]>
Commit | Line | Data |
---|---|---|
0578f103 | 1 | /* CPP Library - lexical analysis. |
cec7bb47 | 2 | Copyright (C) 2000-2013 Free Software Foundation, Inc. |
0578f103 | 3 | Contributed by Per Bothner, 1994-95. |
4 | Based on CCCP program by Paul Rubin, June 1986 | |
5 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
6 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
7 | ||
8 | This program is free software; you can redistribute it and/or modify it | |
9 | under the terms of the GNU General Public License as published by the | |
6bc9506f | 10 | Free Software Foundation; either version 3, or (at your option) any |
0578f103 | 11 | later version. |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
6bc9506f | 19 | along with this program; see the file COPYING3. If not see |
20 | <http://www.gnu.org/licenses/>. */ | |
0578f103 | 21 | |
22 | #include "config.h" | |
23 | #include "system.h" | |
0578f103 | 24 | #include "cpplib.h" |
d856c8a6 | 25 | #include "internal.h" |
0578f103 | 26 | |
79bd622b | 27 | enum spell_type |
241e762e | 28 | { |
79bd622b | 29 | SPELL_OPERATOR = 0, |
79bd622b | 30 | SPELL_IDENT, |
4970d4c2 | 31 | SPELL_LITERAL, |
79bd622b | 32 | SPELL_NONE |
241e762e | 33 | }; |
34 | ||
79bd622b | 35 | struct token_spelling |
241e762e | 36 | { |
79bd622b | 37 | enum spell_type category; |
38 | const unsigned char *name; | |
241e762e | 39 | }; |
40 | ||
0ca849f9 | 41 | static const unsigned char *const digraph_spellings[] = |
924bbf02 | 42 | { UC"%:", UC"%:%:", UC"<:", UC":>", UC"<%", UC"%>" }; |
79bd622b | 43 | |
924bbf02 | 44 | #define OP(e, s) { SPELL_OPERATOR, UC s }, |
45 | #define TK(e, s) { SPELL_ ## s, UC #e }, | |
0ca849f9 | 46 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; |
79bd622b | 47 | #undef OP |
48 | #undef TK | |
49 | ||
50 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
51 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
e2f9a79f | 52 | |
f7fdd7a1 | 53 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); |
54 | static int skip_line_comment (cpp_reader *); | |
55 | static void skip_whitespace (cpp_reader *, cppchar_t); | |
f7fdd7a1 | 56 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); |
57 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); | |
956c6108 | 58 | static void store_comment (cpp_reader *, cpp_token *); |
f7fdd7a1 | 59 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, |
60 | unsigned int, enum cpp_ttype); | |
61 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); | |
62 | static int name_p (cpp_reader *, const cpp_string *); | |
f7fdd7a1 | 63 | static tokenrun *next_tokenrun (tokenrun *); |
64 | ||
f7fdd7a1 | 65 | static _cpp_buff *new_buff (size_t); |
bce8e0c0 | 66 | |
e920deaf | 67 | |
f80e83a9 | 68 | /* Utility routine: |
2c63d6c8 | 69 | |
76faa4c0 | 70 | Compares, the token TOKEN to the NUL-terminated string STRING. |
71 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
f80e83a9 | 72 | int |
f7fdd7a1 | 73 | cpp_ideq (const cpp_token *token, const char *string) |
f80e83a9 | 74 | { |
76faa4c0 | 75 | if (token->type != CPP_NAME) |
f80e83a9 | 76 | return 0; |
76faa4c0 | 77 | |
2ee04baa | 78 | return !ustrcmp (NODE_NAME (token->val.node.node), (const uchar *) string); |
bce8e0c0 | 79 | } |
50fd6b48 | 80 | |
a54e0bf8 | 81 | /* Record a note TYPE at byte POS into the current cleaned logical |
82 | line. */ | |
1e0ef2fd | 83 | static void |
f7fdd7a1 | 84 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) |
338fa5f7 | 85 | { |
a54e0bf8 | 86 | if (buffer->notes_used == buffer->notes_cap) |
87 | { | |
88 | buffer->notes_cap = buffer->notes_cap * 2 + 200; | |
720aca92 | 89 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes, |
90 | buffer->notes_cap); | |
a54e0bf8 | 91 | } |
338fa5f7 | 92 | |
a54e0bf8 | 93 | buffer->notes[buffer->notes_used].pos = pos; |
94 | buffer->notes[buffer->notes_used].type = type; | |
95 | buffer->notes_used++; | |
338fa5f7 | 96 | } |
97 | ||
2431e8ba | 98 | \f |
99 | /* Fast path to find line special characters using optimized character | |
100 | scanning algorithms. Anything complicated falls back to the slow | |
101 | path below. Since this loop is very hot it's worth doing these kinds | |
102 | of optimizations. | |
103 | ||
104 | One of the paths through the ifdefs should provide | |
105 | ||
106 | const uchar *search_line_fast (const uchar *s, const uchar *end); | |
107 | ||
108 | Between S and END, search for \n, \r, \\, ?. Return a pointer to | |
109 | the found character. | |
110 | ||
111 | Note that the last character of the buffer is *always* a newline, | |
112 | as forced by _cpp_convert_input. This fact can be used to avoid | |
113 | explicitly looking for the end of the buffer. */ | |
114 | ||
115 | /* Configure gives us an ifdef test. */ | |
116 | #ifndef WORDS_BIGENDIAN | |
117 | #define WORDS_BIGENDIAN 0 | |
118 | #endif | |
119 | ||
120 | /* We'd like the largest integer that fits into a register. There's nothing | |
121 | in <stdint.h> that gives us that. For most hosts this is unsigned long, | |
122 | but MS decided on an LLP64 model. Thankfully when building with GCC we | |
123 | can get the "real" word size. */ | |
124 | #ifdef __GNUC__ | |
125 | typedef unsigned int word_type __attribute__((__mode__(__word__))); | |
126 | #else | |
127 | typedef unsigned long word_type; | |
128 | #endif | |
129 | ||
130 | /* The code below is only expecting sizes 4 or 8. | |
131 | Die at compile-time if this expectation is violated. */ | |
132 | typedef char check_word_type_size | |
133 | [(sizeof(word_type) == 8 || sizeof(word_type) == 4) * 2 - 1]; | |
134 | ||
135 | /* Return X with the first N bytes forced to values that won't match one | |
136 | of the interesting characters. Note that NUL is not interesting. */ | |
137 | ||
138 | static inline word_type | |
139 | acc_char_mask_misalign (word_type val, unsigned int n) | |
140 | { | |
141 | word_type mask = -1; | |
142 | if (WORDS_BIGENDIAN) | |
143 | mask >>= n * 8; | |
144 | else | |
145 | mask <<= n * 8; | |
146 | return val & mask; | |
147 | } | |
148 | ||
149 | /* Return X replicated to all byte positions within WORD_TYPE. */ | |
150 | ||
151 | static inline word_type | |
152 | acc_char_replicate (uchar x) | |
153 | { | |
154 | word_type ret; | |
155 | ||
156 | ret = (x << 24) | (x << 16) | (x << 8) | x; | |
157 | if (sizeof(word_type) == 8) | |
