]>
Commit | Line | Data |
---|---|---|
45b966db | 1 | /* CPP Library - lexical analysis. |
246a2fcb | 2 | Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009, 2010 |
748086b7 | 3 | Free Software Foundation, Inc. |
45b966db ZW |
4 | Contributed by Per Bothner, 1994-95. |
5 | Based on CCCP program by Paul Rubin, June 1986 | |
6 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
7 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
8 | ||
9 | This program is free software; you can redistribute it and/or modify it | |
10 | under the terms of the GNU General Public License as published by the | |
748086b7 | 11 | Free Software Foundation; either version 3, or (at your option) any |
45b966db ZW |
12 | later version. |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
748086b7 JJ |
20 | along with this program; see the file COPYING3. If not see |
21 | <http://www.gnu.org/licenses/>. */ | |
45b966db ZW |
22 | |
23 | #include "config.h" | |
24 | #include "system.h" | |
45b966db | 25 | #include "cpplib.h" |
4f4e53dd | 26 | #include "internal.h" |
45b966db | 27 | |
93c80368 | 28 | enum spell_type |
f9a0e96c | 29 | { |
93c80368 | 30 | SPELL_OPERATOR = 0, |
93c80368 | 31 | SPELL_IDENT, |
6338b358 | 32 | SPELL_LITERAL, |
93c80368 | 33 | SPELL_NONE |
f9a0e96c ZW |
34 | }; |
35 | ||
93c80368 | 36 | struct token_spelling |
f9a0e96c | 37 | { |
93c80368 NB |
38 | enum spell_type category; |
39 | const unsigned char *name; | |
f9a0e96c ZW |
40 | }; |
41 | ||
8206c799 | 42 | static const unsigned char *const digraph_spellings[] = |
b6baa67d | 43 | { UC"%:", UC"%:%:", UC"<:", UC":>", UC"<%", UC"%>" }; |
93c80368 | 44 | |
b6baa67d KVH |
45 | #define OP(e, s) { SPELL_OPERATOR, UC s }, |
46 | #define TK(e, s) { SPELL_ ## s, UC #e }, | |
8206c799 | 47 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; |
93c80368 NB |
48 | #undef OP |
49 | #undef TK | |
50 | ||
51 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
52 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
f2d5f0cc | 53 | |
6cf87ca4 ZW |
54 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); |
55 | static int skip_line_comment (cpp_reader *); | |
56 | static void skip_whitespace (cpp_reader *, cppchar_t); | |
6cf87ca4 ZW |
57 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); |
58 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); | |
631d0d36 | 59 | static void store_comment (cpp_reader *, cpp_token *); |
6cf87ca4 ZW |
60 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, |
61 | unsigned int, enum cpp_ttype); | |
62 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); | |
63 | static int name_p (cpp_reader *, const cpp_string *); | |
6cf87ca4 ZW |
64 | static tokenrun *next_tokenrun (tokenrun *); |
65 | ||
6cf87ca4 | 66 | static _cpp_buff *new_buff (size_t); |
15dad1d9 | 67 | |
9d10c9a9 | 68 | |
041c3194 | 69 | /* Utility routine: |
9e62c811 | 70 | |
bfb9dc7f ZW |
71 | Compares, the token TOKEN to the NUL-terminated string STRING. |
72 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
041c3194 | 73 | int |
6cf87ca4 | 74 | cpp_ideq (const cpp_token *token, const char *string) |
041c3194 | 75 | { |
bfb9dc7f | 76 | if (token->type != CPP_NAME) |
041c3194 | 77 | return 0; |
bfb9dc7f | 78 | |
9a0c6187 | 79 | return !ustrcmp (NODE_NAME (token->val.node.node), (const uchar *) string); |
15dad1d9 | 80 | } |
1368ee70 | 81 | |
26aea073 NB |
82 | /* Record a note TYPE at byte POS into the current cleaned logical |
83 | line. */ | |
87062813 | 84 | static void |
6cf87ca4 | 85 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) |
0d9f234d | 86 | { |
26aea073 NB |
87 | if (buffer->notes_used == buffer->notes_cap) |
88 | { | |
89 | buffer->notes_cap = buffer->notes_cap * 2 + 200; | |
c3f829c1 GDR |
90 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes, |
91 | buffer->notes_cap); | |
26aea073 | 92 | } |
0d9f234d | 93 | |
26aea073 NB |
94 | buffer->notes[buffer->notes_used].pos = pos; |
95 | buffer->notes[buffer->notes_used].type = type; | |
96 | buffer->notes_used++; | |
0d9f234d NB |
97 | } |
98 | ||
246a2fcb RH |
99 | \f |
100 | /* Fast path to find line special characters using optimized character | |
101 | scanning algorithms. Anything complicated falls back to the slow | |
102 | path below. Since this loop is very hot it's worth doing these kinds | |
103 | of optimizations. | |
104 | ||
105 | One of the paths through the ifdefs should provide | |
106 | ||
107 | const uchar *search_line_fast (const uchar *s, const uchar *end); | |
108 | ||
109 | Between S and END, search for \n, \r, \\, ?. Return a pointer to | |
110 | the found character. | |
111 | ||
112 | Note that the last character of the buffer is *always* a newline, | |
113 | as forced by _cpp_convert_input. This fact can be used to avoid | |
114 | explicitly looking for the end of the buffer. */ | |
115 | ||
116 | /* Configure gives us an ifdef test. */ | |
117 | #ifndef WORDS_BIGENDIAN | |
118 | #define WORDS_BIGENDIAN 0 | |
119 | #endif | |
120 | ||
121 | /* We'd like the largest integer that fits into a register. There's nothing | |
122 | in <stdint.h> that gives us that. For most hosts this is unsigned long, | |
123 | but MS decided on an LLP64 model. Thankfully when building with GCC we | |
124 | can get the "real" word size. */ | |
125 | #ifdef __GNUC__ | |
126 | typedef unsigned int word_type __attribute__((__mode__(__word__))); | |
127 | #else | |
128 | typedef unsigned long word_type; | |
129 | #endif | |
130 | ||
131 | /* The code below is only expecting sizes 4 or 8. | |
132 | Die at compile-time if this expectation is violated. */ | |
133 | typedef char check_word_type_size | |
134 | [(sizeof(word_type) == 8 || sizeof(word_type) == 4) * 2 - 1]; | |
135 | ||
136 | /* Return X with the first N bytes forced to values that won't match one | |
137 | of the interesting characters. Note that NUL is not interesting. */ | |
138 | ||
139 | static inline word_type | |
140 | acc_char_mask_misalign (word_type val, unsigned int n) | |
141 | { | |
142 | word_type mask = -1; | |
143 | if (WORDS_BIGENDIAN) | |
144 | mask >>= n * 8; | |
145 | else | |
146 | mask <<= n * 8; | |
147 | return val & mask; | |
148 | } | |
149 | ||
150 | /* Return X replicated to all byte positions within WORD_TYPE. */ | |
151 | ||
152 | static inline word_type | |
153 | acc_char_replicate (uchar x) | |
154 | { | |
155 | word_type ret; | |
156 | ||
157 | ret = (x << 24) | (x << 16) | (x << 8) | x; | |
158 | if (sizeof(word_type) == 8) | |
159 | ret = (ret << 16 << 16) | ret; | |
160 | return ret; | |
161 | } | |
162 | ||
163 | /* Return non-zero if some byte of VAL is (probably) C. */ | |
164 | ||
165 | static inline word_type | |
166 | acc_char_cmp (word_type val, word_type c) | |
167 | { | |
168 | #if defined(__GNUC__) && defined(__alpha__) | |
169 | /* We can get exact results using a compare-bytes instruction. | |
170 | Get (val == c) via (0 >= (val ^ c)). */ | |
171 | return __builtin_alpha_cmpbge (0, val ^ c); | |
172 | #else | |
173 | word_type magic = 0x7efefefeU; | |
174 | if (sizeof(word_type) == 8) | |
175 | magic = (magic << 16 << 16) | 0xfefefefeU; | |
176 | magic |= 1; | |
177 | ||
178 | val ^= c; | |
179 | return ((val + magic) ^ ~val) & ~magic; | |
180 | #endif | |
181 | } | |
182 | ||
183 | /* Given the result of acc_char_cmp is non-zero, return the index of | |
184 | the found character. If this was a false positive, return -1. */ | |
185 | ||
186 | static inline int | |
187 | acc_char_index (word_type cmp ATTRIBUTE_UNUSED, | |
188 | word_type val ATTRIBUTE_UNUSED) | |
189 | { | |
190 | #if defined(__GNUC__) && defined(__alpha__) && !WORDS_BIGENDIAN | |
191 | /* The cmpbge instruction sets *bits* of the result corresponding to | |
192 | matches in the bytes with no false positives. */ | |
193 | return __builtin_ctzl (cmp); | |
194 | #else | |
195 | unsigned int i; | |
196 | ||
197 | /* ??? It would be nice to force unrolling here, | |
198 | and have all of these constants folded. */ | |
199 | for (i = 0; i < sizeof(word_type); ++i) | |
200 | { | |
201 | uchar c; | |
202 | if (WORDS_BIGENDIAN) | |
203 | c = (val >> (sizeof(word_type) - i - 1) * 8) & 0xff; | |
204 | else | |
205 | c = (val >> i * 8) & 0xff; | |
206 | ||
207 | if (c == '\n' || c == '\r' || c == '\\' || c == '?') | |
208 | return i; | |
209 | } | |
210 | ||
211 | return -1; | |
212 | #endif | |
213 | } | |
214 | ||
215 | /* A version of the fast scanner using bit fiddling techniques. | |
216 | ||
217 | For 32-bit words, one would normally perform 16 comparisons and | |
218 | 16 branches. With this algorithm one performs 24 arithmetic | |
219 | operations and one branch. Whether this is faster with a 32-bit | |
220 | word size is going to be somewhat system dependent. | |
221 | ||
222 | For 64-bit words, we eliminate twice the number of comparisons | |
223 | and branches without increasing the number of arithmetic operations. | |
224 | It's almost certainly going to be a win with 64-bit word size. */ | |
225 | ||
226 | static const uchar * search_line_acc_char (const uchar *, const uchar *) | |
227 | ATTRIBUTE_UNUSED; | |
228 | ||
229 | static const uchar * | |
230 | search_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
231 | { | |
232 | const word_type repl_nl = acc_char_replicate ('\n'); | |
233 | const word_type repl_cr = acc_char_replicate ('\r'); | |
234 | const word_type repl_bs = acc_char_replicate ('\\'); | |
235 | const word_type repl_qm = acc_char_replicate ('?'); | |
236 | ||
237 | unsigned int misalign; | |
238 | const word_type *p; | |
239 | word_type val, t; | |
240 | ||
241 | /* Align the buffer. Mask out any bytes from before the beginning. */ | |
242 | p = (word_type *)((uintptr_t)s & -sizeof(word_type)); | |
243 | val = *p; | |
244 | misalign = (uintptr_t)s & (sizeof(word_type) - 1); | |
245 | if (misalign) | |
246 | val = acc_char_mask_misalign (val, misalign); | |
247 | ||
248 | /* Main loop. */ | |
249 | while (1) | |
250 | { | |
251 | t = acc_char_cmp (val, repl_nl); | |
252 | t |= acc_char_cmp (val, repl_cr); | |
253 | t |= acc_char_cmp (val, repl_bs); | |
254 | t |= acc_char_cmp (val, repl_qm); | |
255 | ||
256 | if (__builtin_expect (t != 0, 0)) | |
257 | { | |
258 | int i = acc_char_index (t, val); | |
259 | if (i >= 0) | |
260 | return (const uchar *)p + i; | |
261 | } | |
262 | ||
263 | val = *++p; | |
264 | } | |
265 | } | |
266 | ||
789d73cb RO |
267 | /* Disable on Solaris 2/x86 until the following problems can be properly |
268 | autoconfed: | |
269 | ||
270 | The Solaris 8 assembler cannot assemble SSE2/SSE4.2 insns. | |
271 | The Solaris 9 assembler cannot assemble SSE4.2 insns. | |
272 | Before Solaris 9 Update 6, SSE insns cannot be executed. | |
273 | The Solaris 10+ assembler tags objects with the instruction set | |
274 | extensions used, so SSE4.2 executables cannot run on machines that | |
275 | don't support that extension. */ | |
276 | ||
277 | #if (GCC_VERSION >= 4005) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__)) | |
246a2fcb RH |
278 | |
279 | /* Replicated character data to be shared between implementations. | |
280 | Recall that outside of a context with vector support we can't | |
281 | define compatible vector types, therefore these are all defined | |
282 | in terms of raw characters. */ | |
283 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { | |
284 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
285 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, | |
286 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
287 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, | |
288 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
289 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, | |
290 | { '?', '?', '?', '?', '?', '?', '?', '?', | |
291 | '?', '?', '?', '?', '?', '?', '?', '?' }, | |
292 | }; | |
293 | ||
294 | /* A version of the fast scanner using MMX vectorized byte compare insns. | |
295 | ||
296 | This uses the PMOVMSKB instruction which was introduced with "MMX2", | |
ef230b38 | 297 | which was packaged into SSE1; it is also present in the AMD MMX |
246a2fcb RH |
298 | extension. Mark the function as using "sse" so that we emit a real |
299 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ | |
300 | ||
301 | static const uchar * | |
302 | #ifndef __SSE__ | |
303 | __attribute__((__target__("sse"))) | |
304 | #endif | |
305 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
306 | { | |
307 | typedef char v8qi __attribute__ ((__vector_size__ (8))); | |
308 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); | |
309 | ||
310 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; | |
311 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; | |
312 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; | |
313 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; | |
314 | ||
315 | unsigned int misalign, found, mask; | |
316 | const v8qi *p; | |
317 | v8qi data, t, c; | |
318 | ||
319 | /* Align the source pointer. While MMX doesn't generate unaligned data | |
320 | faults, this allows us to safely scan to the end of the buffer without | |
321 | reading beyond the end of the last page. */ | |
322 | misalign = (uintptr_t)s & 7; | |
323 | p = (const v8qi *)((uintptr_t)s & -8); | |
324 | data = *p; | |
325 | ||
326 | /* Create a mask for the bytes that are valid within the first | |
327 | 16-byte block. The Idea here is that the AND with the mask | |
328 | within the loop is "free", since we need some AND or TEST | |
329 | insn in order to set the flags for the branch anyway. */ | |
330 | mask = -1u << misalign; | |
331 | ||
