]>
Commit | Line | Data |
---|---|---|
45b966db | 1 | /* CPP Library - lexical analysis. |
7adcbafe | 2 | Copyright (C) 2000-2022 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: | |
530b1d68 | 394 | t = data == repl_nl; |
395 | t |= data == repl_cr; | |
396 | t |= data == repl_bs; | |
397 | t |= data == repl_qm; | |
246a2fcb RH |
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 | ||
d00b1b02 | 534 | #elif (GCC_VERSION >= 4005) && 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 | |
d00b1b02 | 538 | the same as the AltiVec version. */ |
0ccaaab0 | 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. */ | |
082a7b23 | 1065 | if (*s == '\r' && s + 1 != buffer->rlimit && s[1] == '\n') |
26aea073 NB |
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 | ||
51c50026 MP |
1167 | namespace bidi { |
1168 | enum class kind { | |
1169 | NONE, LRE, RLE, LRO, RLO, LRI, RLI, FSI, PDF, PDI, LTR, RTL | |
1170 | }; | |
1171 | ||
1172 | /* All the UTF-8 encodings of bidi characters start with E2. */ | |
1173 | constexpr uchar utf8_start = 0xe2; | |
1174 | ||
bef32d4a DM |
1175 | struct context |
1176 | { | |
1177 | context () {} | |
1178 | context (location_t loc, kind k, bool pdf, bool ucn) | |
1179 | : m_loc (loc), m_kind (k), m_pdf (pdf), m_ucn (ucn) | |
1180 | { | |
1181 | } | |
1182 | ||
1183 | kind get_pop_kind () const | |
1184 | { | |
1185 | return m_pdf ? kind::PDF : kind::PDI; | |
1186 | } | |
1187 | bool ucn_p () const | |
1188 | { | |
1189 | return m_ucn; | |
1190 | } | |
1191 | ||
1192 | location_t m_loc; | |
1193 | kind m_kind; | |
1194 | unsigned m_pdf : 1; | |
1195 | unsigned m_ucn : 1; | |
1196 | }; | |
1197 | ||
51c50026 MP |
1198 | /* A vector holding currently open bidi contexts. We use a char for |
1199 | each context, its LSB is 1 if it represents a PDF context, 0 if it | |
1200 | represents a PDI context. The next bit is 1 if this context was open | |
1201 | by a bidi character written as a UCN, and 0 when it was UTF-8. */ | |
bef32d4a | 1202 | semi_embedded_vec <context, 16> vec; |
51c50026 MP |
1203 | |
1204 | /* Close the whole comment/identifier/string literal/character constant | |
1205 | context. */ | |
1206 | void on_close () | |
1207 | { | |
1208 | vec.truncate (0); | |
1209 | } | |
1210 | ||
1211 | /* Pop the last element in the vector. */ | |
1212 | void pop () | |
1213 | { | |
1214 | unsigned int len = vec.count (); | |
1215 | gcc_checking_assert (len > 0); | |
1216 | vec.truncate (len - 1); | |
1217 | } | |
1218 | ||
bef32d4a DM |
1219 | /* Return the pop kind of the context of the Ith element. */ |
1220 | kind pop_kind_at (unsigned int i) | |
51c50026 | 1221 | { |
bef32d4a | 1222 | return vec[i].get_pop_kind (); |
51c50026 MP |
1223 | } |
1224 | ||
bef32d4a | 1225 | /* Return the pop kind of the context that is currently opened. */ |
51c50026 MP |
1226 | kind current_ctx () |
1227 | { | |
1228 | unsigned int len = vec.count (); | |
1229 | if (len == 0) | |
1230 | return kind::NONE; | |
bef32d4a | 1231 | return vec[len - 1].get_pop_kind (); |
51c50026 MP |
1232 | } |
1233 | ||
1234 | /* Return true if the current context comes from a UCN origin, that is, | |
1235 | the bidi char which started this bidi context was written as a UCN. */ | |
1236 | bool current_ctx_ucn_p () | |
1237 | { | |
1238 | unsigned int len = vec.count (); | |
1239 | gcc_checking_assert (len > 0); | |
bef32d4a | 1240 | return vec[len - 1].m_ucn; |
51c50026 MP |
1241 | } |
1242 | ||
bef32d4a DM |
1243 | location_t current_ctx_loc () |
1244 | { | |
1245 | unsigned int len = vec.count (); | |
1246 | gcc_checking_assert (len > 0); | |
1247 | return vec[len - 1].m_loc; | |
1248 | } | |
1249 | ||
1250 | /* We've read a bidi char, update the current vector as necessary. | |
1251 | LOC is only valid when K is not kind::NONE. */ | |
1252 | void on_char (kind k, bool ucn_p, location_t loc) | |
51c50026 MP |
1253 | { |
1254 | switch (k) | |
1255 | { | |
1256 | case kind::LRE: | |
1257 | case kind::RLE: | |
1258 | case kind::LRO: | |
1259 | case kind::RLO: | |
bef32d4a | 1260 | vec.push (context (loc, k, true, ucn_p)); |
51c50026 MP |
1261 | break; |
1262 | case kind::LRI: | |
1263 | case kind::RLI: | |
1264 | case kind::FSI: | |
bef32d4a | 1265 | vec.push (context (loc, k, false, ucn_p)); |
51c50026 MP |
1266 | break; |
1267 | /* PDF terminates the scope of the last LRE, RLE, LRO, or RLO | |
1268 | whose scope has not yet been terminated. */ | |
1269 | case kind::PDF: | |
1270 | if (current_ctx () == kind::PDF) | |
1271 | pop (); | |
1272 | break; | |
1273 | /* PDI terminates the scope of the last LRI, RLI, or FSI whose | |
1274 | scope has not yet been terminated, as well as the scopes of | |
1275 | any subsequent LREs, RLEs, LROs, or RLOs whose scopes have not | |
1276 | yet been terminated. */ | |
1277 | case kind::PDI: | |
1278 | for (int i = vec.count () - 1; i >= 0; --i) | |
bef32d4a | 1279 | if (pop_kind_at (i) == kind::PDI) |
51c50026 MP |
1280 | { |
1281 | vec.truncate (i); | |
1282 | break; | |
1283 | } | |
1284 | break; | |
1285 | case kind::LTR: | |
1286 | case kind::RTL: | |
1287 | /* These aren't popped by a PDF/PDI. */ | |
1288 | break; | |
630686f9 | 1289 | ATTR_LIKELY case kind::NONE: |
51c50026 MP |
1290 | break; |
1291 | default: | |
1292 | abort (); | |
1293 | } | |
1294 | } | |
1295 | ||
1296 | /* Return a descriptive string for K. */ | |
1297 | const char *to_str (kind k) | |
1298 | { | |
1299 | switch (k) | |
1300 | { | |
1301 | case kind::LRE: | |
1302 | return "U+202A (LEFT-TO-RIGHT EMBEDDING)"; | |
1303 | case kind::RLE: | |
1304 | return "U+202B (RIGHT-TO-LEFT EMBEDDING)"; | |
1305 | case kind::LRO: | |
1306 | return "U+202D (LEFT-TO-RIGHT OVERRIDE)"; | |
1307 | case kind::RLO: | |
1308 | return "U+202E (RIGHT-TO-LEFT OVERRIDE)"; | |
1309 | case kind::LRI: | |
1310 | return "U+2066 (LEFT-TO-RIGHT ISOLATE)"; | |
1311 | case kind::RLI: | |
1312 | return "U+2067 (RIGHT-TO-LEFT ISOLATE)"; | |
1313 | case kind::FSI: | |
1314 | return "U+2068 (FIRST STRONG ISOLATE)"; | |
1315 | case kind::PDF: | |
1316 | return "U+202C (POP DIRECTIONAL FORMATTING)"; | |
1317 | case kind::PDI: | |
1318 | return "U+2069 (POP DIRECTIONAL ISOLATE)"; | |
1319 | case kind::LTR: | |
1320 | return "U+200E (LEFT-TO-RIGHT MARK)"; | |
1321 | case kind::RTL: | |
1322 | return "U+200F (RIGHT-TO-LEFT MARK)"; | |
1323 | default: | |
1324 | abort (); | |
1325 | } | |
1326 | } | |
1327 | } | |
1328 | ||
bef32d4a DM |
1329 | /* Get location_t for the range of bytes [START, START + NUM_BYTES) |
1330 | within the current line in FILE, with the caret at START. */ | |
1331 | ||
1332 | static location_t | |
1333 | get_location_for_byte_range_in_cur_line (cpp_reader *pfile, | |
1334 | const unsigned char *const start, | |
1335 | size_t num_bytes) | |
1336 | { | |
1337 | gcc_checking_assert (num_bytes > 0); | |
1338 | ||
1339 | /* CPP_BUF_COLUMN and linemap_position_for_column both refer | |
1340 | to offsets in bytes, but CPP_BUF_COLUMN is 0-based, | |
1341 | whereas linemap_position_for_column is 1-based. */ | |
1342 | ||
1343 | /* Get 0-based offsets within the line. */ | |
1344 | size_t start_offset = CPP_BUF_COLUMN (pfile->buffer, start); | |
1345 | size_t end_offset = start_offset + num_bytes - 1; | |
1346 | ||
1347 | /* Now convert to location_t, where "columns" are 1-based byte offsets. */ | |
1348 | location_t start_loc = linemap_position_for_column (pfile->line_table, | |
1349 | start_offset + 1); | |
1350 | location_t end_loc = linemap_position_for_column (pfile->line_table, | |
1351 | end_offset + 1); | |
1352 | ||
1353 | if (start_loc == end_loc) | |
1354 | return start_loc; | |
1355 | ||
1356 | source_range src_range; | |
1357 | src_range.m_start = start_loc; | |
1358 | src_range.m_finish = end_loc; | |
1359 | location_t combined_loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
1360 | start_loc, | |
1361 | src_range, | |
1362 | NULL); | |
1363 | return combined_loc; | |
1364 | } | |
1365 | ||
51c50026 MP |
1366 | /* Parse a sequence of 3 bytes starting with P and return its bidi code. */ |
1367 | ||
1368 | static bidi::kind | |
bef32d4a | 1369 | get_bidi_utf8_1 (const unsigned char *const p) |
51c50026 MP |
1370 | { |
1371 | gcc_checking_assert (p[0] == bidi::utf8_start); | |
1372 | ||
1373 | if (p[1] == 0x80) | |
1374 | switch (p[2]) | |
1375 | { | |
1376 | case 0xaa: | |
1377 | return bidi::kind::LRE; | |
1378 | case 0xab: | |
1379 | return bidi::kind::RLE; | |
1380 | case 0xac: | |
1381 | return bidi::kind::PDF; | |
1382 | case 0xad: | |
1383 | return bidi::kind::LRO; | |
1384 | case 0xae: | |
1385 | return bidi::kind::RLO; | |
1386 | case 0x8e: | |
1387 | return bidi::kind::LTR; | |
1388 | case 0x8f: | |
1389 | return bidi::kind::RTL; | |
1390 | default: | |
1391 | break; | |
1392 | } | |
1393 | else if (p[1] == 0x81) | |
1394 | switch (p[2]) | |
1395 | { | |
1396 | case 0xa6: | |
1397 | return bidi::kind::LRI; | |
1398 | case 0xa7: | |
1399 | return bidi::kind::RLI; | |
1400 | case 0xa8: | |
1401 | return bidi::kind::FSI; | |
1402 | case 0xa9: | |
1403 | return bidi::kind::PDI; | |
1404 | default: | |
1405 | break; | |
1406 | } | |
1407 | ||
1408 | return bidi::kind::NONE; | |
1409 | } | |
1410 | ||
bef32d4a DM |
1411 | /* Parse a sequence of 3 bytes starting with P and return its bidi code. |
1412 | If the kind is not NONE, write the location to *OUT.*/ | |
1413 | ||
1414 | static bidi::kind | |
1415 | get_bidi_utf8 (cpp_reader *pfile, const unsigned char *const p, location_t *out) | |
1416 | { | |
1417 | bidi::kind result = get_bidi_utf8_1 (p); | |
1418 | if (result != bidi::kind::NONE) | |
1419 | { | |
1420 | /* We have a sequence of 3 bytes starting at P. */ | |
1421 | *out = get_location_for_byte_range_in_cur_line (pfile, p, 3); | |
1422 | } | |
1423 | return result; | |
1424 | } | |
1425 | ||
51c50026 MP |
1426 | /* Parse a UCN where P points just past \u or \U and return its bidi code. */ |
1427 | ||
1428 | static bidi::kind | |
bef32d4a | 1429 | get_bidi_ucn_1 (const unsigned char *p, bool is_U) |
51c50026 MP |
1430 | { |
1431 | /* 6.4.3 Universal Character Names | |
1432 | \u hex-quad | |
1433 | \U hex-quad hex-quad | |
1434 | where \unnnn means \U0000nnnn. */ | |
1435 | ||
1436 | if (is_U) | |
1437 | { | |
1438 | if (p[0] != '0' || p[1] != '0' || p[2] != '0' || p[3] != '0') | |
1439 | return bidi::kind::NONE; | |
1440 | /* Skip 4B so we can treat \u and \U the same below. */ | |
1441 | p += 4; | |
1442 | } | |
1443 | ||
1444 | /* All code points we are looking for start with 20xx. */ | |
1445 | if (p[0] != '2' || p[1] != '0') | |
1446 | return bidi::kind::NONE; | |
1447 | else if (p[2] == '2') | |
1448 | switch (p[3]) | |
1449 | { | |
1450 | case 'a': | |
1451 | case 'A': | |
1452 | return bidi::kind::LRE; | |
1453 | case 'b': | |
1454 | case 'B': | |
1455 | return bidi::kind::RLE; | |
1456 | case 'c': | |
1457 | case 'C': | |
1458 | return bidi::kind::PDF; | |
1459 | case 'd': | |
1460 | case 'D': | |
1461 | return bidi::kind::LRO; | |
1462 | case 'e': | |
1463 | case 'E': | |
1464 | return bidi::kind::RLO; | |
1465 | default: | |
1466 | break; | |
1467 | } | |
1468 | else if (p[2] == '6') | |
1469 | switch (p[3]) | |
1470 | { | |
1471 | case '6': | |
1472 | return bidi::kind::LRI; | |
1473 | case '7': | |
1474 | return bidi::kind::RLI; | |
1475 | case '8': | |
1476 | return bidi::kind::FSI; | |
1477 | case '9': | |
1478 | return bidi::kind::PDI; | |
1479 | default: | |
1480 | break; | |
1481 | } | |
1482 | else if (p[2] == '0') | |
1483 | switch (p[3]) | |
1484 | { | |
1485 | case 'e': | |
1486 | case 'E': | |
1487 | return bidi::kind::LTR; | |
1488 | case 'f': | |
1489 | case 'F': | |
1490 | return bidi::kind::RTL; | |
1491 | default: | |
1492 | break; | |
1493 | } | |
1494 | ||
1495 | return bidi::kind::NONE; | |
1496 | } | |
1497 | ||
bef32d4a DM |
1498 | /* Parse a UCN where P points just past \u or \U and return its bidi code. |
1499 | If the kind is not NONE, write the location to *OUT.*/ | |
1500 | ||
1501 | static bidi::kind | |
1502 | get_bidi_ucn (cpp_reader *pfile, const unsigned char *p, bool is_U, | |
1503 | location_t *out) | |
1504 | { | |
1505 | bidi::kind result = get_bidi_ucn_1 (p, is_U); | |
1506 | if (result != bidi::kind::NONE) | |
1507 | { | |
1508 | const unsigned char *start = p - 2; | |
1509 | size_t num_bytes = 2 + (is_U ? 8 : 4); | |
1510 | *out = get_location_for_byte_range_in_cur_line (pfile, start, num_bytes); | |
1511 | } | |
1512 | return result; | |
1513 | } | |
1514 | ||
1515 | /* Subclass of rich_location for reporting on unpaired UTF-8 | |
1516 | bidirectional control character(s). | |
1517 | Escape the source lines on output, and show all unclosed | |
1518 | bidi context, labelling everything. */ | |
1519 | ||
1520 | class unpaired_bidi_rich_location : public rich_location | |
1521 | { | |
1522 | public: | |
1523 | class custom_range_label : public range_label | |
1524 | { | |
1525 | public: | |
1526 | label_text get_text (unsigned range_idx) const FINAL OVERRIDE | |
1527 | { | |
1528 | /* range 0 is the primary location; each subsequent range i + 1 | |
1529 | is for bidi::vec[i]. */ | |
1530 | if (range_idx > 0) | |
1531 | { | |
1532 | const bidi::context &ctxt (bidi::vec[range_idx - 1]); | |
1533 | return label_text::borrow (bidi::to_str (ctxt.m_kind)); | |
1534 | } | |
1535 | else | |
1536 | return label_text::borrow (_("end of bidirectional context")); | |
1537 | } | |
1538 | }; | |
1539 | ||
1540 | unpaired_bidi_rich_location (cpp_reader *pfile, location_t loc) | |
1541 | : rich_location (pfile->line_table, loc, &m_custom_label) | |
1542 | { | |
1543 | set_escape_on_output (true); | |
1544 | for (unsigned i = 0; i < bidi::vec.count (); i++) | |
1545 | add_range (bidi::vec[i].m_loc, | |
1546 | SHOW_RANGE_WITHOUT_CARET, | |
1547 | &m_custom_label); | |
1548 | } | |
1549 | ||
1550 | private: | |
1551 | custom_range_label m_custom_label; | |
1552 | }; | |
1553 | ||
51c50026 MP |
1554 | /* We're closing a bidi context, that is, we've encountered a newline, |
1555 | are closing a C-style comment, or are at the end of a string literal, | |
1556 | character constant, or identifier. Warn if this context was not | |
1557 | properly terminated by a PDI or PDF. P points to the last character | |
1558 | in this context. */ | |
1559 | ||
1560 | static void | |
1561 | maybe_warn_bidi_on_close (cpp_reader *pfile, const uchar *p) | |
1562 | { | |
1563 | if (CPP_OPTION (pfile, cpp_warn_bidirectional) == bidirectional_unpaired | |
1564 | && bidi::vec.count () > 0) | |
1565 | { | |
1566 | const location_t loc | |
1567 | = linemap_position_for_column (pfile->line_table, | |
1568 | CPP_BUF_COLUMN (pfile->buffer, p)); | |
bef32d4a DM |
1569 | unpaired_bidi_rich_location rich_loc (pfile, loc); |
1570 | /* cpp_callbacks doesn't yet have a way to handle singular vs plural | |
1571 | forms of a diagnostic, so fake it for now. */ | |
1572 | if (bidi::vec.count () > 1) | |
1573 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, | |
1574 | "unpaired UTF-8 bidirectional control characters " | |
1575 | "detected"); | |
1576 | else | |
1577 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, | |
1578 | "unpaired UTF-8 bidirectional control character " | |
1579 | "detected"); | |
51c50026 MP |
1580 | } |
1581 | /* We're done with this context. */ | |
1582 | bidi::on_close (); | |
1583 | } | |
1584 | ||
1585 | /* We're at the beginning or in the middle of an identifier/comment/string | |
1586 | literal/character constant. Warn if we've encountered a bidi character. | |
bef32d4a DM |
1587 | KIND says which bidi control character it was; UCN_P is true iff this bidi |
1588 | control character was written as a UCN. LOC is the location of the | |
1589 | character, but is only valid if KIND != bidi::kind::NONE. */ | |
51c50026 MP |
1590 | |
1591 | static void | |
bef32d4a DM |
1592 | maybe_warn_bidi_on_char (cpp_reader *pfile, bidi::kind kind, |
1593 | bool ucn_p, location_t loc) | |
51c50026 MP |
1594 | { |
1595 | if (__builtin_expect (kind == bidi::kind::NONE, 1)) | |
1596 | return; | |
1597 | ||
1598 | const auto warn_bidi = CPP_OPTION (pfile, cpp_warn_bidirectional); | |
1599 | ||
1600 | if (warn_bidi != bidirectional_none) | |
1601 | { | |
1a7f2c07 DM |
1602 | rich_location rich_loc (pfile->line_table, loc); |
1603 | rich_loc.set_escape_on_output (true); | |
1604 | ||
51c50026 MP |
1605 | /* It seems excessive to warn about a PDI/PDF that is closing |
1606 | an opened context because we've already warned about the | |
1607 | opening character. Except warn when we have a UCN x UTF-8 | |
1608 | mismatch. */ | |
1609 | if (kind == bidi::current_ctx ()) | |
1610 | { | |
1611 | if (warn_bidi == bidirectional_unpaired | |
1612 | && bidi::current_ctx_ucn_p () != ucn_p) | |
bef32d4a DM |
1613 | { |
1614 | rich_loc.add_range (bidi::current_ctx_loc ()); | |
1615 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, | |
1616 | "UTF-8 vs UCN mismatch when closing " | |
1617 | "a context by \"%s\"", bidi::to_str (kind)); | |
1618 | } | |
51c50026 MP |
1619 | } |
1620 | else if (warn_bidi == bidirectional_any) | |
1621 | { | |
1622 | if (kind == bidi::kind::PDF || kind == bidi::kind::PDI) | |
1a7f2c07 DM |
1623 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
1624 | "\"%s\" is closing an unopened context", | |
1625 | bidi::to_str (kind)); | |
51c50026 | 1626 | else |
1a7f2c07 DM |
1627 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
1628 | "found problematic Unicode character \"%s\"", | |
1629 | bidi::to_str (kind)); | |
51c50026 MP |
1630 | } |
1631 | } | |
1632 | /* We're done with this context. */ | |
bef32d4a | 1633 | bidi::on_char (kind, ucn_p, loc); |
51c50026 MP |
1634 | } |
1635 | ||
0d9f234d NB |
1636 | /* Skip a C-style block comment. We find the end of the comment by |
1637 | seeing if an asterisk is before every '/' we encounter. Returns | |
6f572ac2 NB |
1638 | nonzero if comment terminated by EOF, zero otherwise. |
1639 | ||
1640 | Buffer->cur points to the initial asterisk of the comment. */ | |
26aea073 | 1641 | bool |
6cf87ca4 | 1642 | _cpp_skip_block_comment (cpp_reader *pfile) |
45b966db | 1643 | { |
041c3194 | 1644 | cpp_buffer *buffer = pfile->buffer; |
d08dcf87 ZW |
1645 | const uchar *cur = buffer->cur; |
1646 | uchar c; | |
51c50026 | 1647 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
0d9f234d | 1648 | |
d08dcf87 ZW |
1649 | cur++; |
1650 | if (*cur == '/') | |
1651 | cur++; | |
0d9f234d | 1652 | |
26aea073 NB |
1653 | for (;;) |
1654 | { | |
0d9f234d NB |
1655 | /* People like decorating comments with '*', so check for '/' |
1656 | instead for efficiency. */ | |
d08dcf87 ZW |
1657 | c = *cur++; |
1658 | ||
041c3194 | 1659 | if (c == '/') |
45b966db | 1660 | { |
d08dcf87 | 1661 | if (cur[-2] == '*') |
51c50026 MP |
1662 | { |
1663 | if (warn_bidi_p) | |
1664 | maybe_warn_bidi_on_close (pfile, cur); | |
1665 | break; | |
1666 | } | |
041c3194 | 1667 | |
0d9f234d | 1668 | /* Warn about potential nested comments, but not if the '/' |
a1f300c0 | 1669 | comes immediately before the true comment delimiter. |
041c3194 | 1670 | Don't bother to get it right across escaped newlines. */ |
0d9f234d | 1671 | if (CPP_OPTION (pfile, warn_comments) |
d08dcf87 ZW |
1672 | && cur[0] == '*' && cur[1] != '/') |
1673 | { | |
1674 | buffer->cur = cur; | |
87cf0651 SB |
1675 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
1676 | pfile->line_table->highest_line, | |
1677 | CPP_BUF_COL (buffer), | |
1678 | "\"/*\" within comment"); | |
d08dcf87 | 1679 | } |
45b966db | 1680 | } |
26aea073 NB |
1681 | else if (c == '\n') |
1682 | { | |
12f9df4e | 1683 | unsigned int cols; |
d08dcf87 | 1684 | buffer->cur = cur - 1; |
51c50026 MP |
1685 | if (warn_bidi_p) |
1686 | maybe_warn_bidi_on_close (pfile, cur); | |
26aea073 NB |
1687 | _cpp_process_line_notes (pfile, true); |
1688 | if (buffer->next_line >= buffer->rlimit) | |
1689 | return true; | |
1690 | _cpp_clean_line (pfile); | |
12f9df4e PB |