158 | ret = (ret << 16 << 16) | ret; | |
159 | return ret; | |
160 | } | |
161 | ||
162 | /* Return non-zero if some byte of VAL is (probably) C. */ | |
163 | ||
164 | static inline word_type | |
165 | acc_char_cmp (word_type val, word_type c) | |
166 | { | |
167 | #if defined(__GNUC__) && defined(__alpha__) | |
168 | /* We can get exact results using a compare-bytes instruction. | |
169 | Get (val == c) via (0 >= (val ^ c)). */ | |
170 | return __builtin_alpha_cmpbge (0, val ^ c); | |
171 | #else | |
172 | word_type magic = 0x7efefefeU; | |
173 | if (sizeof(word_type) == 8) | |
174 | magic = (magic << 16 << 16) | 0xfefefefeU; | |
175 | magic |= 1; | |
176 | ||
177 | val ^= c; | |
178 | return ((val + magic) ^ ~val) & ~magic; | |
179 | #endif | |
180 | } | |
181 | ||
182 | /* Given the result of acc_char_cmp is non-zero, return the index of | |
183 | the found character. If this was a false positive, return -1. */ | |
184 | ||
185 | static inline int | |
186 | acc_char_index (word_type cmp ATTRIBUTE_UNUSED, | |
187 | word_type val ATTRIBUTE_UNUSED) | |
188 | { | |
189 | #if defined(__GNUC__) && defined(__alpha__) && !WORDS_BIGENDIAN | |
190 | /* The cmpbge instruction sets *bits* of the result corresponding to | |
191 | matches in the bytes with no false positives. */ | |
192 | return __builtin_ctzl (cmp); | |
193 | #else | |
194 | unsigned int i; | |
195 | ||
196 | /* ??? It would be nice to force unrolling here, | |
197 | and have all of these constants folded. */ | |
198 | for (i = 0; i < sizeof(word_type); ++i) | |
199 | { | |
200 | uchar c; | |
201 | if (WORDS_BIGENDIAN) | |
202 | c = (val >> (sizeof(word_type) - i - 1) * 8) & 0xff; | |
203 | else | |
204 | c = (val >> i * 8) & 0xff; | |
205 | ||
206 | if (c == '\n' || c == '\r' || c == '\\' || c == '?') | |
207 | return i; | |
208 | } | |
209 | ||
210 | return -1; | |
211 | #endif | |
212 | } | |
213 | ||
214 | /* A version of the fast scanner using bit fiddling techniques. | |
215 | ||
216 | For 32-bit words, one would normally perform 16 comparisons and | |
217 | 16 branches. With this algorithm one performs 24 arithmetic | |
218 | operations and one branch. Whether this is faster with a 32-bit | |
219 | word size is going to be somewhat system dependent. | |
220 | ||
221 | For 64-bit words, we eliminate twice the number of comparisons | |
222 | and branches without increasing the number of arithmetic operations. | |
223 | It's almost certainly going to be a win with 64-bit word size. */ | |
224 | ||
225 | static const uchar * search_line_acc_char (const uchar *, const uchar *) | |
226 | ATTRIBUTE_UNUSED; | |
227 | ||
228 | static const uchar * | |
229 | search_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
230 | { | |
231 | const word_type repl_nl = acc_char_replicate ('\n'); | |
232 | const word_type repl_cr = acc_char_replicate ('\r'); | |
233 | const word_type repl_bs = acc_char_replicate ('\\'); | |
234 | const word_type repl_qm = acc_char_replicate ('?'); | |
235 | ||
236 | unsigned int misalign; | |
237 | const word_type *p; | |
238 | word_type val, t; | |
239 | ||
240 | /* Align the buffer. Mask out any bytes from before the beginning. */ | |
241 | p = (word_type *)((uintptr_t)s & -sizeof(word_type)); | |
242 | val = *p; | |
243 | misalign = (uintptr_t)s & (sizeof(word_type) - 1); | |
244 | if (misalign) | |
245 | val = acc_char_mask_misalign (val, misalign); | |
246 | ||
247 | /* Main loop. */ | |
248 | while (1) | |
249 | { | |
250 | t = acc_char_cmp (val, repl_nl); | |
251 | t |= acc_char_cmp (val, repl_cr); | |
252 | t |= acc_char_cmp (val, repl_bs); | |
253 | t |= acc_char_cmp (val, repl_qm); | |
254 | ||
255 | if (__builtin_expect (t != 0, 0)) | |
256 | { | |
257 | int i = acc_char_index (t, val); | |
258 | if (i >= 0) | |
259 | return (const uchar *)p + i; | |
260 | } | |
261 | ||
262 | val = *++p; | |
263 | } | |
264 | } | |
265 | ||
8ab29ece | 266 | /* Disable on Solaris 2/x86 until the following problems can be properly |
267 | autoconfed: | |
268 | ||
8ab29ece | 269 | The Solaris 9 assembler cannot assemble SSE4.2 insns. |
270 | Before Solaris 9 Update 6, SSE insns cannot be executed. | |
271 | The Solaris 10+ assembler tags objects with the instruction set | |
272 | extensions used, so SSE4.2 executables cannot run on machines that | |
273 | don't support that extension. */ | |
274 | ||
275 | #if (GCC_VERSION >= 4005) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__)) | |
2431e8ba | 276 | |
277 | /* Replicated character data to be shared between implementations. | |
278 | Recall that outside of a context with vector support we can't | |
279 | define compatible vector types, therefore these are all defined | |
280 | in terms of raw characters. */ | |
281 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { | |
282 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
283 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, | |
284 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
285 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, | |
286 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
287 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, | |
288 | { '?', '?', '?', '?', '?', '?', '?', '?', | |
289 | '?', '?', '?', '?', '?', '?', '?', '?' }, | |
290 | }; | |
291 | ||
292 | /* A version of the fast scanner using MMX vectorized byte compare insns. | |
293 | ||
294 | This uses the PMOVMSKB instruction which was introduced with "MMX2", | |
cef70b8a | 295 | which was packaged into SSE1; it is also present in the AMD MMX |
2431e8ba | 296 | extension. Mark the function as using "sse" so that we emit a real |
297 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ | |
298 | ||
299 | static const uchar * | |
300 | #ifndef __SSE__ | |
301 | __attribute__((__target__("sse"))) | |
302 | #endif | |
303 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
304 | { | |
305 | typedef char v8qi __attribute__ ((__vector_size__ (8))); | |
306 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); | |
307 | ||
308 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; | |
309 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; | |
310 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; | |
311 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; | |
312 | ||
313 | unsigned int misalign, found, mask; | |
314 | const v8qi *p; | |
315 | v8qi data, t, c; | |