332 | /* Main loop processing 8 bytes at a time. */ | |
333 | goto start; | |
334 | do | |
335 | { | |
336 | data = *++p; | |
337 | mask = -1; | |
338 | ||
339 | start: | |
340 | t = __builtin_ia32_pcmpeqb(data, repl_nl); | |
341 | c = __builtin_ia32_pcmpeqb(data, repl_cr); | |
342 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
343 | c = __builtin_ia32_pcmpeqb(data, repl_bs); | |
344 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
345 | c = __builtin_ia32_pcmpeqb(data, repl_qm); | |
346 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
347 | found = __builtin_ia32_pmovmskb (t); | |
348 | found &= mask; | |
349 | } | |
350 | while (!found); | |
351 | ||
352 | __builtin_ia32_emms (); | |
353 | ||
354 | /* FOUND contains 1 in bits for which we matched a relevant | |
355 | character. Conversion to the byte index is trivial. */ | |
356 | found = __builtin_ctz(found); | |
357 | return (const uchar *)p + found; | |
358 | } | |
359 | ||
360 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ | |
361 | ||
362 | static const uchar * | |
363 | #ifndef __SSE2__ | |
364 | __attribute__((__target__("sse2"))) | |
365 | #endif | |
366 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
367 | { | |
368 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
369 | ||
370 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; | |
371 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; | |
372 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; | |
373 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; | |
374 | ||
375 | unsigned int misalign, found, mask; | |
376 | const v16qi *p; | |
377 | v16qi data, t; | |
378 | ||
379 | /* Align the source pointer. */ | |
380 | misalign = (uintptr_t)s & 15; | |
381 | p = (const v16qi *)((uintptr_t)s & -16); | |
382 | data = *p; | |
383 | ||
384 | /* Create a mask for the bytes that are valid within the first | |
385 | 16-byte block. The Idea here is that the AND with the mask | |
386 | within the loop is "free", since we need some AND or TEST | |
387 | insn in order to set the flags for the branch anyway. */ | |
388 | mask = -1u << misalign; | |
389 | ||
390 | /* Main loop processing 16 bytes at a time. */ | |
391 | goto start; | |
392 | do | |
393 | { | |
394 | data = *++p; | |
395 | mask = -1; | |
396 | ||
397 | start: | |
398 | t = __builtin_ia32_pcmpeqb128(data, repl_nl); | |
399 | t |= __builtin_ia32_pcmpeqb128(data, repl_cr); | |
400 | t |= __builtin_ia32_pcmpeqb128(data, repl_bs); | |
401 | t |= __builtin_ia32_pcmpeqb128(data, repl_qm); | |
402 | found = __builtin_ia32_pmovmskb128 (t); | |
403 | found &= mask; | |
404 | } | |
405 | while (!found); | |
406 | ||
407 | /* FOUND contains 1 in bits for which we matched a relevant | |
408 | character. Conversion to the byte index is trivial. */ | |
409 | found = __builtin_ctz(found); | |
410 | return (const uchar *)p + found; | |
411 | } | |
412 | ||
6f173e52 | 413 | #ifdef HAVE_SSE4 |
246a2fcb RH |
414 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
415 | ||
416 | static const uchar * | |
417 | #ifndef __SSE4_2__ | |
418 | __attribute__((__target__("sse4.2"))) | |
419 | #endif | |
420 | search_line_sse42 (const uchar *s, const uchar *end) | |
421 | { | |
422 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
423 | static const v16qi search = { '\n', '\r', '?', '\\' }; | |
424 | ||
425 | uintptr_t si = (uintptr_t)s; | |
426 | uintptr_t index; | |
427 | ||
428 | /* Check for unaligned input. */ | |
429 | if (si & 15) | |
430 | { | |
431 | if (__builtin_expect (end - s < 16, 0) | |
432 | && __builtin_expect ((si & 0xfff) > 0xff0, 0)) | |
433 | { | |
434 | /* There are less than 16 bytes left in the buffer, and less | |
435 | than 16 bytes left on the page. Reading 16 bytes at this | |
436 | point might generate a spurious page fault. Defer to the | |
437 | SSE2 implementation, which already handles alignment. */ | |
438 | return search_line_sse2 (s, end); | |
439 | } | |
440 | ||
441 | /* ??? The builtin doesn't understand that the PCMPESTRI read from | |
442 | memory need not be aligned. */ | |
443 | __asm ("%vpcmpestri $0, (%1), %2" | |
444 | : "=c"(index) : "r"(s), "x"(search), "a"(4), "d"(16)); | |
445 | if (__builtin_expect (index < 16, 0)) | |
446 | goto found; | |
447 | ||
448 | /* Advance the pointer to an aligned address. We will re-scan a | |
449 | few bytes, but we no longer need care for reading past the | |
450 | end of a page, since we're guaranteed a match. */ | |
451 | s = (const uchar *)((si + 16) & -16); | |
452 | } | |
453 | ||
454 | /* Main loop, processing 16 bytes at a time. By doing the whole loop | |
455 | in inline assembly, we can make proper use of the flags set. */ | |
456 | __asm ( "sub $16, %1\n" | |
457 | " .balign 16\n" | |
458 | "0: add $16, %1\n" | |
459 | " %vpcmpestri $0, (%1), %2\n" | |
460 | " jnc 0b" | |
461 | : "=&c"(index), "+r"(s) | |
462 | : "x"(search), "a"(4), "d"(16)); | |
463 | ||
464 | found: | |
465 | return s + index; | |
466 | } | |
467 | ||
6f173e52 RH |
468 | #else |
469 | /* Work around out-dated assemblers without sse4 support. */ | |
470 | #define search_line_sse42 search_line_sse2 | |
471 | #endif | |
472 | ||
246a2fcb RH |
473 | /* Check the CPU capabilities. */ |
474 | ||
475 | #include "../gcc/config/i386/cpuid.h" | |
476 | ||
477 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); | |
478 | static search_line_fast_type search_line_fast; | |
479 | ||
480 | static void __attribute__((constructor)) | |
481 | init_vectorized_lexer (void) | |
482 | { | |
483 | unsigned dummy, ecx = 0, edx = 0; | |
484 | search_line_fast_type impl = search_line_acc_char; | |
485 | int minimum = 0; | |
486 | ||
487 | #if defined(__SSE4_2__) | |
488 | minimum = 3; | |
489 | #elif defined(__SSE2__) | |
490 | minimum = 2; | |
ef230b38 | 491 | #elif defined(__SSE__) |
246a2fcb RH |
492 | minimum = 1; |
493 | #endif | |
494 | ||
495 | if (minimum == 3) | |
496 | impl = search_line_sse42; | |
497 | else if (__get_cpuid (1, &dummy, &dummy, &ecx, &edx) || minimum == 2) | |
498 | { | |
499 | if (minimum == 3 || (ecx & bit_SSE4_2)) | |
500 | impl = search_line_sse42; | |
501 | else if (minimum == 2 || (edx & bit_SSE2)) | |
502 | impl = search_line_sse2; | |
503 | else if (minimum == 1 || (edx & bit_SSE)) | |
504 | impl = search_line_mmx; | |
505 | } | |
506 | else if (__get_cpuid (0x80000001, &dummy, &dummy, &dummy, &edx)) | |
507 | { | |
5e70c0b5 UB |
508 | if (minimum == 1 |
509 | || (edx & (bit_MMXEXT | bit_CMOV)) == (bit_MMXEXT | bit_CMOV)) | |
246a2fcb RH |
510 | impl = search_line_mmx; |
511 | } | |
512 | ||
513 | search_line_fast = impl; | |
514 | } | |
515 | ||
01956319 | 516 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) |
246a2fcb RH |
517 | |
518 | /* A vection of the fast scanner using AltiVec vectorized byte compares. */ | |
519 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, | |
520 | so we can't compile this function without -maltivec on the command line | |
521 | (or implied by some other switch). */ | |
522 | ||
523 | static const uchar * | |
524 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
525 | { | |
526 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
527 | ||
528 | const vc repl_nl = { | |
529 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
530 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
531 | }; | |
532 | const vc repl_cr = { | |
533 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
534 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
535 | }; | |
536 | const vc repl_bs = { | |
537 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
538 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
539 | }; | |
540 | const vc repl_qm = { | |
541 | '?', '?', '?', '?', '?', '?', '?', '?', | |
542 | '?', '?', '?', '?', '?', '?', '?', '?', | |
543 | }; | |
544 | const vc ones = { | |
545 | -1, -1, -1, -1, -1, -1, -1, -1, | |
546 | -1, -1, -1, -1, -1, -1, -1, -1, | |
547 | }; | |
548 | const vc zero = { 0 }; | |
549 | ||
550 | vc data, mask, t; | |
551 | ||
552 | /* Altivec loads automatically mask addresses with -16. This lets us | |
553 | issue the first load as early as possible. */ | |
554 | data = __builtin_vec_ld(0, (const vc *)s); | |
555 | ||
556 | /* Discard bytes before the beginning of the buffer. Do this by | |
557 | beginning with all ones and shifting in zeros according to the | |
558 | mis-alignment. The LVSR instruction pulls the exact shift we | |
559 | want from the address. */ | |
560 | mask = __builtin_vec_lvsr(0, s); | |
561 | mask = __builtin_vec_perm(zero, ones, mask); | |
562 | data &= mask; | |
563 | ||
564 | /* While altivec loads mask addresses, we still need to align S so | |
565 | that the offset we compute at the end is correct. */ | |
566 | s = (const uchar *)((uintptr_t)s & -16); | |
567 | ||
568 | /* Main loop processing 16 bytes at a time. */ | |
569 | goto start; | |
570 | do | |
571 | { | |
572 | vc m_nl, m_cr, m_bs, m_qm; | |
573 | ||
574 | s += 16; | |
575 | data = __builtin_vec_ld(0, (const vc *)s); | |
576 | ||
577 | start: | |
578 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
579 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
580 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
581 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
582 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
583 | ||
584 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
585 | characters. We want to exit the loop if any byte in T is non-zero. | |
586 | Below is the expansion of vec_any_ne(t, zero). */ | |
587 | } | |
588 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
589 | ||
590 | { | |
591 | #define N (sizeof(vc) / sizeof(long)) | |
592 | ||
593 | typedef char check_count[(N == 2 || N == 4) * 2 - 1]; | |
594 | union { | |
595 | vc v; | |
596 | unsigned long l[N]; | |
597 | } u; | |
598 | unsigned long l, i = 0; | |
599 | ||
600 | u.v = t; | |
601 | ||
602 | /* Find the first word of T that is non-zero. */ | |
603 | switch (N) | |
604 | { | |
605 | case 4: | |
606 | l = u.l[i++]; | |
607 | if (l != 0) | |
608 | break; | |
609 | s += sizeof(unsigned long); | |
610 | l = u.l[i++]; | |
611 | if (l != 0) | |
612 | break; | |
613 | s += sizeof(unsigned long); | |
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 | l = __builtin_clzl(l) >> 3; | |
626 | return s + l; | |
627 | ||
628 | #undef N | |
629 | } | |
630 | } | |
631 | ||
632 | #else | |
633 | ||
634 | /* We only have one accellerated alternative. Use a direct call so that | |
635 | we encourage inlining. */ | |
636 | ||
637 | #define search_line_fast search_line_acc_char | |
638 | ||
639 | #endif | |
640 | ||
26aea073 NB |
641 | /* Returns with a logical line that contains no escaped newlines or |
642 | trigraphs. This is a time-critical inner loop. */ | |
643 | void | |
6cf87ca4 | 644 | _cpp_clean_line (cpp_reader *pfile) |
45b966db | 645 | { |
26aea073 NB |
646 | cpp_buffer *buffer; |
647 | const uchar *s; | |
648 | uchar c, *d, *p; | |
87062813 | 649 | |
26aea073 NB |
650 | buffer = pfile->buffer; |
651 | buffer->cur_note = buffer->notes_used = 0; | |
652 | buffer->cur = buffer->line_base = buffer->next_line; | |
653 | buffer->need_line = false; | |
246a2fcb | 654 | s = buffer->next_line; |
87062813 | 655 | |
26aea073 | 656 | if (!buffer->from_stage3) |
45b966db | 657 | { |
7af45bd4 ILT |
658 | const uchar *pbackslash = NULL; |
659 | ||
246a2fcb | 660 | /* Fast path. This is the common case of an un-escaped line with |
d08dcf87 ZW |
661 | no trigraphs. The primary win here is by not writing any |
662 | data back to memory until we have to. */ | |
246a2fcb | 663 | while (1) |
d08dcf87 | 664 | { |
246a2fcb RH |
665 | /* Perform an optimized search for \n, \r, \\, ?. */ |
666 | s = search_line_fast (s, buffer->rlimit); | |
d08dcf87 | 667 | |
246a2fcb RH |
668 | c = *s; |
669 | if (c == '\\') | |
670 | { | |
671 | /* Record the location of the backslash and continue. */ | |
672 | pbackslash = s++; | |
d08dcf87 | 673 | } |
246a2fcb | 674 | else if (__builtin_expect (c == '?', 0)) |
d08dcf87 | 675 | { |
246a2fcb RH |
676 | if (__builtin_expect (s[1] == '?', false) |
677 | && _cpp_trigraph_map[s[2]]) | |
d08dcf87 | 678 | { |
246a2fcb RH |
679 | /* Have a trigraph. We may or may not have to convert |
680 | it. Add a line note regardless, for -Wtrigraphs. */ | |
681 | add_line_note (buffer, s, s[2]); | |
682 | if (CPP_OPTION (pfile, trigraphs)) | |
683 | { | |
684 | /* We do, and that means we have to switch to the | |
685 | slow path. */ | |
686 | d = (uchar *) s; | |
687 | *d = _cpp_trigraph_map[s[2]]; | |
688 | s += 2; | |
689 | goto slow_path; | |
690 | } | |
d08dcf87 | 691 | } |
246a2fcb RH |
692 | /* Not a trigraph. Continue on fast-path. */ |
693 | s++; | |
d08dcf87 | 694 | } |
246a2fcb RH |
695 | else |
696 | break; | |
d08dcf87 ZW |
697 | } |
698 | ||
246a2fcb RH |
699 | /* This must be \r or \n. We're either done, or we'll be forced |
700 | to write back to the buffer and continue on the slow path. */ | |
701 | d = (uchar *) s; | |
702 | ||
703 | if (__builtin_expect (s == buffer->rlimit, false)) | |
704 | goto done; | |
705 | ||
706 | /* DOS line ending? */ | |
707 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
708 | { | |
709 | s++; | |
710 | if (s == buffer->rlimit) | |
711 | goto done; | |
712 | } | |
713 | ||
714 | if (__builtin_expect (pbackslash == NULL, true)) | |
715 | goto done; | |
716 | ||
717 | /* Check for escaped newline. */ | |
718 | p = d; | |
719 | while (is_nvspace (p[-1])) | |
720 | p--; | |
721 | if (p - 1 != pbackslash) | |
722 | goto done; | |
723 | ||
724 | /* Have an escaped newline; process it and proceed to | |