1691 | |
1692 | cols = buffer->next_line - buffer->line_base; | |
1693 | CPP_INCREMENT_LINE (pfile, cols); | |
1694 | ||
d08dcf87 | 1695 | cur = buffer->cur; |
26aea073 | 1696 | } |
51c50026 MP |
1697 | /* If this is a beginning of a UTF-8 encoding, it might be |
1698 | a bidirectional control character. */ | |
1699 | else if (__builtin_expect (c == bidi::utf8_start, 0) && warn_bidi_p) | |
1700 | { | |
bef32d4a DM |
1701 | location_t loc; |
1702 | bidi::kind kind = get_bidi_utf8 (pfile, cur - 1, &loc); | |
1703 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); | |
51c50026 | 1704 | } |
45b966db | 1705 | } |
041c3194 | 1706 | |
d08dcf87 | 1707 | buffer->cur = cur; |
a8eb6044 | 1708 | _cpp_process_line_notes (pfile, true); |
26aea073 | 1709 | return false; |
45b966db ZW |
1710 | } |
1711 | ||
480709cc | 1712 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
da7d8304 | 1713 | terminating newline. Handles escaped newlines. Returns nonzero |
480709cc | 1714 | if a multiline comment. */ |
041c3194 | 1715 | static int |
6cf87ca4 | 1716 | skip_line_comment (cpp_reader *pfile) |
45b966db | 1717 | { |
cbcff6df | 1718 | cpp_buffer *buffer = pfile->buffer; |
620e594b | 1719 | location_t orig_line = pfile->line_table->highest_line; |
51c50026 | 1720 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
041c3194 | 1721 | |
51c50026 MP |
1722 | if (!warn_bidi_p) |
1723 | while (*buffer->cur != '\n') | |
1724 | buffer->cur++; | |
1725 | else | |
1726 | { | |
1727 | while (*buffer->cur != '\n' | |
1728 | && *buffer->cur != bidi::utf8_start) | |
1729 | buffer->cur++; | |
1730 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0)) | |
1731 | { | |
1732 | while (*buffer->cur != '\n') | |
1733 | { | |
1734 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0)) | |
1735 | { | |
bef32d4a DM |
1736 | location_t loc; |
1737 | bidi::kind kind = get_bidi_utf8 (pfile, buffer->cur, &loc); | |
1738 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); | |
51c50026 MP |
1739 | } |
1740 | buffer->cur++; | |
1741 | } | |
1742 | maybe_warn_bidi_on_close (pfile, buffer->cur); | |
1743 | } | |
1744 | } | |
480709cc | 1745 | |
26aea073 | 1746 | _cpp_process_line_notes (pfile, true); |
500bee0a | 1747 | return orig_line != pfile->line_table->highest_line; |
041c3194 | 1748 | } |
45b966db | 1749 | |
26aea073 | 1750 | /* Skips whitespace, saving the next non-whitespace character. */ |
52fadca8 | 1751 | static void |
6cf87ca4 | 1752 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
041c3194 ZW |
1753 | { |
1754 | cpp_buffer *buffer = pfile->buffer; | |
f7d151fb | 1755 | bool saw_NUL = false; |
45b966db | 1756 | |
0d9f234d | 1757 | do |
041c3194 | 1758 | { |
91fcd158 | 1759 | /* Horizontal space always OK. */ |
26aea073 | 1760 | if (c == ' ' || c == '\t') |
0d9f234d | 1761 | ; |
0d9f234d | 1762 | /* Just \f \v or \0 left. */ |
91fcd158 | 1763 | else if (c == '\0') |
f7d151fb | 1764 | saw_NUL = true; |
93c80368 | 1765 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
500bee0a | 1766 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
ebef4e8c NB |
1767 | CPP_BUF_COL (buffer), |
1768 | "%s in preprocessing directive", | |
1769 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d | 1770 | |
0d9f234d | 1771 | c = *buffer->cur++; |
45b966db | 1772 | } |
ec5c56db | 1773 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
1774 | while (is_nvspace (c)); |
1775 | ||
f7d151fb | 1776 | if (saw_NUL) |
bd5e882c DM |
1777 | { |
1778 | encoding_rich_location rich_loc (pfile); | |
1779 | cpp_error_at (pfile, CPP_DL_WARNING, &rich_loc, | |
1780 | "null character(s) ignored"); | |
1781 | } | |
f7d151fb | 1782 | |
480709cc | 1783 | buffer->cur--; |
041c3194 | 1784 | } |
45b966db | 1785 | |
93c80368 NB |
1786 | /* See if the characters of a number token are valid in a name (no |
1787 | '.', '+' or '-'). */ | |
1788 | static int | |
6cf87ca4 | 1789 | name_p (cpp_reader *pfile, const cpp_string *string) |
93c80368 NB |
1790 | { |
1791 | unsigned int i; | |
1792 | ||
1793 | for (i = 0; i < string->len; i++) | |
1794 | if (!is_idchar (string->text[i])) | |
1795 | return 0; | |
1796 | ||
df383483 | 1797 | return 1; |
93c80368 NB |
1798 | } |
1799 | ||
50668cf6 GK |
1800 | /* After parsing an identifier or other sequence, produce a warning about |
1801 | sequences not in NFC/NFKC. */ | |
1802 | static void | |
1803 | warn_about_normalization (cpp_reader *pfile, | |
1804 | const cpp_token *token, | |
1805 | const struct normalize_state *s) | |
1806 | { | |
1807 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1808 | && !pfile->state.skipping) | |
1809 | { | |
bd5e882c DM |
1810 | location_t loc = token->src_loc; |
1811 | ||
1812 | /* If possible, create a location range for the token. */ | |
1813 | if (loc >= RESERVED_LOCATION_COUNT | |
1814 | && token->type != CPP_EOF | |
1815 | /* There must be no line notes to process. */ | |
1816 | && (!(pfile->buffer->cur | |
1817 | >= pfile->buffer->notes[pfile->buffer->cur_note].pos | |
1818 | && !pfile->overlaid_buffer))) | |
1819 | { | |
1820 | source_range tok_range; | |
1821 | tok_range.m_start = loc; | |
1822 | tok_range.m_finish | |
1823 | = linemap_position_for_column (pfile->line_table, | |
1824 | CPP_BUF_COLUMN (pfile->buffer, | |
1825 | pfile->buffer->cur)); | |
1826 | loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
1827 | loc, tok_range, NULL); | |
1828 | } | |
1829 | ||
1830 | encoding_rich_location rich_loc (pfile, loc); | |
1831 | ||
50668cf6 GK |
1832 | /* Make sure that the token is printed using UCNs, even |
1833 | if we'd otherwise happily print UTF-8. */ | |
c3f829c1 | 1834 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
50668cf6 GK |
1835 | size_t sz; |
1836 | ||
1837 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1838 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
bd5e882c DM |
1839 | cpp_warning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
1840 | "`%.*s' is not in NFKC", (int) sz, buf); | |
c264208e | 1841 | else if (CPP_OPTION (pfile, cplusplus)) |
bd5e882c | 1842 | cpp_pedwarning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
c4d6dcac | 1843 | "`%.*s' is not in NFC", (int) sz, buf); |
50668cf6 | 1844 | else |
bd5e882c DM |
1845 | cpp_warning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
1846 | "`%.*s' is not in NFC", (int) sz, buf); | |
55e7f907 | 1847 | free (buf); |
50668cf6 GK |
1848 | } |
1849 | } | |
1850 | ||
7d112d66 LH |
1851 | static const cppchar_t utf8_signifier = 0xC0; |
1852 | ||
1853 | /* Returns TRUE if the sequence starting at buffer->cur is valid in | |
1613e52b | 1854 | an identifier. FIRST is TRUE if this starts an identifier. */ |
51c50026 | 1855 | |
bced6edf | 1856 | static bool |
50668cf6 GK |
1857 | forms_identifier_p (cpp_reader *pfile, int first, |
1858 | struct normalize_state *state) | |
bced6edf | 1859 | { |
1613e52b | 1860 | cpp_buffer *buffer = pfile->buffer; |
51c50026 | 1861 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
1613e52b NB |
1862 | |
1863 | if (*buffer->cur == '$') | |
1864 | { | |
1865 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1866 | return false; | |
1867 | ||
1868 | buffer->cur++; | |
78b8811a | 1869 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
1613e52b | 1870 | { |
78b8811a | 1871 | CPP_OPTION (pfile, warn_dollars) = 0; |
0527bc4e | 1872 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
1613e52b NB |
1873 | } |
1874 | ||
1875 | return true; | |
1876 | } | |
bced6edf | 1877 | |
7d112d66 LH |
1878 | /* Is this a syntactically valid UCN or a valid UTF-8 char? */ |
1879 | if (CPP_OPTION (pfile, extended_identifiers)) | |
bced6edf | 1880 | { |
fbb22910 | 1881 | cppchar_t s; |
7d112d66 LH |
1882 | if (*buffer->cur >= utf8_signifier) |
1883 | { | |
51c50026 MP |
1884 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0) |
1885 | && warn_bidi_p) | |
1886 | { | |
bef32d4a DM |
1887 | location_t loc; |
1888 | bidi::kind kind = get_bidi_utf8 (pfile, buffer->cur, &loc); | |
1889 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); | |
51c50026 | 1890 | } |
7d112d66 LH |
1891 | if (_cpp_valid_utf8 (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1892 | state, &s)) | |
1893 | return true; | |
1894 | } | |
1895 | else if (*buffer->cur == '\\' | |
1896 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) | |
1897 | { | |
1898 | buffer->cur += 2; | |
51c50026 MP |
1899 | if (warn_bidi_p) |
1900 | { | |
bef32d4a DM |
1901 | location_t loc; |
1902 | bidi::kind kind = get_bidi_ucn (pfile, | |
1903 | buffer->cur, | |
1904 | buffer->cur[-1] == 'U', | |
1905 | &loc); | |
1906 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/true, loc); | |
51c50026 | 1907 | } |
7d112d66 LH |
1908 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1909 | state, &s, NULL, NULL)) | |
1910 | return true; | |
1911 | buffer->cur -= 2; | |
1912 | } | |
bced6edf | 1913 | } |
bced6edf | 1914 | |
1613e52b | 1915 | return false; |
bced6edf NB |
1916 | } |
1917 | ||
fb771b9d TT |
1918 | /* Helper function to issue error about improper __VA_OPT__ use. */ |
1919 | static void | |
1920 | maybe_va_opt_error (cpp_reader *pfile) | |
1921 | { | |
1922 | if (CPP_PEDANTIC (pfile) && !CPP_OPTION (pfile, va_opt)) | |
1923 | { | |
1924 | /* __VA_OPT__ should not be accepted at all, but allow it in | |
1925 | system headers. */ | |
bf425849 | 1926 | if (!_cpp_in_system_header (pfile)) |
fb771b9d | 1927 | cpp_error (pfile, CPP_DL_PEDWARN, |
b04445d4 | 1928 | "__VA_OPT__ is not available until C++20"); |
fb771b9d TT |
1929 | } |
1930 | else if (!pfile->state.va_args_ok) | |
1931 | { | |
1932 | /* __VA_OPT__ should only appear in the replacement list of a | |
1933 | variadic macro. */ | |
1934 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1935 | "__VA_OPT__ can only appear in the expansion" | |
b04445d4 | 1936 | " of a C++20 variadic macro"); |
fb771b9d TT |
1937 | } |
1938 | } | |
1939 | ||
17e7cb85 KT |
1940 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1941 | static cpp_hashnode * | |
1942 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1943 | { | |
1944 | cpp_hashnode *result; | |
1945 | const uchar *cur; | |
1946 | unsigned int len; | |
1947 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1948 | ||
1949 | cur = base + 1; | |
1950 | while (ISIDNUM (*cur)) | |
1951 | { | |
1952 | hash = HT_HASHSTEP (hash, *cur); | |
1953 | cur++; | |
1954 | } | |
1955 | len = cur - base; | |
1956 | hash = HT_HASHFINISH (hash, len); | |
1957 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1958 | base, len, hash, HT_ALLOC)); | |
1959 | ||
1960 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1961 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1962 | && !pfile->state.skipping, 0)) | |
1963 | { | |
1964 | /* It is allowed to poison the same identifier twice. */ | |
1965 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1966 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1967 | NODE_NAME (result)); | |
1968 | ||
1969 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1970 | replacement list of a variadic macro. */ | |
1971 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1972 | && !pfile->state.va_args_ok) | |
3976796b ESR |
1973 | { |
1974 | if (CPP_OPTION (pfile, cplusplus)) | |
1975 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1976 | "__VA_ARGS__ can only appear in the expansion" | |
1977 | " of a C++11 variadic macro"); | |
1978 | else | |
1979 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1980 | "__VA_ARGS__ can only appear in the expansion" | |
1981 | " of a C99 variadic macro"); | |
1982 | } | |
17e7cb85 | 1983 | |
fb771b9d TT |
1984 | if (result == pfile->spec_nodes.n__VA_OPT__) |
1985 | maybe_va_opt_error (pfile); | |
1986 | ||
17e7cb85 KT |
1987 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
1988 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1989 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1990 | "identifier \"%s\" is a special operator name in C++", | |
1991 | NODE_NAME (result)); | |
17e7cb85 KT |
1992 | } |
1993 | ||
1994 | return result; | |
1995 | } | |
1996 | ||
1997 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1998 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1999 | cpp_hashnode * | |
2000 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
2001 | { | |
2002 | cpp_hashnode *result; | |
2003 | result = lex_identifier_intern (pfile, (uchar *) name); | |
2004 | return result; | |
2005 | } | |
2006 | ||
bced6edf | 2007 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
0d9f234d | 2008 | static cpp_hashnode * |
50668cf6 | 2009 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
be5ffc59 | 2010 | struct normalize_state *nst, cpp_hashnode **spelling) |
45b966db | 2011 | { |
93c80368 | 2012 | cpp_hashnode *result; |
47e20491 | 2013 | const uchar *cur; |
c6e83800 ZW |
2014 | unsigned int len; |
2015 | unsigned int hash = HT_HASHSTEP (0, *base); | |
51c50026 | 2016 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
2c3fcba6 | 2017 | |
c6e83800 | 2018 | cur = pfile->buffer->cur; |
47e20491 | 2019 | if (! starts_ucn) |
d3f4ff8b JM |
2020 | { |
2021 | while (ISIDNUM (*cur)) | |
2022 | { | |
2023 | hash = HT_HASHSTEP (hash, *cur); | |
2024 | cur++; | |
2025 | } | |
2026 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *(cur - 1)); | |
2027 | } | |
47e20491 | 2028 | pfile->buffer->cur = cur; |
50668cf6 | 2029 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
10cf9bde | 2030 | { |
7d112d66 LH |
2031 | /* Slower version for identifiers containing UCNs |
2032 | or extended chars (including $). */ | |
47e20491 GK |
2033 | do { |
2034 | while (ISIDNUM (*pfile->buffer->cur)) | |
50668cf6 | 2035 | { |
d3f4ff8b | 2036 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *pfile->buffer->cur); |
50668cf6 | 2037 | pfile->buffer->cur++; |
50668cf6 GK |
2038 | } |
2039 | } while (forms_identifier_p (pfile, false, nst)); | |
51c50026 MP |
2040 | if (warn_bidi_p) |
2041 | maybe_warn_bidi_on_close (pfile, pfile->buffer->cur); | |
47e20491 GK |
2042 | result = _cpp_interpret_identifier (pfile, base, |
2043 | pfile->buffer->cur - base); | |
be5ffc59 | 2044 | *spelling = cpp_lookup (pfile, base, pfile->buffer->cur - base); |
2c3fcba6 | 2045 | } |
47e20491 GK |
2046 | else |
2047 | { | |
2048 | len = cur - base; | |
2049 | hash = HT_HASHFINISH (hash, len); | |
bced6edf | 2050 | |
2bf41bf0 TT |
2051 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
2052 | base, len, hash, HT_ALLOC)); | |
be5ffc59 | 2053 | *spelling = result; |
47e20491 | 2054 | } |
2c3fcba6 | 2055 | |
bced6edf | 2056 | /* Rarely, identifiers require diagnostics when lexed. */ |
2c3fcba6 ZW |
2057 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
2058 | && !pfile->state.skipping, 0)) | |
2059 | { | |
2060 | /* It is allowed to poison the same identifier twice. */ | |
2061 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
0527bc4e | 2062 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
2c3fcba6 ZW |
2063 | NODE_NAME (result)); |
2064 | ||
2065 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
2066 | replacement list of a variadic macro. */ | |
2067 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
2068 | && !pfile->state.va_args_ok) | |
3976796b ESR |
2069 | { |
2070 | if (CPP_OPTION (pfile, cplusplus)) | |
2071 | cpp_error (pfile, CPP_DL_PEDWARN, | |
2072 | "__VA_ARGS__ can only appear in the expansion" | |
2073 | " of a C++11 variadic macro"); | |
2074 | else | |
2075 | cpp_error (pfile, CPP_DL_PEDWARN, | |
2076 | "__VA_ARGS__ can only appear in the expansion" | |
2077 | " of a C99 variadic macro"); | |
2078 | } | |
3d8b2a98 | 2079 | |
fb771b9d TT |
2080 | /* __VA_OPT__ should only appear in the replacement list of a |
2081 | variadic macro. */ | |
2082 | if (result == pfile->spec_nodes.n__VA_OPT__) | |
2083 | maybe_va_opt_error (pfile); | |
2084 | ||
3d8b2a98 ILT |
2085 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
2086 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
2087 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
2088 | "identifier \"%s\" is a special operator name in C++", | |
2089 | NODE_NAME (result)); | |
2c3fcba6 ZW |
2090 | } |
2091 | ||
2092 | return result; | |
2093 | } | |
2094 | ||
bced6edf | 2095 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
45b966db | 2096 | static void |
50668cf6 GK |
2097 | lex_number (cpp_reader *pfile, cpp_string *number, |
2098 | struct normalize_state *nst) | |
45b966db | 2099 | { |
562a5c27 | 2100 | const uchar *cur; |
bced6edf NB |
2101 | const uchar *base; |
2102 | uchar *dest; | |
45b966db | 2103 | |
bced6edf NB |
2104 | base = pfile->buffer->cur - 1; |
2105 | do | |
041c3194 | 2106 | { |
8f51cf38 | 2107 | const uchar *adj_digit_sep = NULL; |
bced6edf | 2108 | cur = pfile->buffer->cur; |
0d9f234d | 2109 | |
bced6edf | 2110 | /* N.B. ISIDNUM does not include $. */ |
8f51cf38 JM |
2111 | while (ISIDNUM (*cur) |
2112 | || (*cur == '.' && !DIGIT_SEP (cur[-1])) | |
2113 | || DIGIT_SEP (*cur) | |
2114 | || (VALID_SIGN (*cur, cur[-1]) && !DIGIT_SEP (cur[-2]))) | |
50668cf6 | 2115 | { |
d3f4ff8b | 2116 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *cur); |
8f51cf38 JM |
2117 | /* Adjacent digit separators do not form part of the pp-number syntax. |
2118 | However, they can safely be diagnosed here as an error, since '' is | |
2119 | not a valid preprocessing token. */ | |
2120 | if (DIGIT_SEP (*cur) && DIGIT_SEP (cur[-1]) && !adj_digit_sep) | |
2121 | adj_digit_sep = cur; | |
50668cf6 | 2122 | cur++; |
50668cf6 | 2123 | } |
a5858a3d ESR |
2124 | /* A number can't end with a digit separator. */ |
2125 | while (cur > pfile->buffer->cur && DIGIT_SEP (cur[-1])) | |
2126 | --cur; | |
8f51cf38 JM |
2127 | if (adj_digit_sep && adj_digit_sep < cur) |
2128 | cpp_error (pfile, CPP_DL_ERROR, "adjacent digit separators"); | |
45b966db | 2129 | |
10cf9bde | 2130 | pfile->buffer->cur = cur; |
45b966db | 2131 | } |
50668cf6 | 2132 | while (forms_identifier_p (pfile, false, nst)); |
93c80368 | 2133 | |
bced6edf NB |
2134 | number->len = cur - base; |
2135 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
2136 | memcpy (dest, base, number->len); | |
2137 | dest[number->len] = '\0'; | |
2138 | number->text = dest; | |
93c80368 NB |
2139 | } |
2140 | ||
6338b358 NB |
2141 | /* Create a token of type TYPE with a literal spelling. */ |
2142 | static void | |
6cf87ca4 ZW |
2143 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
2144 | unsigned int len, enum cpp_ttype type) | |
6338b358 | 2145 | { |
6338b358 NB |
2146 | token->type = type; |
2147 | token->val.str.len = len; | |
13f93cf5 NS |
2148 | token->val.str.text = cpp_alloc_token_string (pfile, base, len); |
2149 | } | |
2150 | ||
2151 | const uchar * | |
2152 | cpp_alloc_token_string (cpp_reader *pfile, | |
2153 | const unsigned char *ptr, unsigned len) | |
2154 | { | |
2155 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
2156 | ||
2157 | dest[len] = 0; | |
2158 | memcpy (dest, ptr, len); | |
2159 | return dest; | |
6338b358 NB |
2160 | } |
2161 | ||
ed63c387 NS |
2162 | /* A pair of raw buffer pointers. The currently open one is [1], the |
2163 | first one is [0]. Used for string literal lexing. */ | |
2164 | struct lit_accum { | |
2165 | _cpp_buff *first; | |
2166 | _cpp_buff *last; | |
2167 | const uchar *rpos; | |
2168 | size_t accum; | |
2169 | ||
2170 | lit_accum () | |
2171 | : first (NULL), last (NULL), rpos (0), accum (0) | |
2172 | { | |
2173 | } | |
2174 | ||
2175 | void append (cpp_reader *, const uchar *, size_t); | |
2176 | ||
2177 | void read_begin (cpp_reader *); | |
2178 | bool reading_p () const | |
2179 | { | |
2180 | return rpos != NULL; | |
2181 | } | |
2182 | char read_char () | |
2183 | { | |
2184 | char c = *rpos++; | |
2185 | if (rpos == BUFF_FRONT (last)) | |
2186 | rpos = NULL; | |
2187 | return c; | |
2188 | } | |
2189 | }; | |
2190 | ||
00a81b8b JM |
2191 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