316 | ||
317 | /* Align the source pointer. While MMX doesn't generate unaligned data | |
318 | faults, this allows us to safely scan to the end of the buffer without | |
319 | reading beyond the end of the last page. */ | |
320 | misalign = (uintptr_t)s & 7; | |
321 | p = (const v8qi *)((uintptr_t)s & -8); | |
322 | data = *p; | |
323 | ||
324 | /* Create a mask for the bytes that are valid within the first | |
325 | 16-byte block. The Idea here is that the AND with the mask | |
326 | within the loop is "free", since we need some AND or TEST | |
327 | insn in order to set the flags for the branch anyway. */ | |
328 | mask = -1u << misalign; | |
329 | ||
330 | /* Main loop processing 8 bytes at a time. */ | |
331 | goto start; | |
332 | do | |
333 | { | |
334 | data = *++p; | |
335 | mask = -1; | |
336 | ||
337 | start: | |
338 | t = __builtin_ia32_pcmpeqb(data, repl_nl); | |
339 | c = __builtin_ia32_pcmpeqb(data, repl_cr); | |
340 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
341 | c = __builtin_ia32_pcmpeqb(data, repl_bs); | |
342 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
343 | c = __builtin_ia32_pcmpeqb(data, repl_qm); | |
344 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
345 | found = __builtin_ia32_pmovmskb (t); | |
346 | found &= mask; | |
347 | } | |
348 | while (!found); | |
349 | ||
350 | __builtin_ia32_emms (); | |
351 | ||
352 | /* FOUND contains 1 in bits for which we matched a relevant | |
353 | character. Conversion to the byte index is trivial. */ | |
354 | found = __builtin_ctz(found); | |
355 | return (const uchar *)p + found; | |
356 | } | |
357 | ||
358 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ | |
359 | ||
360 | static const uchar * | |
361 | #ifndef __SSE2__ | |
362 | __attribute__((__target__("sse2"))) | |
363 | #endif | |
364 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
365 | { | |
366 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
367 | ||
368 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; | |
369 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; | |
370 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; | |
371 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; | |
372 | ||
373 | unsigned int misalign, found, mask; | |
374 | const v16qi *p; | |
375 | v16qi data, t; | |
376 | ||
377 | /* Align the source pointer. */ | |
378 | misalign = (uintptr_t)s & 15; | |
379 | p = (const v16qi *)((uintptr_t)s & -16); | |
380 | data = *p; | |
381 | ||
382 | /* Create a mask for the bytes that are valid within the first | |
383 | 16-byte block. The Idea here is that the AND with the mask | |
384 | within the loop is "free", since we need some AND or TEST | |
385 | insn in order to set the flags for the branch anyway. */ | |
386 | mask = -1u << misalign; | |
387 | ||
388 | /* Main loop processing 16 bytes at a time. */ | |
389 | goto start; | |
390 | do | |
391 | { | |
392 | data = *++p; | |
393 | mask = -1; | |
394 | ||
395 | start: | |
396 | t = __builtin_ia32_pcmpeqb128(data, repl_nl); | |
397 | t |= __builtin_ia32_pcmpeqb128(data, repl_cr); | |
398 | t |= __builtin_ia32_pcmpeqb128(data, repl_bs); | |
399 | t |= __builtin_ia32_pcmpeqb128(data, repl_qm); | |
400 | found = __builtin_ia32_pmovmskb128 (t); | |
401 | found &= mask; | |
402 | } | |
403 | while (!found); | |
404 | ||
405 | /* FOUND contains 1 in bits for which we matched a relevant | |
406 | character. Conversion to the byte index is trivial. */ | |
407 | found = __builtin_ctz(found); | |
408 | return (const uchar *)p + found; | |
409 | } | |
410 | ||
b315ae35 | 411 | #ifdef HAVE_SSE4 |
2431e8ba | 412 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
413 | ||
414 | static const uchar * | |
415 | #ifndef __SSE4_2__ | |
416 | __attribute__((__target__("sse4.2"))) | |
417 | #endif | |
418 | search_line_sse42 (const uchar *s, const uchar *end) | |
419 | { | |
420 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
421 | static const v16qi search = { '\n', '\r', '?', '\\' }; | |
422 | ||
423 | uintptr_t si = (uintptr_t)s; | |
424 | uintptr_t index; | |
425 | ||
426 | /* Check for unaligned input. */ | |
427 | if (si & 15) | |
428 | { | |
112f073c | 429 | v16qi sv; |
430 | ||
2431e8ba | 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. */ | |
112f073c | 443 | sv = __builtin_ia32_loaddqu ((const char *) s); |
444 | index = __builtin_ia32_pcmpestri128 (search, 4, sv, 16, 0); | |
445 | ||
2431e8ba | 446 | if (__builtin_expect (index < 16, 0)) |
447 | goto found; | |
448 | ||
449 | /* Advance the pointer to an aligned address. We will re-scan a | |
450 | few bytes, but we no longer need care for reading past the | |
451 | end of a page, since we're guaranteed a match. */ | |
452 | s = (const uchar *)((si + 16) & -16); | |
453 | } | |
454 | ||
455 | /* Main loop, processing 16 bytes at a time. By doing the whole loop | |
456 | in inline assembly, we can make proper use of the flags set. */ | |
457 | __asm ( "sub $16, %1\n" | |
458 | " .balign 16\n" | |
459 | "0: add $16, %1\n" | |
460 | " %vpcmpestri $0, (%1), %2\n" | |
461 | " jnc 0b" | |
462 | : "=&c"(index), "+r"(s) | |
463 | : "x"(search), "a"(4), "d"(16)); | |
464 | ||
465 | found: | |
466 | return s + index; | |
467 | } | |
468 | ||
b315ae35 | 469 | #else |
470 | /* Work around out-dated assemblers without sse4 support. */ | |
471 | #define search_line_sse42 search_line_sse2 | |
472 | #endif | |
473 | ||
2431e8ba | 474 | /* Check the CPU capabilities. */ |
475 | ||
476 | #include "../gcc/config/i386/cpuid.h" | |
477 | ||
478 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); | |
479 | static search_line_fast_type search_line_fast; | |
480 | ||
b735cc56 | 481 | #define HAVE_init_vectorized_lexer 1 |
482 | static inline void | |
2431e8ba | 483 | init_vectorized_lexer (void) |
484 | { | |
485 | unsigned dummy, ecx = 0, edx = 0; | |
486 | search_line_fast_type impl = search_line_acc_char; | |
487 | int minimum = 0; | |
488 | ||
489 | #if defined(__SSE4_2__) | |
490 | minimum = 3; | |
491 | #elif defined(__SSE2__) | |
492 | minimum = 2; | |
cef70b8a | 493 | #elif defined(__SSE__) |
2431e8ba | 494 | minimum = 1; |
495 | #endif | |
496 | ||
497 | if (minimum == 3) | |
498 | impl = search_line_sse42; | |
499 | else if (__get_cpuid (1, &dummy, &dummy, &ecx, &edx) || minimum == 2) | |
500 | { | |
501 | if (minimum == 3 || (ecx & bit_SSE4_2)) | |