725 | the slow path. */ | |
726 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
727 | d = p - 2; | |
728 | buffer->next_line = p - 1; | |
26aea073 | 729 | |
246a2fcb RH |
730 | slow_path: |
731 | while (1) | |
4a5b68a2 | 732 | { |
26aea073 NB |
733 | c = *++s; |
734 | *++d = c; | |
735 | ||
736 | if (c == '\n' || c == '\r') | |
737 | { | |
246a2fcb | 738 | /* Handle DOS line endings. */ |
26aea073 NB |
739 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
740 | s++; | |
741 | if (s == buffer->rlimit) | |
742 | break; | |
743 | ||
744 | /* Escaped? */ | |
745 | p = d; | |
746 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
747 | p--; | |
748 | if (p == buffer->next_line || p[-1] != '\\') | |
749 | break; | |
750 | ||
41c32c98 | 751 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
26aea073 NB |
752 | d = p - 2; |
753 | buffer->next_line = p - 1; | |
754 | } | |
755 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
756 | { | |
757 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
41c32c98 | 758 | add_line_note (buffer, d, s[2]); |
26aea073 NB |
759 | if (CPP_OPTION (pfile, trigraphs)) |
760 | { | |
761 | *d = _cpp_trigraph_map[s[2]]; | |
762 | s += 2; | |
763 | } | |
764 | } | |
4a5b68a2 | 765 | } |
45b966db | 766 | } |
26aea073 NB |
767 | else |
768 | { | |
246a2fcb | 769 | while (*s != '\n' && *s != '\r') |
26aea073 | 770 | s++; |
26aea073 NB |
771 | d = (uchar *) s; |
772 | ||
773 | /* Handle DOS line endings. */ | |
774 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
775 | s++; | |
776 | } | |
0d9f234d | 777 | |
d08dcf87 | 778 | done: |
26aea073 | 779 | *d = '\n'; |
41c32c98 NB |
780 | /* A sentinel note that should never be processed. */ |
781 | add_line_note (buffer, d + 1, '\n'); | |
26aea073 | 782 | buffer->next_line = s + 1; |
45b966db ZW |
783 | } |
784 | ||
a8eb6044 NB |
785 | /* Return true if the trigraph indicated by NOTE should be warned |
786 | about in a comment. */ | |
787 | static bool | |
6cf87ca4 | 788 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
a8eb6044 NB |
789 | { |
790 | const uchar *p; | |
791 | ||
792 | /* Within comments we don't warn about trigraphs, unless the | |
793 | trigraph forms an escaped newline, as that may change | |
6356f892 | 794 | behavior. */ |
a8eb6044 NB |
795 | if (note->type != '/') |
796 | return false; | |
797 | ||
798 | /* If -trigraphs, then this was an escaped newline iff the next note | |
799 | is coincident. */ | |
800 | if (CPP_OPTION (pfile, trigraphs)) | |
801 | return note[1].pos == note->pos; | |
802 | ||
803 | /* Otherwise, see if this forms an escaped newline. */ | |
804 | p = note->pos + 3; | |
805 | while (is_nvspace (*p)) | |
806 | p++; | |
807 | ||
808 | /* There might have been escaped newlines between the trigraph and the | |
809 | newline we found. Hence the position test. */ | |
810 | return (*p == '\n' && p < note[1].pos); | |
811 | } | |
812 | ||
26aea073 NB |
813 | /* Process the notes created by add_line_note as far as the current |
814 | location. */ | |
815 | void | |
6cf87ca4 | 816 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
45b966db | 817 | { |
29401c30 NB |
818 | cpp_buffer *buffer = pfile->buffer; |
819 | ||
26aea073 | 820 | for (;;) |
041c3194 | 821 | { |
26aea073 NB |
822 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
823 | unsigned int col; | |
a5c3cccd | 824 | |
26aea073 NB |
825 | if (note->pos > buffer->cur) |
826 | break; | |
a5c3cccd | 827 | |
26aea073 NB |
828 | buffer->cur_note++; |
829 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
4d6baafa | 830 | |
41c32c98 | 831 | if (note->type == '\\' || note->type == ' ') |
26aea073 | 832 | { |
41c32c98 | 833 | if (note->type == ' ' && !in_comment) |
500bee0a | 834 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
26aea073 | 835 | "backslash and newline separated by space"); |
41c32c98 | 836 | |
26aea073 | 837 | if (buffer->next_line > buffer->rlimit) |
87062813 | 838 | { |
500bee0a | 839 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
26aea073 NB |
840 | "backslash-newline at end of file"); |
841 | /* Prevent "no newline at end of file" warning. */ | |
842 | buffer->next_line = buffer->rlimit; | |
87062813 | 843 | } |
26aea073 NB |
844 | |
845 | buffer->line_base = note->pos; | |
12f9df4e | 846 | CPP_INCREMENT_LINE (pfile, 0); |
0d9f234d | 847 | } |
41c32c98 NB |
848 | else if (_cpp_trigraph_map[note->type]) |
849 | { | |
a8eb6044 NB |
850 | if (CPP_OPTION (pfile, warn_trigraphs) |
851 | && (!in_comment || warn_in_comment (pfile, note))) | |
41c32c98 NB |
852 | { |
853 | if (CPP_OPTION (pfile, trigraphs)) | |
87cf0651 SB |
854 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
855 | pfile->line_table->highest_line, col, | |
856 | "trigraph ??%c converted to %c", | |
857 | note->type, | |
858 | (int) _cpp_trigraph_map[note->type]); | |
41c32c98 | 859 | else |
905bd7b5 | 860 | { |
87cf0651 SB |
861 | cpp_warning_with_line |
862 | (pfile, CPP_W_TRIGRAPHS, | |
863 | pfile->line_table->highest_line, col, | |
905bd7b5 GK |
864 | "trigraph ??%c ignored, use -trigraphs to enable", |
865 | note->type); | |
866 | } | |
41c32c98 NB |
867 | } |
868 | } | |
00a81b8b JM |
869 | else if (note->type == 0) |
870 | /* Already processed in lex_raw_string. */; | |
41c32c98 NB |
871 | else |
872 | abort (); | |
041c3194 | 873 | } |
45b966db ZW |
874 | } |
875 | ||
0d9f234d NB |
876 | /* Skip a C-style block comment. We find the end of the comment by |
877 | seeing if an asterisk is before every '/' we encounter. Returns | |
6f572ac2 NB |
878 | nonzero if comment terminated by EOF, zero otherwise. |
879 | ||
880 | Buffer->cur points to the initial asterisk of the comment. */ | |
26aea073 | 881 | bool |
6cf87ca4 | 882 | _cpp_skip_block_comment (cpp_reader *pfile) |
45b966db | 883 | { |
041c3194 | 884 | cpp_buffer *buffer = pfile->buffer; |
d08dcf87 ZW |
885 | const uchar *cur = buffer->cur; |
886 | uchar c; | |
0d9f234d | 887 | |
d08dcf87 ZW |
888 | cur++; |
889 | if (*cur == '/') | |
890 | cur++; | |
0d9f234d | 891 | |
26aea073 NB |
892 | for (;;) |
893 | { | |
0d9f234d NB |
894 | /* People like decorating comments with '*', so check for '/' |
895 | instead for efficiency. */ | |
d08dcf87 ZW |
896 | c = *cur++; |
897 | ||
041c3194 | 898 | if (c == '/') |
45b966db | 899 | { |
d08dcf87 | 900 | if (cur[-2] == '*') |
0d9f234d | 901 | break; |
041c3194 | 902 | |
0d9f234d | 903 | /* Warn about potential nested comments, but not if the '/' |
a1f300c0 | 904 | comes immediately before the true comment delimiter. |
041c3194 | 905 | Don't bother to get it right across escaped newlines. */ |
0d9f234d | 906 | if (CPP_OPTION (pfile, warn_comments) |
d08dcf87 ZW |
907 | && cur[0] == '*' && cur[1] != '/') |
908 | { | |
909 | buffer->cur = cur; | |
87cf0651 SB |
910 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
911 | pfile->line_table->highest_line, | |
912 | CPP_BUF_COL (buffer), | |
913 | "\"/*\" within comment"); | |
d08dcf87 | 914 | } |
45b966db | 915 | } |
26aea073 NB |
916 | else if (c == '\n') |
917 | { | |
12f9df4e | 918 | unsigned int cols; |
d08dcf87 | 919 | buffer->cur = cur - 1; |
26aea073 NB |
920 | _cpp_process_line_notes (pfile, true); |
921 | if (buffer->next_line >= buffer->rlimit) | |
922 | return true; | |
923 | _cpp_clean_line (pfile); | |
12f9df4e PB |
924 | |
925 | cols = buffer->next_line - buffer->line_base; | |
926 | CPP_INCREMENT_LINE (pfile, cols); | |
927 | ||
d08dcf87 | 928 | cur = buffer->cur; |
26aea073 | 929 | } |
45b966db | 930 | } |
041c3194 | 931 | |
d08dcf87 | 932 | buffer->cur = cur; |
a8eb6044 | 933 | _cpp_process_line_notes (pfile, true); |
26aea073 | 934 | return false; |
45b966db ZW |
935 | } |
936 | ||
480709cc | 937 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
da7d8304 | 938 | terminating newline. Handles escaped newlines. Returns nonzero |
480709cc | 939 | if a multiline comment. */ |
041c3194 | 940 | static int |
6cf87ca4 | 941 | skip_line_comment (cpp_reader *pfile) |
45b966db | 942 | { |
cbcff6df | 943 | cpp_buffer *buffer = pfile->buffer; |
1bb64668 | 944 | source_location orig_line = pfile->line_table->highest_line; |
041c3194 | 945 | |
26aea073 NB |
946 | while (*buffer->cur != '\n') |
947 | buffer->cur++; | |
480709cc | 948 | |
26aea073 | 949 | _cpp_process_line_notes (pfile, true); |
500bee0a | 950 | return orig_line != pfile->line_table->highest_line; |
041c3194 | 951 | } |
45b966db | 952 | |
26aea073 | 953 | /* Skips whitespace, saving the next non-whitespace character. */ |
52fadca8 | 954 | static void |
6cf87ca4 | 955 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
041c3194 ZW |
956 | { |
957 | cpp_buffer *buffer = pfile->buffer; | |
f7d151fb | 958 | bool saw_NUL = false; |
45b966db | 959 | |
0d9f234d | 960 | do |
041c3194 | 961 | { |
91fcd158 | 962 | /* Horizontal space always OK. */ |
26aea073 | 963 | if (c == ' ' || c == '\t') |
0d9f234d | 964 | ; |
0d9f234d | 965 | /* Just \f \v or \0 left. */ |
91fcd158 | 966 | else if (c == '\0') |
f7d151fb | 967 | saw_NUL = true; |
93c80368 | 968 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
500bee0a | 969 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
ebef4e8c NB |
970 | CPP_BUF_COL (buffer), |
971 | "%s in preprocessing directive", | |
972 | c == '\f' ? "form feed" : "vertical tab"); | |
0d9f234d | 973 | |
0d9f234d | 974 | c = *buffer->cur++; |
45b966db | 975 | } |
ec5c56db | 976 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
0d9f234d NB |
977 | while (is_nvspace (c)); |
978 | ||
f7d151fb | 979 | if (saw_NUL) |
0527bc4e | 980 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
f7d151fb | 981 | |
480709cc | 982 | buffer->cur--; |
041c3194 | 983 | } |
45b966db | 984 | |
93c80368 NB |
985 | /* See if the characters of a number token are valid in a name (no |
986 | '.', '+' or '-'). */ | |
987 | static int | |
6cf87ca4 | 988 | name_p (cpp_reader *pfile, const cpp_string *string) |
93c80368 NB |
989 | { |
990 | unsigned int i; | |
991 | ||
992 | for (i = 0; i < string->len; i++) | |
993 | if (!is_idchar (string->text[i])) | |
994 | return 0; | |
995 | ||
df383483 | 996 | return 1; |
93c80368 NB |
997 | } |
998 | ||
50668cf6 GK |
999 | /* After parsing an identifier or other sequence, produce a warning about |
1000 | sequences not in NFC/NFKC. */ | |
1001 | static void | |
1002 | warn_about_normalization (cpp_reader *pfile, | |
1003 | const cpp_token *token, | |
1004 | const struct normalize_state *s) | |
1005 | { | |
1006 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1007 | && !pfile->state.skipping) | |
1008 | { | |
1009 | /* Make sure that the token is printed using UCNs, even | |
1010 | if we'd otherwise happily print UTF-8. */ | |
c3f829c1 | 1011 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
50668cf6 GK |
1012 | size_t sz; |
1013 | ||
1014 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1015 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
87cf0651 SB |
1016 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1017 | "`%.*s' is not in NFKC", (int) sz, buf); | |
50668cf6 | 1018 | else |
87cf0651 SB |
1019 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1020 | "`%.*s' is not in NFC", (int) sz, buf); | |
50668cf6 GK |
1021 | } |
1022 | } | |
1023 | ||
bced6edf | 1024 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in |
1613e52b | 1025 | an identifier. FIRST is TRUE if this starts an identifier. */ |
bced6edf | 1026 | static bool |
50668cf6 GK |
1027 | forms_identifier_p (cpp_reader *pfile, int first, |
1028 | struct normalize_state *state) | |
bced6edf | 1029 | { |
1613e52b NB |
1030 | cpp_buffer *buffer = pfile->buffer; |
1031 | ||
1032 | if (*buffer->cur == '$') | |
1033 | { | |
1034 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1035 | return false; | |
1036 | ||
1037 | buffer->cur++; | |
78b8811a | 1038 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
1613e52b | 1039 | { |
78b8811a | 1040 | CPP_OPTION (pfile, warn_dollars) = 0; |
0527bc4e | 1041 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
1613e52b NB |
1042 | } |
1043 | ||
1044 | return true; | |
1045 | } | |
bced6edf | 1046 | |
1613e52b | 1047 | /* Is this a syntactically valid UCN? */ |
af15a2fe | 1048 | if (CPP_OPTION (pfile, extended_identifiers) |
6baba9bb | 1049 | && *buffer->cur == '\\' |
1613e52b | 1050 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) |
bced6edf | 1051 | { |
1613e52b | 1052 | buffer->cur += 2; |
50668cf6 GK |
1053 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1054 | state)) | |
1613e52b NB |
1055 | return true; |
1056 | buffer->cur -= 2; | |
bced6edf | 1057 | } |
bced6edf | 1058 | |
1613e52b | 1059 | return false; |
bced6edf NB |
1060 | } |
1061 | ||
17e7cb85 KT |
1062 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1063 | static cpp_hashnode * | |
1064 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1065 | { | |
1066 | cpp_hashnode *result; | |
1067 | const uchar *cur; | |
1068 | unsigned int len; | |
1069 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1070 | ||
1071 | cur = base + 1; | |
1072 | while (ISIDNUM (*cur)) | |
1073 | { | |
1074 | hash = HT_HASHSTEP (hash, *cur); | |
1075 | cur++; | |
1076 | } | |
1077 | len = cur - base; | |
1078 | hash = HT_HASHFINISH (hash, len); | |
1079 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1080 | base, len, hash, HT_ALLOC)); | |
1081 | ||
1082 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1083 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1084 | && !pfile->state.skipping, 0)) | |
1085 | { | |
1086 | /* It is allowed to poison the same identifier twice. */ | |