2192 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
2193 | ||
ed63c387 NS |
2194 | void |
2195 | lit_accum::append (cpp_reader *pfile, const uchar *base, size_t len) | |
00a81b8b | 2196 | { |
ed63c387 NS |
2197 | if (!last) |
2198 | /* Starting. */ | |
2199 | first = last = _cpp_get_buff (pfile, len); | |
2200 | else if (len > BUFF_ROOM (last)) | |
00a81b8b | 2201 | { |
ed63c387 NS |
2202 | /* There is insufficient room in the buffer. Copy what we can, |
2203 | and then either extend or create a new one. */ | |
2204 | size_t room = BUFF_ROOM (last); | |
2205 | memcpy (BUFF_FRONT (last), base, room); | |
2206 | BUFF_FRONT (last) += room; | |
00a81b8b JM |
2207 | base += room; |
2208 | len -= room; | |
ed63c387 | 2209 | accum += room; |
00a81b8b | 2210 | |
ed63c387 NS |
2211 | gcc_checking_assert (!rpos); |
2212 | ||
2213 | last = _cpp_append_extend_buff (pfile, last, len); | |
2214 | } | |
00a81b8b | 2215 | |
ed63c387 NS |
2216 | memcpy (BUFF_FRONT (last), base, len); |
2217 | BUFF_FRONT (last) += len; | |
2218 | accum += len; | |
00a81b8b JM |
2219 | } |
2220 | ||
ed63c387 NS |
2221 | void |
2222 | lit_accum::read_begin (cpp_reader *pfile) | |
2223 | { | |
2224 | /* We never accumulate more than 4 chars to read. */ | |
2225 | if (BUFF_ROOM (last) < 4) | |
2226 | ||
2227 | last = _cpp_append_extend_buff (pfile, last, 4); | |
2228 | rpos = BUFF_FRONT (last); | |
2229 | } | |
c865f923 ESR |
2230 | |
2231 | /* Returns true if a macro has been defined. | |
2232 | This might not work if compile with -save-temps, | |
2233 | or preprocess separately from compilation. */ | |
2234 | ||
2235 | static bool | |
2236 | is_macro(cpp_reader *pfile, const uchar *base) | |
2237 | { | |
2238 | const uchar *cur = base; | |
2239 | if (! ISIDST (*cur)) | |
2240 | return false; | |
2241 | unsigned int hash = HT_HASHSTEP (0, *cur); | |
2242 | ++cur; | |
2243 | while (ISIDNUM (*cur)) | |
2244 | { | |
2245 | hash = HT_HASHSTEP (hash, *cur); | |
2246 | ++cur; | |
2247 | } | |
2248 | hash = HT_HASHFINISH (hash, cur - base); | |
2249 | ||
2250 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
2251 | base, cur - base, hash, HT_NO_INSERT)); | |
2252 | ||
3f6677f4 | 2253 | return result && cpp_macro_p (result); |
c865f923 ESR |
2254 | } |
2255 | ||
b44f8ad8 JW |
2256 | /* Returns true if a literal suffix does not have the expected form |
2257 | and is defined as a macro. */ | |
2258 | ||
2259 | static bool | |
2260 | is_macro_not_literal_suffix(cpp_reader *pfile, const uchar *base) | |
2261 | { | |
2262 | /* User-defined literals outside of namespace std must start with a single | |
2263 | underscore, so assume anything of that form really is a UDL suffix. | |
2264 | We don't need to worry about UDLs defined inside namespace std because | |
2265 | their names are reserved, so cannot be used as macro names in valid | |
2266 | programs. */ | |
2267 | if (base[0] == '_' && base[1] != '_') | |
2268 | return false; | |
2269 | return is_macro (pfile, base); | |
2270 | } | |
c865f923 | 2271 | |
ed63c387 NS |
2272 | /* Lexes a raw string. The stored string contains the spelling, |
2273 | including double quotes, delimiter string, '(' and ')', any leading | |
2274 | 'L', 'u', 'U' or 'u8' and 'R' modifier. The created token contains | |
2275 | the type of the literal, or CPP_OTHER if it was not properly | |
2276 | terminated. | |
2277 | ||
2278 | BASE is the start of the token. Updates pfile->buffer->cur to just | |
2279 | after the lexed string. | |
2c6e3f55 JJ |
2280 | |
2281 | The spelling is NUL-terminated, but it is not guaranteed that this | |
2282 | is the first NUL since embedded NULs are preserved. */ | |
2283 | ||
2284 | static void | |
ed63c387 | 2285 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
2c6e3f55 | 2286 | { |
ed63c387 | 2287 | const uchar *pos = base; |
51c50026 | 2288 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
ed63c387 NS |
2289 | |
2290 | /* 'tis a pity this information isn't passed down from the lexer's | |
2291 | initial categorization of the token. */ | |
2292 | enum cpp_ttype type = CPP_STRING; | |
2293 | ||
2294 | if (*pos == 'L') | |
2295 | { | |
2296 | type = CPP_WSTRING; | |
2297 | pos++; | |
2298 | } | |
2299 | else if (*pos == 'U') | |
2300 | { | |
2301 | type = CPP_STRING32; | |
2302 | pos++; | |
2303 | } | |
2304 | else if (*pos == 'u') | |
2305 | { | |
2306 | if (pos[1] == '8') | |
2307 | { | |
2308 | type = CPP_UTF8STRING; | |
2309 | pos++; | |
2310 | } | |
2311 | else | |
2312 | type = CPP_STRING16; | |
2313 | pos++; | |
2314 | } | |
2315 | ||
2316 | gcc_checking_assert (pos[0] == 'R' && pos[1] == '"'); | |
2317 | pos += 2; | |
2318 | ||
00a81b8b | 2319 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 2320 | |
ed63c387 NS |
2321 | /* Skip notes before the ". */ |
2322 | while (note->pos < pos) | |
2323 | ++note; | |
2324 | ||
2325 | lit_accum accum; | |
2326 | ||
2327 | uchar prefix[17]; | |
2328 | unsigned prefix_len = 0; | |
2329 | enum Phase | |
2330 | { | |
2331 | PHASE_PREFIX = -2, | |
2332 | PHASE_NONE = -1, | |
2333 | PHASE_SUFFIX = 0 | |
2334 | } phase = PHASE_PREFIX; | |
2335 | ||
8cf88735 JJ |
2336 | for (;;) |
2337 | { | |
ed63c387 | 2338 | gcc_checking_assert (note->pos >= pos); |
00a81b8b | 2339 | |
ed63c387 NS |
2340 | /* Undo any escaped newlines and trigraphs. */ |
2341 | if (!accum.reading_p () && note->pos == pos) | |
2342 | switch (note->type) | |
2343 | { | |
2344 | case '\\': | |
2345 | case ' ': | |
2346 | /* Restore backslash followed by newline. */ | |
2347 | accum.append (pfile, base, pos - base); | |
2348 | base = pos; | |
2349 | accum.read_begin (pfile); | |
2350 | accum.append (pfile, UC"\\", 1); | |
2351 | ||
2352 | after_backslash: | |
2353 | if (note->type == ' ') | |
2354 | /* GNU backslash whitespace newline extension. FIXME | |
2355 | could be any sequence of non-vertical space. When we | |
2356 | can properly restore any such sequence, we should | |
2357 | mark this note as handled so _cpp_process_line_notes | |
2358 | doesn't warn. */ | |
2359 | accum.append (pfile, UC" ", 1); | |
2360 | ||
2361 | accum.append (pfile, UC"\n", 1); | |
2362 | note++; | |
2363 | break; | |
00a81b8b | 2364 | |
ed63c387 NS |
2365 | case '\n': |
2366 | /* This can happen for ??/<NEWLINE> when trigraphs are not | |
2367 | being interpretted. */ | |
2368 | gcc_checking_assert (!CPP_OPTION (pfile, trigraphs)); | |
2369 | note->type = 0; | |
2370 | note++; | |
2371 | break; | |
00a81b8b | 2372 | |
ed63c387 NS |
2373 | default: |
2374 | gcc_checking_assert (_cpp_trigraph_map[note->type]); | |
2375 | ||
2376 | /* Don't warn about this trigraph in | |
2377 | _cpp_process_line_notes, since trigraphs show up as | |
2378 | trigraphs in raw strings. */ | |
2379 | uchar type = note->type; | |
2380 | note->type = 0; | |
2381 | ||
2382 | if (CPP_OPTION (pfile, trigraphs)) | |
2383 | { | |
2384 | accum.append (pfile, base, pos - base); | |
2385 | base = pos; | |
2386 | accum.read_begin (pfile); | |
2387 | accum.append (pfile, UC"??", 2); | |
2388 | accum.append (pfile, &type, 1); | |
2389 | ||
2390 | /* ??/ followed by newline gets two line notes, one for | |
2391 | the trigraph and one for the backslash/newline. */ | |
2392 | if (type == '/' && note[1].pos == pos) | |
2393 | { | |
2394 | note++; | |
2395 | gcc_assert (note->type == '\\' || note->type == ' '); | |
2396 | goto after_backslash; | |
2397 | } | |
2398 | /* Skip the replacement character. */ | |
2399 | base = ++pos; | |
2400 | } | |
2401 | ||
2402 | note++; | |
2403 | break; | |
2404 | } | |
2405 | ||
2406 | /* Now get a char to process. Either from an expanded note, or | |
2407 | from the line buffer. */ | |
2408 | bool read_note = accum.reading_p (); | |
2409 | char c = read_note ? accum.read_char () : *pos++; | |
2c6e3f55 | 2410 | |
ed63c387 | 2411 | if (phase == PHASE_PREFIX) |
2c6e3f55 | 2412 | { |
ed63c387 | 2413 | if (c == '(') |
8cf88735 | 2414 | { |
ed63c387 NS |
2415 | /* Done. */ |
2416 | phase = PHASE_NONE; | |
2417 | prefix[prefix_len++] = '"'; | |
2418 | } | |
2419 | else if (prefix_len < 16 | |
2420 | /* Prefix chars are any of the basic character set, | |
2421 | [lex.charset] except for ' | |
2422 | ()\\\t\v\f\n'. Optimized for a contiguous | |
2423 | alphabet. */ | |
2424 | /* Unlike a switch, this collapses down to one or | |
2425 | two shift and bitmask operations on an ASCII | |
2426 | system, with an outlier or two. */ | |
2427 | && (('Z' - 'A' == 25 | |
2428 | ? ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) | |
2429 | : ISIDST (c)) | |
2430 | || (c >= '0' && c <= '9') | |
2431 | || c == '_' || c == '{' || c == '}' | |
2432 | || c == '[' || c == ']' || c == '#' | |
2433 | || c == '<' || c == '>' || c == '%' | |
2434 | || c == ':' || c == ';' || c == '.' || c == '?' | |
2435 | || c == '*' || c == '+' || c == '-' || c == '/' | |
2436 | || c == '^' || c == '&' || c == '|' || c == '~' | |
2437 | || c == '!' || c == '=' || c == ',' | |
2438 | || c == '"' || c == '\'')) | |
2439 | prefix[prefix_len++] = c; | |
2440 | else | |
2441 | { | |
2442 | /* Something is wrong. */ | |
2443 | int col = CPP_BUF_COLUMN (pfile->buffer, pos) + read_note; | |
2444 | if (prefix_len == 16) | |
2445 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
2446 | col, "raw string delimiter longer " | |
2447 | "than 16 characters"); | |
2448 | else if (c == '\n') | |
2449 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
2450 | col, "invalid new-line in raw " | |
2451 | "string delimiter"); | |
2452 | else | |
2453 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
2454 | col, "invalid character '%c' in " | |
2455 | "raw string delimiter", c); | |
2456 | type = CPP_OTHER; | |
2457 | phase = PHASE_NONE; | |
2458 | /* Continue until we get a close quote, that's probably | |
2459 | the best failure mode. */ | |
2460 | prefix_len = 0; | |
8cf88735 | 2461 | } |
ae49af94 JJ |
2462 | if (c != '\n') |
2463 | continue; | |
8cf88735 | 2464 | } |
ed63c387 NS |
2465 | |
2466 | if (phase != PHASE_NONE) | |
8cf88735 | 2467 | { |
ed63c387 NS |
2468 | if (prefix[phase] != c) |
2469 | phase = PHASE_NONE; | |
2470 | else if (unsigned (phase + 1) == prefix_len) | |
8cf88735 | 2471 | break; |
ed63c387 NS |
2472 | else |
2473 | { | |
2474 | phase = Phase (phase + 1); | |
2475 | continue; | |
2476 | } | |
2c6e3f55 | 2477 | } |
ed63c387 NS |
2478 | |
2479 | if (!prefix_len && c == '"') | |
2480 | /* Failure mode lexing. */ | |
2481 | goto out; | |
2482 | else if (prefix_len && c == ')') | |
2483 | phase = PHASE_SUFFIX; | |
2484 | else if (!read_note && c == '\n') | |
2c6e3f55 | 2485 | { |
ed63c387 NS |
2486 | pos--; |
2487 | pfile->buffer->cur = pos; | |
2c6e3f55 | 2488 | if (pfile->state.in_directive |
d5e48350 JJ |
2489 | || (pfile->state.parsing_args |
2490 | && pfile->buffer->next_line >= pfile->buffer->rlimit)) | |
2c6e3f55 | 2491 | { |
2c6e3f55 JJ |
2492 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, |
2493 | "unterminated raw string"); | |
ed63c387 NS |
2494 | type = CPP_OTHER; |
2495 | goto out; | |
2c6e3f55 JJ |
2496 | } |
2497 | ||
ed63c387 | 2498 | accum.append (pfile, base, pos - base + 1); |
2a0225e4 NS |
2499 | _cpp_process_line_notes (pfile, false); |
2500 | ||
2501 | if (pfile->buffer->next_line < pfile->buffer->rlimit) | |
2c6e3f55 JJ |
2502 | CPP_INCREMENT_LINE (pfile, 0); |
2503 | pfile->buffer->need_line = true; | |
2504 | ||
2505 | if (!_cpp_get_fresh_line (pfile)) | |
2506 | { | |
ed63c387 | 2507 | /* We ran out of file and failed to get a line. */ |
620e594b | 2508 | location_t src_loc = token->src_loc; |
2c6e3f55 JJ |
2509 | token->type = CPP_EOF; |
2510 | /* Tell the compiler the line number of the EOF token. */ | |
2511 | token->src_loc = pfile->line_table->highest_line; | |
2512 | token->flags = BOL; | |
ed63c387 NS |
2513 | if (accum.first) |
2514 | _cpp_release_buff (pfile, accum.first); | |
2c6e3f55 JJ |
2515 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, |
2516 | "unterminated raw string"); | |
2a0225e4 NS |
2517 | /* Now pop the buffer that _cpp_get_fresh_line did not. */ |
2518 | _cpp_pop_buffer (pfile); | |
2c6e3f55 JJ |
2519 | return; |
2520 | } | |
2521 | ||
ed63c387 | 2522 | pos = base = pfile->buffer->cur; |
00a81b8b | 2523 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 2524 | } |
51c50026 MP |
2525 | else if (__builtin_expect ((unsigned char) c == bidi::utf8_start, 0) |
2526 | && warn_bidi_p) | |
bef32d4a DM |
2527 | { |
2528 | location_t loc; | |
2529 | bidi::kind kind = get_bidi_utf8 (pfile, pos - 1, &loc); | |
2530 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); | |
2531 | } | |
2c6e3f55 JJ |
2532 | } |
2533 | ||
51c50026 MP |
2534 | if (warn_bidi_p) |
2535 | maybe_warn_bidi_on_close (pfile, pos); | |
2536 | ||
3ce4f9e4 ESR |
2537 | if (CPP_OPTION (pfile, user_literals)) |
2538 | { | |
c865f923 ESR |
2539 | /* If a string format macro, say from inttypes.h, is placed touching |
2540 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 | 2541 | literal thus breaking the program. */ |
ed63c387 | 2542 | if (is_macro_not_literal_suffix (pfile, pos)) |
7f5f5f98 | 2543 | { |
112448b4 | 2544 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 2545 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
2546 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
2547 | token->src_loc, 0, | |
2548 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 2549 | "a space between literal and string macro"); |
7f5f5f98 | 2550 | } |
3ce4f9e4 | 2551 | /* Grab user defined literal suffix. */ |
ed63c387 | 2552 | else if (ISIDST (*pos)) |
3ce4f9e4 ESR |
2553 | { |
2554 | type = cpp_userdef_string_add_type (type); | |
ed63c387 | 2555 | ++pos; |
7f5f5f98 | 2556 | |
ed63c387 NS |
2557 | while (ISIDNUM (*pos)) |
2558 | ++pos; | |
3ce4f9e4 | 2559 | } |
3ce4f9e4 ESR |
2560 | } |
2561 | ||
ed63c387 NS |
2562 | out: |
2563 | pfile->buffer->cur = pos; | |
2564 | if (!accum.accum) | |
2565 | create_literal (pfile, token, base, pos - base, type); | |
2c6e3f55 JJ |
2566 | else |
2567 | { | |
ed63c387 NS |
2568 | size_t extra_len = pos - base; |
2569 | uchar *dest = _cpp_unaligned_alloc (pfile, accum.accum + extra_len + 1); | |
2c6e3f55 JJ |
2570 | |
2571 | token->type = type; | |
ed63c387 | 2572 | token->val.str.len = accum.accum + extra_len; |
2c6e3f55 | 2573 | token->val.str.text = dest; |
ed63c387 | 2574 | for (_cpp_buff *buf = accum.first; buf; buf = buf->next) |
2c6e3f55 | 2575 | { |
ed63c387 NS |
2576 | size_t len = BUFF_FRONT (buf) - buf->base; |
2577 | memcpy (dest, buf->base, len); | |
2578 | dest += len; | |
2c6e3f55 | 2579 | } |
ed63c387 NS |
2580 | _cpp_release_buff (pfile, accum.first); |
2581 | memcpy (dest, base, extra_len); | |
2582 | dest[extra_len] = '\0'; | |
2c6e3f55 JJ |
2583 | } |
2584 | } | |
2585 | ||
bced6edf | 2586 | /* Lexes a string, character constant, or angle-bracketed header file |
6338b358 | 2587 | name. The stored string contains the spelling, including opening |
2c6e3f55 JJ |
2588 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
2589 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
2590 | if it was not properly terminated, or CPP_LESS for an unterminated | |
2591 | header name which must be relexed as normal tokens. | |
6338b358 NB |
2592 | |
2593 | The spelling is NUL-terminated, but it is not guaranteed that this | |
2594 | is the first NUL since embedded NULs are preserved. */ | |
041c3194 | 2595 | static void |
6cf87ca4 | 2596 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
45b966db | 2597 | { |
6338b358 NB |
2598 | bool saw_NUL = false; |
2599 | const uchar *cur; | |
bced6edf | 2600 | cppchar_t terminator; |
6338b358 NB |
2601 | enum cpp_ttype type; |
2602 | ||
2603 | cur = base; | |
2604 | terminator = *cur++; | |
2c6e3f55 | 2605 | if (terminator == 'L' || terminator == 'U') |
6338b358 | 2606 | terminator = *cur++; |
2c6e3f55 JJ |
2607 | else if (terminator == 'u') |
2608 | { | |
2609 | terminator = *cur++; | |
2610 | if (terminator == '8') | |
2611 | terminator = *cur++; | |
2612 | } | |
2613 | if (terminator == 'R') | |
2614 | { | |
ed63c387 | 2615 | lex_raw_string (pfile, token, base); |
2c6e3f55 JJ |
2616 | return; |
2617 | } | |
2618 | if (terminator == '"') | |
b6baa67d KVH |
2619 | type = (*base == 'L' ? CPP_WSTRING : |
2620 | *base == 'U' ? CPP_STRING32 : | |
2c6e3f55 JJ |
2621 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
2622 | : CPP_STRING); | |
6338b358 | 2623 | else if (terminator == '\'') |
b6baa67d KVH |
2624 | type = (*base == 'L' ? CPP_WCHAR : |
2625 | *base == 'U' ? CPP_CHAR32 : | |
fe95b036 ESR |
2626 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8CHAR : CPP_CHAR16) |
2627 | : CPP_CHAR); | |
6338b358 NB |
2628 | else |
2629 | terminator = '>', type = CPP_HEADER_NAME; | |
93c80368 | 2630 | |
51c50026 | 2631 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
0d9f234d | 2632 | for (;;) |
45b966db | 2633 | { |
6338b358 | 2634 | cppchar_t c = *cur++; |
7868b4a2 | 2635 | |
6f572ac2 | 2636 | /* In #include-style directives, terminators are not escapable. */ |
6338b358 | 2637 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
51c50026 MP |
2638 | { |
2639 | if ((cur[0] == 'u' || cur[0] == 'U') && warn_bidi_p) | |
2640 | { | |
bef32d4a DM |
2641 | location_t loc; |
2642 | bidi::kind kind = get_bidi_ucn (pfile, cur + 1, cur[0] == 'U', | |
2643 | &loc); | |
2644 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/true, loc); | |
51c50026 MP |
2645 | } |
2646 | cur++; | |
2647 | } | |
6338b358 | 2648 | else if (c == terminator) |
51c50026 MP |
2649 | { |
2650 | if (warn_bidi_p) | |
2651 | maybe_warn_bidi_on_close (pfile, cur - 1); | |
2652 | break; | |
2653 | } | |
6338b358 | 2654 | else if (c == '\n') |
0d9f234d | 2655 | { |
6338b358 | 2656 | cur--; |
4bb09c26 JM |
2657 | /* Unmatched quotes always yield undefined behavior, but |
2658 | greedy lexing means that what appears to be an unterminated | |
2659 | header name may actually be a legitimate sequence of tokens. */ | |
2660 | if (terminator == '>') | |
2661 | { | |
2662 | token->type = CPP_LESS; | |
2663 | return; | |
2664 | } | |
6338b358 NB |
2665 | type = CPP_OTHER; |
2666 | break; | |
45b966db | 2667 | } |
6338b358 NB |
2668 | else if (c == '\0') |
2669 | saw_NUL = true; | |
51c50026 MP |
2670 | else if (__builtin_expect (c == bidi::utf8_start, 0) && warn_bidi_p) |
2671 | { | |
bef32d4a DM |
2672 | location_t loc; |
2673 | bidi::kind kind = get_bidi_utf8 (pfile, cur - 1, &loc); | |
2674 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); | |
51c50026 | 2675 | } |
45b966db ZW |
2676 | } |
2677 | ||
6338b358 | 2678 | if (saw_NUL && !pfile->state.skipping) |
0527bc4e JDA |
2679 | cpp_error (pfile, CPP_DL_WARNING, |
2680 | "null character(s) preserved in literal"); | |
45b966db | 2681 | |
c663e301 JM |
2682 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
2683 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
2684 | (int) terminator); | |
2685 | ||
3ce4f9e4 ESR |
2686 | if (CPP_OPTION (pfile, user_literals)) |
2687 | { | |
c865f923 ESR |
2688 | /* If a string format macro, say from inttypes.h, is placed touching |
2689 | a string literal it could be parsed as a C++11 user-defined string | |
b44f8ad8 JW |
2690 | literal thus breaking the program. */ |
2691 | if (is_macro_not_literal_suffix (pfile, cur)) | |
7f5f5f98 | 2692 | { |
112448b4 | 2693 | /* Raise a warning, but do not consume subsequent tokens. */ |
7aee8646 | 2694 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
7f5f5f98 OW |
2695 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
2696 | token->src_loc, 0, | |
2697 | "invalid suffix on literal; C++11 requires " | |
c865f923 | 2698 | "a space between literal and string macro"); |
7f5f5f98 | 2699 | } |
3ce4f9e4 | 2700 | /* Grab user defined literal suffix. */ |
561f7fc7 | 2701 | else if (ISIDST (*cur)) |
3ce4f9e4 ESR |
2702 | { |
2703 | type = cpp_userdef_char_add_type (type); | |
2704 | type = cpp_userdef_string_add_type (type); | |