502 | impl = search_line_sse42; | |
503 | else if (minimum == 2 || (edx & bit_SSE2)) | |
504 | impl = search_line_sse2; | |
505 | else if (minimum == 1 || (edx & bit_SSE)) | |
506 | impl = search_line_mmx; | |
507 | } | |
508 | else if (__get_cpuid (0x80000001, &dummy, &dummy, &dummy, &edx)) | |
509 | { | |
85303bd5 | 510 | if (minimum == 1 |
511 | || (edx & (bit_MMXEXT | bit_CMOV)) == (bit_MMXEXT | bit_CMOV)) | |
2431e8ba | 512 | impl = search_line_mmx; |
513 | } | |
514 | ||
515 | search_line_fast = impl; | |
516 | } | |
517 | ||
70ebee13 | 518 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) |
2431e8ba | 519 | |
520 | /* A vection of the fast scanner using AltiVec vectorized byte compares. */ | |
521 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, | |
522 | so we can't compile this function without -maltivec on the command line | |
523 | (or implied by some other switch). */ | |
524 | ||
525 | static const uchar * | |
526 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
527 | { | |
528 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
529 | ||
530 | const vc repl_nl = { | |
531 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
532 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
533 | }; | |
534 | const vc repl_cr = { | |
535 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
536 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
537 | }; | |
538 | const vc repl_bs = { | |
539 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
540 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
541 | }; | |
542 | const vc repl_qm = { | |
543 | '?', '?', '?', '?', '?', '?', '?', '?', | |
544 | '?', '?', '?', '?', '?', '?', '?', '?', | |
545 | }; | |
546 | const vc ones = { | |
547 | -1, -1, -1, -1, -1, -1, -1, -1, | |
548 | -1, -1, -1, -1, -1, -1, -1, -1, | |
549 | }; | |
550 | const vc zero = { 0 }; | |
551 | ||
552 | vc data, mask, t; | |
553 | ||
554 | /* Altivec loads automatically mask addresses with -16. This lets us | |
555 | issue the first load as early as possible. */ | |
556 | data = __builtin_vec_ld(0, (const vc *)s); | |
557 | ||
558 | /* Discard bytes before the beginning of the buffer. Do this by | |
559 | beginning with all ones and shifting in zeros according to the | |
560 | mis-alignment. The LVSR instruction pulls the exact shift we | |
561 | want from the address. */ | |
562 | mask = __builtin_vec_lvsr(0, s); | |
563 | mask = __builtin_vec_perm(zero, ones, mask); | |
564 | data &= mask; | |
565 | ||
566 | /* While altivec loads mask addresses, we still need to align S so | |
567 | that the offset we compute at the end is correct. */ | |
568 | s = (const uchar *)((uintptr_t)s & -16); | |
569 | ||
570 | /* Main loop processing 16 bytes at a time. */ | |
571 | goto start; | |
572 | do | |
573 | { | |
574 | vc m_nl, m_cr, m_bs, m_qm; | |
575 | ||
576 | s += 16; | |
577 | data = __builtin_vec_ld(0, (const vc *)s); | |
578 | ||
579 | start: | |
580 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
581 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
582 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
583 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
584 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
585 | ||
586 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
587 | characters. We want to exit the loop if any byte in T is non-zero. | |
588 | Below is the expansion of vec_any_ne(t, zero). */ | |
589 | } | |
590 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
591 | ||
592 | { | |
593 | #define N (sizeof(vc) / sizeof(long)) | |
594 | ||
2431e8ba | 595 | union { |
596 | vc v; | |
5fe44548 | 597 | /* Statically assert that N is 2 or 4. */ |
598 | unsigned long l[(N == 2 || N == 4) ? N : -1]; | |
2431e8ba | 599 | } u; |
600 | unsigned long l, i = 0; | |
601 | ||
602 | u.v = t; | |
603 | ||
604 | /* Find the first word of T that is non-zero. */ | |
605 | switch (N) | |
606 | { | |
607 | case 4: | |
608 | l = u.l[i++]; | |
609 | if (l != 0) | |
610 | break; | |
611 | s += sizeof(unsigned long); | |
612 | l = u.l[i++]; | |
613 | if (l != 0) | |
614 | break; | |
615 | s += sizeof(unsigned long); | |
616 | case 2: | |
617 | l = u.l[i++]; | |
618 | if (l != 0) | |
619 | break; | |
620 | s += sizeof(unsigned long); | |
621 | l = u.l[i]; | |
622 | } | |
623 | ||
624 | /* L now contains 0xff in bytes for which we matched one of the | |
625 | relevant characters. We can find the byte index by finding | |
626 | its bit index and dividing by 8. */ | |
627 | l = __builtin_clzl(l) >> 3; | |
628 | return s + l; | |
629 | ||
630 | #undef N | |
631 | } | |
632 | } | |
633 | ||
c434932e | 634 | #elif defined (__ARM_NEON__) |
635 | #include "arm_neon.h" | |
636 | ||
637 | static const uchar * | |
638 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
639 | { | |
640 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); | |
641 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); | |
642 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); | |
643 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); | |
644 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); | |
645 | ||
646 | unsigned int misalign, found, mask; | |
647 | const uint8_t *p; | |
648 | uint8x16_t data; | |
649 | ||
650 | /* Align the source pointer. */ | |
651 | misalign = (uintptr_t)s & 15; | |
652 | p = (const uint8_t *)((uintptr_t)s & -16); | |
653 | data = vld1q_u8 (p); | |
654 | ||
655 | /* Create a mask for the bytes that are valid within the first | |
656 | 16-byte block. The Idea here is that the AND with the mask | |
657 | within the loop is "free", since we need some AND or TEST | |
658 | insn in order to set the flags for the branch anyway. */ | |
659 | mask = (-1u << misalign) & 0xffff; | |
660 | ||
661 | /* Main loop, processing 16 bytes at a time. */ | |
662 | goto start; | |
663 | ||
664 | do | |
665 | { | |
666 | uint8x8_t l; | |
667 | uint16x4_t m; | |
668 | uint32x2_t n; | |
669 | uint8x16_t t, u, v, w; | |
670 | ||
671 | p += 16; | |
672 | data = vld1q_u8 (p); | |
673 | mask = 0xffff; | |
674 | ||
675 | start: | |
676 | t = vceqq_u8 (data, repl_nl); | |
677 | u = vceqq_u8 (data, repl_cr); | |
678 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
679 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
680 | t = vandq_u8 (vorrq_u8 (v, w), xmask); | |
