | |
3729 | whilst expanding macros). It therefore must not be used by any | |
3730 | code that they might call: specifically the lexer and the guts of | |
3731 | the macro expander. | |
3732 | ||
3733 | All existing other uses clearly fit this restriction: storing | |
3734 | registered pragmas during initialization. */ | |
93c80368 | 3735 | unsigned char * |
6cf87ca4 | 3736 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
3fef5b2b | 3737 | { |
8c3b2693 NB |
3738 | _cpp_buff *buff = pfile->a_buff; |
3739 | unsigned char *result = buff->cur; | |
3fef5b2b | 3740 | |
8c3b2693 | 3741 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 3742 | { |
8c3b2693 NB |
3743 | buff = _cpp_get_buff (pfile, len); |
3744 | buff->next = pfile->a_buff; | |
3745 | pfile->a_buff = buff; | |
3746 | result = buff->cur; | |
3fef5b2b | 3747 | } |
041c3194 | 3748 | |
8c3b2693 | 3749 | buff->cur = result + len; |
93c80368 | 3750 | return result; |
041c3194 | 3751 | } |
d8044160 | 3752 | |
10f04917 NS |
3753 | /* Commit or allocate storage from a buffer. */ |
3754 | ||
3755 | void * | |
3756 | _cpp_commit_buff (cpp_reader *pfile, size_t size) | |
3757 | { | |
3758 | void *ptr = BUFF_FRONT (pfile->a_buff); | |
3759 | ||
3760 | if (pfile->hash_table->alloc_subobject) | |
3761 | { | |
3762 | void *copy = pfile->hash_table->alloc_subobject (size); | |
3763 | memcpy (copy, ptr, size); | |
3764 | ptr = copy; | |
3765 | } | |
3766 | else | |
3767 | BUFF_FRONT (pfile->a_buff) += size; | |
3768 | ||
3769 | return ptr; | |
3770 | } | |
3771 | ||
d8044160 GK |
3772 | /* Say which field of TOK is in use. */ |
3773 | ||
3774 | enum cpp_token_fld_kind | |
c26302d5 | 3775 | cpp_token_val_index (const cpp_token *tok) |
d8044160 GK |
3776 | { |
3777 | switch (TOKEN_SPELL (tok)) | |
3778 | { | |
3779 | case SPELL_IDENT: | |
3780 | return CPP_TOKEN_FLD_NODE; | |
3781 | case SPELL_LITERAL: | |
3782 | return CPP_TOKEN_FLD_STR; | |
aa508502 | 3783 | case SPELL_OPERATOR: |
3f23e487 AP |
3784 | /* Operands which were originally spelled as ident keep around |
3785 | the node for the exact spelling. */ | |
3786 | if (tok->flags & NAMED_OP) | |
3787 | return CPP_TOKEN_FLD_NODE; | |
3788 | else if (tok->type == CPP_PASTE) | |
9a0c6187 | 3789 | return CPP_TOKEN_FLD_TOKEN_NO; |
aa508502 JM |
3790 | else |
3791 | return CPP_TOKEN_FLD_NONE; | |
d8044160 GK |
3792 | case SPELL_NONE: |
3793 | if (tok->type == CPP_MACRO_ARG) | |
3794 | return CPP_TOKEN_FLD_ARG_NO; | |
3795 | else if (tok->type == CPP_PADDING) | |
3796 | return CPP_TOKEN_FLD_SOURCE; | |
21b11495 | 3797 | else if (tok->type == CPP_PRAGMA) |
bc4071dd | 3798 | return CPP_TOKEN_FLD_PRAGMA; |
191816a3 | 3799 | /* fall through */ |
d8044160 GK |
3800 | default: |
3801 | return CPP_TOKEN_FLD_NONE; | |
3802 | } | |
3803 | } | |
e3dfef44 | 3804 | |
f3f6029d | 3805 | /* All tokens lexed in R after calling this function will be forced to |
620e594b | 3806 | have their location_t to be P, until |
e3dfef44 GC |
3807 | cpp_stop_forcing_token_locations is called for R. */ |
3808 | ||
3809 | void | |
620e594b | 3810 | cpp_force_token_locations (cpp_reader *r, location_t loc) |
e3dfef44 | 3811 | { |
f3f6029d | 3812 | r->forced_token_location = loc; |
e3dfef44 GC |
3813 | } |
3814 | ||
3815 | /* Go back to assigning locations naturally for lexed tokens. */ | |
3816 | ||
3817 | void | |
3818 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
3819 | { | |
f3f6029d | 3820 | r->forced_token_location = 0; |
e3dfef44 | 3821 | } |