2705 | ++cur; | |
7f5f5f98 OW |
2706 | |
2707 | while (ISIDNUM (*cur)) | |
2708 | ++cur; | |
3ce4f9e4 | 2709 | } |
3ce4f9e4 | 2710 | } |
fe191308 JM |
2711 | else if (CPP_OPTION (pfile, cpp_warn_cxx11_compat) |
2712 | && is_macro (pfile, cur) | |
2713 | && !pfile->state.skipping) | |
2714 | cpp_warning_with_line (pfile, CPP_W_CXX11_COMPAT, | |
2715 | token->src_loc, 0, "C++11 requires a space " | |
2716 | "between string literal and macro"); | |
3ce4f9e4 | 2717 | |
6338b358 NB |
2718 | pfile->buffer->cur = cur; |
2719 | create_literal (pfile, token, base, cur - base, type); | |
0d9f234d | 2720 | } |
041c3194 | 2721 | |
631d0d36 MG |
2722 | /* Return the comment table. The client may not make any assumption |
2723 | about the ordering of the table. */ | |
2724 | cpp_comment_table * | |
2725 | cpp_get_comments (cpp_reader *pfile) | |
2726 | { | |
2727 | return &pfile->comments; | |
2728 | } | |
2729 | ||
2730 | /* Append a comment to the end of the comment table. */ | |
2731 | static void | |
2732 | store_comment (cpp_reader *pfile, cpp_token *token) | |
2733 | { | |
2734 | int len; | |
2735 | ||
2736 | if (pfile->comments.allocated == 0) | |
2737 | { | |
2738 | pfile->comments.allocated = 256; | |
2739 | pfile->comments.entries = (cpp_comment *) xmalloc | |
2740 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
2741 | } | |
2742 | ||
2743 | if (pfile->comments.count == pfile->comments.allocated) | |
2744 | { | |
2745 | pfile->comments.allocated *= 2; | |
2746 | pfile->comments.entries = (cpp_comment *) xrealloc | |
2747 | (pfile->comments.entries, | |
2748 | pfile->comments.allocated * sizeof (cpp_comment)); | |
2749 | } | |
2750 | ||
2751 | len = token->val.str.len; | |
2752 | ||
2753 | /* Copy comment. Note, token may not be NULL terminated. */ | |
2754 | pfile->comments.entries[pfile->comments.count].comment = | |
2755 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
2756 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
2757 | token->val.str.text, len); | |
2758 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
2759 | ||
2760 | /* Set source location. */ | |
2761 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
2762 | ||
2763 | /* Increment the count of entries in the comment table. */ | |
2764 | pfile->comments.count++; | |
2765 | } | |
2766 | ||
93c80368 | 2767 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 2768 | static void |
6cf87ca4 ZW |
2769 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
2770 | cppchar_t type) | |
9e62c811 | 2771 | { |
041c3194 | 2772 | unsigned char *buffer; |
651a20b5 | 2773 | unsigned int len, clen, i; |
df383483 | 2774 | |
1c6d33ef | 2775 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
480709cc | 2776 | |
3542203b NB |
2777 | /* C++ comments probably (not definitely) have moved past a new |
2778 | line, which we don't want to save in the comment. */ | |
480709cc | 2779 | if (is_vspace (pfile->buffer->cur[-1])) |
3542203b | 2780 | len--; |
477cdac7 | 2781 | |
651a20b5 KT |
2782 | /* If we are currently in a directive or in argument parsing, then |
2783 | we need to store all C++ comments as C comments internally, and | |
2784 | so we need to allocate a little extra space in that case. | |
477cdac7 JT |
2785 | |
2786 | Note that the only time we encounter a directive here is | |
2787 | when we are saving comments in a "#define". */ | |
651a20b5 KT |
2788 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
2789 | && type == '/') ? len + 2 : len; | |
477cdac7 JT |
2790 | |
2791 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
df383483 | 2792 | |
041c3194 | 2793 | token->type = CPP_COMMENT; |
477cdac7 | 2794 | token->val.str.len = clen; |
0d9f234d | 2795 | token->val.str.text = buffer; |
45b966db | 2796 | |
1c6d33ef NB |
2797 | buffer[0] = '/'; |
2798 | memcpy (buffer + 1, from, len - 1); | |
477cdac7 | 2799 | |
1eeeb6a4 | 2800 | /* Finish conversion to a C comment, if necessary. */ |
651a20b5 | 2801 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
477cdac7 JT |
2802 | { |
2803 | buffer[1] = '*'; | |
2804 | buffer[clen - 2] = '*'; | |
2805 | buffer[clen - 1] = '/'; | |
651a20b5 KT |
2806 | /* As there can be in a C++ comments illegal sequences for C comments |
2807 | we need to filter them out. */ | |
2808 | for (i = 2; i < (clen - 2); i++) | |
2809 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
2810 | buffer[i] = '|'; | |
477cdac7 | 2811 | } |
631d0d36 MG |
2812 | |
2813 | /* Finally store this comment for use by clients of libcpp. */ | |
2814 | store_comment (pfile, token); | |
0d9f234d | 2815 | } |
45b966db | 2816 | |
81fea426 MP |
2817 | /* Returns true if comment at COMMENT_START is a recognized FALLTHROUGH |
2818 | comment. */ | |
2819 | ||
2820 | static bool | |
2821 | fallthrough_comment_p (cpp_reader *pfile, const unsigned char *comment_start) | |
2822 | { | |
2823 | const unsigned char *from = comment_start + 1; | |
70f6d5e1 JJ |
2824 | |
2825 | switch (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough)) | |
2826 | { | |
2827 | /* For both -Wimplicit-fallthrough=0 and -Wimplicit-fallthrough=5 we | |
2828 | don't recognize any comments. The latter only checks attributes, | |
2829 | the former doesn't warn. */ | |
2830 | case 0: | |
2831 | default: | |
2832 | return false; | |
2833 | /* -Wimplicit-fallthrough=1 considers any comment, no matter what | |
2834 | content it has. */ | |
2835 | case 1: | |
2836 | return true; | |
2837 | case 2: | |
2838 | /* -Wimplicit-fallthrough=2 looks for (case insensitive) | |
2839 | .*falls?[ \t-]*thr(u|ough).* regex. */ | |
2840 | for (; (size_t) (pfile->buffer->cur - from) >= sizeof "fallthru" - 1; | |
2841 | from++) | |
2842 | { | |
2843 | /* Is there anything like strpbrk with upper boundary, or | |
2844 | memchr looking for 2 characters rather than just one? */ | |
2845 | if (from[0] != 'f' && from[0] != 'F') | |
2846 | continue; | |
2847 | if (from[1] != 'a' && from[1] != 'A') | |
2848 | continue; | |
2849 | if (from[2] != 'l' && from[2] != 'L') | |
2850 | continue; | |
2851 | if (from[3] != 'l' && from[3] != 'L') | |
2852 | continue; | |
2853 | from += sizeof "fall" - 1; | |
2854 | if (from[0] == 's' || from[0] == 'S') | |
2855 | from++; | |
2856 | while (*from == ' ' || *from == '\t' || *from == '-') | |
2857 | from++; | |
2858 | if (from[0] != 't' && from[0] != 'T') | |
2859 | continue; | |
2860 | if (from[1] != 'h' && from[1] != 'H') | |
2861 | continue; | |
2862 | if (from[2] != 'r' && from[2] != 'R') | |
2863 | continue; | |
2864 | if (from[3] == 'u' || from[3] == 'U') | |
2865 | return true; | |
2866 | if (from[3] != 'o' && from[3] != 'O') | |
2867 | continue; | |
2868 | if (from[4] != 'u' && from[4] != 'U') | |
2869 | continue; | |
2870 | if (from[5] != 'g' && from[5] != 'G') | |
2871 | continue; | |
2872 | if (from[6] != 'h' && from[6] != 'H') | |
2873 | continue; | |
2874 | return true; | |
2875 | } | |
2876 | return false; | |
2877 | case 3: | |
2878 | case 4: | |
2879 | break; | |
2880 | } | |
2881 | ||
81fea426 MP |
2882 | /* Whole comment contents: |
2883 | -fallthrough | |
2884 | @fallthrough@ | |
2885 | */ | |
2886 | if (*from == '-' || *from == '@') | |
2887 | { | |
2888 | size_t len = sizeof "fallthrough" - 1; | |
2889 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2890 | return false; | |
2891 | if (memcmp (from + 1, "fallthrough", len)) | |
2892 | return false; | |
2893 | if (*from == '@') | |
2894 | { | |
2895 | if (from[len + 1] != '@') | |
2896 | return false; | |
2897 | len++; | |
2898 | } | |
2899 | from += 1 + len; | |
2900 | } | |
2901 | /* Whole comment contents (regex): | |
70f6d5e1 | 2902 | lint -fallthrough[ \t]* |
81b02905 JJ |
2903 | */ |
2904 | else if (*from == 'l') | |
2905 | { | |
2906 | size_t len = sizeof "int -fallthrough" - 1; | |
2907 | if ((size_t) (pfile->buffer->cur - from - 1) < len) | |
2908 | return false; | |
2909 | if (memcmp (from + 1, "int -fallthrough", len)) | |
70f6d5e1 | 2910 | return false; |
81b02905 | 2911 | from += 1 + len; |
70f6d5e1 JJ |
2912 | while (*from == ' ' || *from == '\t') |
2913 | from++; | |
2914 | } | |
2915 | /* Whole comment contents (regex): | |
2916 | [ \t]*FALLTHR(U|OUGH)[ \t]* | |
2917 | */ | |
2918 | else if (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough) == 4) | |
2919 | { | |
2920 | while (*from == ' ' || *from == '\t') | |
2921 | from++; | |
2922 | if ((size_t) (pfile->buffer->cur - from) < sizeof "FALLTHRU" - 1) | |
2923 | return false; | |
2924 | if (memcmp (from, "FALLTHR", sizeof "FALLTHR" - 1)) | |
2925 | return false; | |
2926 | from += sizeof "FALLTHR" - 1; | |
2927 | if (*from == 'U') | |
2928 | from++; | |
2929 | else if ((size_t) (pfile->buffer->cur - from) < sizeof "OUGH" - 1) | |
2930 | return false; | |
2931 | else if (memcmp (from, "OUGH", sizeof "OUGH" - 1)) | |
2932 | return false; | |
2933 | else | |
2934 | from += sizeof "OUGH" - 1; | |
2935 | while (*from == ' ' || *from == '\t') | |
2936 | from++; | |
81b02905 JJ |
2937 | } |
2938 | /* Whole comment contents (regex): | |
ee19ef45 JJ |
2939 | [ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)? |
2940 | [ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
2941 | [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)? | |
81fea426 MP |
2942 | */ |
2943 | else | |
2944 | { | |
81b02905 | 2945 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 MP |
2946 | from++; |
2947 | unsigned char f = *from; | |
81b02905 JJ |
2948 | bool all_upper = false; |
2949 | if (f == 'E' || f == 'e') | |
70f6d5e1 JJ |
2950 | { |
2951 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2952 | < sizeof "else fallthru" - 1) |
2953 | return false; | |
ee19ef45 | 2954 | if (f == 'E' && memcmp (from + 1, "LSE", sizeof "LSE" - 1) == 0) |
81b02905 | 2955 | all_upper = true; |
ee19ef45 JJ |
2956 | else if (memcmp (from + 1, "lse", sizeof "lse" - 1)) |
2957 | return false; | |
2958 | from += sizeof "else" - 1; | |
2959 | if (*from == ',') | |
2960 | from++; | |
70f6d5e1 | 2961 | if (*from != ' ') |
ee19ef45 JJ |
2962 | return false; |
2963 | from++; | |
2964 | if (all_upper && *from == 'f') | |
81b02905 | 2965 | return false; |
81b02905 JJ |
2966 | if (f == 'e' && *from == 'F') |
2967 | return false; | |
2968 | f = *from; | |
70f6d5e1 | 2969 | } |
81b02905 | 2970 | else if (f == 'I' || f == 'i') |
70f6d5e1 JJ |
2971 | { |
2972 | if ((size_t) (pfile->buffer->cur - from) | |
81b02905 JJ |
2973 | < sizeof "intentional fallthru" - 1) |
2974 | return false; | |
2975 | if (f == 'I' && memcmp (from + 1, "NTENTIONAL", | |
2976 | sizeof "NTENTIONAL" - 1) == 0) | |
2977 | all_upper = true; | |
2978 | else if (memcmp (from + 1, "ntentional", | |
2979 | sizeof "ntentional" - 1)) | |
2980 | return false; | |
2981 | from += sizeof "intentional" - 1; | |
2982 | if (*from == ' ') | |
2983 | { | |
2984 | from++; | |
2985 | if (all_upper && *from == 'f') | |
2986 | return false; | |
2987 | } | |
2988 | else if (all_upper) | |
2989 | { | |
2990 | if (memcmp (from, "LY F", sizeof "LY F" - 1)) | |
2991 | return false; | |
2992 | from += sizeof "LY " - 1; | |
2993 | } | |
2994 | else | |
2995 | { | |
2996 | if (memcmp (from, "ly ", sizeof "ly " - 1)) | |
2997 | return false; | |
2998 | from += sizeof "ly " - 1; | |
2999 | } | |
3000 | if (f == 'i' && *from == 'F') | |
3001 | return false; | |
3002 | f = *from; | |
70f6d5e1 | 3003 | } |
81fea426 MP |
3004 | if (f != 'F' && f != 'f') |
3005 | return false; | |
7bad794a | 3006 | if ((size_t) (pfile->buffer->cur - from) < sizeof "fallthru" - 1) |
81fea426 | 3007 | return false; |
81fea426 MP |
3008 | if (f == 'F' && memcmp (from + 1, "ALL", sizeof "ALL" - 1) == 0) |
3009 | all_upper = true; | |
81b02905 JJ |
3010 | else if (all_upper) |
3011 | return false; | |
81fea426 MP |
3012 | else if (memcmp (from + 1, "all", sizeof "all" - 1)) |
3013 | return false; | |
7bad794a JJ |
3014 | from += sizeof "fall" - 1; |
3015 | if (*from == (all_upper ? 'S' : 's') && from[1] == ' ') | |
3016 | from += 2; | |
3017 | else if (*from == ' ' || *from == '-') | |
3018 | from++; | |
3019 | else if (*from != (all_upper ? 'T' : 't')) | |
81fea426 | 3020 | return false; |
81fea426 MP |
3021 | if ((f == 'f' || *from != 'T') && (all_upper || *from != 't')) |
3022 | return false; | |
7bad794a | 3023 | if ((size_t) (pfile->buffer->cur - from) < sizeof "thru" - 1) |
81fea426 MP |
3024 | return false; |
3025 | if (memcmp (from + 1, all_upper ? "HRU" : "hru", sizeof "hru" - 1)) | |
3026 | { | |
7bad794a | 3027 | if ((size_t) (pfile->buffer->cur - from) < sizeof "through" - 1) |
81fea426 MP |
3028 | return false; |
3029 | if (memcmp (from + 1, all_upper ? "HROUGH" : "hrough", | |
3030 | sizeof "hrough" - 1)) | |
3031 | return false; | |
3032 | from += sizeof "through" - 1; | |
3033 | } | |
3034 | else | |
3035 | from += sizeof "thru" - 1; | |
81b02905 | 3036 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
81fea426 | 3037 | from++; |
81b02905 JJ |
3038 | if (*from == '-') |
3039 | { | |
3040 | from++; | |
3041 | if (*comment_start == '*') | |
3042 | { | |
3043 | do | |
3044 | { | |
3045 | while (*from && *from != '*' | |
3046 | && *from != '\n' && *from != '\r') | |
3047 | from++; | |
3048 | if (*from != '*' || from[1] == '/') | |
3049 | break; | |
3050 | from++; | |
3051 | } | |
3052 | while (1); | |
3053 | } | |
3054 | else | |
3055 | while (*from && *from != '\n' && *from != '\r') | |
3056 | from++; | |
3057 | } | |
81fea426 MP |
3058 | } |
3059 | /* C block comment. */ | |
3060 | if (*comment_start == '*') | |
3061 | { | |
3062 | if (*from != '*' || from[1] != '/') | |
3063 | return false; | |
3064 | } | |
3065 | /* C++ line comment. */ | |
3066 | else if (*from != '\n') | |
3067 | return false; | |
3068 | ||
3069 | return true; | |
3070 | } | |
3071 | ||
5fddcffc NB |
3072 | /* Allocate COUNT tokens for RUN. */ |
3073 | void | |
6cf87ca4 | 3074 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
5fddcffc | 3075 | { |
72bb2c39 | 3076 | run->base = XNEWVEC (cpp_token, count); |
5fddcffc NB |
3077 | run->limit = run->base + count; |
3078 | run->next = NULL; | |
3079 | } | |
3080 | ||
3081 | /* Returns the next tokenrun, or creates one if there is none. */ | |
3082 | static tokenrun * | |
6cf87ca4 | 3083 | next_tokenrun (tokenrun *run) |
5fddcffc NB |
3084 | { |
3085 | if (run->next == NULL) | |
3086 | { | |
72bb2c39 | 3087 | run->next = XNEW (tokenrun); |
bdcbe496 | 3088 | run->next->prev = run; |
5fddcffc NB |
3089 | _cpp_init_tokenrun (run->next, 250); |
3090 | } | |
3091 | ||
3092 | return run->next; | |
3093 | } | |
3094 | ||
ad2305ad | 3095 | /* Return the number of not yet processed token in a given |
92582b75 TT |
3096 | context. */ |
3097 | int | |
ad2305ad | 3098 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
92582b75 | 3099 | { |
92582b75 | 3100 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
cbbcf655 | 3101 | return (LAST (context).token - FIRST (context).token); |
92582b75 TT |
3102 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
3103 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
cbbcf655 | 3104 | return (LAST (context).ptoken - FIRST (context).ptoken); |
92582b75 TT |
3105 | else |
3106 | abort (); | |
3107 | } | |
3108 | ||
ad2305ad DS |
3109 | /* Returns the token present at index INDEX in a given context. If |
3110 | INDEX is zero, the next token to be processed is returned. */ | |
92582b75 | 3111 | static const cpp_token* |
ad2305ad | 3112 | _cpp_token_from_context_at (cpp_context *context, int index) |
92582b75 | 3113 | { |
92582b75 TT |
3114 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
3115 | return &(FIRST (context).token[index]); | |
3116 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
3117 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
3118 | return FIRST (context).ptoken[index]; | |
3119 | else | |
3120 | abort (); | |
3121 | } | |
3122 | ||
5950c3c9 BE |
3123 | /* Look ahead in the input stream. */ |
3124 | const cpp_token * | |
3125 | cpp_peek_token (cpp_reader *pfile, int index) | |
3126 | { | |
3127 | cpp_context *context = pfile->context; | |
3128 | const cpp_token *peektok; | |
3129 | int count; | |
3130 | ||
3131 | /* First, scan through any pending cpp_context objects. */ | |
3132 | while (context->prev) | |
3133 | { | |
ad2305ad | 3134 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
5950c3c9 BE |
3135 | |
3136 | if (index < (int) sz) | |
ad2305ad | 3137 | return _cpp_token_from_context_at (context, index); |
5950c3c9 BE |
3138 | index -= (int) sz; |
3139 | context = context->prev; | |
3140 | } | |
3141 | ||
3142 | /* We will have to read some new tokens after all (and do so | |
3143 | without invalidating preceding tokens). */ | |
3144 | count = index; | |
3145 | pfile->keep_tokens++; | |
3146 | ||
b8cd77f4 JJ |
3147 | /* For peeked tokens temporarily disable line_change reporting, |
3148 | until the tokens are parsed for real. */ | |
3149 | void (*line_change) (cpp_reader *, const cpp_token *, int) | |
3150 | = pfile->cb.line_change; | |
3151 | pfile->cb.line_change = NULL; | |
3152 | ||
5950c3c9 BE |
3153 | do |
3154 | { | |
3155 | peektok = _cpp_lex_token (pfile); | |
3156 | if (peektok->type == CPP_EOF) | |
e4b33ee5 JJ |
3157 | { |
3158 | index--; | |
3159 | break; | |
3160 | } | |
8bd9a00f NS |
3161 | else if (peektok->type == CPP_PRAGMA) |
3162 | { | |
3163 | /* Don't peek past a pragma. */ | |
3164 | if (peektok == &pfile->directive_result) | |
3165 | /* Save the pragma in the buffer. */ | |
3166 | *pfile->cur_token++ = *peektok; | |
3167 | index--; | |
3168 | break; | |
3169 | } | |
5950c3c9 BE |
3170 | } |
3171 | while (index--); | |
3172 | ||
e4b33ee5 | 3173 | _cpp_backup_tokens_direct (pfile, count - index); |
5950c3c9 | 3174 | pfile->keep_tokens--; |
b8cd77f4 | 3175 | pfile->cb.line_change = line_change; |
5950c3c9 BE |
3176 | |
3177 | return peektok; | |
3178 | } | |
3179 | ||
4ed5bcfb NB |
3180 | /* Allocate a single token that is invalidated at the same time as the |
3181 | rest of the tokens on the line. Has its line and col set to the | |
3182 | same as the last lexed token, so that diagnostics appear in the | |
3183 | right place. */ | |
3184 | cpp_token * | |
6cf87ca4 | 3185 | _cpp_temp_token (cpp_reader *pfile) |
4ed5bcfb NB |
3186 | { |
3187 | cpp_token *old, *result; | |
5950c3c9 BE |
3188 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
3189 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
4ed5bcfb NB |
3190 | |
3191 | old = pfile->cur_token - 1; | |
5950c3c9 BE |
3192 | /* Any pre-existing lookaheads must not be clobbered. */ |
3193 | if (la) | |
3194 | { | |
3195 | if (sz <= la) | |
3196 | { | |
3197 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
3198 | ||
3199 | if (sz < la) | |
3200 | memmove (next->base + 1, next->base, | |
3201 | (la - sz) * sizeof (cpp_token)); | |
3202 | ||
3203 | next->base[0] = pfile->cur_run->limit[-1]; | |
3204 | } | |
3205 | ||
3206 | if (sz > 1) | |
3207 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
3208 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
3209 | } | |
3210 | ||
3211 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
4ed5bcfb NB |
3212 | { |
3213 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
3214 | pfile->cur_token = pfile->cur_run->base; | |
3215 | } | |
3216 | ||
3217 | result = pfile->cur_token++; | |
12f9df4e | 3218 | result->src_loc = old->src_loc; |
4ed5bcfb NB |
3219 | return result; |
3220 | } | |
3221 | ||
c9c3d5f2 NS |
3222 | /* We're at the beginning of a logical line (so not in |
3223 | directives-mode) and RESULT is a CPP_NAME with NODE_MODULE set. See | |
3224 | if we should enter deferred_pragma mode to tokenize the rest of the | |
3225 | line as a module control-line. */ | |
3226 | ||
3227 | static void | |
3228 | cpp_maybe_module_directive (cpp_reader *pfile, cpp_token *result) | |
3229 | { | |
3230 | unsigned backup = 0; /* Tokens we peeked. */ | |
3231 | cpp_hashnode *node = result->val.node.node; | |