681 | l = vpadd_u8 (vget_low_u8 (t), vget_high_u8 (t)); | |
682 | m = vpaddl_u8 (l); | |
683 | n = vpaddl_u16 (m); | |
684 | ||
685 | found = vget_lane_u32 ((uint32x2_t) vorr_u64 ((uint64x1_t) n, | |
686 | vshr_n_u64 ((uint64x1_t) n, 24)), 0); | |
687 | found &= mask; | |
688 | } | |
689 | while (!found); | |
690 | ||
691 | /* FOUND contains 1 in bits for which we matched a relevant | |
692 | character. Conversion to the byte index is trivial. */ | |
693 | found = __builtin_ctz (found); | |
694 | return (const uchar *)p + found; | |
695 | } | |
696 | ||
2431e8ba | 697 | #else |
698 | ||
699 | /* We only have one accellerated alternative. Use a direct call so that | |
700 | we encourage inlining. */ | |
701 | ||
702 | #define search_line_fast search_line_acc_char | |
703 | ||
704 | #endif | |
705 | ||
b735cc56 | 706 | /* Initialize the lexer if needed. */ |
707 | ||
708 | void | |
709 | _cpp_init_lexer (void) | |
710 | { | |
711 | #ifdef HAVE_init_vectorized_lexer | |
712 | init_vectorized_lexer (); | |
713 | #endif | |
714 | } | |
715 | ||
a54e0bf8 | 716 | /* Returns with a logical line that contains no escaped newlines or |
717 | trigraphs. This is a time-critical inner loop. */ | |
718 | void | |
f7fdd7a1 | 719 | _cpp_clean_line (cpp_reader *pfile) |
0578f103 | 720 | { |
a54e0bf8 | 721 | cpp_buffer *buffer; |
722 | const uchar *s; | |
723 | uchar c, *d, *p; | |
1e0ef2fd | 724 | |
a54e0bf8 | 725 | buffer = pfile->buffer; |
726 | buffer->cur_note = buffer->notes_used = 0; | |
727 | buffer->cur = buffer->line_base = buffer->next_line; | |
728 | buffer->need_line = false; | |
2431e8ba | 729 | s = buffer->next_line; |
1e0ef2fd | 730 | |
a54e0bf8 | 731 | if (!buffer->from_stage3) |
0578f103 | 732 | { |
5008f5c5 | 733 | const uchar *pbackslash = NULL; |
734 | ||
2431e8ba | 735 | /* Fast path. This is the common case of an un-escaped line with |
54d3be91 | 736 | no trigraphs. The primary win here is by not writing any |
737 | data back to memory until we have to. */ | |
2431e8ba | 738 | while (1) |
54d3be91 | 739 | { |
2431e8ba | 740 | /* Perform an optimized search for \n, \r, \\, ?. */ |
741 | s = search_line_fast (s, buffer->rlimit); | |
54d3be91 | 742 | |
2431e8ba | 743 | c = *s; |
744 | if (c == '\\') | |
745 | { | |
746 | /* Record the location of the backslash and continue. */ | |
747 | pbackslash = s++; | |
54d3be91 | 748 | } |
2431e8ba | 749 | else if (__builtin_expect (c == '?', 0)) |
54d3be91 | 750 | { |
2431e8ba | 751 | if (__builtin_expect (s[1] == '?', false) |
752 | && _cpp_trigraph_map[s[2]]) | |
54d3be91 | 753 | { |
2431e8ba | 754 | /* Have a trigraph. We may or may not have to convert |
755 | it. Add a line note regardless, for -Wtrigraphs. */ | |
756 | add_line_note (buffer, s, s[2]); | |
757 | if (CPP_OPTION (pfile, trigraphs)) | |
758 | { | |
759 | /* We do, and that means we have to switch to the | |
760 | slow path. */ | |
761 | d = (uchar *) s; | |
762 | *d = _cpp_trigraph_map[s[2]]; | |
763 | s += 2; | |
764 | goto slow_path; | |
765 | } | |
54d3be91 | 766 | } |
2431e8ba | 767 | /* Not a trigraph. Continue on fast-path. */ |
768 | s++; | |
54d3be91 | 769 | } |
2431e8ba | 770 | else |
771 | break; | |
54d3be91 | 772 | } |
773 | ||
2431e8ba | 774 | /* This must be \r or \n. We're either done, or we'll be forced |
775 | to write back to the buffer and continue on the slow path. */ | |
776 | d = (uchar *) s; | |
777 | ||
778 | if (__builtin_expect (s == buffer->rlimit, false)) | |
779 | goto done; | |
780 | ||
781 | /* DOS line ending? */ | |
782 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
783 | { | |
784 | s++; | |
785 | if (s == buffer->rlimit) | |
786 | goto done; | |
787 | } | |
788 | ||
789 | if (__builtin_expect (pbackslash == NULL, true)) | |
790 | goto done; | |
791 | ||
792 | /* Check for escaped newline. */ | |
793 | p = d; | |
794 | while (is_nvspace (p[-1])) | |
795 | p--; | |
796 | if (p - 1 != pbackslash) | |
797 | goto done; | |
798 | ||
799 | /* Have an escaped newline; process it and proceed to | |
800 | the slow path. */ | |
801 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
802 | d = p - 2; | |
803 | buffer->next_line = p - 1; | |
a54e0bf8 | 804 | |
2431e8ba | 805 | slow_path: |
806 | while (1) | |
4b912310 | 807 | { |
a54e0bf8 | 808 | c = *++s; |
809 | *++d = c; | |
810 | ||
811 | if (c == '\n' || c == '\r') | |
812 | { | |
2431e8ba | 813 | /* Handle DOS line endings. */ |
a54e0bf8 | 814 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
815 | s++; | |
816 | if (s == buffer->rlimit) | |
817 | break; | |
818 | ||
819 | /* Escaped? */ | |
820 | p = d; | |
821 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
822 | p--; | |
823 | if (p == buffer->next_line || p[-1] != '\\') | |
824 | break; | |
825 | ||
aad4a87f | 826 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
a54e0bf8 | 827 | d = p - 2; |
828 | buffer->next_line = p - 1; | |
829 | } | |
830 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
831 | { | |
832 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
aad4a87f | 833 | add_line_note (buffer, d, s[2]); |
a54e0bf8 | 834 | if (CPP_OPTION (pfile, trigraphs)) |
835 | { | |
836 | *d = _cpp_trigraph_map[s[2]]; | |
837 | s += 2; | |
838 | } | |
839 | } | |
4b912310 | 840 | } |
0578f103 | 841 | } |
a54e0bf8 | 842 | else |
843 | { | |
2431e8ba | 844 | while (*s != '\n' && *s != '\r') |
a54e0bf8 | 845 | s++; |
a54e0bf8 | 846 | d = (uchar *) s; |
847 | ||
848 | /* Handle DOS line endings. */ | |
849 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
850 | s++; | |
851 | } | |
338fa5f7 | 852 | |
54d3be91 | 853 | done: |
a54e0bf8 | 854 | *d = '\n'; |
aad4a87f | 855 | /* A sentinel note that should never be processed. */ |
856 | add_line_note (buffer, d + 1, '\n'); | |
a54e0bf8 | 857 | buffer->next_line = s + 1; |
0578f103 | 858 | } |
859 | ||
3078f2b2 | 860 | /* Return true if the trigraph indicated by NOTE should be warned |
861 | about in a comment. */ | |
862 | static bool | |
f7fdd7a1 | 863 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
3078f2b2 | 864 | { |
865 | const uchar *p; | |
866 | ||
867 | /* Within comments we don't warn about trigraphs, unless the | |
868 | trigraph forms an escaped newline, as that may change | |
7ef5b942 | 869 | behavior. */ |
3078f2b2 | 870 | if (note->type != '/') |
871 | return false; | |
872 | ||
873 | /* If -trigraphs, then this was an escaped newline iff the next note | |