1087 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1088 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1089 | NODE_NAME (result)); | |
1090 | ||
1091 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1092 | replacement list of a variadic macro. */ | |
1093 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1094 | && !pfile->state.va_args_ok) | |
1095 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1096 | "__VA_ARGS__ can only appear in the expansion" | |
1097 | " of a C99 variadic macro"); | |
1098 | ||
1099 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1100 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1101 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1102 | "identifier \"%s\" is a special operator name in C++", | |
1103 | NODE_NAME (result)); | |
17e7cb85 KT |
1104 | } |
1105 | ||
1106 | return result; | |
1107 | } | |
1108 | ||
1109 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1110 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1111 | cpp_hashnode * | |
1112 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1113 | { | |
1114 | cpp_hashnode *result; | |
1115 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1116 | return result; | |
1117 | } | |
1118 | ||
bced6edf | 1119 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
0d9f234d | 1120 | static cpp_hashnode * |
50668cf6 GK |
1121 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
1122 | struct normalize_state *nst) | |
45b966db | 1123 | { |
93c80368 | 1124 | cpp_hashnode *result; |
47e20491 | 1125 | const uchar *cur; |
c6e83800 ZW |
1126 | unsigned int len; |
1127 | unsigned int hash = HT_HASHSTEP (0, *base); | |
2c3fcba6 | 1128 | |
c6e83800 | 1129 | cur = pfile->buffer->cur; |
47e20491 GK |
1130 | if (! starts_ucn) |
1131 | while (ISIDNUM (*cur)) | |
1132 | { | |
1133 | hash = HT_HASHSTEP (hash, *cur); | |
1134 | cur++; | |
1135 | } | |
1136 | pfile->buffer->cur = cur; | |
50668cf6 | 1137 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
10cf9bde | 1138 | { |
47e20491 GK |
1139 | /* Slower version for identifiers containing UCNs (or $). */ |
1140 | do { | |
1141 | while (ISIDNUM (*pfile->buffer->cur)) | |
50668cf6 GK |
1142 | { |
1143 | pfile->buffer->cur++; | |
1144 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1145 | } | |
1146 | } while (forms_identifier_p (pfile, false, nst)); | |
47e20491 GK |
1147 | result = _cpp_interpret_identifier (pfile, base, |
1148 | pfile->buffer->cur - base); | |
2c3fcba6 | 1149 | } |
47e20491 GK |
1150 | else |
1151 | { | |
1152 | len = cur - base; | |
1153 | hash = HT_HASHFINISH (hash, len); | |
bced6edf | 1154 | |
2bf41bf0 TT |
1155 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1156 | base, len, hash, HT_ALLOC)); | |
47e20491 | 1157 | } |
2c3fcba6 | 1158 | |
bced6edf | 1159 | /* Rarely, identifiers require diagnostics when lexed. */ |
2c3fcba6 ZW |
1160 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1161 | && !pfile->state.skipping, 0)) | |
1162 | { | |
1163 | /* It is allowed to poison the same identifier twice. */ | |
1164 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
0527bc4e | 1165 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
2c3fcba6 ZW |
1166 | NODE_NAME (result)); |
1167 | ||
1168 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1169 | replacement list of a variadic macro. */ | |
1170 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1171 | && !pfile->state.va_args_ok) | |
0527bc4e | 1172 | cpp_error (pfile, CPP_DL_PEDWARN, |
6cf87ca4 ZW |
1173 | "__VA_ARGS__ can only appear in the expansion" |
1174 | " of a C99 variadic macro"); | |
3d8b2a98 ILT |
1175 | |
1176 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1177 | if (result->flags & NODE_WARN_OPERATOR) | |
87cf0651 SB |
1178 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1179 | "identifier \"%s\" is a special operator name in C++", | |
1180 | NODE_NAME (result)); | |
2c3fcba6 ZW |
1181 | } |
1182 | ||
1183 | return result; | |
1184 | } | |
1185 | ||
bced6edf | 1186 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
45b966db | 1187 | static void |
50668cf6 GK |
1188 | lex_number (cpp_reader *pfile, cpp_string *number, |
1189 | struct normalize_state *nst) | |
45b966db | 1190 | { |
562a5c27 | 1191 | const uchar *cur; |
bced6edf NB |
1192 | const uchar *base; |
1193 | uchar *dest; | |
45b966db | 1194 | |
bced6edf NB |
1195 | base = pfile->buffer->cur - 1; |
1196 | do | |
041c3194 | 1197 | { |
bced6edf | 1198 | cur = pfile->buffer->cur; |
0d9f234d | 1199 | |
bced6edf NB |
1200 | /* N.B. ISIDNUM does not include $. */ |
1201 | while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1])) | |
50668cf6 GK |
1202 | { |
1203 | cur++; | |
1204 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1205 | } | |
45b966db | 1206 | |
10cf9bde | 1207 | pfile->buffer->cur = cur; |
45b966db | 1208 | } |
50668cf6 | 1209 | while (forms_identifier_p (pfile, false, nst)); |
93c80368 | 1210 | |
bced6edf NB |
1211 | number->len = cur - base; |
1212 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1213 | memcpy (dest, base, number->len); | |
1214 | dest[number->len] = '\0'; | |
1215 | number->text = dest; | |
93c80368 NB |
1216 | } |
1217 | ||
6338b358 NB |
1218 | /* Create a token of type TYPE with a literal spelling. */ |
1219 | static void | |
6cf87ca4 ZW |
1220 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1221 | unsigned int len, enum cpp_ttype type) | |
6338b358 NB |
1222 | { |
1223 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1224 | ||
1225 | memcpy (dest, base, len); | |
1226 | dest[len] = '\0'; | |
1227 | token->type = type; | |
1228 | token->val.str.len = len; | |
1229 | token->val.str.text = dest; | |
1230 | } | |
1231 | ||
00a81b8b JM |
1232 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1233 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1234 | ||
1235 | static void | |
1236 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1237 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1238 | { | |
1239 | _cpp_buff *first_buff = *first_buff_p; | |
1240 | _cpp_buff *last_buff = *last_buff_p; | |
1241 | ||
1242 | if (first_buff == NULL) | |
1243 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1244 | else if (len > BUFF_ROOM (last_buff)) | |
1245 | { | |
1246 | size_t room = BUFF_ROOM (last_buff); | |
1247 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1248 | BUFF_FRONT (last_buff) += room; | |
1249 | base += room; | |
1250 | len -= room; | |
1251 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1252 | } | |
1253 | ||
1254 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1255 | BUFF_FRONT (last_buff) += len; | |
1256 | ||
1257 | *first_buff_p = first_buff; | |
1258 | *last_buff_p = last_buff; | |
1259 | } | |
1260 | ||
2c6e3f55 | 1261 | /* Lexes a raw string. The stored string contains the spelling, including |
00a81b8b | 1262 | double quotes, delimiter string, '(' and ')', any leading |
2c6e3f55 JJ |
1263 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1264 | literal, or CPP_OTHER if it was not properly terminated. | |
1265 | ||
1266 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1267 | is the first NUL since embedded NULs are preserved. */ | |
1268 | ||
1269 | static void | |
1270 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1271 | const uchar *cur) | |
1272 | { | |
2c6e3f55 JJ |
1273 | const uchar *raw_prefix; |
1274 | unsigned int raw_prefix_len = 0; | |
1275 | enum cpp_ttype type; | |
1276 | size_t total_len = 0; | |
1277 | _cpp_buff *first_buff = NULL, *last_buff = NULL; | |
00a81b8b | 1278 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 JJ |
1279 | |
1280 | type = (*base == 'L' ? CPP_WSTRING : | |
1281 | *base == 'U' ? CPP_STRING32 : | |
1282 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1283 | : CPP_STRING); | |
1284 | ||
1285 | raw_prefix = cur + 1; | |
1286 | while (raw_prefix_len < 16) | |
1287 | { | |
1288 | switch (raw_prefix[raw_prefix_len]) | |
1289 | { | |
52150625 | 1290 | case ' ': case '(': case ')': case '\\': case '\t': |
2c6e3f55 JJ |
1291 | case '\v': case '\f': case '\n': default: |
1292 | break; | |
1293 | /* Basic source charset except the above chars. */ | |
1294 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1295 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1296 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1297 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1298 | case 'y': case 'z': | |
1299 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1300 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1301 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1302 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1303 | case 'Y': case 'Z': | |
1304 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1305 | case '6': case '7': case '8': case '9': | |
52150625 | 1306 | case '_': case '{': case '}': case '#': case '[': case ']': |
2c6e3f55 JJ |
1307 | case '<': case '>': case '%': case ':': case ';': case '.': |
1308 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1309 | case '&': case '|': case '~': case '!': case '=': case ',': | |
52150625 | 1310 | case '"': case '\'': |
2c6e3f55 JJ |
1311 | raw_prefix_len++; |
1312 | continue; | |
1313 | } | |
1314 | break; | |
1315 | } | |
1316 | ||
52150625 | 1317 | if (raw_prefix[raw_prefix_len] != '(') |
2c6e3f55 JJ |
1318 | { |
1319 | int col = CPP_BUF_COLUMN (pfile->buffer, raw_prefix + raw_prefix_len) | |
1320 | + 1; | |
1321 | if (raw_prefix_len == 16) | |
1322 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1323 | "raw string delimiter longer than 16 characters"); | |
1324 | else | |
1325 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1326 | "invalid character '%c' in raw string delimiter", | |
1327 | (int) raw_prefix[raw_prefix_len]); | |
1328 | pfile->buffer->cur = raw_prefix - 1; | |
1329 | create_literal (pfile, token, base, raw_prefix - 1 - base, CPP_OTHER); | |
1330 | return; | |
1331 | } | |
1332 | ||
1333 | cur = raw_prefix + raw_prefix_len + 1; | |
1334 | for (;;) | |
1335 | { | |
00a81b8b JM |
1336 | #define BUF_APPEND(STR,LEN) \ |
1337 | do { \ | |
1338 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1339 | &first_buff, &last_buff); \ | |
1340 | total_len += (LEN); \ | |
1341 | } while (0); | |
1342 | ||
1343 | cppchar_t c; | |
1344 | ||
1345 | /* If we previously performed any trigraph or line splicing | |
1346 | transformations, undo them within the body of the raw string. */ | |
1347 | while (note->pos < cur) | |
1348 | ++note; | |
1349 | for (; note->pos == cur; ++note) | |
1350 | { | |
1351 | switch (note->type) | |
1352 | { | |
1353 | case '\\': | |
1354 | case ' ': | |
1355 | /* Restore backslash followed by newline. */ | |
1356 | BUF_APPEND (base, cur - base); | |
1357 | base = cur; | |
1358 | BUF_APPEND ("\\", 1); | |
1359 | after_backslash: | |
1360 | if (note->type == ' ') | |
1361 | { | |
1362 | /* GNU backslash whitespace newline extension. FIXME | |
1363 | could be any sequence of non-vertical space. When we | |
1364 | can properly restore any such sequence, we should mark | |
1365 | this note as handled so _cpp_process_line_notes | |
1366 | doesn't warn. */ | |
1367 | BUF_APPEND (" ", 1); | |
1368 | } | |
1369 | ||
1370 | BUF_APPEND ("\n", 1); | |
1371 | break; | |
1372 | ||
1373 | case 0: | |
1374 | /* Already handled. */ | |
1375 | break; | |
1376 | ||
1377 | default: | |
1378 | if (_cpp_trigraph_map[note->type]) | |
1379 | { | |
1380 | /* Don't warn about this trigraph in | |
1381 | _cpp_process_line_notes, since trigraphs show up as | |
1382 | trigraphs in raw strings. */ | |
d947ada0 | 1383 | uchar type = note->type; |
00a81b8b JM |
1384 | note->type = 0; |
1385 | ||
1386 | if (!CPP_OPTION (pfile, trigraphs)) | |
1387 | /* If we didn't convert the trigraph in the first | |
1388 | place, don't do anything now either. */ | |
1389 | break; | |
1390 | ||
1391 | BUF_APPEND (base, cur - base); | |
1392 | base = cur; | |
1393 | BUF_APPEND ("??", 2); | |
1394 | ||
1395 | /* ??/ followed by newline gets two line notes, one for | |
1396 | the trigraph and one for the backslash/newline. */ | |
1397 | if (type == '/' && note[1].pos == cur) | |
1398 | { | |
1399 | if (note[1].type != '\\' | |
1400 | && note[1].type != ' ') | |
1401 | abort (); | |
1402 | BUF_APPEND ("/", 1); | |
1403 | ++note; | |
1404 | goto after_backslash; | |
1405 | } | |
1406 | /* The ) from ??) could be part of the suffix. */ | |
1407 | else if (type == ')' | |
1408 | && strncmp ((const char *) cur+1, | |
1409 | (const char *) raw_prefix, | |
1410 | raw_prefix_len) == 0 | |
1411 | && cur[raw_prefix_len+1] == '"') | |
1412 | { | |
6cfae070 JJ |
1413 | BUF_APPEND (")", 1); |
1414 | base++; | |
1415 | cur += raw_prefix_len + 2; | |
00a81b8b JM |
1416 | goto break_outer_loop; |
1417 | } | |
1418 | else | |
1419 | { | |
1420 | /* Skip the replacement character. */ | |
1421 | base = ++cur; | |
1422 | BUF_APPEND (&type, 1); | |
1423 | } | |
1424 | } | |
1425 | else | |
1426 | abort (); | |
1427 | break; | |
1428 | } | |
1429 | } | |
1430 | c = *cur++; | |
2c6e3f55 | 1431 | |
52150625 | 1432 | if (c == ')' |
2c6e3f55 JJ |
1433 | && strncmp ((const char *) cur, (const char *) raw_prefix, |
1434 | raw_prefix_len) == 0 | |
1435 | && cur[raw_prefix_len] == '"') | |
1436 | { | |
1437 | cur += raw_prefix_len + 1; | |
1438 | break; | |
1439 | } | |
1440 | else if (c == '\n') | |
1441 | { | |
1442 | if (pfile->state.in_directive | |
1443 | || pfile->state.parsing_args | |
1444 | || pfile->state.in_deferred_pragma) | |
1445 | { | |
1446 | cur--; | |
1447 | type = CPP_OTHER; | |
1448 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1449 | "unterminated raw string"); | |