3232 | cpp_token *peek = result; | |
3233 | cpp_token *keyword = peek; | |
3234 | cpp_hashnode *(&n_modules)[spec_nodes::M_HWM][2] = pfile->spec_nodes.n_modules; | |
3235 | int header_count = 0; | |
3236 | ||
3237 | /* Make sure the incoming state is as we expect it. This way we | |
3238 | can restore it using constants. */ | |
3239 | gcc_checking_assert (!pfile->state.in_deferred_pragma | |
3240 | && !pfile->state.skipping | |
3241 | && !pfile->state.parsing_args | |
3242 | && !pfile->state.angled_headers | |
3243 | && (pfile->state.save_comments | |
3244 | == !CPP_OPTION (pfile, discard_comments))); | |
3245 | ||
3246 | /* Enter directives mode sufficiently for peeking. We don't have | |
3247 | to actually set in_directive. */ | |
3248 | pfile->state.in_deferred_pragma = true; | |
3249 | ||
3250 | /* These two fields are needed to process tokenization in deferred | |
3251 | pragma mode. They are not used outside deferred pragma mode or | |
3252 | directives mode. */ | |
3253 | pfile->state.pragma_allow_expansion = true; | |
3254 | pfile->directive_line = result->src_loc; | |
3255 | ||
3256 | /* Saving comments is incompatible with directives mode. */ | |
3257 | pfile->state.save_comments = 0; | |
3258 | ||
3259 | if (node == n_modules[spec_nodes::M_EXPORT][0]) | |
3260 | { | |
3261 | peek = _cpp_lex_direct (pfile); | |
3262 | keyword = peek; | |
3263 | backup++; | |
3264 | if (keyword->type != CPP_NAME) | |
3265 | goto not_module; | |
3266 | node = keyword->val.node.node; | |
3267 | if (!(node->flags & NODE_MODULE)) | |
3268 | goto not_module; | |
3269 | } | |
3270 | ||
3271 | if (node == n_modules[spec_nodes::M__IMPORT][0]) | |
3272 | /* __import */ | |
3273 | header_count = backup + 2 + 16; | |
3274 | else if (node == n_modules[spec_nodes::M_IMPORT][0]) | |
3275 | /* import */ | |
3276 | header_count = backup + 2 + (CPP_OPTION (pfile, preprocessed) ? 16 : 0); | |
3277 | else if (node == n_modules[spec_nodes::M_MODULE][0]) | |
3278 | ; /* module */ | |
3279 | else | |
3280 | goto not_module; | |
3281 | ||
3282 | /* We've seen [export] {module|import|__import}. Check the next token. */ | |
3283 | if (header_count) | |
3284 | /* After '{,__}import' a header name may appear. */ | |
3285 | pfile->state.angled_headers = true; | |
3286 | peek = _cpp_lex_direct (pfile); | |
3287 | backup++; | |
3288 | ||
3289 | /* ... import followed by identifier, ':', '<' or | |
3290 | header-name preprocessing tokens, or module | |
3291 | followed by cpp-identifier, ':' or ';' preprocessing | |
3292 | tokens. C++ keywords are not yet relevant. */ | |
3293 | if (peek->type == CPP_NAME | |
3294 | || peek->type == CPP_COLON | |
3295 | || (header_count | |
3296 | ? (peek->type == CPP_LESS | |
3297 | || (peek->type == CPP_STRING && peek->val.str.text[0] != 'R') | |
3298 | || peek->type == CPP_HEADER_NAME) | |
3299 | : peek->type == CPP_SEMICOLON)) | |
3300 | { | |
3301 | pfile->state.pragma_allow_expansion = !CPP_OPTION (pfile, preprocessed); | |
3302 | if (!pfile->state.pragma_allow_expansion) | |
3303 | pfile->state.prevent_expansion++; | |
3304 | ||
3305 | if (!header_count && linemap_included_from | |
3306 | (LINEMAPS_LAST_ORDINARY_MAP (pfile->line_table))) | |
3307 | cpp_error_with_line (pfile, CPP_DL_ERROR, keyword->src_loc, 0, | |
3308 | "module control-line cannot be in included file"); | |
3309 | ||
3310 | /* The first one or two tokens cannot be macro names. */ | |
3311 | for (int ix = backup; ix--;) | |
3312 | { | |
3313 | cpp_token *tok = ix ? keyword : result; | |
3314 | cpp_hashnode *node = tok->val.node.node; | |
3315 | ||
3316 | /* Don't attempt to expand the token. */ | |
3317 | tok->flags |= NO_EXPAND; | |
3318 | if (_cpp_defined_macro_p (node) | |
13f93cf5 | 3319 | && _cpp_maybe_notify_macro_use (pfile, node, tok->src_loc) |
c9c3d5f2 NS |
3320 | && !cpp_fun_like_macro_p (node)) |
3321 | cpp_error_with_line (pfile, CPP_DL_ERROR, tok->src_loc, 0, | |
3322 | "module control-line \"%s\" cannot be" | |
3323 | " an object-like macro", | |
3324 | NODE_NAME (node)); | |
3325 | } | |
3326 | ||
3327 | /* Map to underbar variants. */ | |
3328 | keyword->val.node.node = n_modules[header_count | |
3329 | ? spec_nodes::M_IMPORT | |
3330 | : spec_nodes::M_MODULE][1]; | |
3331 | if (backup != 1) | |
3332 | result->val.node.node = n_modules[spec_nodes::M_EXPORT][1]; | |
3333 | ||
3334 | /* Maybe tell the tokenizer we expect a header-name down the | |
3335 | road. */ | |
3336 | pfile->state.directive_file_token = header_count; | |
3337 | } | |
3338 | else | |
3339 | { | |
3340 | not_module: | |
3341 | /* Drop out of directive mode. */ | |
3342 | /* We aaserted save_comments had this value upon entry. */ | |
3343 | pfile->state.save_comments | |
3344 | = !CPP_OPTION (pfile, discard_comments); | |
3345 | pfile->state.in_deferred_pragma = false; | |
3346 | /* Do not let this remain on. */ | |
3347 | pfile->state.angled_headers = false; | |
3348 | } | |
3349 | ||
3350 | /* In either case we want to backup the peeked tokens. */ | |
3351 | if (backup) | |
3352 | { | |
3353 | /* If we saw EOL, we should drop it, because this isn't a module | |
3354 | control-line after all. */ | |
3355 | bool eol = peek->type == CPP_PRAGMA_EOL; | |
3356 | if (!eol || backup > 1) | |
3357 | { | |
3358 | /* Put put the peeked tokens back */ | |
3359 | _cpp_backup_tokens_direct (pfile, backup); | |
3360 | /* But if the last one was an EOL, forget it. */ | |
3361 | if (eol) | |
3362 | pfile->lookaheads--; | |
3363 | } | |
3364 | } | |
3365 | } | |
3366 | ||
14baae01 NB |
3367 | /* Lex a token into RESULT (external interface). Takes care of issues |
3368 | like directive handling, token lookahead, multiple include | |
a1f300c0 | 3369 | optimization and skipping. */ |
345894b4 | 3370 | const cpp_token * |
6cf87ca4 | 3371 | _cpp_lex_token (cpp_reader *pfile) |
5fddcffc | 3372 | { |
bdcbe496 | 3373 | cpp_token *result; |
5fddcffc | 3374 | |
bdcbe496 | 3375 | for (;;) |
5fddcffc | 3376 | { |
bdcbe496 | 3377 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 3378 | { |
bdcbe496 NB |
3379 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
3380 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc | 3381 | } |
ee380365 TT |
3382 | /* We assume that the current token is somewhere in the current |
3383 | run. */ | |
3384 | if (pfile->cur_token < pfile->cur_run->base | |
3385 | || pfile->cur_token >= pfile->cur_run->limit) | |
3386 | abort (); | |
5fddcffc | 3387 | |
bdcbe496 | 3388 | if (pfile->lookaheads) |
14baae01 NB |
3389 | { |
3390 | pfile->lookaheads--; | |
3391 | result = pfile->cur_token++; | |
3392 | } | |
bdcbe496 | 3393 | else |
14baae01 | 3394 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
3395 | |
3396 | if (result->flags & BOL) | |
5fddcffc | 3397 | { |
bdcbe496 NB |
3398 | /* Is this a directive. If _cpp_handle_directive returns |
3399 | false, it is an assembler #. */ | |
3400 | if (result->type == CPP_HASH | |
e808ec9c NB |
3401 | /* 6.10.3 p 11: Directives in a list of macro arguments |
3402 | gives undefined behavior. This implementation | |
3403 | handles the directive as normal. */ | |
bc4071dd | 3404 | && pfile->state.parsing_args != 1) |
21b11495 | 3405 | { |
bc4071dd | 3406 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
21b11495 | 3407 | { |
bc4071dd RH |
3408 | if (pfile->directive_result.type == CPP_PADDING) |
3409 | continue; | |
21b11495 | 3410 | result = &pfile->directive_result; |
21b11495 ZW |
3411 | } |
3412 | } | |
bc4071dd RH |
3413 | else if (pfile->state.in_deferred_pragma) |
3414 | result = &pfile->directive_result; | |
c9c3d5f2 NS |
3415 | else if (result->type == CPP_NAME |
3416 | && (result->val.node.node->flags & NODE_MODULE) | |
3417 | && !pfile->state.skipping | |
3418 | /* Unlike regular directives, we do not deal with | |
3419 | tokenizing module directives as macro arguments. | |
3420 | That's not permitted. */ | |
3421 | && !pfile->state.parsing_args) | |
3422 | { | |
3423 | /* P1857. Before macro expansion, At start of logical | |
3424 | line ... */ | |
3425 | /* We don't have to consider lookaheads at this point. */ | |
3426 | gcc_checking_assert (!pfile->lookaheads); | |
3427 | ||
3428 | cpp_maybe_module_directive (pfile, result); | |
3429 | } | |
21b11495 | 3430 | |
97293897 | 3431 | if (pfile->cb.line_change && !pfile->state.skipping) |
6cf87ca4 | 3432 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
5fddcffc | 3433 | } |
5fddcffc | 3434 | |
bdcbe496 | 3435 | /* We don't skip tokens in directives. */ |
bc4071dd | 3436 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
bdcbe496 | 3437 | break; |
5fddcffc | 3438 | |
bdcbe496 | 3439 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 3440 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
6356f892 | 3441 | get here and MI optimization works. */ |
5fddcffc | 3442 | pfile->mi_valid = false; |
bdcbe496 NB |
3443 | |
3444 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
3445 | break; | |
5fddcffc NB |
3446 | } |
3447 | ||
345894b4 | 3448 | return result; |
5fddcffc NB |
3449 | } |
3450 | ||
26aea073 NB |
3451 | /* Returns true if a fresh line has been loaded. */ |
3452 | bool | |
6cf87ca4 | 3453 | _cpp_get_fresh_line (cpp_reader *pfile) |
004cb263 | 3454 | { |
26aea073 NB |
3455 | /* We can't get a new line until we leave the current directive. */ |
3456 | if (pfile->state.in_directive) | |
3457 | return false; | |
df383483 | 3458 | |
26aea073 | 3459 | for (;;) |
1a76916c | 3460 | { |
26aea073 | 3461 | cpp_buffer *buffer = pfile->buffer; |
1a76916c | 3462 | |
26aea073 NB |
3463 | if (!buffer->need_line) |
3464 | return true; | |
3465 | ||
3466 | if (buffer->next_line < buffer->rlimit) | |
004cb263 | 3467 | { |
26aea073 NB |
3468 | _cpp_clean_line (pfile); |
3469 | return true; | |
3470 | } | |
004cb263 | 3471 | |
26aea073 NB |
3472 | /* First, get out of parsing arguments state. */ |
3473 | if (pfile->state.parsing_args) | |
3474 | return false; | |
3475 | ||
3476 | /* End of buffer. Non-empty files should end in a newline. */ | |
3477 | if (buffer->buf != buffer->rlimit | |
3478 | && buffer->next_line > buffer->rlimit | |
3479 | && !buffer->from_stage3) | |
3480 | { | |
ed0e74e0 | 3481 | /* Clip to buffer size. */ |
26aea073 | 3482 | buffer->next_line = buffer->rlimit; |
26aea073 | 3483 | } |
22234f56 | 3484 | |
2a0225e4 NS |
3485 | if (buffer->prev && !buffer->return_at_eof) |
3486 | _cpp_pop_buffer (pfile); | |
3487 | else | |
3488 | { | |
3489 | /* End of translation. Do not pop the buffer yet. Increment | |
3490 | line number so that the EOF token is on a line of its own | |
3491 | (_cpp_lex_direct doesn't increment in that case, because | |
3492 | it's hard for it to distinguish this special case). */ | |
3493 | CPP_INCREMENT_LINE (pfile, 0); | |
3494 | return false; | |
3495 | } | |
26aea073 | 3496 | } |
004cb263 NB |
3497 | } |
3498 | ||
6f572ac2 NB |
3499 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
3500 | do \ | |
3501 | { \ | |
3502 | result->type = ELSE_TYPE; \ | |
3503 | if (*buffer->cur == CHAR) \ | |
3504 | buffer->cur++, result->type = THEN_TYPE; \ | |
3505 | } \ | |
3506 | while (0) | |
480709cc | 3507 | |
14baae01 NB |
3508 | /* Lex a token into pfile->cur_token, which is also incremented, to |
3509 | get diagnostics pointing to the correct location. | |
3510 | ||
3511 | Does not handle issues such as token lookahead, multiple-include | |
f1ba665b | 3512 | optimization, directives, skipping etc. This function is only |
14baae01 NB |
3513 | suitable for use by _cpp_lex_token, and in special cases like |
3514 | lex_expansion_token which doesn't care for any of these issues. | |
3515 | ||
3516 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
3517 | otherwise returns to the start of the token buffer if permissible. | |
3518 | Returns the location of the lexed token. */ | |
3519 | cpp_token * | |
6cf87ca4 | 3520 | _cpp_lex_direct (cpp_reader *pfile) |
45b966db | 3521 | { |
0d9f234d | 3522 | cppchar_t c; |
adb84b42 | 3523 | cpp_buffer *buffer; |
7bad794a JJ |
3524 | const unsigned char *comment_start; |
3525 | bool fallthrough_comment = false; | |
14baae01 | 3526 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 3527 | |
5fddcffc | 3528 | fresh_line: |
26aea073 | 3529 | result->flags = 0; |
2be570f9 | 3530 | buffer = pfile->buffer; |
a506c55c | 3531 | if (buffer->need_line) |
26aea073 | 3532 | { |
55dfce4d JJ |
3533 | if (pfile->state.in_deferred_pragma) |
3534 | { | |
3535 | /* This can happen in cases like: | |
3536 | #define loop(x) whatever | |
3537 | #pragma omp loop | |
3538 | where when trying to expand loop we need to peek | |
3539 | next token after loop, but aren't still in_deferred_pragma | |
3540 | mode but are in in_directive mode, so buffer->need_line | |
3541 | is set, a CPP_EOF is peeked. */ | |
3542 | result->type = CPP_PRAGMA_EOL; | |
3543 | pfile->state.in_deferred_pragma = false; | |
3544 | if (!pfile->state.pragma_allow_expansion) | |
3545 | pfile->state.prevent_expansion--; | |
3546 | return result; | |
3547 | } | |
26aea073 NB |
3548 | if (!_cpp_get_fresh_line (pfile)) |
3549 | { | |
3550 | result->type = CPP_EOF; | |
dbcc6b15 NS |
3551 | /* Not a real EOF in a directive or arg parsing -- we refuse |
3552 | to advance to the next file now, and will once we're out | |
3553 | of those modes. */ | |
3554 | if (!pfile->state.in_directive && !pfile->state.parsing_args) | |
9ff7868d NB |
3555 | { |
3556 | /* Tell the compiler the line number of the EOF token. */ | |
500bee0a | 3557 | result->src_loc = pfile->line_table->highest_line; |
9ff7868d | 3558 | result->flags = BOL; |
2a0225e4 NS |
3559 | /* Now pop the buffer that _cpp_get_fresh_line did not. */ |
3560 | _cpp_pop_buffer (pfile); | |
9ff7868d | 3561 | } |
26aea073 NB |
3562 | return result; |
3563 | } | |
81fea426 | 3564 | if (buffer != pfile->buffer) |
7bad794a | 3565 | fallthrough_comment = false; |
26aea073 NB |
3566 | if (!pfile->keep_tokens) |
3567 | { | |
3568 | pfile->cur_run = &pfile->base_run; | |
3569 | result = pfile->base_run.base; | |
3570 | pfile->cur_token = result + 1; | |
3571 | } | |
3572 | result->flags = BOL; | |
3573 | if (pfile->state.parsing_args == 2) | |
3574 | result->flags |= PREV_WHITE; | |
3575 | } | |
a506c55c | 3576 | buffer = pfile->buffer; |
5fddcffc | 3577 | update_tokens_line: |
500bee0a | 3578 | result->src_loc = pfile->line_table->highest_line; |
041c3194 | 3579 | |
5fddcffc | 3580 | skipped_white: |
26aea073 NB |
3581 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
3582 | && !pfile->overlaid_buffer) | |
3583 | { | |
3584 | _cpp_process_line_notes (pfile, false); | |
500bee0a | 3585 | result->src_loc = pfile->line_table->highest_line; |
26aea073 | 3586 | } |
480709cc | 3587 | c = *buffer->cur++; |
12f9df4e | 3588 | |
f3f6029d NS |
3589 | if (pfile->forced_token_location) |
3590 | result->src_loc = pfile->forced_token_location; | |
e3dfef44 GC |
3591 | else |
3592 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
3593 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
5fddcffc | 3594 | |
0d9f234d | 3595 | switch (c) |
45b966db | 3596 | { |
4d6baafa NB |
3597 | case ' ': case '\t': case '\f': case '\v': case '\0': |
3598 | result->flags |= PREV_WHITE; | |
26aea073 NB |
3599 | skip_whitespace (pfile, c); |
3600 | goto skipped_white; | |
0d9f234d | 3601 | |
26aea073 | 3602 | case '\n': |
056f95ec NS |
3603 | /* Increment the line, unless this is the last line ... */ |
3604 | if (buffer->cur < buffer->rlimit | |
3605 | /* ... or this is a #include, (where _cpp_stack_file needs to | |
3606 | unwind by one line) ... */ | |
3607 | || (pfile->state.in_directive > 1 | |
3608 | /* ... except traditional-cpp increments this elsewhere. */ | |
3609 | && !CPP_OPTION (pfile, traditional))) | |
12f9df4e | 3610 | CPP_INCREMENT_LINE (pfile, 0); |
26aea073 | 3611 | buffer->need_line = true; |
8bd9a00f NS |
3612 | if (pfile->state.in_deferred_pragma) |
3613 | { | |
3614 | /* Produce the PRAGMA_EOL on this line. File reading | |
3615 | ensures there is always a \n at end of the buffer, thus | |
3616 | in a deferred pragma we always see CPP_PRAGMA_EOL before | |
3617 | any CPP_EOF. */ | |
3618 | result->type = CPP_PRAGMA_EOL; | |
3619 | result->flags &= ~PREV_WHITE; | |
3620 | pfile->state.in_deferred_pragma = false; | |
3621 | if (!pfile->state.pragma_allow_expansion) | |
3622 | pfile->state.prevent_expansion--; | |
3623 | return result; | |
3624 | } | |
26aea073 | 3625 | goto fresh_line; |
46d07497 | 3626 | |
0d9f234d NB |
3627 | case '0': case '1': case '2': case '3': case '4': |
3628 | case '5': case '6': case '7': case '8': case '9': | |
50668cf6 GK |
3629 | { |
3630 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
3631 | result->type = CPP_NUMBER; | |
3632 | lex_number (pfile, &result->val.str, &nst); | |
3633 | warn_about_normalization (pfile, result, &nst); | |
3634 | break; | |
3635 | } | |
46d07497 | 3636 | |
0abc6a6a | 3637 | case 'L': |
b6baa67d KVH |
3638 | case 'u': |
3639 | case 'U': | |
2c6e3f55 JJ |
3640 | case 'R': |
3641 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
3642 | wide strings or raw strings. */ | |
a48e3dd1 JM |
3643 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
3644 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
bced6edf | 3645 | { |
2c6e3f55 JJ |
3646 | if ((*buffer->cur == '\'' && c != 'R') |
3647 | || *buffer->cur == '"' | |
3648 | || (*buffer->cur == 'R' | |
3649 | && c != 'R' | |
3650 | && buffer->cur[1] == '"' | |
a48e3dd1 | 3651 | && CPP_OPTION (pfile, rliterals)) |
2c6e3f55 JJ |
3652 | || (*buffer->cur == '8' |
3653 | && c == 'u' | |
fe95b036 ESR |
3654 | && ((buffer->cur[1] == '"' || (buffer->cur[1] == '\'' |
3655 | && CPP_OPTION (pfile, utf8_char_literals))) | |
a48e3dd1 JM |
3656 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
3657 | && CPP_OPTION (pfile, rliterals))))) | |
b6baa67d KVH |
3658 | { |
3659 | lex_string (pfile, result, buffer->cur - 1); | |
3660 | break; | |
3661 | } | |
bced6edf | 3662 | } |
df383483 | 3663 | /* Fall through. */ |
0abc6a6a | 3664 | |
0d9f234d NB |
3665 | case '_': |
3666 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
3667 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
3668 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
b6baa67d | 3669 | case 's': case 't': case 'v': case 'w': case 'x': |
0d9f234d NB |
3670 | case 'y': case 'z': |
3671 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
0abc6a6a | 3672 | case 'G': case 'H': case 'I': case 'J': case 'K': |
2c6e3f55 | 3673 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
b6baa67d | 3674 | case 'S': case 'T': case 'V': case 'W': case 'X': |
0d9f234d NB |
3675 | case 'Y': case 'Z': |
3676 | result->type = CPP_NAME; | |
50668cf6 GK |
3677 | { |
3678 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
9a0c6187 | 3679 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
be5ffc59 JM |
3680 | &nst, |
3681 | &result->val.node.spelling); | |
50668cf6 GK |
3682 | warn_about_normalization (pfile, result, &nst); |
3683 | } | |
0d9f234d | 3684 | |
0d9f234d | 3685 | /* Convert named operators to their proper types. */ |
9a0c6187 | 3686 | if (result->val.node.node->flags & NODE_OPERATOR) |
0d9f234d NB |
3687 | { |
3688 | result->flags |= NAMED_OP; | |
9a0c6187 | 3689 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
0d9f234d | 3690 | } |
81fea426 MP |
3691 | |
3692 | /* Signal FALLTHROUGH comment followed by another token. */ | |
7bad794a | 3693 | if (fallthrough_comment) |
81fea426 | 3694 | result->flags |= PREV_FALLTHROUGH; |
0d9f234d NB |
3695 | break; |
3696 | ||
3697 | case '\'': | |
3698 | case '"': | |
6338b358 | 3699 | lex_string (pfile, result, buffer->cur - 1); |
0d9f234d | 3700 | break; |
041c3194 | 3701 | |
0d9f234d | 3702 | case '/': |
1c6d33ef NB |
3703 | /* A potential block or line comment. */ |
3704 | comment_start = buffer->cur; | |
6f572ac2 NB |
3705 | c = *buffer->cur; |
3706 | ||
1c6d33ef NB |
3707 | if (c == '*') |
3708 | { | |