874 | is coincident. */ | |
875 | if (CPP_OPTION (pfile, trigraphs)) | |
876 | return note[1].pos == note->pos; | |
877 | ||
878 | /* Otherwise, see if this forms an escaped newline. */ | |
879 | p = note->pos + 3; | |
880 | while (is_nvspace (*p)) | |
881 | p++; | |
882 | ||
883 | /* There might have been escaped newlines between the trigraph and the | |
884 | newline we found. Hence the position test. */ | |
885 | return (*p == '\n' && p < note[1].pos); | |
886 | } | |
887 | ||
a54e0bf8 | 888 | /* Process the notes created by add_line_note as far as the current |
889 | location. */ | |
890 | void | |
f7fdd7a1 | 891 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
0578f103 | 892 | { |
c808d026 | 893 | cpp_buffer *buffer = pfile->buffer; |
894 | ||
a54e0bf8 | 895 | for (;;) |
f80e83a9 | 896 | { |
a54e0bf8 | 897 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
898 | unsigned int col; | |
396ffa86 | 899 | |
a54e0bf8 | 900 | if (note->pos > buffer->cur) |
901 | break; | |
396ffa86 | 902 | |
a54e0bf8 | 903 | buffer->cur_note++; |
904 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
435fb09b | 905 | |
aad4a87f | 906 | if (note->type == '\\' || note->type == ' ') |
a54e0bf8 | 907 | { |
aad4a87f | 908 | if (note->type == ' ' && !in_comment) |
dbddc569 | 909 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
a54e0bf8 | 910 | "backslash and newline separated by space"); |
aad4a87f | 911 | |
a54e0bf8 | 912 | if (buffer->next_line > buffer->rlimit) |
1e0ef2fd | 913 | { |
dbddc569 | 914 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
a54e0bf8 | 915 | "backslash-newline at end of file"); |
916 | /* Prevent "no newline at end of file" warning. */ | |
917 | buffer->next_line = buffer->rlimit; | |
1e0ef2fd | 918 | } |
a54e0bf8 | 919 | |
920 | buffer->line_base = note->pos; | |
610625e3 | 921 | CPP_INCREMENT_LINE (pfile, 0); |
338fa5f7 | 922 | } |
aad4a87f | 923 | else if (_cpp_trigraph_map[note->type]) |
924 | { | |
3078f2b2 | 925 | if (CPP_OPTION (pfile, warn_trigraphs) |
926 | && (!in_comment || warn_in_comment (pfile, note))) | |
aad4a87f | 927 | { |
928 | if (CPP_OPTION (pfile, trigraphs)) | |
3a79f5da | 929 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
930 | pfile->line_table->highest_line, col, | |
931 | "trigraph ??%c converted to %c", | |
932 | note->type, | |
933 | (int) _cpp_trigraph_map[note->type]); | |
aad4a87f | 934 | else |
1542b1ef | 935 | { |
3a79f5da | 936 | cpp_warning_with_line |
937 | (pfile, CPP_W_TRIGRAPHS, | |
938 | pfile->line_table->highest_line, col, | |
1542b1ef | 939 | "trigraph ??%c ignored, use -trigraphs to enable", |
940 | note->type); | |
941 | } | |
aad4a87f | 942 | } |
943 | } | |
3a45011c | 944 | else if (note->type == 0) |
945 | /* Already processed in lex_raw_string. */; | |
aad4a87f | 946 | else |
947 | abort (); | |
f80e83a9 | 948 | } |
0578f103 | 949 | } |
950 | ||
338fa5f7 | 951 | /* Skip a C-style block comment. We find the end of the comment by |
952 | seeing if an asterisk is before every '/' we encounter. Returns | |
edaf8cb5 | 953 | nonzero if comment terminated by EOF, zero otherwise. |
954 | ||
955 | Buffer->cur points to the initial asterisk of the comment. */ | |
a54e0bf8 | 956 | bool |
f7fdd7a1 | 957 | _cpp_skip_block_comment (cpp_reader *pfile) |
0578f103 | 958 | { |
f80e83a9 | 959 | cpp_buffer *buffer = pfile->buffer; |
54d3be91 | 960 | const uchar *cur = buffer->cur; |
961 | uchar c; | |
338fa5f7 | 962 | |
54d3be91 | 963 | cur++; |
964 | if (*cur == '/') | |
965 | cur++; | |
338fa5f7 | 966 | |
a54e0bf8 | 967 | for (;;) |
968 | { | |
338fa5f7 | 969 | /* People like decorating comments with '*', so check for '/' |
970 | instead for efficiency. */ | |
54d3be91 | 971 | c = *cur++; |
972 | ||
f80e83a9 | 973 | if (c == '/') |
0578f103 | 974 | { |
54d3be91 | 975 | if (cur[-2] == '*') |
338fa5f7 | 976 | break; |
f80e83a9 | 977 | |
338fa5f7 | 978 | /* Warn about potential nested comments, but not if the '/' |
3fb1e43b | 979 | comes immediately before the true comment delimiter. |
f80e83a9 | 980 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 981 | if (CPP_OPTION (pfile, warn_comments) |
54d3be91 | 982 | && cur[0] == '*' && cur[1] != '/') |
983 | { | |
984 | buffer->cur = cur; | |
3a79f5da | 985 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
986 | pfile->line_table->highest_line, | |
987 | CPP_BUF_COL (buffer), | |
988 | "\"/*\" within comment"); | |
54d3be91 | 989 | } |
0578f103 | 990 | } |
a54e0bf8 | 991 | else if (c == '\n') |
992 | { | |
610625e3 | 993 | unsigned int cols; |
54d3be91 | 994 | buffer->cur = cur - 1; |
a54e0bf8 | 995 | _cpp_process_line_notes (pfile, true); |
996 | if (buffer->next_line >= buffer->rlimit) | |
997 | return true; | |
998 | _cpp_clean_line (pfile); | |
610625e3 | 999 | |
1000 | cols = buffer->next_line - buffer->line_base; | |
1001 | CPP_INCREMENT_LINE (pfile, cols); | |
1002 | ||
54d3be91 | 1003 | cur = buffer->cur; |
a54e0bf8 | 1004 | } |
0578f103 | 1005 | } |
f80e83a9 | 1006 | |
54d3be91 | 1007 | buffer->cur = cur; |
3078f2b2 | 1008 | _cpp_process_line_notes (pfile, true); |
a54e0bf8 | 1009 | return false; |
0578f103 | 1010 | } |
1011 | ||
1c124f85 | 1012 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
d10cfa8d | 1013 | terminating newline. Handles escaped newlines. Returns nonzero |
1c124f85 | 1014 | if a multiline comment. */ |
f80e83a9 | 1015 | static int |
f7fdd7a1 | 1016 | skip_line_comment (cpp_reader *pfile) |
0578f103 | 1017 | { |
f669338a | 1018 | cpp_buffer *buffer = pfile->buffer; |
4999c35b | 1019 | source_location orig_line = pfile->line_table->highest_line; |
f80e83a9 | 1020 | |
a54e0bf8 | 1021 | while (*buffer->cur != '\n') |
1022 | buffer->cur++; | |
1c124f85 | 1023 | |
a54e0bf8 | 1024 | _cpp_process_line_notes (pfile, true); |
dbddc569 | 1025 | return orig_line != pfile->line_table->highest_line; |
f80e83a9 | 1026 | } |
0578f103 | 1027 | |
a54e0bf8 | 1028 | /* Skips whitespace, saving the next non-whitespace character. */ |
b86584f6 | 1029 | static void |
f7fdd7a1 | 1030 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
f80e83a9 | 1031 | { |
1032 | cpp_buffer *buffer = pfile->buffer; | |
fe9eb18b | 1033 | bool saw_NUL = false; |
0578f103 | 1034 | |
338fa5f7 | 1035 | do |
f80e83a9 | 1036 | { |
78719282 | 1037 | /* Horizontal space always OK. */ |
a54e0bf8 | 1038 | if (c == ' ' || c == '\t') |