1450 | break; | |
1451 | } | |
1452 | ||
00a81b8b | 1453 | BUF_APPEND (base, cur - base); |
2c6e3f55 JJ |
1454 | |
1455 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1456 | CPP_INCREMENT_LINE (pfile, 0); | |
1457 | pfile->buffer->need_line = true; | |
1458 | ||
00a81b8b JM |
1459 | pfile->buffer->cur = cur-1; |
1460 | _cpp_process_line_notes (pfile, false); | |
2c6e3f55 JJ |
1461 | if (!_cpp_get_fresh_line (pfile)) |
1462 | { | |
1463 | source_location src_loc = token->src_loc; | |
1464 | token->type = CPP_EOF; | |
1465 | /* Tell the compiler the line number of the EOF token. */ | |
1466 | token->src_loc = pfile->line_table->highest_line; | |
1467 | token->flags = BOL; | |
1468 | if (first_buff != NULL) | |
1469 | _cpp_release_buff (pfile, first_buff); | |
1470 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1471 | "unterminated raw string"); | |
1472 | return; | |
1473 | } | |
1474 | ||
1475 | cur = base = pfile->buffer->cur; | |
00a81b8b | 1476 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
2c6e3f55 | 1477 | } |
2c6e3f55 | 1478 | } |
00a81b8b | 1479 | break_outer_loop: |
2c6e3f55 | 1480 | |
2c6e3f55 JJ |
1481 | pfile->buffer->cur = cur; |
1482 | if (first_buff == NULL) | |
1483 | create_literal (pfile, token, base, cur - base, type); | |
1484 | else | |
1485 | { | |
1486 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1487 | ||
1488 | token->type = type; | |
1489 | token->val.str.len = total_len + (cur - base); | |
1490 | token->val.str.text = dest; | |
1491 | last_buff = first_buff; | |
1492 | while (last_buff != NULL) | |
1493 | { | |
1494 | memcpy (dest, last_buff->base, | |
1495 | BUFF_FRONT (last_buff) - last_buff->base); | |
1496 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1497 | last_buff = last_buff->next; | |
1498 | } | |
1499 | _cpp_release_buff (pfile, first_buff); | |
1500 | memcpy (dest, base, cur - base); | |
1501 | dest[cur - base] = '\0'; | |
1502 | } | |
1503 | } | |
1504 | ||
bced6edf | 1505 | /* Lexes a string, character constant, or angle-bracketed header file |
6338b358 | 1506 | name. The stored string contains the spelling, including opening |
2c6e3f55 JJ |
1507 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1508 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1509 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1510 | header name which must be relexed as normal tokens. | |
6338b358 NB |
1511 | |
1512 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1513 | is the first NUL since embedded NULs are preserved. */ | |
041c3194 | 1514 | static void |
6cf87ca4 | 1515 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
45b966db | 1516 | { |
6338b358 NB |
1517 | bool saw_NUL = false; |
1518 | const uchar *cur; | |
bced6edf | 1519 | cppchar_t terminator; |
6338b358 NB |
1520 | enum cpp_ttype type; |
1521 | ||
1522 | cur = base; | |
1523 | terminator = *cur++; | |
2c6e3f55 | 1524 | if (terminator == 'L' || terminator == 'U') |
6338b358 | 1525 | terminator = *cur++; |
2c6e3f55 JJ |
1526 | else if (terminator == 'u') |
1527 | { | |
1528 | terminator = *cur++; | |
1529 | if (terminator == '8') | |
1530 | terminator = *cur++; | |
1531 | } | |
1532 | if (terminator == 'R') | |
1533 | { | |
1534 | lex_raw_string (pfile, token, base, cur); | |
1535 | return; | |
1536 | } | |
1537 | if (terminator == '"') | |
b6baa67d KVH |
1538 | type = (*base == 'L' ? CPP_WSTRING : |
1539 | *base == 'U' ? CPP_STRING32 : | |
2c6e3f55 JJ |
1540 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
1541 | : CPP_STRING); | |
6338b358 | 1542 | else if (terminator == '\'') |
b6baa67d KVH |
1543 | type = (*base == 'L' ? CPP_WCHAR : |
1544 | *base == 'U' ? CPP_CHAR32 : | |
1545 | *base == 'u' ? CPP_CHAR16 : CPP_CHAR); | |
6338b358 NB |
1546 | else |
1547 | terminator = '>', type = CPP_HEADER_NAME; | |
93c80368 | 1548 | |
0d9f234d | 1549 | for (;;) |
45b966db | 1550 | { |
6338b358 | 1551 | cppchar_t c = *cur++; |
7868b4a2 | 1552 | |
6f572ac2 | 1553 | /* In #include-style directives, terminators are not escapable. */ |
6338b358 NB |
1554 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
1555 | cur++; | |
1556 | else if (c == terminator) | |
bced6edf | 1557 | break; |
6338b358 | 1558 | else if (c == '\n') |
0d9f234d | 1559 | { |
6338b358 | 1560 | cur--; |
4bb09c26 JM |
1561 | /* Unmatched quotes always yield undefined behavior, but |
1562 | greedy lexing means that what appears to be an unterminated | |
1563 | header name may actually be a legitimate sequence of tokens. */ | |
1564 | if (terminator == '>') | |
1565 | { | |
1566 | token->type = CPP_LESS; | |
1567 | return; | |
1568 | } | |
6338b358 NB |
1569 | type = CPP_OTHER; |
1570 | break; | |
45b966db | 1571 | } |
6338b358 NB |
1572 | else if (c == '\0') |
1573 | saw_NUL = true; | |
45b966db ZW |
1574 | } |
1575 | ||
6338b358 | 1576 | if (saw_NUL && !pfile->state.skipping) |
0527bc4e JDA |
1577 | cpp_error (pfile, CPP_DL_WARNING, |
1578 | "null character(s) preserved in literal"); | |
45b966db | 1579 | |
c663e301 JM |
1580 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
1581 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
1582 | (int) terminator); | |
1583 | ||
6338b358 NB |
1584 | pfile->buffer->cur = cur; |
1585 | create_literal (pfile, token, base, cur - base, type); | |
0d9f234d | 1586 | } |
041c3194 | 1587 | |
631d0d36 MG |
1588 | /* Return the comment table. The client may not make any assumption |
1589 | about the ordering of the table. */ | |
1590 | cpp_comment_table * | |
1591 | cpp_get_comments (cpp_reader *pfile) | |
1592 | { | |
1593 | return &pfile->comments; | |
1594 | } | |
1595 | ||
1596 | /* Append a comment to the end of the comment table. */ | |
1597 | static void | |
1598 | store_comment (cpp_reader *pfile, cpp_token *token) | |
1599 | { | |
1600 | int len; | |
1601 | ||
1602 | if (pfile->comments.allocated == 0) | |
1603 | { | |
1604 | pfile->comments.allocated = 256; | |
1605 | pfile->comments.entries = (cpp_comment *) xmalloc | |
1606 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
1607 | } | |
1608 | ||
1609 | if (pfile->comments.count == pfile->comments.allocated) | |
1610 | { | |
1611 | pfile->comments.allocated *= 2; | |
1612 | pfile->comments.entries = (cpp_comment *) xrealloc | |
1613 | (pfile->comments.entries, | |
1614 | pfile->comments.allocated * sizeof (cpp_comment)); | |
1615 | } | |
1616 | ||
1617 | len = token->val.str.len; | |
1618 | ||
1619 | /* Copy comment. Note, token may not be NULL terminated. */ | |
1620 | pfile->comments.entries[pfile->comments.count].comment = | |
1621 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
1622 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
1623 | token->val.str.text, len); | |
1624 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
1625 | ||
1626 | /* Set source location. */ | |
1627 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
1628 | ||
1629 | /* Increment the count of entries in the comment table. */ | |
1630 | pfile->comments.count++; | |
1631 | } | |
1632 | ||
93c80368 | 1633 | /* The stored comment includes the comment start and any terminator. */ |
9e62c811 | 1634 | static void |
6cf87ca4 ZW |
1635 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
1636 | cppchar_t type) | |
9e62c811 | 1637 | { |
041c3194 | 1638 | unsigned char *buffer; |
651a20b5 | 1639 | unsigned int len, clen, i; |
df383483 | 1640 | |
1c6d33ef | 1641 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
480709cc | 1642 | |
3542203b NB |
1643 | /* C++ comments probably (not definitely) have moved past a new |
1644 | line, which we don't want to save in the comment. */ | |
480709cc | 1645 | if (is_vspace (pfile->buffer->cur[-1])) |
3542203b | 1646 | len--; |
477cdac7 | 1647 | |
651a20b5 KT |
1648 | /* If we are currently in a directive or in argument parsing, then |
1649 | we need to store all C++ comments as C comments internally, and | |
1650 | so we need to allocate a little extra space in that case. | |
477cdac7 JT |
1651 | |
1652 | Note that the only time we encounter a directive here is | |
1653 | when we are saving comments in a "#define". */ | |
651a20b5 KT |
1654 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
1655 | && type == '/') ? len + 2 : len; | |
477cdac7 JT |
1656 | |
1657 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
df383483 | 1658 | |
041c3194 | 1659 | token->type = CPP_COMMENT; |
477cdac7 | 1660 | token->val.str.len = clen; |
0d9f234d | 1661 | token->val.str.text = buffer; |
45b966db | 1662 | |
1c6d33ef NB |
1663 | buffer[0] = '/'; |
1664 | memcpy (buffer + 1, from, len - 1); | |
477cdac7 | 1665 | |
1eeeb6a4 | 1666 | /* Finish conversion to a C comment, if necessary. */ |
651a20b5 | 1667 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
477cdac7 JT |
1668 | { |
1669 | buffer[1] = '*'; | |
1670 | buffer[clen - 2] = '*'; | |
1671 | buffer[clen - 1] = '/'; | |
651a20b5 KT |
1672 | /* As there can be in a C++ comments illegal sequences for C comments |
1673 | we need to filter them out. */ | |
1674 | for (i = 2; i < (clen - 2); i++) | |
1675 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
1676 | buffer[i] = '|'; | |
477cdac7 | 1677 | } |
631d0d36 MG |
1678 | |
1679 | /* Finally store this comment for use by clients of libcpp. */ | |
1680 | store_comment (pfile, token); | |
0d9f234d | 1681 | } |
45b966db | 1682 | |
5fddcffc NB |
1683 | /* Allocate COUNT tokens for RUN. */ |
1684 | void | |
6cf87ca4 | 1685 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
5fddcffc | 1686 | { |
72bb2c39 | 1687 | run->base = XNEWVEC (cpp_token, count); |
5fddcffc NB |
1688 | run->limit = run->base + count; |
1689 | run->next = NULL; | |
1690 | } | |
1691 | ||
1692 | /* Returns the next tokenrun, or creates one if there is none. */ | |
1693 | static tokenrun * | |
6cf87ca4 | 1694 | next_tokenrun (tokenrun *run) |
5fddcffc NB |
1695 | { |
1696 | if (run->next == NULL) | |
1697 | { | |
72bb2c39 | 1698 | run->next = XNEW (tokenrun); |
bdcbe496 | 1699 | run->next->prev = run; |
5fddcffc NB |
1700 | _cpp_init_tokenrun (run->next, 250); |
1701 | } | |
1702 | ||
1703 | return run->next; | |
1704 | } | |
1705 | ||
5950c3c9 BE |
1706 | /* Look ahead in the input stream. */ |
1707 | const cpp_token * | |
1708 | cpp_peek_token (cpp_reader *pfile, int index) | |
1709 | { | |
1710 | cpp_context *context = pfile->context; | |
1711 | const cpp_token *peektok; | |
1712 | int count; | |
1713 | ||
1714 | /* First, scan through any pending cpp_context objects. */ | |
1715 | while (context->prev) | |
1716 | { | |
1717 | ptrdiff_t sz = (context->direct_p | |
1718 | ? LAST (context).token - FIRST (context).token | |
1719 | : LAST (context).ptoken - FIRST (context).ptoken); | |
1720 | ||
1721 | if (index < (int) sz) | |
1722 | return (context->direct_p | |
1723 | ? FIRST (context).token + index | |
1724 | : *(FIRST (context).ptoken + index)); | |
1725 | ||
1726 | index -= (int) sz; | |
1727 | context = context->prev; | |
1728 | } | |
1729 | ||
1730 | /* We will have to read some new tokens after all (and do so | |
1731 | without invalidating preceding tokens). */ | |
1732 | count = index; | |
1733 | pfile->keep_tokens++; | |
1734 | ||
1735 | do | |
1736 | { | |
1737 | peektok = _cpp_lex_token (pfile); | |
1738 | if (peektok->type == CPP_EOF) | |
1739 | return peektok; | |
1740 | } | |
1741 | while (index--); | |
1742 | ||
1743 | _cpp_backup_tokens_direct (pfile, count + 1); | |
1744 | pfile->keep_tokens--; | |
1745 | ||
1746 | return peektok; | |
1747 | } | |
1748 | ||
4ed5bcfb NB |
1749 | /* Allocate a single token that is invalidated at the same time as the |
1750 | rest of the tokens on the line. Has its line and col set to the | |
1751 | same as the last lexed token, so that diagnostics appear in the | |
1752 | right place. */ | |
1753 | cpp_token * | |
6cf87ca4 | 1754 | _cpp_temp_token (cpp_reader *pfile) |
4ed5bcfb NB |
1755 | { |
1756 | cpp_token *old, *result; | |
5950c3c9 BE |
1757 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
1758 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
4ed5bcfb NB |
1759 | |
1760 | old = pfile->cur_token - 1; | |
5950c3c9 BE |
1761 | /* Any pre-existing lookaheads must not be clobbered. */ |
1762 | if (la) | |
1763 | { | |
1764 | if (sz <= la) | |
1765 | { | |
1766 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
1767 | ||
1768 | if (sz < la) | |
1769 | memmove (next->base + 1, next->base, | |
1770 | (la - sz) * sizeof (cpp_token)); | |
1771 | ||
1772 | next->base[0] = pfile->cur_run->limit[-1]; | |
1773 | } | |
1774 | ||
1775 | if (sz > 1) | |
1776 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
1777 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
1778 | } | |
1779 | ||
1780 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
4ed5bcfb NB |
1781 | { |
1782 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
1783 | pfile->cur_token = pfile->cur_run->base; | |
1784 | } | |
1785 | ||
1786 | result = pfile->cur_token++; | |
12f9df4e | 1787 | result->src_loc = old->src_loc; |
4ed5bcfb NB |
1788 | return result; |
1789 | } | |
1790 | ||
14baae01 NB |
1791 | /* Lex a token into RESULT (external interface). Takes care of issues |
1792 | like directive handling, token lookahead, multiple include | |
a1f300c0 | 1793 | optimization and skipping. */ |
345894b4 | 1794 | const cpp_token * |
6cf87ca4 | 1795 | _cpp_lex_token (cpp_reader *pfile) |
5fddcffc | 1796 | { |
bdcbe496 | 1797 | cpp_token *result; |
5fddcffc | 1798 | |
bdcbe496 | 1799 | for (;;) |
5fddcffc | 1800 | { |
bdcbe496 | 1801 | if (pfile->cur_token == pfile->cur_run->limit) |
5fddcffc | 1802 | { |
bdcbe496 NB |
1803 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