26aea073 | 3709 | if (_cpp_skip_block_comment (pfile)) |
0527bc4e | 3710 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
0d9f234d | 3711 | } |
909eb89c | 3712 | else if (c == '/' && ! CPP_OPTION (pfile, traditional)) |
0d9f234d | 3713 | { |
909eb89c | 3714 | /* Don't warn for system headers. */ |
bf425849 | 3715 | if (_cpp_in_system_header (pfile)) |
909eb89c | 3716 | ; |
f3bede71 | 3717 | /* Warn about comments if pedantically GNUC89, and not |
bdb05a7b | 3718 | in system headers. */ |
909eb89c MP |
3719 | else if (CPP_OPTION (pfile, lang) == CLK_GNUC89 |
3720 | && CPP_PEDANTIC (pfile) | |
3721 | && ! buffer->warned_cplusplus_comments) | |
041c3194 | 3722 | { |
0c86a39d JJ |
3723 | if (cpp_error (pfile, CPP_DL_PEDWARN, |
3724 | "C++ style comments are not allowed in ISO C90")) | |
3725 | cpp_error (pfile, CPP_DL_NOTE, | |
3726 | "(this will be reported only once per input file)"); | |
1c6d33ef NB |
3727 | buffer->warned_cplusplus_comments = 1; |
3728 | } | |
f3bede71 | 3729 | /* Or if specifically desired via -Wc90-c99-compat. */ |
177cce46 | 3730 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
dd3ff077 | 3731 | && ! CPP_OPTION (pfile, cplusplus) |
f3bede71 MP |
3732 | && ! buffer->warned_cplusplus_comments) |
3733 | { | |
0c86a39d JJ |
3734 | if (cpp_error (pfile, CPP_DL_WARNING, |
3735 | "C++ style comments are incompatible with C90")) | |
3736 | cpp_error (pfile, CPP_DL_NOTE, | |
3737 | "(this will be reported only once per input file)"); | |
f3bede71 MP |
3738 | buffer->warned_cplusplus_comments = 1; |
3739 | } | |
909eb89c MP |
3740 | /* In C89/C94, C++ style comments are forbidden. */ |
3741 | else if ((CPP_OPTION (pfile, lang) == CLK_STDC89 | |
3742 | || CPP_OPTION (pfile, lang) == CLK_STDC94)) | |
3743 | { | |
3744 | /* But don't be confused about valid code such as | |
3745 | - // immediately followed by *, | |
3746 | - // in a preprocessing directive, | |
3747 | - // in an #if 0 block. */ | |
3748 | if (buffer->cur[1] == '*' | |
3749 | || pfile->state.in_directive | |
3750 | || pfile->state.skipping) | |
3751 | { | |
3752 | result->type = CPP_DIV; | |
3753 | break; | |
3754 | } | |
3755 | else if (! buffer->warned_cplusplus_comments) | |
3756 | { | |
0c86a39d JJ |
3757 | if (cpp_error (pfile, CPP_DL_ERROR, |
3758 | "C++ style comments are not allowed in " | |
3759 | "ISO C90")) | |
3760 | cpp_error (pfile, CPP_DL_NOTE, | |
3761 | "(this will be reported only once per input " | |
3762 | "file)"); | |
909eb89c MP |
3763 | buffer->warned_cplusplus_comments = 1; |
3764 | } | |
3765 | } | |
01ef6563 | 3766 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
87cf0651 | 3767 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
1c6d33ef | 3768 | } |
480709cc NB |
3769 | else if (c == '=') |
3770 | { | |
6f572ac2 | 3771 | buffer->cur++; |
480709cc NB |
3772 | result->type = CPP_DIV_EQ; |
3773 | break; | |
3774 | } | |
3775 | else | |
3776 | { | |
480709cc NB |
3777 | result->type = CPP_DIV; |
3778 | break; | |
3779 | } | |
0d9f234d | 3780 | |
7bad794a JJ |
3781 | if (fallthrough_comment_p (pfile, comment_start)) |
3782 | fallthrough_comment = true; | |
3783 | ||
05945a1b DM |
3784 | if (pfile->cb.comment) |
3785 | { | |
3786 | size_t len = pfile->buffer->cur - comment_start; | |
3787 | pfile->cb.comment (pfile, result->src_loc, comment_start - 1, | |
3788 | len + 1); | |
3789 | } | |
3790 | ||
1c6d33ef NB |
3791 | if (!pfile->state.save_comments) |
3792 | { | |
3793 | result->flags |= PREV_WHITE; | |
5fddcffc | 3794 | goto update_tokens_line; |
0d9f234d | 3795 | } |
1c6d33ef | 3796 | |
7bad794a | 3797 | if (fallthrough_comment) |
81fea426 MP |
3798 | result->flags |= PREV_FALLTHROUGH; |
3799 | ||
1c6d33ef | 3800 | /* Save the comment as a token in its own right. */ |
477cdac7 | 3801 | save_comment (pfile, result, comment_start, c); |
bdcbe496 | 3802 | break; |
0d9f234d NB |
3803 | |
3804 | case '<': | |
3805 | if (pfile->state.angled_headers) | |
3806 | { | |
6338b358 | 3807 | lex_string (pfile, result, buffer->cur - 1); |
4bb09c26 JM |
3808 | if (result->type != CPP_LESS) |
3809 | break; | |
0d9f234d | 3810 | } |
45b966db | 3811 | |
6f572ac2 NB |
3812 | result->type = CPP_LESS; |
3813 | if (*buffer->cur == '=') | |
b7689b96 JM |
3814 | { |
3815 | buffer->cur++, result->type = CPP_LESS_EQ; | |
3816 | if (*buffer->cur == '>' | |
3817 | && CPP_OPTION (pfile, cplusplus) | |
b04445d4 | 3818 | && CPP_OPTION (pfile, lang) >= CLK_GNUCXX20) |
b7689b96 JM |
3819 | buffer->cur++, result->type = CPP_SPACESHIP; |
3820 | } | |
6f572ac2 | 3821 | else if (*buffer->cur == '<') |
0d9f234d | 3822 | { |
6f572ac2 NB |
3823 | buffer->cur++; |
3824 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
0d9f234d | 3825 | } |
6f572ac2 | 3826 | else if (CPP_OPTION (pfile, digraphs)) |
480709cc | 3827 | { |
6f572ac2 NB |
3828 | if (*buffer->cur == ':') |
3829 | { | |
1582c677 PC |
3830 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
3831 | three characters are <:: and the subsequent character | |
3832 | is neither : nor >, the < is treated as a preprocessor | |
3833 | token by itself". */ | |
3834 | if (CPP_OPTION (pfile, cplusplus) | |
61949153 PC |
3835 | && CPP_OPTION (pfile, lang) != CLK_CXX98 |
3836 | && CPP_OPTION (pfile, lang) != CLK_GNUCXX | |
1582c677 PC |
3837 | && buffer->cur[1] == ':' |
3838 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
3839 | break; | |
3840 | ||
6f572ac2 NB |
3841 | buffer->cur++; |
3842 | result->flags |= DIGRAPH; | |
3843 | result->type = CPP_OPEN_SQUARE; | |
3844 | } | |
3845 | else if (*buffer->cur == '%') | |
3846 | { | |
3847 | buffer->cur++; | |
3848 | result->flags |= DIGRAPH; | |
3849 | result->type = CPP_OPEN_BRACE; | |
3850 | } | |
480709cc | 3851 | } |
0d9f234d NB |
3852 | break; |
3853 | ||
3854 | case '>': | |
6f572ac2 NB |
3855 | result->type = CPP_GREATER; |
3856 | if (*buffer->cur == '=') | |
3857 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
3858 | else if (*buffer->cur == '>') | |
0d9f234d | 3859 | { |
6f572ac2 NB |
3860 | buffer->cur++; |
3861 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
3862 | } | |
0d9f234d NB |
3863 | break; |
3864 | ||
cbcff6df | 3865 | case '%': |
6f572ac2 NB |
3866 | result->type = CPP_MOD; |
3867 | if (*buffer->cur == '=') | |
3868 | buffer->cur++, result->type = CPP_MOD_EQ; | |
3869 | else if (CPP_OPTION (pfile, digraphs)) | |
480709cc | 3870 | { |
6f572ac2 | 3871 | if (*buffer->cur == ':') |
480709cc | 3872 | { |
6f572ac2 NB |
3873 | buffer->cur++; |
3874 | result->flags |= DIGRAPH; | |
3875 | result->type = CPP_HASH; | |
3876 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
9a0c6187 | 3877 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
6f572ac2 NB |
3878 | } |
3879 | else if (*buffer->cur == '>') | |
3880 | { | |
3881 | buffer->cur++; | |
3882 | result->flags |= DIGRAPH; | |
3883 | result->type = CPP_CLOSE_BRACE; | |
480709cc | 3884 | } |
480709cc | 3885 | } |
0d9f234d NB |
3886 | break; |
3887 | ||
cbcff6df | 3888 | case '.': |
480709cc | 3889 | result->type = CPP_DOT; |
6f572ac2 | 3890 | if (ISDIGIT (*buffer->cur)) |
480709cc | 3891 | { |
50668cf6 | 3892 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
480709cc | 3893 | result->type = CPP_NUMBER; |
50668cf6 GK |
3894 | lex_number (pfile, &result->val.str, &nst); |
3895 | warn_about_normalization (pfile, result, &nst); | |
480709cc | 3896 | } |
6f572ac2 NB |
3897 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
3898 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
3899 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
3900 | buffer->cur++, result->type = CPP_DOT_STAR; | |
0d9f234d | 3901 | break; |
45b966db | 3902 | |
0d9f234d | 3903 | case '+': |
6f572ac2 NB |
3904 | result->type = CPP_PLUS; |
3905 | if (*buffer->cur == '+') | |
3906 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
3907 | else if (*buffer->cur == '=') | |
3908 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
0d9f234d | 3909 | break; |
04e3ec78 | 3910 | |
0d9f234d | 3911 | case '-': |
6f572ac2 NB |
3912 | result->type = CPP_MINUS; |
3913 | if (*buffer->cur == '>') | |
0d9f234d | 3914 | { |
6f572ac2 | 3915 | buffer->cur++; |
480709cc | 3916 | result->type = CPP_DEREF; |
6f572ac2 NB |
3917 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
3918 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
480709cc | 3919 | } |
6f572ac2 NB |
3920 | else if (*buffer->cur == '-') |
3921 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
3922 | else if (*buffer->cur == '=') | |
3923 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
0d9f234d | 3924 | break; |
45b966db | 3925 | |
0d9f234d | 3926 | case '&': |
6f572ac2 NB |
3927 | result->type = CPP_AND; |
3928 | if (*buffer->cur == '&') | |
3929 | buffer->cur++, result->type = CPP_AND_AND; | |
3930 | else if (*buffer->cur == '=') | |
3931 | buffer->cur++, result->type = CPP_AND_EQ; | |
0d9f234d | 3932 | break; |
df383483 | 3933 | |
0d9f234d | 3934 | case '|': |
6f572ac2 NB |
3935 | result->type = CPP_OR; |
3936 | if (*buffer->cur == '|') | |
3937 | buffer->cur++, result->type = CPP_OR_OR; | |
3938 | else if (*buffer->cur == '=') | |
3939 | buffer->cur++, result->type = CPP_OR_EQ; | |
0d9f234d | 3940 | break; |
45b966db | 3941 | |
0d9f234d | 3942 | case ':': |
6f572ac2 | 3943 | result->type = CPP_COLON; |
93313b94 | 3944 | if (*buffer->cur == ':' && CPP_OPTION (pfile, scope)) |
6f572ac2 NB |
3945 | buffer->cur++, result->type = CPP_SCOPE; |
3946 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
0d9f234d | 3947 | { |
6f572ac2 | 3948 | buffer->cur++; |
0d9f234d | 3949 | result->flags |= DIGRAPH; |
480709cc NB |
3950 | result->type = CPP_CLOSE_SQUARE; |
3951 | } | |
0d9f234d | 3952 | break; |
45b966db | 3953 | |
480709cc NB |
3954 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
3955 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
3956 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
3957 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
9a0c6187 | 3958 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
480709cc | 3959 | |
26aea073 | 3960 | case '?': result->type = CPP_QUERY; break; |
0d9f234d NB |
3961 | case '~': result->type = CPP_COMPL; break; |
3962 | case ',': result->type = CPP_COMMA; break; | |
3963 | case '(': result->type = CPP_OPEN_PAREN; break; | |
3964 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
3965 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
3966 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
3967 | case '{': result->type = CPP_OPEN_BRACE; break; | |
3968 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
3969 | case ';': result->type = CPP_SEMICOLON; break; | |
3970 | ||
40f03658 | 3971 | /* @ is a punctuator in Objective-C. */ |
cc937581 | 3972 | case '@': result->type = CPP_ATSIGN; break; |
0d9f234d | 3973 | |
7d112d66 | 3974 | default: |
1613e52b NB |
3975 | { |
3976 | const uchar *base = --buffer->cur; | |
0abc6a6a | 3977 | |
7d112d66 LH |
3978 | /* Check for an extended identifier ($ or UCN or UTF-8). */ |
3979 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
50668cf6 | 3980 | if (forms_identifier_p (pfile, true, &nst)) |
1613e52b NB |
3981 | { |
3982 | result->type = CPP_NAME; | |
be5ffc59 JM |
3983 | result->val.node.node = lex_identifier (pfile, base, true, &nst, |
3984 | &result->val.node.spelling); | |
50668cf6 | 3985 | warn_about_normalization (pfile, result, &nst); |
1613e52b NB |
3986 | break; |
3987 | } | |
7d112d66 LH |
3988 | |
3989 | /* Otherwise this will form a CPP_OTHER token. Parse valid UTF-8 as a | |
3990 | single token. */ | |
1613e52b | 3991 | buffer->cur++; |
7d112d66 LH |
3992 | if (c >= utf8_signifier) |
3993 | { | |
3994 | const uchar *pstr = base; | |
3995 | cppchar_t s; | |
3996 | if (_cpp_valid_utf8 (pfile, &pstr, buffer->rlimit, 0, NULL, &s)) | |
3997 | buffer->cur = pstr; | |
3998 | } | |
3999 | create_literal (pfile, result, base, buffer->cur - base, CPP_OTHER); | |
4000 | break; | |
1067694a | 4001 | } |
1613e52b | 4002 | |
0d9f234d | 4003 | } |
bdcbe496 | 4004 | |
a3998c2f DM |
4005 | /* Potentially convert the location of the token to a range. */ |
4006 | if (result->src_loc >= RESERVED_LOCATION_COUNT | |
4007 | && result->type != CPP_EOF) | |
4008 | { | |
4009 | /* Ensure that any line notes are processed, so that we have the | |
4010 | correct physical line/column for the end-point of the token even | |
4011 | when a logical line is split via one or more backslashes. */ | |
4012 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos | |
4013 | && !pfile->overlaid_buffer) | |
4014 | _cpp_process_line_notes (pfile, false); | |
4015 | ||
4016 | source_range tok_range; | |
4017 | tok_range.m_start = result->src_loc; | |
4018 | tok_range.m_finish | |
4019 | = linemap_position_for_column (pfile->line_table, | |
4020 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
4021 | ||
4022 | result->src_loc = COMBINE_LOCATION_DATA (pfile->line_table, | |
4023 | result->src_loc, | |
4024 | tok_range, NULL); | |
4025 | } | |
ebedc9a3 | 4026 | |
bdcbe496 | 4027 | return result; |
0d9f234d NB |
4028 | } |
4029 | ||
59325650 NB |
4030 | /* An upper bound on the number of bytes needed to spell TOKEN. |
4031 | Does not include preceding whitespace. */ | |
93c80368 | 4032 | unsigned int |
6cf87ca4 | 4033 | cpp_token_len (const cpp_token *token) |
0d9f234d | 4034 | { |
93c80368 | 4035 | unsigned int len; |
6d2c2047 | 4036 | |
93c80368 | 4037 | switch (TOKEN_SPELL (token)) |
041c3194 | 4038 | { |
cc955282 | 4039 | default: len = 6; break; |
6338b358 | 4040 | case SPELL_LITERAL: len = token->val.str.len; break; |
9a0c6187 | 4041 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
041c3194 | 4042 | } |
59325650 NB |
4043 | |
4044 | return len; | |
6d2c2047 ZW |
4045 | } |
4046 | ||
47e20491 GK |
4047 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
4048 | Return the number of bytes read out of NAME. (There are always | |
4049 | 10 bytes written to BUFFER.) */ | |
4050 | ||
4051 | static size_t | |
4052 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
4053 | { | |
4054 | int j; | |
4055 | int ucn_len = 0; | |
4056 | int ucn_len_c; | |
4057 | unsigned t; | |
4058 | unsigned long utf32; | |
4059 | ||
4060 | /* Compute the length of the UTF-8 sequence. */ | |
4061 | for (t = *name; t & 0x80; t <<= 1) | |
4062 | ucn_len++; | |
4063 | ||
4064 | utf32 = *name & (0x7F >> ucn_len); | |
4065 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
4066 | { | |
4067 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
4068 | ||
4069 | /* Ill-formed UTF-8. */ | |
4070 | if ((*name & ~0x3F) != 0x80) | |
4071 | abort (); | |
4072 | } | |
4073 | ||
4074 | *buffer++ = '\\'; | |
4075 | *buffer++ = 'U'; | |
4076 | for (j = 7; j >= 0; j--) | |
4077 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
4078 | return ucn_len; | |
4079 | } | |
4080 | ||
cfc93532 MLI |
4081 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
4082 | the spelling of the digraph. */ | |
4083 | static const unsigned char * | |
4084 | cpp_digraph2name (enum cpp_ttype type) | |
4085 | { | |
4086 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
4087 | } | |
47e20491 | 4088 | |
be5ffc59 JM |
4089 | /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER. |
4090 | The buffer must already contain the enough space to hold the | |
4091 | token's spelling. Returns a pointer to the character after the | |
4092 | last character written. */ | |
4093 | unsigned char * | |
4094 | _cpp_spell_ident_ucns (unsigned char *buffer, cpp_hashnode *ident) | |
4095 | { | |
4096 | size_t i; | |
4097 | const unsigned char *name = NODE_NAME (ident); | |
4098 | ||
4099 | for (i = 0; i < NODE_LEN (ident); i++) | |
4100 | if (name[i] & ~0x7F) | |
4101 | { | |
4102 | i += utf8_to_ucn (buffer, name + i) - 1; | |
4103 | buffer += 10; | |
4104 | } | |
4105 | else | |
4106 | *buffer++ = name[i]; | |
4107 | ||
4108 | return buffer; | |
4109 | } | |
4110 | ||
041c3194 | 4111 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 | 4112 | already contain the enough space to hold the token's spelling. |
6cf87ca4 | 4113 | Returns a pointer to the character after the last character written. |
47e20491 | 4114 | FORSTRING is true if this is to be the spelling after translation |
be5ffc59 JM |
4115 | phase 1 (with the original spelling of extended identifiers), false |
4116 | if extended identifiers should always be written using UCNs (there is | |
4117 | no option for always writing them in the internal UTF-8 form). | |
6cf87ca4 | 4118 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
93c80368 | 4119 | unsigned char * |
6cf87ca4 | 4120 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
47e20491 | 4121 | unsigned char *buffer, bool forstring) |
041c3194 | 4122 | { |
96be6998 | 4123 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
4124 | { |
4125 | case SPELL_OPERATOR: | |
4126 | { | |
4127 | const unsigned char *spelling; | |
4128 | unsigned char c; | |
d6d5f795 | 4129 | |
041c3194 | 4130 | if (token->flags & DIGRAPH) |
cfc93532 | 4131 | spelling = cpp_digraph2name (token->type); |
92936ecf ZW |
4132 | else if (token->flags & NAMED_OP) |
4133 | goto spell_ident; | |
041c3194 | 4134 | else |
96be6998 | 4135 | spelling = TOKEN_NAME (token); |
df383483 | 4136 | |
041c3194 ZW |
4137 | while ((c = *spelling++) != '\0') |
4138 | *buffer++ = c; | |
4139 | } | |
4140 | break; | |
d6d5f795 | 4141 | |
47ad4138 | 4142 | spell_ident: |
041c3194 | 4143 | case SPELL_IDENT: |
47e20491 GK |
4144 | if (forstring) |
4145 | { | |
be5ffc59 JM |
4146 | memcpy (buffer, NODE_NAME (token->val.node.spelling), |
4147 | NODE_LEN (token->val.node.spelling)); | |
4148 | buffer += NODE_LEN (token->val.node.spelling); | |
47e20491 GK |
4149 | } |
4150 | else | |
be5ffc59 | 4151 | buffer = _cpp_spell_ident_ucns (buffer, token->val.node.node); |
041c3194 | 4152 | break; |
d6d5f795 | 4153 | |
6338b358 | 4154 | case SPELL_LITERAL: |
47ad4138 ZW |
4155 | memcpy (buffer, token->val.str.text, token->val.str.len); |
4156 | buffer += token->val.str.len; | |
4157 | break; | |
4158 | ||
041c3194 | 4159 | case SPELL_NONE: |
0527bc4e JDA |
4160 | cpp_error (pfile, CPP_DL_ICE, |
4161 | "unspellable token %s", TOKEN_NAME (token)); | |
041c3194 ZW |
4162 | break; |
4163 | } | |
d6d5f795 | 4164 | |
041c3194 ZW |
4165 | return buffer; |
4166 | } | |
d6d5f795 | 4167 | |
5d8ebbd8 NB |
4168 | /* Returns TOKEN spelt as a null-terminated string. The string is |
4169 | freed when the reader is destroyed. Useful for diagnostics. */ | |
93c80368 | 4170 | unsigned char * |
6cf87ca4 | 4171 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
59325650 NB |
4172 | { |
4173 | unsigned int len = cpp_token_len (token) + 1; | |
ece54d54 | 4174 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 4175 | |
47e20491 | 4176 | end = cpp_spell_token (pfile, token, start, false); |
93c80368 | 4177 | end[0] = '\0'; |
c5a04734 | 4178 | |
93c80368 NB |
4179 | return start; |
4180 | } | |
c5a04734 | 4181 | |
cfc93532 MLI |
4182 | /* Returns a pointer to a string which spells the token defined by |
4183 | TYPE and FLAGS. Used by C front ends, which really should move to | |
4184 | using cpp_token_as_text. */ | |
93c80368 | 4185 | const char * |
cfc93532 | 4186 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
93c80368 | 4187 | { |
cfc93532 MLI |
4188 | if (flags & DIGRAPH) |
4189 | return (const char *) cpp_digraph2name (type); | |
4190 | else if (flags & NAMED_OP) | |
4191 | return cpp_named_operator2name (type); | |
4192 | ||
93c80368 NB |
4193 | return (const char *) token_spellings[type].name; |
4194 | } | |
c5a04734 | 4195 | |
4ed5bcfb NB |
4196 | /* Writes the spelling of token to FP, without any preceding space. |
4197 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
4198 | double-buffering. */ | |
93c80368 | 4199 | void |
6cf87ca4 | 4200 | cpp_output_token (const cpp_token *token, FILE *fp) |
93c80368 | 4201 | { |
93c80368 | 4202 | switch (TOKEN_SPELL (token)) |