338fa5f7 | 1039 | ; |
338fa5f7 | 1040 | /* Just \f \v or \0 left. */ |
78719282 | 1041 | else if (c == '\0') |
fe9eb18b | 1042 | saw_NUL = true; |
79bd622b | 1043 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
dbddc569 | 1044 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
73328dce | 1045 | CPP_BUF_COL (buffer), |
1046 | "%s in preprocessing directive", | |
1047 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 1048 | |
338fa5f7 | 1049 | c = *buffer->cur++; |
0578f103 | 1050 | } |
2c0e001b | 1051 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
338fa5f7 | 1052 | while (is_nvspace (c)); |
1053 | ||
fe9eb18b | 1054 | if (saw_NUL) |
d80d2074 | 1055 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
fe9eb18b | 1056 | |
1c124f85 | 1057 | buffer->cur--; |
f80e83a9 | 1058 | } |
0578f103 | 1059 | |
79bd622b | 1060 | /* See if the characters of a number token are valid in a name (no |
1061 | '.', '+' or '-'). */ | |
1062 | static int | |
f7fdd7a1 | 1063 | name_p (cpp_reader *pfile, const cpp_string *string) |
79bd622b | 1064 | { |
1065 | unsigned int i; | |
1066 | ||
1067 | for (i = 0; i < string->len; i++) | |
1068 | if (!is_idchar (string->text[i])) | |
1069 | return 0; | |
1070 | ||
b1a9ff83 | 1071 | return 1; |
79bd622b | 1072 | } |
1073 | ||
bce47149 | 1074 | /* After parsing an identifier or other sequence, produce a warning about |
1075 | sequences not in NFC/NFKC. */ | |
1076 | static void | |
1077 | warn_about_normalization (cpp_reader *pfile, | |
1078 | const cpp_token *token, | |
1079 | const struct normalize_state *s) | |
1080 | { | |
1081 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1082 | && !pfile->state.skipping) | |
1083 | { | |
1084 | /* Make sure that the token is printed using UCNs, even | |
1085 | if we'd otherwise happily print UTF-8. */ | |
720aca92 | 1086 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
bce47149 | 1087 | size_t sz; |
1088 | ||
1089 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1090 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
3a79f5da | 1091 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1092 | "`%.*s' is not in NFKC", (int) sz, buf); | |
bce47149 | 1093 | else |
3a79f5da | 1094 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1095 | "`%.*s' is not in NFC", (int) sz, buf); | |
15fc692a | 1096 | free (buf); |
bce47149 | 1097 | } |
1098 | } | |
1099 | ||
5bb46c08 | 1100 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in |
2cbf1359 | 1101 | an identifier. FIRST is TRUE if this starts an identifier. */ |
5bb46c08 | 1102 | static bool |
bce47149 | 1103 | forms_identifier_p (cpp_reader *pfile, int first, |
1104 | struct normalize_state *state) | |
5bb46c08 | 1105 | { |
2cbf1359 | 1106 | cpp_buffer *buffer = pfile->buffer; |
1107 | ||
1108 | if (*buffer->cur == '$') | |
1109 | { | |
1110 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1111 | return false; | |
1112 | ||
1113 | buffer->cur++; | |
f0c2775b | 1114 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
2cbf1359 | 1115 | { |
f0c2775b | 1116 | CPP_OPTION (pfile, warn_dollars) = 0; |
d80d2074 | 1117 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
2cbf1359 | 1118 | } |
1119 | ||
1120 | return true; | |
1121 | } | |
5bb46c08 | 1122 | |
2cbf1359 | 1123 | /* Is this a syntactically valid UCN? */ |
865c4e44 | 1124 | if (CPP_OPTION (pfile, extended_identifiers) |
4e9d1e6d | 1125 | && *buffer->cur == '\\' |
2cbf1359 | 1126 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) |
5bb46c08 | 1127 | { |
2cbf1359 | 1128 | buffer->cur += 2; |
bce47149 | 1129 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1130 | state)) | |
2cbf1359 | 1131 | return true; |
1132 | buffer->cur -= 2; | |
5bb46c08 | 1133 | } |
5bb46c08 | 1134 | |
2cbf1359 | 1135 | return false; |
5bb46c08 | 1136 | } |
1137 | ||
038c21f1 | 1138 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1139 | static cpp_hashnode * | |
1140 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1141 | { | |
1142 | cpp_hashnode *result; | |
1143 | const uchar *cur; | |
1144 | unsigned int len; | |
1145 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1146 | ||
1147 | cur = base + 1; | |
1148 | while (ISIDNUM (*cur)) | |
1149 | { | |
1150 | hash = HT_HASHSTEP (hash, *cur); | |
1151 | cur++; | |
1152 | } | |
1153 | len = cur - base; | |
1154 | hash = HT_HASHFINISH (hash, len); | |
1155 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1156 | base, len, hash, HT_ALLOC)); | |
1157 | ||
1158 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1159 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1160 | && !pfile->state.skipping, 0)) | |
1161 | { | |
1162 | /* It is allowed to poison the same identifier twice. */ | |
1163 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1164 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1165 | NODE_NAME (result)); | |
1166 | ||
1167 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1168 | replacement list of a variadic macro. */ | |
1169 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1170 | && !pfile->state.va_args_ok) | |
1171 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1172 | "__VA_ARGS__ can only appear in the expansion" | |
1173 | " of a C99 variadic macro"); | |
1174 | ||
1175 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1176 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1177 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1178 | "identifier \"%s\" is a special operator name in C++", | |
1179 | NODE_NAME (result)); | |
038c21f1 | 1180 | } |
1181 | ||
1182 | return result; | |
1183 | } | |
1184 | ||
1185 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1186 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1187 | cpp_hashnode * | |
1188 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1189 | { | |
1190 | cpp_hashnode *result; | |
1191 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1192 | return result; | |
1193 | } | |
1194 | ||
5bb46c08 | 1195 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
338fa5f7 | 1196 | static cpp_hashnode * |
bce47149 | 1197 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
1198 | struct normalize_state *nst) | |
0578f103 | 1199 | { |
79bd622b | 1200 | cpp_hashnode *result; |
bb1fa6bb | 1201 | const uchar *cur; |
3eb3f293 | 1202 | unsigned int len; |
1203 | unsigned int hash = HT_HASHSTEP (0, *base); | |