1804 | pfile->cur_token = pfile->cur_run->base; | |
5fddcffc | 1805 | } |
ee380365 TT |
1806 | /* We assume that the current token is somewhere in the current |
1807 | run. */ | |
1808 | if (pfile->cur_token < pfile->cur_run->base | |
1809 | || pfile->cur_token >= pfile->cur_run->limit) | |
1810 | abort (); | |
5fddcffc | 1811 | |
bdcbe496 | 1812 | if (pfile->lookaheads) |
14baae01 NB |
1813 | { |
1814 | pfile->lookaheads--; | |
1815 | result = pfile->cur_token++; | |
1816 | } | |
bdcbe496 | 1817 | else |
14baae01 | 1818 | result = _cpp_lex_direct (pfile); |
bdcbe496 NB |
1819 | |
1820 | if (result->flags & BOL) | |
5fddcffc | 1821 | { |
bdcbe496 NB |
1822 | /* Is this a directive. If _cpp_handle_directive returns |
1823 | false, it is an assembler #. */ | |
1824 | if (result->type == CPP_HASH | |
e808ec9c NB |
1825 | /* 6.10.3 p 11: Directives in a list of macro arguments |
1826 | gives undefined behavior. This implementation | |
1827 | handles the directive as normal. */ | |
bc4071dd | 1828 | && pfile->state.parsing_args != 1) |
21b11495 | 1829 | { |
bc4071dd | 1830 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
21b11495 | 1831 | { |
bc4071dd RH |
1832 | if (pfile->directive_result.type == CPP_PADDING) |
1833 | continue; | |
21b11495 | 1834 | result = &pfile->directive_result; |
21b11495 ZW |
1835 | } |
1836 | } | |
bc4071dd RH |
1837 | else if (pfile->state.in_deferred_pragma) |
1838 | result = &pfile->directive_result; | |
21b11495 | 1839 | |
97293897 | 1840 | if (pfile->cb.line_change && !pfile->state.skipping) |
6cf87ca4 | 1841 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
5fddcffc | 1842 | } |
5fddcffc | 1843 | |
bdcbe496 | 1844 | /* We don't skip tokens in directives. */ |
bc4071dd | 1845 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
bdcbe496 | 1846 | break; |
5fddcffc | 1847 | |
bdcbe496 | 1848 | /* Outside a directive, invalidate controlling macros. At file |
14baae01 | 1849 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
6356f892 | 1850 | get here and MI optimization works. */ |
5fddcffc | 1851 | pfile->mi_valid = false; |
bdcbe496 NB |
1852 | |
1853 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
1854 | break; | |
5fddcffc NB |
1855 | } |
1856 | ||
345894b4 | 1857 | return result; |
5fddcffc NB |
1858 | } |
1859 | ||
26aea073 NB |
1860 | /* Returns true if a fresh line has been loaded. */ |
1861 | bool | |
6cf87ca4 | 1862 | _cpp_get_fresh_line (cpp_reader *pfile) |
004cb263 | 1863 | { |
22234f56 PB |
1864 | int return_at_eof; |
1865 | ||
26aea073 NB |
1866 | /* We can't get a new line until we leave the current directive. */ |
1867 | if (pfile->state.in_directive) | |
1868 | return false; | |
df383483 | 1869 | |
26aea073 | 1870 | for (;;) |
1a76916c | 1871 | { |
26aea073 | 1872 | cpp_buffer *buffer = pfile->buffer; |
1a76916c | 1873 | |
26aea073 NB |
1874 | if (!buffer->need_line) |
1875 | return true; | |
1876 | ||
1877 | if (buffer->next_line < buffer->rlimit) | |
004cb263 | 1878 | { |
26aea073 NB |
1879 | _cpp_clean_line (pfile); |
1880 | return true; | |
1881 | } | |
004cb263 | 1882 | |
26aea073 NB |
1883 | /* First, get out of parsing arguments state. */ |
1884 | if (pfile->state.parsing_args) | |
1885 | return false; | |
1886 | ||
1887 | /* End of buffer. Non-empty files should end in a newline. */ | |
1888 | if (buffer->buf != buffer->rlimit | |
1889 | && buffer->next_line > buffer->rlimit | |
1890 | && !buffer->from_stage3) | |
1891 | { | |
ed0e74e0 | 1892 | /* Clip to buffer size. */ |
26aea073 | 1893 | buffer->next_line = buffer->rlimit; |
26aea073 | 1894 | } |
22234f56 PB |
1895 | |
1896 | return_at_eof = buffer->return_at_eof; | |
26aea073 | 1897 | _cpp_pop_buffer (pfile); |
22234f56 | 1898 | if (pfile->buffer == NULL || return_at_eof) |
a506c55c | 1899 | return false; |
26aea073 | 1900 | } |
004cb263 NB |
1901 | } |
1902 | ||
6f572ac2 NB |
1903 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
1904 | do \ | |
1905 | { \ | |
1906 | result->type = ELSE_TYPE; \ | |
1907 | if (*buffer->cur == CHAR) \ | |
1908 | buffer->cur++, result->type = THEN_TYPE; \ | |
1909 | } \ | |
1910 | while (0) | |
480709cc | 1911 | |
14baae01 NB |
1912 | /* Lex a token into pfile->cur_token, which is also incremented, to |
1913 | get diagnostics pointing to the correct location. | |
1914 | ||
1915 | Does not handle issues such as token lookahead, multiple-include | |
f1ba665b | 1916 | optimization, directives, skipping etc. This function is only |
14baae01 NB |
1917 | suitable for use by _cpp_lex_token, and in special cases like |
1918 | lex_expansion_token which doesn't care for any of these issues. | |
1919 | ||
1920 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
1921 | otherwise returns to the start of the token buffer if permissible. | |
1922 | Returns the location of the lexed token. */ | |
1923 | cpp_token * | |
6cf87ca4 | 1924 | _cpp_lex_direct (cpp_reader *pfile) |
45b966db | 1925 | { |
0d9f234d | 1926 | cppchar_t c; |
adb84b42 | 1927 | cpp_buffer *buffer; |
0d9f234d | 1928 | const unsigned char *comment_start; |
14baae01 | 1929 | cpp_token *result = pfile->cur_token++; |
9ec7291f | 1930 | |
5fddcffc | 1931 | fresh_line: |
26aea073 | 1932 | result->flags = 0; |
2be570f9 | 1933 | buffer = pfile->buffer; |
a506c55c | 1934 | if (buffer->need_line) |
26aea073 | 1935 | { |
bc4071dd RH |
1936 | if (pfile->state.in_deferred_pragma) |
1937 | { | |
1938 | result->type = CPP_PRAGMA_EOL; | |
1939 | pfile->state.in_deferred_pragma = false; | |
1940 | if (!pfile->state.pragma_allow_expansion) | |
1941 | pfile->state.prevent_expansion--; | |
1942 | return result; | |
1943 | } | |
26aea073 NB |
1944 | if (!_cpp_get_fresh_line (pfile)) |
1945 | { | |
1946 | result->type = CPP_EOF; | |
9ff7868d NB |
1947 | if (!pfile->state.in_directive) |
1948 | { | |
1949 | /* Tell the compiler the line number of the EOF token. */ | |
500bee0a | 1950 | result->src_loc = pfile->line_table->highest_line; |
9ff7868d NB |
1951 | result->flags = BOL; |
1952 | } | |
26aea073 NB |
1953 | return result; |
1954 | } | |
1955 | if (!pfile->keep_tokens) | |
1956 | { | |
1957 | pfile->cur_run = &pfile->base_run; | |
1958 | result = pfile->base_run.base; | |
1959 | pfile->cur_token = result + 1; | |
1960 | } | |
1961 | result->flags = BOL; | |
1962 | if (pfile->state.parsing_args == 2) | |
1963 | result->flags |= PREV_WHITE; | |
1964 | } | |
a506c55c | 1965 | buffer = pfile->buffer; |
5fddcffc | 1966 | update_tokens_line: |
500bee0a | 1967 | result->src_loc = pfile->line_table->highest_line; |
041c3194 | 1968 | |
5fddcffc | 1969 | skipped_white: |
26aea073 NB |
1970 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
1971 | && !pfile->overlaid_buffer) | |
1972 | { | |
1973 | _cpp_process_line_notes (pfile, false); | |
500bee0a | 1974 | result->src_loc = pfile->line_table->highest_line; |
26aea073 | 1975 | } |
480709cc | 1976 | c = *buffer->cur++; |
12f9df4e | 1977 | |
3f6ced10 GC |
1978 | result->src_loc = linemap_position_for_column (pfile->line_table, |
1979 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
5fddcffc | 1980 | |
0d9f234d | 1981 | switch (c) |
45b966db | 1982 | { |
4d6baafa NB |
1983 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1984 | result->flags |= PREV_WHITE; | |
26aea073 NB |
1985 | skip_whitespace (pfile, c); |
1986 | goto skipped_white; | |
0d9f234d | 1987 | |
26aea073 | 1988 | case '\n': |
12f9df4e PB |
1989 | if (buffer->cur < buffer->rlimit) |
1990 | CPP_INCREMENT_LINE (pfile, 0); | |
26aea073 NB |
1991 | buffer->need_line = true; |
1992 | goto fresh_line; | |
46d07497 | 1993 | |
0d9f234d NB |
1994 | case '0': case '1': case '2': case '3': case '4': |
1995 | case '5': case '6': case '7': case '8': case '9': | |
50668cf6 GK |
1996 | { |
1997 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
1998 | result->type = CPP_NUMBER; | |
1999 | lex_number (pfile, &result->val.str, &nst); | |
2000 | warn_about_normalization (pfile, result, &nst); | |
2001 | break; | |
2002 | } | |
46d07497 | 2003 | |
0abc6a6a | 2004 | case 'L': |
b6baa67d KVH |
2005 | case 'u': |
2006 | case 'U': | |
2c6e3f55 JJ |
2007 | case 'R': |
2008 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2009 | wide strings or raw strings. */ | |
b6baa67d | 2010 | if (c == 'L' || CPP_OPTION (pfile, uliterals)) |
bced6edf | 2011 | { |
2c6e3f55 JJ |
2012 | if ((*buffer->cur == '\'' && c != 'R') |
2013 | || *buffer->cur == '"' | |
2014 | || (*buffer->cur == 'R' | |
2015 | && c != 'R' | |
2016 | && buffer->cur[1] == '"' | |
2017 | && CPP_OPTION (pfile, uliterals)) | |
2018 | || (*buffer->cur == '8' | |
2019 | && c == 'u' | |
2020 | && (buffer->cur[1] == '"' | |
2021 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"')))) | |
b6baa67d KVH |
2022 | { |
2023 | lex_string (pfile, result, buffer->cur - 1); | |
2024 | break; | |
2025 | } | |
bced6edf | 2026 | } |
df383483 | 2027 | /* Fall through. */ |
0abc6a6a | 2028 | |
0d9f234d NB |
2029 | case '_': |
2030 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2031 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2032 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
b6baa67d | 2033 | case 's': case 't': case 'v': case 'w': case 'x': |
0d9f234d NB |
2034 | case 'y': case 'z': |
2035 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
0abc6a6a | 2036 | case 'G': case 'H': case 'I': case 'J': case 'K': |
2c6e3f55 | 2037 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
b6baa67d | 2038 | case 'S': case 'T': case 'V': case 'W': case 'X': |
0d9f234d NB |
2039 | case 'Y': case 'Z': |
2040 | result->type = CPP_NAME; | |
50668cf6 GK |
2041 | { |
2042 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
9a0c6187 JM |
2043 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
2044 | &nst); | |
50668cf6 GK |
2045 | warn_about_normalization (pfile, result, &nst); |
2046 | } | |
0d9f234d | 2047 | |
0d9f234d | 2048 | /* Convert named operators to their proper types. */ |
9a0c6187 | 2049 | if (result->val.node.node->flags & NODE_OPERATOR) |
0d9f234d NB |
2050 | { |
2051 | result->flags |= NAMED_OP; | |
9a0c6187 | 2052 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
0d9f234d NB |
2053 | } |
2054 | break; | |
2055 | ||
2056 | case '\'': | |
2057 | case '"': | |
6338b358 | 2058 | lex_string (pfile, result, buffer->cur - 1); |
0d9f234d | 2059 | break; |
041c3194 | 2060 | |
0d9f234d | 2061 | case '/': |
1c6d33ef NB |
2062 | /* A potential block or line comment. */ |
2063 | comment_start = buffer->cur; | |
6f572ac2 NB |
2064 | c = *buffer->cur; |
2065 | ||
1c6d33ef NB |
2066 | if (c == '*') |
2067 | { | |
26aea073 | 2068 | if (_cpp_skip_block_comment (pfile)) |
0527bc4e | 2069 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
0d9f234d | 2070 | } |
480709cc | 2071 | else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments) |
12f9df4e | 2072 | || cpp_in_system_header (pfile))) |
0d9f234d | 2073 | { |
bdb05a7b NB |
2074 | /* Warn about comments only if pedantically GNUC89, and not |
2075 | in system headers. */ | |
2076 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
a94c1199 | 2077 | && ! buffer->warned_cplusplus_comments) |
041c3194 | 2078 | { |
0527bc4e | 2079 | cpp_error (pfile, CPP_DL_PEDWARN, |
56508306 | 2080 | "C++ style comments are not allowed in ISO C90"); |
0527bc4e | 2081 | cpp_error (pfile, CPP_DL_PEDWARN, |
ebef4e8c | 2082 | "(this will be reported only once per input file)"); |
1c6d33ef NB |
2083 | buffer->warned_cplusplus_comments = 1; |
2084 | } | |
0d9f234d | 2085 | |
01ef6563 | 2086 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
87cf0651 | 2087 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
1c6d33ef | 2088 | } |
480709cc NB |
2089 | else if (c == '=') |
2090 | { | |
6f572ac2 | 2091 | buffer->cur++; |
480709cc NB |
2092 | result->type = CPP_DIV_EQ; |
2093 | break; | |
2094 | } | |
2095 | else | |
2096 | { | |
480709cc NB |
2097 | result->type = CPP_DIV; |
2098 | break; | |
2099 | } | |
0d9f234d | 2100 | |
1c6d33ef NB |
2101 | if (!pfile->state.save_comments) |
2102 | { | |
2103 | result->flags |= PREV_WHITE; | |
5fddcffc | 2104 | goto update_tokens_line; |
0d9f234d | 2105 | } |
1c6d33ef NB |
2106 | |
2107 | /* Save the comment as a token in its own right. */ | |
477cdac7 | 2108 | save_comment (pfile, result, comment_start, c); |
bdcbe496 | 2109 | break; |
0d9f234d NB |
2110 | |
2111 | case '<': | |
2112 | if (pfile->state.angled_headers) | |
2113 | { | |
6338b358 | 2114 | lex_string (pfile, result, buffer->cur - 1); |
4bb09c26 JM |
2115 | if (result->type != CPP_LESS) |
2116 | break; | |
0d9f234d | 2117 | } |
45b966db | 2118 | |
6f572ac2 NB |
2119 | result->type = CPP_LESS; |
2120 | if (*buffer->cur == '=') | |
2121 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2122 | else if (*buffer->cur == '<') | |
0d9f234d | 2123 | { |
6f572ac2 NB |
2124 | buffer->cur++; |
2125 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
0d9f234d | 2126 | } |
6f572ac2 | 2127 | else if (CPP_OPTION (pfile, digraphs)) |
480709cc | 2128 | { |
6f572ac2 NB |
2129 | if (*buffer->cur == ':') |
2130 | { | |
2131 | buffer->cur++; | |
2132 | result->flags |= DIGRAPH; | |
2133 | result->type = CPP_OPEN_SQUARE; | |
2134 | } | |
2135 | else if (*buffer->cur == '%') | |
2136 | { | |
2137 | buffer->cur++; | |
2138 | result->flags |= DIGRAPH; | |
2139 | result->type = CPP_OPEN_BRACE; | |
2140 | } | |
480709cc | 2141 | } |
0d9f234d NB |
2142 | break; |
2143 | ||
2144 | case '>': | |
6f572ac2 NB |
2145 | result->type = CPP_GREATER; |
2146 | if (*buffer->cur == '=') | |
2147 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
2148 | else if (*buffer->cur == '>') | |