c5a04734 | 4203 | { |
93c80368 NB |
4204 | case SPELL_OPERATOR: |
4205 | { | |
4206 | const unsigned char *spelling; | |
3b681e9d | 4207 | int c; |
c5a04734 | 4208 | |
93c80368 | 4209 | if (token->flags & DIGRAPH) |
cfc93532 | 4210 | spelling = cpp_digraph2name (token->type); |
93c80368 NB |
4211 | else if (token->flags & NAMED_OP) |
4212 | goto spell_ident; | |
4213 | else | |
4214 | spelling = TOKEN_NAME (token); | |
041c3194 | 4215 | |
3b681e9d ZW |
4216 | c = *spelling; |
4217 | do | |
4218 | putc (c, fp); | |
4219 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
4220 | } |
4221 | break; | |
041c3194 | 4222 | |
93c80368 NB |
4223 | spell_ident: |
4224 | case SPELL_IDENT: | |
47e20491 GK |
4225 | { |
4226 | size_t i; | |
9a0c6187 | 4227 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 4228 | |
9a0c6187 | 4229 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
4230 | if (name[i] & ~0x7F) |
4231 | { | |
4232 | unsigned char buffer[10]; | |
4233 | i += utf8_to_ucn (buffer, name + i) - 1; | |
4234 | fwrite (buffer, 1, 10, fp); | |
4235 | } | |
4236 | else | |
9a0c6187 | 4237 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
47e20491 GK |
4238 | } |
4239 | break; | |
041c3194 | 4240 | |
6338b358 | 4241 | case SPELL_LITERAL: |
c9c3d5f2 NS |
4242 | if (token->type == CPP_HEADER_NAME) |
4243 | fputc ('"', fp); | |
47ad4138 | 4244 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
c9c3d5f2 NS |
4245 | if (token->type == CPP_HEADER_NAME) |
4246 | fputc ('"', fp); | |
47ad4138 ZW |
4247 | break; |
4248 | ||
93c80368 NB |
4249 | case SPELL_NONE: |
4250 | /* An error, most probably. */ | |
4251 | break; | |
041c3194 | 4252 | } |
c5a04734 ZW |
4253 | } |
4254 | ||
93c80368 NB |
4255 | /* Compare two tokens. */ |
4256 | int | |
6cf87ca4 | 4257 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
c5a04734 | 4258 | { |
93c80368 NB |
4259 | if (a->type == b->type && a->flags == b->flags) |
4260 | switch (TOKEN_SPELL (a)) | |
4261 | { | |
4262 | default: /* Keep compiler happy. */ | |
4263 | case SPELL_OPERATOR: | |
9a0c6187 | 4264 | /* token_no is used to track where multiple consecutive ## |
aa508502 | 4265 | tokens were originally located. */ |
9a0c6187 | 4266 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
93c80368 | 4267 | case SPELL_NONE: |
9a0c6187 | 4268 | return (a->type != CPP_MACRO_ARG |
be5ffc59 JM |
4269 | || (a->val.macro_arg.arg_no == b->val.macro_arg.arg_no |
4270 | && a->val.macro_arg.spelling == b->val.macro_arg.spelling)); | |
93c80368 | 4271 | case SPELL_IDENT: |
be5ffc59 JM |
4272 | return (a->val.node.node == b->val.node.node |
4273 | && a->val.node.spelling == b->val.node.spelling); | |
6338b358 | 4274 | case SPELL_LITERAL: |
93c80368 NB |
4275 | return (a->val.str.len == b->val.str.len |
4276 | && !memcmp (a->val.str.text, b->val.str.text, | |
4277 | a->val.str.len)); | |
4278 | } | |
c5a04734 | 4279 | |
041c3194 ZW |
4280 | return 0; |
4281 | } | |
4282 | ||
93c80368 NB |
4283 | /* Returns nonzero if a space should be inserted to avoid an |
4284 | accidental token paste for output. For simplicity, it is | |
4285 | conservative, and occasionally advises a space where one is not | |
4286 | needed, e.g. "." and ".2". */ | |
93c80368 | 4287 | int |
6cf87ca4 ZW |
4288 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
4289 | const cpp_token *token2) | |
c5a04734 | 4290 | { |
93c80368 NB |
4291 | enum cpp_ttype a = token1->type, b = token2->type; |
4292 | cppchar_t c; | |
c5a04734 | 4293 | |
93c80368 NB |
4294 | if (token1->flags & NAMED_OP) |
4295 | a = CPP_NAME; | |
4296 | if (token2->flags & NAMED_OP) | |
4297 | b = CPP_NAME; | |
c5a04734 | 4298 | |
93c80368 NB |
4299 | c = EOF; |
4300 | if (token2->flags & DIGRAPH) | |
37b8524c | 4301 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
4302 | else if (token_spellings[b].category == SPELL_OPERATOR) |
4303 | c = token_spellings[b].name[0]; | |
c5a04734 | 4304 | |
93c80368 | 4305 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 4306 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 4307 | return 1; |
c5a04734 | 4308 | |
93c80368 | 4309 | switch (a) |
c5a04734 | 4310 | { |
b52dbbf8 SE |
4311 | case CPP_GREATER: return c == '>'; |
4312 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
93c80368 NB |
4313 | case CPP_PLUS: return c == '+'; |
4314 | case CPP_MINUS: return c == '-' || c == '>'; | |
4315 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
4316 | case CPP_MOD: return c == ':' || c == '>'; | |
4317 | case CPP_AND: return c == '&'; | |
4318 | case CPP_OR: return c == '|'; | |
4319 | case CPP_COLON: return c == ':' || c == '>'; | |
4320 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 4321 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 | 4322 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
170c850e | 4323 | case CPP_PRAGMA: |
93c80368 NB |
4324 | case CPP_NAME: return ((b == CPP_NUMBER |
4325 | && name_p (pfile, &token2->val.str)) | |
4326 | || b == CPP_NAME | |
4327 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
4328 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
3e3fdf3d | 4329 | || b == CPP_CHAR |
93c80368 | 4330 | || c == '.' || c == '+' || c == '-'); |
1613e52b | 4331 | /* UCNs */ |
1067694a NB |
4332 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
4333 | && b == CPP_NAME) | |
1613e52b | 4334 | || (CPP_OPTION (pfile, objc) |
1067694a | 4335 | && token1->val.str.text[0] == '@' |
1613e52b | 4336 | && (b == CPP_NAME || b == CPP_STRING))); |
b7689b96 | 4337 | case CPP_LESS_EQ: return c == '>'; |
87e356ba JJ |
4338 | case CPP_STRING: |
4339 | case CPP_WSTRING: | |
4340 | case CPP_UTF8STRING: | |
4341 | case CPP_STRING16: | |
4342 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals) | |
4343 | && (b == CPP_NAME | |
4344 | || (TOKEN_SPELL (token2) == SPELL_LITERAL | |
4345 | && ISIDST (token2->val.str.text[0])))); | |
4346 | ||
93c80368 | 4347 | default: break; |
c5a04734 | 4348 | } |
c5a04734 | 4349 | |
417f3e3a | 4350 | return 0; |
c5a04734 ZW |
4351 | } |
4352 | ||
93c80368 | 4353 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
4354 | character, to FP. Leading whitespace is removed. If there are |
4355 | macros, special token padding is not performed. */ | |
c5a04734 | 4356 | void |
6cf87ca4 | 4357 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
c5a04734 | 4358 | { |
4ed5bcfb | 4359 | const cpp_token *token; |
96be6998 | 4360 | |
4ed5bcfb NB |
4361 | token = cpp_get_token (pfile); |
4362 | while (token->type != CPP_EOF) | |
96be6998 | 4363 | { |
4ed5bcfb NB |
4364 | cpp_output_token (token, fp); |
4365 | token = cpp_get_token (pfile); | |
4366 | if (token->flags & PREV_WHITE) | |
4367 | putc (' ', fp); | |
96be6998 ZW |
4368 | } |
4369 | ||
93c80368 | 4370 | putc ('\n', fp); |
041c3194 | 4371 | } |
c5a04734 | 4372 | |
5d6342eb TT |
4373 | /* Return a string representation of all the remaining tokens on the |
4374 | current line. The result is allocated using xmalloc and must be | |
4375 | freed by the caller. */ | |
4376 | unsigned char * | |
4377 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
4378 | { | |
4379 | const cpp_token *token; | |
4380 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
4381 | unsigned int alloced = 120 + out; | |
4382 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
4383 | ||
4384 | /* If DIR_NAME is empty, there are no initial contents. */ | |
4385 | if (dir_name) | |
4386 | { | |
4387 | sprintf ((char *) result, "#%s ", dir_name); | |
4388 | out += 2; | |
4389 | } | |
4390 | ||
4391 | token = cpp_get_token (pfile); | |
4392 | while (token->type != CPP_EOF) | |
4393 | { | |
4394 | unsigned char *last; | |
4395 | /* Include room for a possible space and the terminating nul. */ | |
4396 | unsigned int len = cpp_token_len (token) + 2; | |
4397 | ||
4398 | if (out + len > alloced) | |
4399 | { | |
4400 | alloced *= 2; | |
4401 | if (out + len > alloced) | |
4402 | alloced = out + len; | |
4403 | result = (unsigned char *) xrealloc (result, alloced); | |
4404 | } | |
4405 | ||
4406 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
4407 | out = last - result; | |
4408 | ||
4409 | token = cpp_get_token (pfile); | |
4410 | if (token->flags & PREV_WHITE) | |
4411 | result[out++] = ' '; | |
4412 | } | |
4413 | ||
4414 | result[out] = '\0'; | |
4415 | return result; | |
4416 | } | |
4417 | ||
1e013d2e NB |
4418 | /* Memory buffers. Changing these three constants can have a dramatic |
4419 | effect on performance. The values here are reasonable defaults, | |
4420 | but might be tuned. If you adjust them, be sure to test across a | |
4421 | range of uses of cpplib, including heavy nested function-like macro | |
4422 | expansion. Also check the change in peak memory usage (NJAMD is a | |
4423 | good tool for this). */ | |
4424 | #define MIN_BUFF_SIZE 8000 | |
87062813 | 4425 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
1e013d2e NB |
4426 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
4427 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 4428 | |
87062813 NB |
4429 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
4430 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
4431 | #endif | |
4432 | ||
c9e7a609 NB |
4433 | /* Create a new allocation buffer. Place the control block at the end |
4434 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 | 4435 | static _cpp_buff * |
6cf87ca4 | 4436 | new_buff (size_t len) |
b8af0ca5 NB |
4437 | { |
4438 | _cpp_buff *result; | |
ece54d54 | 4439 | unsigned char *base; |
b8af0ca5 | 4440 | |
1e013d2e NB |
4441 | if (len < MIN_BUFF_SIZE) |
4442 | len = MIN_BUFF_SIZE; | |
c70f6ed3 | 4443 | len = CPP_ALIGN (len); |
b8af0ca5 | 4444 | |
ceb17928 | 4445 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
4446 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff |
4447 | struct first. */ | |
4448 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
4449 | base = XNEWVEC (unsigned char, len + slen); | |
4450 | result = (_cpp_buff *) base; | |
4451 | base += slen; | |
4452 | #else | |
c3f829c1 | 4453 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
b8af0ca5 | 4454 | result = (_cpp_buff *) (base + len); |
1a80db97 | 4455 | #endif |
b8af0ca5 NB |
4456 | result->base = base; |
4457 | result->cur = base; | |
4458 | result->limit = base + len; | |
4459 | result->next = NULL; | |
4460 | return result; | |
4461 | } | |
4462 | ||
4463 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
4464 | void | |
6cf87ca4 | 4465 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
b8af0ca5 NB |
4466 | { |
4467 | _cpp_buff *end = buff; | |
4468 | ||
4469 | while (end->next) | |
4470 | end = end->next; | |
4471 | end->next = pfile->free_buffs; | |
4472 | pfile->free_buffs = buff; | |
4473 | } | |
4474 | ||
4475 | /* Return a free buffer of size at least MIN_SIZE. */ | |
4476 | _cpp_buff * | |
6cf87ca4 | 4477 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
b8af0ca5 NB |
4478 | { |
4479 | _cpp_buff *result, **p; | |
4480 | ||
4481 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
4482 | { | |
6142088c | 4483 | size_t size; |
1e013d2e NB |
4484 | |
4485 | if (*p == NULL) | |
b8af0ca5 | 4486 | return new_buff (min_size); |
1e013d2e NB |
4487 | result = *p; |
4488 | size = result->limit - result->base; | |
4489 | /* Return a buffer that's big enough, but don't waste one that's | |
4490 | way too big. */ | |
34f5271d | 4491 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
b8af0ca5 NB |
4492 | break; |
4493 | } | |
4494 | ||
4495 | *p = result->next; | |
4496 | result->next = NULL; | |
4497 | result->cur = result->base; | |
4498 | return result; | |
4499 | } | |
4500 | ||
4fe9b91c | 4501 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
4502 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
4503 | the excess bytes to the new buffer. Chains the new buffer after | |
4504 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 4505 | _cpp_buff * |
6cf87ca4 | 4506 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
b8af0ca5 | 4507 | { |
6142088c | 4508 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 4509 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 4510 | |
8c3b2693 NB |
4511 | buff->next = new_buff; |
4512 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
4513 | return new_buff; | |
4514 | } | |
4515 | ||
4fe9b91c | 4516 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
4517 | remaining bytes of the buffer pointed to by BUFF, and at least |
4518 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
4519 | Chains the new buffer before the buffer pointed to by BUFF, and | |
4520 | updates the pointer to point to the new buffer. */ | |
4521 | void | |
6cf87ca4 | 4522 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
8c3b2693 NB |
4523 | { |
4524 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
4525 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
4526 | ||
4527 | new_buff = _cpp_get_buff (pfile, size); | |
4528 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
4529 | new_buff->next = old_buff; | |
4530 | *pbuff = new_buff; | |
b8af0ca5 NB |
4531 | } |
4532 | ||
4533 | /* Free a chain of buffers starting at BUFF. */ | |
4534 | void | |
5671bf27 | 4535 | _cpp_free_buff (_cpp_buff *buff) |
b8af0ca5 NB |
4536 | { |
4537 | _cpp_buff *next; | |
4538 | ||
4539 | for (; buff; buff = next) | |
4540 | { | |
4541 | next = buff->next; | |
ceb17928 | 4542 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
1a80db97 JJ |
4543 | free (buff); |
4544 | #else | |
b8af0ca5 | 4545 | free (buff->base); |
1a80db97 | 4546 | #endif |
b8af0ca5 NB |
4547 | } |
4548 | } | |
417f3e3a | 4549 | |
ece54d54 NB |
4550 | /* Allocate permanent, unaligned storage of length LEN. */ |
4551 | unsigned char * | |
6cf87ca4 | 4552 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
ece54d54 NB |
4553 | { |
4554 | _cpp_buff *buff = pfile->u_buff; | |
4555 | unsigned char *result = buff->cur; | |
4556 | ||
4557 | if (len > (size_t) (buff->limit - result)) | |
4558 | { | |
4559 | buff = _cpp_get_buff (pfile, len); | |
4560 | buff->next = pfile->u_buff; | |
4561 | pfile->u_buff = buff; | |
4562 | result = buff->cur; | |
4563 | } | |
4564 | ||
4565 | buff->cur = result + len; | |
4566 | return result; | |
4567 | } | |
4568 | ||
87062813 NB |
4569 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
4570 | That buffer is used for growing allocations when saving macro | |
4571 | replacement lists in a #define, and when parsing an answer to an | |
4572 | assertion in #assert, #unassert or #if (and therefore possibly | |
4573 | whilst expanding macros). It therefore must not be used by any | |
4574 | code that they might call: specifically the lexer and the guts of | |
4575 | the macro expander. | |
4576 | ||
4577 | All existing other uses clearly fit this restriction: storing | |
4578 | registered pragmas during initialization. */ | |
93c80368 | 4579 | unsigned char * |
6cf87ca4 | 4580 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
3fef5b2b | 4581 | { |
8c3b2693 NB |
4582 | _cpp_buff *buff = pfile->a_buff; |
4583 | unsigned char *result = buff->cur; | |
3fef5b2b | 4584 | |
8c3b2693 | 4585 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 4586 | { |
8c3b2693 NB |
4587 | buff = _cpp_get_buff (pfile, len); |
4588 | buff->next = pfile->a_buff; | |
4589 | pfile->a_buff = buff; | |
4590 | result = buff->cur; | |
3fef5b2b | 4591 | } |
041c3194 | 4592 | |
8c3b2693 | 4593 | buff->cur = result + len; |
93c80368 | 4594 | return result; |
041c3194 | 4595 | } |
d8044160 | 4596 | |
10f04917 NS |
4597 | /* Commit or allocate storage from a buffer. */ |
4598 | ||
4599 | void * | |
4600 | _cpp_commit_buff (cpp_reader *pfile, size_t size) | |
4601 | { | |
4602 | void *ptr = BUFF_FRONT (pfile->a_buff); | |
4603 | ||
4604 | if (pfile->hash_table->alloc_subobject) | |
4605 | { | |
4606 | void *copy = pfile->hash_table->alloc_subobject (size); | |
4607 | memcpy (copy, ptr, size); | |
4608 | ptr = copy; | |
4609 | } | |
4610 | else | |
4611 | BUFF_FRONT (pfile->a_buff) += size; | |
4612 | ||
4613 | return ptr; | |
4614 | } | |
4615 | ||
d8044160 GK |
4616 | /* Say which field of TOK is in use. */ |
4617 | ||
4618 | enum cpp_token_fld_kind | |
c26302d5 | 4619 | cpp_token_val_index (const cpp_token *tok) |
d8044160 GK |
4620 | { |
4621 | switch (TOKEN_SPELL (tok)) | |
4622 | { | |
4623 | case SPELL_IDENT: | |
4624 | return CPP_TOKEN_FLD_NODE; | |
4625 | case SPELL_LITERAL: | |
4626 | return CPP_TOKEN_FLD_STR; | |
aa508502 | 4627 | case SPELL_OPERATOR: |
3f23e487 AP |
4628 | /* Operands which were originally spelled as ident keep around |
4629 | the node for the exact spelling. */ | |
4630 | if (tok->flags & NAMED_OP) | |
4631 | return CPP_TOKEN_FLD_NODE; | |
4632 | else if (tok->type == CPP_PASTE) | |
9a0c6187 | 4633 | return CPP_TOKEN_FLD_TOKEN_NO; |
aa508502 JM |
4634 | else |
4635 | return CPP_TOKEN_FLD_NONE; | |
d8044160 GK |
4636 | case SPELL_NONE: |
4637 | if (tok->type == CPP_MACRO_ARG) | |
4638 | return CPP_TOKEN_FLD_ARG_NO; | |
4639 | else if (tok->type == CPP_PADDING) | |
4640 | return CPP_TOKEN_FLD_SOURCE; | |
21b11495 | 4641 | else if (tok->type == CPP_PRAGMA) |
bc4071dd | 4642 | return CPP_TOKEN_FLD_PRAGMA; |
191816a3 | 4643 | /* fall through */ |
d8044160 GK |
4644 | default: |
4645 | return CPP_TOKEN_FLD_NONE; | |
4646 | } | |
4647 | } | |
e3dfef44 | 4648 | |
f3f6029d | 4649 | /* All tokens lexed in R after calling this function will be forced to |
620e594b | 4650 | have their location_t to be P, until |
e3dfef44 GC |
4651 | cpp_stop_forcing_token_locations is called for R. */ |
4652 | ||
4653 | void | |
620e594b | 4654 | cpp_force_token_locations (cpp_reader *r, location_t loc) |
e3dfef44 | 4655 | { |
f3f6029d | 4656 | r->forced_token_location = loc; |
e3dfef44 GC |
4657 | } |
4658 | ||
4659 | /* Go back to assigning locations naturally for lexed tokens. */ | |
4660 | ||
4661 | void | |
4662 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
4663 | { | |
f3f6029d | 4664 | r->forced_token_location = 0; |
e3dfef44 | 4665 | } |
b224c376 NS |
4666 | |
4667 | /* We're looking at \, if it's escaping EOL, look past it. If at | |
4668 | LIMIT, don't advance. */ | |
4669 | ||
4670 | static const unsigned char * | |
4671 | do_peek_backslash (const unsigned char *peek, const unsigned char *limit) | |
4672 | { | |
4673 | const unsigned char *probe = peek; | |
4674 | ||
4675 | if (__builtin_expect (peek[1] == '\n', true)) | |
4676 | { | |
4677 | eol: | |
4678 | probe += 2; | |
4679 | if (__builtin_expect (probe < limit, true)) | |
4680 | { | |
4681 | peek = probe; | |
4682 | if (*peek == '\\') | |
4683 | /* The user might be perverse. */ | |
4684 | return do_peek_backslash (peek, limit); | |
4685 | } | |
4686 | } | |
4687 | else if (__builtin_expect (peek[1] == '\r', false)) | |
4688 | { | |
4689 | if (probe[2] == '\n') | |
4690 | probe++; | |
4691 | goto eol; | |
4692 | } | |
4693 | ||
4694 | return peek; | |
4695 | } | |
4696 | ||
4697 | static const unsigned char * | |
4698 | do_peek_next (const unsigned char *peek, const unsigned char *limit) | |
4699 | { | |
4700 | if (__builtin_expect (*peek == '\\', false)) | |
4701 | peek = do_peek_backslash (peek, limit); | |
4702 | return peek; | |
4703 | } | |
4704 | ||
4705 | static const unsigned char * | |
4706 | do_peek_prev (const unsigned char *peek, const unsigned char *bound) | |
4707 | { | |
4708 | if (peek == bound) | |
4709 | return NULL; | |
4710 | ||
4711 | unsigned char c = *--peek; | |
4712 | if (__builtin_expect (c == '\n', false) | |
4713 | || __builtin_expect (c == 'r', false)) | |
4714 | { | |
4715 | if (peek == bound) | |
4716 | return peek; | |
4717 | int ix = -1; | |
4718 | if (c == '\n' && peek[ix] == '\r') | |
4719 | { | |
4720 | if (peek + ix == bound) | |
4721 | return peek; | |
4722 | ix--; | |
4723 | } | |
4724 | ||
4725 | if (peek[ix] == '\\') | |
4726 | return do_peek_prev (peek + ix, bound); | |