0d9f234d | 2149 | { |
6f572ac2 NB |
2150 | buffer->cur++; |
2151 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
2152 | } | |
0d9f234d NB |
2153 | break; |
2154 | ||
cbcff6df | 2155 | case '%': |
6f572ac2 NB |
2156 | result->type = CPP_MOD; |
2157 | if (*buffer->cur == '=') | |
2158 | buffer->cur++, result->type = CPP_MOD_EQ; | |
2159 | else if (CPP_OPTION (pfile, digraphs)) | |
480709cc | 2160 | { |
6f572ac2 | 2161 | if (*buffer->cur == ':') |
480709cc | 2162 | { |
6f572ac2 NB |
2163 | buffer->cur++; |
2164 | result->flags |= DIGRAPH; | |
2165 | result->type = CPP_HASH; | |
2166 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
9a0c6187 | 2167 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
6f572ac2 NB |
2168 | } |
2169 | else if (*buffer->cur == '>') | |
2170 | { | |
2171 | buffer->cur++; | |
2172 | result->flags |= DIGRAPH; | |
2173 | result->type = CPP_CLOSE_BRACE; | |
480709cc | 2174 | } |
480709cc | 2175 | } |
0d9f234d NB |
2176 | break; |
2177 | ||
cbcff6df | 2178 | case '.': |
480709cc | 2179 | result->type = CPP_DOT; |
6f572ac2 | 2180 | if (ISDIGIT (*buffer->cur)) |
480709cc | 2181 | { |
50668cf6 | 2182 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
480709cc | 2183 | result->type = CPP_NUMBER; |
50668cf6 GK |
2184 | lex_number (pfile, &result->val.str, &nst); |
2185 | warn_about_normalization (pfile, result, &nst); | |
480709cc | 2186 | } |
6f572ac2 NB |
2187 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
2188 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
2189 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
2190 | buffer->cur++, result->type = CPP_DOT_STAR; | |
0d9f234d | 2191 | break; |
45b966db | 2192 | |
0d9f234d | 2193 | case '+': |
6f572ac2 NB |
2194 | result->type = CPP_PLUS; |
2195 | if (*buffer->cur == '+') | |
2196 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
2197 | else if (*buffer->cur == '=') | |
2198 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
0d9f234d | 2199 | break; |
04e3ec78 | 2200 | |
0d9f234d | 2201 | case '-': |
6f572ac2 NB |
2202 | result->type = CPP_MINUS; |
2203 | if (*buffer->cur == '>') | |
0d9f234d | 2204 | { |
6f572ac2 | 2205 | buffer->cur++; |
480709cc | 2206 | result->type = CPP_DEREF; |
6f572ac2 NB |
2207 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
2208 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
480709cc | 2209 | } |
6f572ac2 NB |
2210 | else if (*buffer->cur == '-') |
2211 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
2212 | else if (*buffer->cur == '=') | |
2213 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
0d9f234d | 2214 | break; |
45b966db | 2215 | |
0d9f234d | 2216 | case '&': |
6f572ac2 NB |
2217 | result->type = CPP_AND; |
2218 | if (*buffer->cur == '&') | |
2219 | buffer->cur++, result->type = CPP_AND_AND; | |
2220 | else if (*buffer->cur == '=') | |
2221 | buffer->cur++, result->type = CPP_AND_EQ; | |
0d9f234d | 2222 | break; |
df383483 | 2223 | |
0d9f234d | 2224 | case '|': |
6f572ac2 NB |
2225 | result->type = CPP_OR; |
2226 | if (*buffer->cur == '|') | |
2227 | buffer->cur++, result->type = CPP_OR_OR; | |
2228 | else if (*buffer->cur == '=') | |
2229 | buffer->cur++, result->type = CPP_OR_EQ; | |
0d9f234d | 2230 | break; |
45b966db | 2231 | |
0d9f234d | 2232 | case ':': |
6f572ac2 NB |
2233 | result->type = CPP_COLON; |
2234 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
2235 | buffer->cur++, result->type = CPP_SCOPE; | |
2236 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
0d9f234d | 2237 | { |
6f572ac2 | 2238 | buffer->cur++; |
0d9f234d | 2239 | result->flags |= DIGRAPH; |
480709cc NB |
2240 | result->type = CPP_CLOSE_SQUARE; |
2241 | } | |
0d9f234d | 2242 | break; |
45b966db | 2243 | |
480709cc NB |
2244 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
2245 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
2246 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
2247 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
9a0c6187 | 2248 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
480709cc | 2249 | |
26aea073 | 2250 | case '?': result->type = CPP_QUERY; break; |
0d9f234d NB |
2251 | case '~': result->type = CPP_COMPL; break; |
2252 | case ',': result->type = CPP_COMMA; break; | |
2253 | case '(': result->type = CPP_OPEN_PAREN; break; | |
2254 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
2255 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
2256 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
2257 | case '{': result->type = CPP_OPEN_BRACE; break; | |
2258 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
2259 | case ';': result->type = CPP_SEMICOLON; break; | |
2260 | ||
40f03658 | 2261 | /* @ is a punctuator in Objective-C. */ |
cc937581 | 2262 | case '@': result->type = CPP_ATSIGN; break; |
0d9f234d | 2263 | |
0abc6a6a | 2264 | case '$': |
1613e52b NB |
2265 | case '\\': |
2266 | { | |
2267 | const uchar *base = --buffer->cur; | |
50668cf6 | 2268 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
0abc6a6a | 2269 | |
50668cf6 | 2270 | if (forms_identifier_p (pfile, true, &nst)) |
1613e52b NB |
2271 | { |
2272 | result->type = CPP_NAME; | |
9a0c6187 | 2273 | result->val.node.node = lex_identifier (pfile, base, true, &nst); |
50668cf6 | 2274 | warn_about_normalization (pfile, result, &nst); |
1613e52b NB |
2275 | break; |
2276 | } | |
2277 | buffer->cur++; | |
1067694a | 2278 | } |
1613e52b | 2279 | |
1067694a | 2280 | default: |
6338b358 NB |
2281 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); |
2282 | break; | |
0d9f234d | 2283 | } |
bdcbe496 NB |
2284 | |
2285 | return result; | |
0d9f234d NB |
2286 | } |
2287 | ||
59325650 NB |
2288 | /* An upper bound on the number of bytes needed to spell TOKEN. |
2289 | Does not include preceding whitespace. */ | |
93c80368 | 2290 | unsigned int |
6cf87ca4 | 2291 | cpp_token_len (const cpp_token *token) |
0d9f234d | 2292 | { |
93c80368 | 2293 | unsigned int len; |
6d2c2047 | 2294 | |
93c80368 | 2295 | switch (TOKEN_SPELL (token)) |
041c3194 | 2296 | { |
cc955282 | 2297 | default: len = 6; break; |
6338b358 | 2298 | case SPELL_LITERAL: len = token->val.str.len; break; |
9a0c6187 | 2299 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
041c3194 | 2300 | } |
59325650 NB |
2301 | |
2302 | return len; | |
6d2c2047 ZW |
2303 | } |
2304 | ||
47e20491 GK |
2305 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
2306 | Return the number of bytes read out of NAME. (There are always | |
2307 | 10 bytes written to BUFFER.) */ | |
2308 | ||
2309 | static size_t | |
2310 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
2311 | { | |
2312 | int j; | |
2313 | int ucn_len = 0; | |
2314 | int ucn_len_c; | |
2315 | unsigned t; | |
2316 | unsigned long utf32; | |
2317 | ||
2318 | /* Compute the length of the UTF-8 sequence. */ | |
2319 | for (t = *name; t & 0x80; t <<= 1) | |
2320 | ucn_len++; | |
2321 | ||
2322 | utf32 = *name & (0x7F >> ucn_len); | |
2323 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
2324 | { | |
2325 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
2326 | ||
2327 | /* Ill-formed UTF-8. */ | |
2328 | if ((*name & ~0x3F) != 0x80) | |
2329 | abort (); | |
2330 | } | |
2331 | ||
2332 | *buffer++ = '\\'; | |
2333 | *buffer++ = 'U'; | |
2334 | for (j = 7; j >= 0; j--) | |
2335 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
2336 | return ucn_len; | |
2337 | } | |
2338 | ||
cfc93532 MLI |
2339 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
2340 | the spelling of the digraph. */ | |
2341 | static const unsigned char * | |
2342 | cpp_digraph2name (enum cpp_ttype type) | |
2343 | { | |
2344 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
2345 | } | |
47e20491 | 2346 | |
041c3194 | 2347 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
cf00a885 | 2348 | already contain the enough space to hold the token's spelling. |
6cf87ca4 | 2349 | Returns a pointer to the character after the last character written. |
47e20491 GK |
2350 | FORSTRING is true if this is to be the spelling after translation |
2351 | phase 1 (this is different for UCNs). | |
6cf87ca4 | 2352 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
93c80368 | 2353 | unsigned char * |
6cf87ca4 | 2354 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
47e20491 | 2355 | unsigned char *buffer, bool forstring) |
041c3194 | 2356 | { |
96be6998 | 2357 | switch (TOKEN_SPELL (token)) |
041c3194 ZW |
2358 | { |
2359 | case SPELL_OPERATOR: | |
2360 | { | |
2361 | const unsigned char *spelling; | |
2362 | unsigned char c; | |
d6d5f795 | 2363 | |
041c3194 | 2364 | if (token->flags & DIGRAPH) |
cfc93532 | 2365 | spelling = cpp_digraph2name (token->type); |
92936ecf ZW |
2366 | else if (token->flags & NAMED_OP) |
2367 | goto spell_ident; | |
041c3194 | 2368 | else |
96be6998 | 2369 | spelling = TOKEN_NAME (token); |
df383483 | 2370 | |
041c3194 ZW |
2371 | while ((c = *spelling++) != '\0') |
2372 | *buffer++ = c; | |
2373 | } | |
2374 | break; | |
d6d5f795 | 2375 | |
47ad4138 | 2376 | spell_ident: |
041c3194 | 2377 | case SPELL_IDENT: |
47e20491 GK |
2378 | if (forstring) |
2379 | { | |
9a0c6187 JM |
2380 | memcpy (buffer, NODE_NAME (token->val.node.node), |
2381 | NODE_LEN (token->val.node.node)); | |
2382 | buffer += NODE_LEN (token->val.node.node); | |
47e20491 GK |
2383 | } |
2384 | else | |
2385 | { | |
2386 | size_t i; | |
9a0c6187 | 2387 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 2388 | |
9a0c6187 | 2389 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
2390 | if (name[i] & ~0x7F) |
2391 | { | |
2392 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2393 | buffer += 10; | |
2394 | } | |
2395 | else | |
9a0c6187 | 2396 | *buffer++ = NODE_NAME (token->val.node.node)[i]; |
47e20491 | 2397 | } |
041c3194 | 2398 | break; |
d6d5f795 | 2399 | |
6338b358 | 2400 | case SPELL_LITERAL: |
47ad4138 ZW |
2401 | memcpy (buffer, token->val.str.text, token->val.str.len); |
2402 | buffer += token->val.str.len; | |
2403 | break; | |
2404 | ||
041c3194 | 2405 | case SPELL_NONE: |
0527bc4e JDA |
2406 | cpp_error (pfile, CPP_DL_ICE, |
2407 | "unspellable token %s", TOKEN_NAME (token)); | |
041c3194 ZW |
2408 | break; |
2409 | } | |
d6d5f795 | 2410 | |
041c3194 ZW |
2411 | return buffer; |
2412 | } | |
d6d5f795 | 2413 | |
5d8ebbd8 NB |
2414 | /* Returns TOKEN spelt as a null-terminated string. The string is |
2415 | freed when the reader is destroyed. Useful for diagnostics. */ | |
93c80368 | 2416 | unsigned char * |
6cf87ca4 | 2417 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
59325650 NB |
2418 | { |
2419 | unsigned int len = cpp_token_len (token) + 1; | |
ece54d54 | 2420 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
c5a04734 | 2421 | |
47e20491 | 2422 | end = cpp_spell_token (pfile, token, start, false); |
93c80368 | 2423 | end[0] = '\0'; |
c5a04734 | 2424 | |
93c80368 NB |
2425 | return start; |
2426 | } | |
c5a04734 | 2427 | |
cfc93532 MLI |
2428 | /* Returns a pointer to a string which spells the token defined by |
2429 | TYPE and FLAGS. Used by C front ends, which really should move to | |
2430 | using cpp_token_as_text. */ | |
93c80368 | 2431 | const char * |
cfc93532 | 2432 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
93c80368 | 2433 | { |
cfc93532 MLI |
2434 | if (flags & DIGRAPH) |
2435 | return (const char *) cpp_digraph2name (type); | |
2436 | else if (flags & NAMED_OP) | |
2437 | return cpp_named_operator2name (type); | |
2438 | ||
93c80368 NB |
2439 | return (const char *) token_spellings[type].name; |
2440 | } | |
c5a04734 | 2441 | |
4ed5bcfb NB |
2442 | /* Writes the spelling of token to FP, without any preceding space. |
2443 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
2444 | double-buffering. */ | |
93c80368 | 2445 | void |
6cf87ca4 | 2446 | cpp_output_token (const cpp_token *token, FILE *fp) |
93c80368 | 2447 | { |
93c80368 | 2448 | switch (TOKEN_SPELL (token)) |
c5a04734 | 2449 | { |
93c80368 NB |
2450 | case SPELL_OPERATOR: |
2451 | { | |
2452 | const unsigned char *spelling; | |
3b681e9d | 2453 | int c; |
c5a04734 | 2454 | |
93c80368 | 2455 | if (token->flags & DIGRAPH) |
cfc93532 | 2456 | spelling = cpp_digraph2name (token->type); |
93c80368 NB |
2457 | else if (token->flags & NAMED_OP) |
2458 | goto spell_ident; | |
2459 | else | |
2460 | spelling = TOKEN_NAME (token); | |
041c3194 | 2461 | |
3b681e9d ZW |
2462 | c = *spelling; |
2463 | do | |
2464 | putc (c, fp); | |
2465 | while ((c = *++spelling) != '\0'); | |
93c80368 NB |
2466 | } |
2467 | break; | |
041c3194 | 2468 | |
93c80368 NB |
2469 | spell_ident: |
2470 | case SPELL_IDENT: | |
47e20491 GK |
2471 | { |
2472 | size_t i; | |
9a0c6187 | 2473 | const unsigned char * name = NODE_NAME (token->val.node.node); |
47e20491 | 2474 | |
9a0c6187 | 2475 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
47e20491 GK |
2476 | if (name[i] & ~0x7F) |
2477 | { | |
2478 | unsigned char buffer[10]; | |
2479 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2480 | fwrite (buffer, 1, 10, fp); | |
2481 | } | |
2482 | else | |
9a0c6187 | 2483 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
47e20491 GK |
2484 | } |
2485 | break; | |
041c3194 | 2486 | |
6338b358 | 2487 | case SPELL_LITERAL: |
47ad4138 ZW |
2488 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
2489 | break; | |
2490 | ||
93c80368 NB |
2491 | case SPELL_NONE: |
2492 | /* An error, most probably. */ | |
2493 | break; | |
041c3194 | 2494 | } |
c5a04734 ZW |
2495 | } |
2496 | ||
93c80368 NB |
2497 | /* Compare two tokens. */ |
2498 | int | |
6cf87ca4 | 2499 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
c5a04734 | 2500 | { |
93c80368 NB |
2501 | if (a->type == b->type && a->flags == b->flags) |
2502 | switch (TOKEN_SPELL (a)) | |
2503 | { | |
2504 | default: /* Keep compiler happy. */ | |
2505 | case SPELL_OPERATOR: | |
9a0c6187 | 2506 | /* token_no is used to track where multiple consecutive ## |
aa508502 | 2507 | tokens were originally located. */ |
9a0c6187 | 2508 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
93c80368 | 2509 | case SPELL_NONE: |
9a0c6187 JM |
2510 | return (a->type != CPP_MACRO_ARG |
2511 | || a->val.macro_arg.arg_no == b->val.macro_arg.arg_no); | |
93c80368 | 2512 | case SPELL_IDENT: |
9a0c6187 | 2513 | return a->val.node.node == b->val.node.node; |
6338b358 | 2514 | case SPELL_LITERAL: |
93c80368 NB |
2515 | return (a->val.str.len == b->val.str.len |
2516 | && !memcmp (a->val.str.text, b->val.str.text, | |
2517 | a->val.str.len)); | |
2518 | } | |
c5a04734 | 2519 | |
041c3194 ZW |
2520 | return 0; |
2521 | } | |
2522 | ||
93c80368 NB |
2523 | /* Returns nonzero if a space should be inserted to avoid an |
2524 | accidental token paste for output. For simplicity, it is | |
2525 | conservative, and occasionally advises a space where one is not | |
2526 | needed, e.g. "." and ".2". */ | |
93c80368 | 2527 | int |
6cf87ca4 ZW |
2528 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
2529 | const cpp_token *token2) | |
c5a04734 | 2530 | { |
93c80368 NB |
2531 | enum cpp_ttype a = token1->type, b = token2->type; |
2532 | cppchar_t c; | |
c5a04734 | 2533 | |
93c80368 NB |
2534 | if (token1->flags & NAMED_OP) |
2535 | a = CPP_NAME; | |
2536 | if (token2->flags & NAMED_OP) | |
2537 | b = CPP_NAME; | |
c5a04734 | 2538 | |
93c80368 NB |
2539 | c = EOF; |
2540 | if (token2->flags & DIGRAPH) | |
37b8524c | 2541 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
93c80368 NB |
2542 | else if (token_spellings[b].category == SPELL_OPERATOR) |
2543 | c = token_spellings[b].name[0]; | |
c5a04734 | 2544 | |
93c80368 | 2545 | /* Quickly get everything that can paste with an '='. */ |
37b8524c | 2546 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
93c80368 | 2547 | return 1; |
c5a04734 | 2548 | |
93c80368 | 2549 | switch (a) |
c5a04734 | 2550 | { |
b52dbbf8 SE |
2551 | case CPP_GREATER: return c == '>'; |
2552 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
93c80368 NB |
2553 | case CPP_PLUS: return c == '+'; |
2554 | case CPP_MINUS: return c == '-' || c == '>'; | |
2555 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
2556 | case CPP_MOD: return c == ':' || c == '>'; | |
2557 | case CPP_AND: return c == '&'; | |
2558 | case CPP_OR: return c == '|'; | |
2559 | case CPP_COLON: return c == ':' || c == '>'; | |
2560 | case CPP_DEREF: return c == '*'; | |
26ec42ee | 2561 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
93c80368 NB |
2562 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
2563 | case CPP_NAME: return ((b == CPP_NUMBER | |
2564 | && name_p (pfile, &token2->val.str)) | |
2565 | || b == CPP_NAME | |
2566 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
2567 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
2568 | || c == '.' || c == '+' || c == '-'); | |
1613e52b | 2569 | /* UCNs */ |
1067694a NB |
2570 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
2571 | && b == CPP_NAME) | |
1613e52b | 2572 | || (CPP_OPTION (pfile, objc) |
1067694a | 2573 | && token1->val.str.text[0] == '@' |
1613e52b | 2574 | && (b == CPP_NAME || b == CPP_STRING))); |
93c80368 | 2575 | default: break; |
c5a04734 | 2576 | } |
c5a04734 | 2577 | |
417f3e3a | 2578 | return 0; |
c5a04734 ZW |
2579 | } |
2580 | ||
93c80368 | 2581 | /* Output all the remaining tokens on the current line, and a newline |
4ed5bcfb NB |
2582 | character, to FP. Leading whitespace is removed. If there are |
2583 | macros, special token padding is not performed. */ | |
c5a04734 | 2584 | void |
6cf87ca4 | 2585 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
c5a04734 | 2586 | { |
4ed5bcfb | 2587 | const cpp_token *token; |
96be6998 | 2588 | |
4ed5bcfb NB |
2589 | token = cpp_get_token (pfile); |
2590 | while (token->type != CPP_EOF) | |
96be6998 | 2591 | { |
4ed5bcfb NB |
2592 | cpp_output_token (token, fp); |
2593 | token = cpp_get_token (pfile); | |
2594 | if (token->flags & PREV_WHITE) | |
2595 | putc (' ', fp); | |
96be6998 ZW |
2596 | } |
2597 | ||
93c80368 | 2598 | putc ('\n', fp); |
041c3194 | 2599 | } |
c5a04734 | 2600 | |
5d6342eb TT |
2601 | /* Return a string representation of all the remaining tokens on the |
2602 | current line. The result is allocated using xmalloc and must be | |
2603 | freed by the caller. */ | |
2604 | unsigned char * | |
2605 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
2606 | { | |
2607 | const cpp_token *token; | |
2608 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
2609 | unsigned int alloced = 120 + out; | |
2610 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
2611 | ||
2612 | /* If DIR_NAME is empty, there are no initial contents. */ | |
2613 | if (dir_name) | |
2614 | { | |
2615 | sprintf ((char *) result, "#%s ", dir_name); | |
2616 | out += 2; | |
2617 | } | |
2618 | ||
2619 | token = cpp_get_token (pfile); | |
2620 | while (token->type != CPP_EOF) | |
2621 | { | |
2622 | unsigned char *last; | |
2623 | /* Include room for a possible space and the terminating nul. */ | |
2624 | unsigned int len = cpp_token_len (token) + 2; | |
2625 | ||
2626 | if (out + len > alloced) | |
2627 | { | |
2628 | alloced *= 2; | |
2629 | if (out + len > alloced) | |
2630 | alloced = out + len; | |
2631 | result = (unsigned char *) xrealloc (result, alloced); | |
2632 | } | |
2633 | ||
2634 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
2635 | out = last - result; | |
2636 | ||
2637 | token = cpp_get_token (pfile); | |
2638 | if (token->flags & PREV_WHITE) | |
2639 | result[out++] = ' '; | |
2640 | } | |
2641 | ||
2642 | result[out] = '\0'; | |
2643 | return result; | |
2644 | } | |
2645 | ||
1e013d2e NB |
2646 | /* Memory buffers. Changing these three constants can have a dramatic |
2647 | effect on performance. The values here are reasonable defaults, | |
2648 | but might be tuned. If you adjust them, be sure to test across a | |
2649 | range of uses of cpplib, including heavy nested function-like macro | |
2650 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2651 | good tool for this). */ | |
2652 | #define MIN_BUFF_SIZE 8000 | |
87062813 | 2653 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
1e013d2e NB |
2654 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
2655 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
417f3e3a | 2656 | |
87062813 NB |
2657 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
2658 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
2659 | #endif | |
2660 | ||
c9e7a609 NB |
2661 | /* Create a new allocation buffer. Place the control block at the end |
2662 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
b8af0ca5 | 2663 | static _cpp_buff * |
6cf87ca4 | 2664 | new_buff (size_t len) |
b8af0ca5 NB |
2665 | { |
2666 | _cpp_buff *result; | |
ece54d54 | 2667 | unsigned char *base; |
b8af0ca5 | 2668 | |
1e013d2e NB |
2669 | if (len < MIN_BUFF_SIZE) |
2670 | len = MIN_BUFF_SIZE; | |
c70f6ed3 | 2671 | len = CPP_ALIGN (len); |
b8af0ca5 | 2672 | |
c3f829c1 | 2673 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
b8af0ca5 NB |
2674 | result = (_cpp_buff *) (base + len); |
2675 | result->base = base; | |
2676 | result->cur = base; | |
2677 | result->limit = base + len; | |
2678 | result->next = NULL; | |
2679 | return result; | |
2680 | } | |
2681 | ||
2682 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2683 | void | |
6cf87ca4 | 2684 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
b8af0ca5 NB |
2685 | { |
2686 | _cpp_buff *end = buff; | |
2687 | ||
2688 | while (end->next) | |
2689 | end = end->next; | |
2690 | end->next = pfile->free_buffs; | |
2691 | pfile->free_buffs = buff; | |
2692 | } | |
2693 | ||
2694 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2695 | _cpp_buff * | |
6cf87ca4 | 2696 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
b8af0ca5 NB |
2697 | { |
2698 | _cpp_buff *result, **p; | |
2699 | ||
2700 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2701 | { | |
6142088c | 2702 | size_t size; |
1e013d2e NB |
2703 | |
2704 | if (*p == NULL) | |
b8af0ca5 | 2705 | return new_buff (min_size); |
1e013d2e NB |
2706 | result = *p; |
2707 | size = result->limit - result->base; | |
2708 | /* Return a buffer that's big enough, but don't waste one that's | |
2709 | way too big. */ | |
34f5271d | 2710 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
b8af0ca5 NB |
2711 | break; |
2712 | } | |
2713 | ||
2714 | *p = result->next; | |
2715 | result->next = NULL; | |
2716 | result->cur = result->base; | |
2717 | return result; | |
2718 | } | |
2719 | ||
4fe9b91c | 2720 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
2721 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
2722 | the excess bytes to the new buffer. Chains the new buffer after | |
2723 | BUFF, and returns the new buffer. */ | |
b8af0ca5 | 2724 | _cpp_buff * |
6cf87ca4 | 2725 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
b8af0ca5 | 2726 | { |
6142088c | 2727 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
8c3b2693 | 2728 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
b8af0ca5 | 2729 | |
8c3b2693 NB |
2730 | buff->next = new_buff; |
2731 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
2732 | return new_buff; | |
2733 | } | |
2734 | ||
4fe9b91c | 2735 | /* Creates a new buffer with enough space to hold the uncommitted |
8c3b2693 NB |
2736 | remaining bytes of the buffer pointed to by BUFF, and at least |
2737 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
2738 | Chains the new buffer before the buffer pointed to by BUFF, and | |
2739 | updates the pointer to point to the new buffer. */ | |
2740 | void | |
6cf87ca4 | 2741 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
8c3b2693 NB |
2742 | { |
2743 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
2744 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
2745 | ||
2746 | new_buff = _cpp_get_buff (pfile, size); | |
2747 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
2748 | new_buff->next = old_buff; | |
2749 | *pbuff = new_buff; | |
b8af0ca5 NB |
2750 | } |
2751 | ||
2752 | /* Free a chain of buffers starting at BUFF. */ | |
2753 | void | |
5671bf27 | 2754 | _cpp_free_buff (_cpp_buff *buff) |
b8af0ca5 NB |
2755 | { |
2756 | _cpp_buff *next; | |
2757 | ||
2758 | for (; buff; buff = next) | |
2759 | { | |
2760 | next = buff->next; | |
2761 | free (buff->base); | |
2762 | } | |
2763 | } | |
417f3e3a | 2764 | |
ece54d54 NB |
2765 | /* Allocate permanent, unaligned storage of length LEN. */ |
2766 | unsigned char * | |
6cf87ca4 | 2767 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
ece54d54 NB |
2768 | { |
2769 | _cpp_buff *buff = pfile->u_buff; | |
2770 | unsigned char *result = buff->cur; | |
2771 | ||
2772 | if (len > (size_t) (buff->limit - result)) | |
2773 | { | |
2774 | buff = _cpp_get_buff (pfile, len); | |
2775 | buff->next = pfile->u_buff; | |
2776 | pfile->u_buff = buff; | |
2777 | result = buff->cur; | |
2778 | } | |
2779 | ||
2780 | buff->cur = result + len; | |
2781 | return result; | |
2782 | } | |
2783 | ||
87062813 NB |
2784 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
2785 | That buffer is used for growing allocations when saving macro | |
2786 | replacement lists in a #define, and when parsing an answer to an | |
2787 | assertion in #assert, #unassert or #if (and therefore possibly | |
2788 | whilst expanding macros). It therefore must not be used by any | |
2789 | code that they might call: specifically the lexer and the guts of | |
2790 | the macro expander. | |
2791 | ||
2792 | All existing other uses clearly fit this restriction: storing | |
2793 | registered pragmas during initialization. */ | |
93c80368 | 2794 | unsigned char * |
6cf87ca4 | 2795 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
3fef5b2b | 2796 | { |
8c3b2693 NB |
2797 | _cpp_buff *buff = pfile->a_buff; |
2798 | unsigned char *result = buff->cur; | |
3fef5b2b | 2799 | |
8c3b2693 | 2800 | if (len > (size_t) (buff->limit - result)) |
3fef5b2b | 2801 | { |
8c3b2693 NB |
2802 | buff = _cpp_get_buff (pfile, len); |
2803 | buff->next = pfile->a_buff; | |
2804 | pfile->a_buff = buff; | |
2805 | result = buff->cur; | |
3fef5b2b | 2806 | } |
041c3194 | 2807 | |
8c3b2693 | 2808 | buff->cur = result + len; |
93c80368 | 2809 | return result; |
041c3194 | 2810 | } |
d8044160 GK |
2811 | |
2812 | /* Say which field of TOK is in use. */ | |
2813 | ||
2814 | enum cpp_token_fld_kind | |
2815 | cpp_token_val_index (cpp_token *tok) | |
2816 | { | |
2817 | switch (TOKEN_SPELL (tok)) | |
2818 | { | |
2819 | case SPELL_IDENT: | |
2820 | return CPP_TOKEN_FLD_NODE; | |
2821 | case SPELL_LITERAL: | |
2822 | return CPP_TOKEN_FLD_STR; | |
aa508502 JM |
2823 | case SPELL_OPERATOR: |
2824 | if (tok->type == CPP_PASTE) | |
9a0c6187 | 2825 | return CPP_TOKEN_FLD_TOKEN_NO; |
aa508502 JM |
2826 | else |
2827 | return CPP_TOKEN_FLD_NONE; | |
d8044160 GK |
2828 | case SPELL_NONE: |
2829 | if (tok->type == CPP_MACRO_ARG) | |
2830 | return CPP_TOKEN_FLD_ARG_NO; | |
2831 | else if (tok->type == CPP_PADDING) | |
2832 | return CPP_TOKEN_FLD_SOURCE; | |
21b11495 | 2833 | else if (tok->type == CPP_PRAGMA) |
bc4071dd | 2834 | return CPP_TOKEN_FLD_PRAGMA; |
d8044160 GK |
2835 | /* else fall through */ |
2836 | default: | |
2837 | return CPP_TOKEN_FLD_NONE; | |
2838 | } | |
2839 | } |