4727 | ||
4728 | return peek; | |
4729 | } | |
4730 | else | |
4731 | return peek; | |
4732 | } | |
4733 | ||
c9c3d5f2 NS |
4734 | /* If PEEK[-1] is identifier MATCH, scan past it and trailing white |
4735 | space. Otherwise return NULL. */ | |
4736 | ||
4737 | static const unsigned char * | |
4738 | do_peek_ident (const char *match, const unsigned char *peek, | |
4739 | const unsigned char *limit) | |
4740 | { | |
4741 | for (; *++match; peek++) | |
4742 | if (*peek != *match) | |
4743 | { | |
4744 | peek = do_peek_next (peek, limit); | |
4745 | if (*peek != *match) | |
4746 | return NULL; | |
4747 | } | |
4748 | ||
4749 | /* Must now not be looking at an identifier char. */ | |
4750 | peek = do_peek_next (peek, limit); | |
4751 | if (ISIDNUM (*peek)) | |
4752 | return NULL; | |
4753 | ||
4754 | /* Skip control-line whitespace. */ | |
4755 | ws: | |
4756 | while (*peek == ' ' || *peek == '\t') | |
4757 | peek++; | |
4758 | if (__builtin_expect (*peek == '\\', false)) | |
4759 | { | |
4760 | peek = do_peek_backslash (peek, limit); | |
4761 | if (*peek != '\\') | |
4762 | goto ws; | |
4763 | } | |
4764 | ||
4765 | return peek; | |
4766 | } | |
4767 | ||
4768 | /* Are we looking at a module control line starting as PEEK - 1? */ | |
4769 | ||
4770 | static bool | |
4771 | do_peek_module (cpp_reader *pfile, unsigned char c, | |
4772 | const unsigned char *peek, const unsigned char *limit) | |
4773 | { | |
4774 | bool import = false; | |
4775 | ||
4776 | if (__builtin_expect (c == 'e', false)) | |
4777 | { | |
4778 | if (!((peek[0] == 'x' || peek[0] == '\\') | |
4779 | && (peek = do_peek_ident ("export", peek, limit)))) | |
4780 | return false; | |
4781 | ||
4782 | /* export, peek for import or module. No need to peek __import | |
4783 | here. */ | |
4784 | if (peek[0] == 'i') | |
4785 | { | |
4786 | if (!((peek[1] == 'm' || peek[1] == '\\') | |
4787 | && (peek = do_peek_ident ("import", peek + 1, limit)))) | |
4788 | return false; | |
4789 | import = true; | |
4790 | } | |
4791 | else if (peek[0] == 'm') | |
4792 | { | |
4793 | if (!((peek[1] == 'o' || peek[1] == '\\') | |
4794 | && (peek = do_peek_ident ("module", peek + 1, limit)))) | |
4795 | return false; | |
4796 | } | |
4797 | else | |
4798 | return false; | |
4799 | } | |
4800 | else if (__builtin_expect (c == 'i', false)) | |
4801 | { | |
4802 | if (!((peek[0] == 'm' || peek[0] == '\\') | |
4803 | && (peek = do_peek_ident ("import", peek, limit)))) | |
4804 | return false; | |
4805 | import = true; | |
4806 | } | |
4807 | else if (__builtin_expect (c == '_', false)) | |
4808 | { | |
4809 | /* Needed for translated includes. */ | |
4810 | if (!((peek[0] == '_' || peek[0] == '\\') | |
4811 | && (peek = do_peek_ident ("__import", peek, limit)))) | |
4812 | return false; | |
4813 | import = true; | |
4814 | } | |
4815 | else if (__builtin_expect (c == 'm', false)) | |
4816 | { | |
4817 | if (!((peek[0] == 'o' || peek[0] == '\\') | |
4818 | && (peek = do_peek_ident ("module", peek, limit)))) | |
4819 | return false; | |
4820 | } | |
4821 | else | |
4822 | return false; | |
4823 | ||
4824 | /* Peek the next character to see if it's good enough. We'll be at | |
4825 | the first non-whitespace char, including skipping an escaped | |
4826 | newline. */ | |
4827 | /* ... import followed by identifier, ':', '<' or header-name | |
4828 | preprocessing tokens, or module followed by identifier, ':' or | |
4829 | ';' preprocessing tokens. */ | |
4830 | unsigned char p = *peek++; | |
4831 | ||
4832 | /* A character literal is ... single quotes, ... optionally preceded | |
4833 | by u8, u, U, or L */ | |
4834 | /* A string-literal is a ... double quotes, optionally prefixed by | |
4835 | R, u8, u8R, u, uR, U, UR, L, or LR */ | |
4836 | if (p == 'u') | |
4837 | { | |
4838 | peek = do_peek_next (peek, limit); | |
4839 | if (*peek == '8') | |
4840 | { | |
4841 | peek++; | |
4842 | goto peek_u8; | |
4843 | } | |
4844 | goto peek_u; | |
4845 | } | |
4846 | else if (p == 'U' || p == 'L') | |
4847 | { | |
4848 | peek_u8: | |
4849 | peek = do_peek_next (peek, limit); | |
4850 | peek_u: | |
4851 | if (*peek == '\"' || *peek == '\'') | |
4852 | return false; | |
4853 | ||
4854 | if (*peek == 'R') | |
4855 | goto peek_R; | |
4856 | /* Identifier. Ok. */ | |
4857 | } | |
4858 | else if (p == 'R') | |
4859 | { | |
4860 | peek_R: | |
4861 | if (CPP_OPTION (pfile, rliterals)) | |
4862 | { | |
4863 | peek = do_peek_next (peek, limit); | |
4864 | if (*peek == '\"') | |
4865 | return false; | |
4866 | } | |
4867 | /* Identifier. Ok. */ | |
4868 | } | |
4869 | else if ('Z' - 'A' == 25 | |
4870 | ? ((p >= 'A' && p <= 'Z') || (p >= 'a' && p <= 'z') || p == '_') | |
4871 | : ISIDST (p)) | |
4872 | { | |
4873 | /* Identifier. Ok. */ | |
4874 | } | |
4875 | else if (p == '<') | |
4876 | { | |
4877 | /* Maybe angle header, ok for import. Reject | |
4878 | '<=', '<<' digraph:'<:'. */ | |
4879 | if (!import) | |
4880 | return false; | |
4881 | peek = do_peek_next (peek, limit); | |
4882 | if (*peek == '=' || *peek == '<' | |
4883 | || (*peek == ':' && CPP_OPTION (pfile, digraphs))) | |
4884 | return false; | |
4885 | } | |
4886 | else if (p == ';') | |
4887 | { | |
4888 | /* SEMICOLON, ok for module. */ | |
4889 | if (import) | |
4890 | return false; | |
4891 | } | |
4892 | else if (p == '"') | |
4893 | { | |
4894 | /* STRING, ok for import. */ | |
4895 | if (!import) | |
4896 | return false; | |
4897 | } | |
4898 | else if (p == ':') | |
4899 | { | |
4900 | /* Maybe COLON, ok. Reject '::', digraph:':>'. */ | |
4901 | peek = do_peek_next (peek, limit); | |
4902 | if (*peek == ':' || (*peek == '>' && CPP_OPTION (pfile, digraphs))) | |
4903 | return false; | |
4904 | } | |
4905 | else | |
4906 | /* FIXME: Detect a unicode character, excluding those not | |
4907 | permitted as the initial character. [lex.name]/1. I presume | |
4908 | we need to check the \[uU] spellings, and directly using | |
4909 | Unicode in say UTF8 form? Or perhaps we do the phase-1 | |
4910 | conversion of UTF8 to universal-character-names? */ | |
4911 | return false; | |
4912 | ||
4913 | return true; | |
4914 | } | |
4915 | ||
b224c376 NS |
4916 | /* Directives-only scanning. Somewhat more relaxed than correct |
4917 | parsing -- some ill-formed programs will not be rejected. */ | |
4918 | ||
4919 | void | |
4920 | cpp_directive_only_process (cpp_reader *pfile, | |
4921 | void *data, | |
4922 | void (*cb) (cpp_reader *, CPP_DO_task, void *, ...)) | |
4923 | { | |
c9c3d5f2 NS |
4924 | bool module_p = CPP_OPTION (pfile, module_directives); |
4925 | ||
b224c376 NS |
4926 | do |
4927 | { | |
4928 | restart: | |
4929 | /* Buffer initialization, but no line cleaning. */ | |
4930 | cpp_buffer *buffer = pfile->buffer; | |
4931 | buffer->cur_note = buffer->notes_used = 0; | |
4932 | buffer->cur = buffer->line_base = buffer->next_line; | |
4933 | buffer->need_line = false; | |
ac16f432 | 4934 | /* Files always end in a newline or carriage return. We rely on this for |
b224c376 | 4935 | character peeking safety. */ |
ac16f432 | 4936 | gcc_assert (buffer->rlimit[0] == '\n' || buffer->rlimit[0] == '\r'); |
b224c376 NS |
4937 | |
4938 | const unsigned char *base = buffer->cur; | |
4939 | unsigned line_count = 0; | |
4940 | const unsigned char *line_start = base; | |
4941 | ||
4942 | bool bol = true; | |
4943 | bool raw = false; | |
4944 | ||
4945 | const unsigned char *lwm = base; | |
4946 | for (const unsigned char *pos = base, *limit = buffer->rlimit; | |
4947 | pos < limit;) | |
4948 | { | |
4949 | unsigned char c = *pos++; | |
4950 | /* This matches the switch in _cpp_lex_direct. */ | |
4951 | switch (c) | |
4952 | { | |
4953 | case ' ': case '\t': case '\f': case '\v': | |
4954 | /* Whitespace, do nothing. */ | |
4955 | break; | |
4956 | ||
4957 | case '\r': /* MAC line ending, or Windows \r\n */ | |
4958 | if (*pos == '\n') | |
4959 | pos++; | |
4960 | /* FALLTHROUGH */ | |
4961 | ||
4962 | case '\n': | |
4963 | bol = true; | |
4964 | ||
4965 | next_line: | |
4966 | CPP_INCREMENT_LINE (pfile, 0); | |
4967 | line_count++; | |
4968 | line_start = pos; | |
4969 | break; | |
4970 | ||
4971 | case '\\': | |
4972 | /* <backslash><newline> is removed, and doesn't undo any | |
4973 | preceeding escape or whatnot. */ | |
4974 | if (*pos == '\n') | |
4975 | { | |
4976 | pos++; | |
4977 | goto next_line; | |
4978 | } | |
4979 | else if (*pos == '\r') | |
4980 | { | |
4981 | if (pos[1] == '\n') | |
4982 | pos++; | |
4983 | pos++; | |
4984 | goto next_line; | |
4985 | } | |
4986 | goto dflt; | |
4987 | ||
4988 | case '#': | |
4989 | if (bol) | |
4990 | { | |
4991 | /* Line directive. */ | |
4992 | if (pos - 1 > base && !pfile->state.skipping) | |
4993 | cb (pfile, CPP_DO_print, data, | |
4994 | line_count, base, pos - 1 - base); | |
4995 | ||
4996 | /* Prep things for directive handling. */ | |
4997 | buffer->next_line = pos; | |
4998 | buffer->need_line = true; | |
2a0225e4 NS |
4999 | bool ok = _cpp_get_fresh_line (pfile); |
5000 | gcc_checking_assert (ok); | |
b224c376 NS |
5001 | |
5002 | /* Ensure proper column numbering for generated | |
5003 | error messages. */ | |
5004 | buffer->line_base -= pos - line_start; | |
5005 | ||
5006 | _cpp_handle_directive (pfile, line_start + 1 != pos); | |
5007 | ||
5008 | /* Sanitize the line settings. Duplicate #include's can | |
5009 | mess things up. */ | |
5010 | // FIXME: Necessary? | |
5011 | pfile->line_table->highest_location | |
5012 | = pfile->line_table->highest_line; | |
5013 | ||
5014 | if (!pfile->state.skipping | |
5015 | && pfile->buffer->next_line < pfile->buffer->rlimit) | |
5016 | cb (pfile, CPP_DO_location, data, | |
5017 | pfile->line_table->highest_line); | |
5018 | ||
5019 | goto restart; | |
5020 | } | |
5021 | goto dflt; | |
5022 | ||
5023 | case '/': | |
5024 | { | |
5025 | const unsigned char *peek = do_peek_next (pos, limit); | |
5026 | if (!(*peek == '/' || *peek == '*')) | |
5027 | goto dflt; | |
5028 | ||
5029 | /* Line or block comment */ | |
5030 | bool is_block = *peek == '*'; | |
5031 | bool star = false; | |
5032 | bool esc = false; | |
5033 | location_t sloc | |
5034 | = linemap_position_for_column (pfile->line_table, | |
5035 | pos - line_start); | |
5036 | ||
5037 | while (pos < limit) | |
5038 | { | |
5039 | char c = *pos++; | |
5040 | switch (c) | |
5041 | { | |
5042 | case '\\': | |
5043 | esc = true; | |
5044 | break; | |
5045 | ||
5046 | case '\r': | |
5047 | if (*pos == '\n') | |
5048 | pos++; | |
5049 | /* FALLTHROUGH */ | |
5050 | ||
5051 | case '\n': | |
5052 | { | |
5053 | CPP_INCREMENT_LINE (pfile, 0); | |
5054 | line_count++; | |
5055 | line_start = pos; | |
5056 | if (!esc && !is_block) | |
5057 | { | |
5058 | bol = true; | |
5059 | goto done_comment; | |
5060 | } | |
5061 | } | |
5062 | if (!esc) | |
5063 | star = false; | |
5064 | esc = false; | |
5065 | break; | |
5066 | ||
5067 | case '*': | |
049f0efe | 5068 | if (pos > peek) |
b224c376 NS |
5069 | star = is_block; |
5070 | esc = false; | |
5071 | break; | |
5072 | ||
5073 | case '/': | |
5074 | if (star) | |
5075 | goto done_comment; | |
5076 | /* FALLTHROUGH */ | |
5077 | ||
5078 | default: | |
5079 | star = false; | |
5080 | esc = false; | |
5081 | break; | |
5082 | } | |
5083 | } | |
d15a2d26 JJ |
5084 | if (pos < limit || is_block) |
5085 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, | |
5086 | "unterminated comment"); | |
b224c376 NS |
5087 | done_comment: |
5088 | lwm = pos; | |
5089 | break; | |
5090 | } | |
5091 | ||
5092 | case '\'': | |
5093 | if (!CPP_OPTION (pfile, digit_separators)) | |
5094 | goto delimited_string; | |
5095 | ||
5096 | /* Possibly a number punctuator. */ | |
5097 | if (!ISIDNUM (*do_peek_next (pos, limit))) | |
5098 | goto delimited_string; | |
5099 | ||
5100 | goto quote_peek; | |
5101 | ||
5102 | case '\"': | |
5103 | if (!CPP_OPTION (pfile, rliterals)) | |
5104 | goto delimited_string; | |
5105 | ||
5106 | quote_peek: | |
5107 | { | |
5108 | /* For ' see if it's a number punctuator | |
5109 | \.?<digit>(<digit>|<identifier-nondigit> | |
5110 | |'<digit>|'<nondigit>|[eEpP]<sign>|\.)* */ | |
5111 | /* For " see if it's a raw string | |
5112 | {U,L,u,u8}R. This includes CPP_NUMBER detection, | |
5113 | because that could be 0e+R. */ | |
5114 | const unsigned char *peek = pos - 1; | |
5115 | bool quote_first = c == '"'; | |
5116 | bool quote_eight = false; | |
5117 | bool maybe_number_start = false; | |
5118 | bool want_number = false; | |
5119 | ||
5120 | while ((peek = do_peek_prev (peek, lwm))) | |
5121 | { | |
5122 | unsigned char p = *peek; | |
5123 | if (quote_first) | |
5124 | { | |
5125 | if (!raw) | |
5126 | { | |
5127 | if (p != 'R') | |
5128 | break; | |
5129 | raw = true; | |
5130 | continue; | |
5131 | } | |
5132 | ||
5133 | quote_first = false; | |
5134 | if (p == 'L' || p == 'U' || p == 'u') | |
5135 | ; | |
5136 | else if (p == '8') | |
5137 | quote_eight = true; | |
5138 | else | |
5139 | goto second_raw; | |
5140 | } | |
5141 | else if (quote_eight) | |
5142 | { | |
5143 | if (p != 'u') | |
5144 | { | |
5145 | raw = false; | |
5146 | break; | |
5147 | } | |
5148 | quote_eight = false; | |
5149 | } | |
5150 | else if (c == '"') | |
5151 | { | |
5152 | second_raw:; | |
5153 | if (!want_number && ISIDNUM (p)) | |
5154 | { | |
5155 | raw = false; | |
5156 | break; | |
5157 | } | |
5158 | } | |
5159 | ||
5160 | if (ISDIGIT (p)) | |
5161 | maybe_number_start = true; | |
5162 | else if (p == '.') | |
5163 | want_number = true; | |
5164 | else if (ISIDNUM (p)) | |
5165 | maybe_number_start = false; | |
5166 | else if (p == '+' || p == '-') | |
5167 | { | |
5168 | if (const unsigned char *peek_prev | |
5169 | = do_peek_prev (peek, lwm)) | |
5170 | { | |
5171 | p = *peek_prev; | |
5172 | if (p == 'e' || p == 'E' | |
5173 | || p == 'p' || p == 'P') | |
5174 | { | |
5175 | want_number = true; | |
5176 | maybe_number_start = false; | |
5177 | } | |
5178 | else | |
5179 | break; | |
5180 | } | |
5181 | else | |
5182 | break; | |
5183 | } | |
5184 | else if (p == '\'' || p == '\"') | |
5185 | { | |
5186 | /* If this is lwm, this must be the end of a | |
5187 | previous string. So this is a trailing | |
5188 | literal type, (a) if those are allowed, | |
5189 | and (b) maybe_start is false. Otherwise | |
5190 | this must be a CPP_NUMBER because we've | |
5191 | met another ', and we'd have checked that | |
5192 | in its own right. */ | |
5193 | if (peek == lwm && CPP_OPTION (pfile, uliterals)) | |
5194 | { | |
5195 | if (!maybe_number_start && !want_number) | |
5196 | /* Must be a literal type. */ | |
5197 | raw = false; | |
5198 | } | |
5199 | else if (p == '\'' | |
5200 | && CPP_OPTION (pfile, digit_separators)) | |
5201 | maybe_number_start = true; | |
5202 | break; | |
5203 | } | |
5204 | else if (c == '\'') | |
5205 | break; | |
5206 | else if (!quote_first && !quote_eight) | |
5207 | break; | |
5208 | } | |
5209 | ||
5210 | if (maybe_number_start) | |
5211 | { | |
5212 | if (c == '\'') | |
5213 | /* A CPP NUMBER. */ | |
5214 | goto dflt; | |
5215 | raw = false; | |
5216 | } | |
5217 | ||
5218 | goto delimited_string; | |
5219 | } | |
5220 | ||
5221 | delimited_string: | |
5222 | { | |
5223 | /* (Possibly raw) string or char literal. */ | |
5224 | unsigned char end = c; | |
5225 | int delim_len = -1; | |
5226 | const unsigned char *delim = NULL; | |
5227 | location_t sloc = linemap_position_for_column (pfile->line_table, | |
5228 | pos - line_start); | |
5229 | int esc = 0; | |
5230 | ||
5231 | if (raw) | |
5232 | { | |
5233 | /* There can be no line breaks in the delimiter. */ | |
5234 | delim = pos; | |
5235 | for (delim_len = 0; (c = *pos++) != '('; delim_len++) | |
5236 | { | |
5237 | if (delim_len == 16) | |
5238 | { | |
5239 | cpp_error_with_line (pfile, CPP_DL_ERROR, | |
5240 | sloc, 0, | |
5241 | "raw string delimiter" | |
5242 | " longer than %d" | |
5243 | " characters", | |
5244 | delim_len); | |
5245 | raw = false; | |
5246 | pos = delim; | |
5247 | break; | |
5248 | } | |
5249 | if (strchr (") \\\t\v\f\n", c)) | |
5250 | { | |
5251 | cpp_error_with_line (pfile, CPP_DL_ERROR, | |
5252 | sloc, 0, | |
5253 | "invalid character '%c'" | |
5254 | " in raw string" | |
5255 | " delimiter", c); | |
5256 | raw = false; | |
5257 | pos = delim; | |
5258 | break; | |
5259 | } | |
5260 | if (pos >= limit) | |
5261 | goto bad_string; | |
5262 | } | |
5263 | } | |
5264 | ||
5265 | while (pos < limit) | |
5266 | { | |
5267 | char c = *pos++; | |
5268 | switch (c) | |
5269 | { | |
5270 | case '\\': | |
5271 | if (!raw) | |
5272 | esc++; | |
5273 | break; | |
5274 | ||
5275 | case '\r': | |
5276 | if (*pos == '\n') | |
5277 | pos++; | |
5278 | /* FALLTHROUGH */ | |
5279 | ||
5280 | case '\n': | |
5281 | { | |
5282 | CPP_INCREMENT_LINE (pfile, 0); | |
5283 | line_count++; | |
5284 | line_start = pos; | |
5285 | } | |
5286 | if (esc) | |
5287 | esc--; | |
5288 | break; | |
5289 | ||
5290 | case ')': | |
5291 | if (raw | |
5292 | && pos + delim_len + 1 < limit | |
5293 | && pos[delim_len] == end | |
5294 | && !memcmp (delim, pos, delim_len)) | |
5295 | { | |
5296 | pos += delim_len + 1; | |
5297 | raw = false; | |
5298 | goto done_string; | |
5299 | } | |
5300 | break; | |
5301 | ||
5302 | default: | |
5303 | if (!raw && !(esc & 1) && c == end) | |
5304 | goto done_string; | |
5305 | esc = 0; | |
5306 | break; | |
5307 | } | |
5308 | } | |
5309 | bad_string: | |
5310 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, | |
5311 | "unterminated literal"); | |
5312 | ||
5313 | done_string: | |
5314 | raw = false; | |
5315 | lwm = pos - 1; | |
5316 | } | |
5317 | goto dflt; | |
5318 | ||
c9c3d5f2 NS |
5319 | case '_': |
5320 | case 'e': | |
5321 | case 'i': | |
5322 | case 'm': | |
5323 | if (bol && module_p && !pfile->state.skipping | |
5324 | && do_peek_module (pfile, c, pos, limit)) | |
5325 | { | |
5326 | /* We've seen the start of a module control line. | |
5327 | Start up the tokenizer. */ | |
5328 | pos--; /* Backup over the first character. */ | |
5329 | ||
5330 | /* Backup over whitespace to start of line. */ | |
5331 | while (pos > line_start | |
5332 | && (pos[-1] == ' ' || pos[-1] == '\t')) | |
5333 | pos--; | |
5334 | ||
5335 | if (pos > base) | |
5336 | cb (pfile, CPP_DO_print, data, line_count, base, pos - base); | |
5337 | ||
5338 | /* Prep things for directive handling. */ | |
5339 | buffer->next_line = pos; | |
5340 | buffer->need_line = true; | |
5341 | ||
5342 | /* Now get tokens until the PRAGMA_EOL. */ | |
5343 | do | |
5344 | { | |
5345 | location_t spelling; | |
5346 | const cpp_token *tok | |
5347 | = cpp_get_token_with_location (pfile, &spelling); | |
5348 | ||
5349 | gcc_assert (pfile->state.in_deferred_pragma | |
5350 | || tok->type == CPP_PRAGMA_EOL); | |
5351 | cb (pfile, CPP_DO_token, data, tok, spelling); | |
5352 | } | |
5353 | while (pfile->state.in_deferred_pragma); | |
5354 | ||
5355 | if (pfile->buffer->next_line < pfile->buffer->rlimit) | |
5356 | cb (pfile, CPP_DO_location, data, | |
5357 | pfile->line_table->highest_line); | |
5358 | ||
5359 | pfile->mi_valid = false; | |
5360 | goto restart; | |
5361 | } | |
5362 | goto dflt; | |
5363 | ||
b224c376 NS |
5364 | default: |
5365 | dflt: | |
5366 | bol = false; | |
5367 | pfile->mi_valid = false; | |
5368 | break; | |
5369 | } | |
5370 | } | |
5371 | ||
5372 | if (buffer->rlimit > base && !pfile->state.skipping) | |
c6b664e2 JJ |
5373 | { |
5374 | const unsigned char *limit = buffer->rlimit; | |
5375 | /* If the file was not newline terminated, add rlimit, which is | |
5376 | guaranteed to point to a newline, to the end of our range. */ | |
5377 | if (limit[-1] != '\n') | |
5378 | { | |
5379 | limit++; | |
5380 | CPP_INCREMENT_LINE (pfile, 0); | |
5381 | line_count++; | |
5382 | } | |
5383 | cb (pfile, CPP_DO_print, data, line_count, base, limit - base); | |
5384 | } | |
b224c376 NS |
5385 | |
5386 | _cpp_pop_buffer (pfile); | |
5387 | } | |
5388 | while (pfile->buffer); | |
5389 | } |