66a5287e | 1204 | |
3eb3f293 | 1205 | cur = pfile->buffer->cur; |
bb1fa6bb | 1206 | if (! starts_ucn) |
1207 | while (ISIDNUM (*cur)) | |
1208 | { | |
1209 | hash = HT_HASHSTEP (hash, *cur); | |
1210 | cur++; | |
1211 | } | |
1212 | pfile->buffer->cur = cur; | |
bce47149 | 1213 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
78a11351 | 1214 | { |
bb1fa6bb | 1215 | /* Slower version for identifiers containing UCNs (or $). */ |
1216 | do { | |
1217 | while (ISIDNUM (*pfile->buffer->cur)) | |
bce47149 | 1218 | { |
1219 | pfile->buffer->cur++; | |
1220 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1221 | } | |
1222 | } while (forms_identifier_p (pfile, false, nst)); | |
bb1fa6bb | 1223 | result = _cpp_interpret_identifier (pfile, base, |
1224 | pfile->buffer->cur - base); | |
66a5287e | 1225 | } |
bb1fa6bb | 1226 | else |
1227 | { | |
1228 | len = cur - base; | |
1229 | hash = HT_HASHFINISH (hash, len); | |
5bb46c08 | 1230 | |
e297899b | 1231 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1232 | base, len, hash, HT_ALLOC)); | |
bb1fa6bb | 1233 | } |
66a5287e | 1234 | |
5bb46c08 | 1235 | /* Rarely, identifiers require diagnostics when lexed. */ |
66a5287e | 1236 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1237 | && !pfile->state.skipping, 0)) | |
1238 | { | |
1239 | /* It is allowed to poison the same identifier twice. */ | |
1240 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
d80d2074 | 1241 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
66a5287e | 1242 | NODE_NAME (result)); |
1243 | ||
1244 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1245 | replacement list of a variadic macro. */ | |
1246 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1247 | && !pfile->state.va_args_ok) | |
d80d2074 | 1248 | cpp_error (pfile, CPP_DL_PEDWARN, |
f7fdd7a1 | 1249 | "__VA_ARGS__ can only appear in the expansion" |
1250 | " of a C99 variadic macro"); | |
2a6a6991 | 1251 | |
1252 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1253 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1254 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1255 | "identifier \"%s\" is a special operator name in C++", | |
1256 | NODE_NAME (result)); | |
66a5287e | 1257 | } |
1258 | ||
1259 | return result; | |
1260 | } | |
1261 | ||
5bb46c08 | 1262 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
0578f103 | 1263 | static void |
bce47149 | 1264 | lex_number (cpp_reader *pfile, cpp_string *number, |
1265 | struct normalize_state *nst) | |
0578f103 | 1266 | { |
b6d18b0a | 1267 | const uchar *cur; |
5bb46c08 | 1268 | const uchar *base; |
1269 | uchar *dest; | |
0578f103 | 1270 | |
5bb46c08 | 1271 | base = pfile->buffer->cur - 1; |
1272 | do | |
f80e83a9 | 1273 | { |
5bb46c08 | 1274 | cur = pfile->buffer->cur; |
338fa5f7 | 1275 | |
5bb46c08 | 1276 | /* N.B. ISIDNUM does not include $. */ |
1277 | while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1])) | |
bce47149 | 1278 | { |
1279 | cur++; | |
1280 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1281 | } | |
0578f103 | 1282 | |
78a11351 | 1283 | pfile->buffer->cur = cur; |
0578f103 | 1284 | } |
bce47149 | 1285 | while (forms_identifier_p (pfile, false, nst)); |
79bd622b | 1286 | |
5bb46c08 | 1287 | number->len = cur - base; |
1288 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1289 | memcpy (dest, base, number->len); | |
1290 | dest[number->len] = '\0'; | |
1291 | number->text = dest; | |
79bd622b | 1292 | } |
1293 | ||
4970d4c2 | 1294 | /* Create a token of type TYPE with a literal spelling. */ |
1295 | static void | |
f7fdd7a1 | 1296 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1297 | unsigned int len, enum cpp_ttype type) | |
4970d4c2 | 1298 | { |
1299 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1300 | ||
1301 | memcpy (dest, base, len); | |
1302 | dest[len] = '\0'; | |
1303 | token->type = type; | |
1304 | token->val.str.len = len; | |
1305 | token->val.str.text = dest; | |
1306 | } | |
1307 | ||
3a45011c | 1308 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1309 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1310 | ||
1311 | static void | |
1312 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1313 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1314 | { | |
1315 | _cpp_buff *first_buff = *first_buff_p; | |
1316 | _cpp_buff *last_buff = *last_buff_p; | |
1317 | ||
1318 | if (first_buff == NULL) | |
1319 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1320 | else if (len > BUFF_ROOM (last_buff)) | |
1321 | { | |
1322 | size_t room = BUFF_ROOM (last_buff); | |
1323 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1324 | BUFF_FRONT (last_buff) += room; | |
1325 | base += room; | |
1326 | len -= room; | |
1327 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1328 | } | |
1329 | ||
1330 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1331 | BUFF_FRONT (last_buff) += len; | |
1332 | ||
1333 | *first_buff_p = first_buff; | |
1334 | *last_buff_p = last_buff; | |
1335 | } | |
1336 | ||
538ba11a | 1337 | /* Lexes a raw string. The stored string contains the spelling, including |
3a45011c | 1338 | double quotes, delimiter string, '(' and ')', any leading |
538ba11a | 1339 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1340 | literal, or CPP_OTHER if it was not properly terminated. | |
1341 | ||
1342 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1343 | is the first NUL since embedded NULs are preserved. */ | |
1344 | ||
1345 | static void | |
1346 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1347 | const uchar *cur) | |
1348 | { | |
538ba11a | 1349 | const uchar *raw_prefix; |
1350 | unsigned int raw_prefix_len = 0; | |
1351 | enum cpp_ttype type; | |
1352 | size_t total_len = 0; | |
1353 | _cpp_buff *first_buff = NULL, *last_buff = NULL; | |
3a45011c | 1354 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1355 | |
1356 | type = (*base == 'L' ? CPP_WSTRING : | |
1357 | *base == 'U' ? CPP_STRING32 : | |
1358 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1359 | : CPP_STRING); | |
1360 | ||
1361 | raw_prefix = cur + 1; | |
1362 | while (raw_prefix_len < 16) | |
1363 | { | |
1364 | switch (raw_prefix[raw_prefix_len]) | |
1365 | { | |
7fe0ef3a | 1366 | case ' ': case '(': case ')': case '\\': case '\t': |
538ba11a | 1367 | case '\v': case '\f': case '\n': default: |
1368 | break; | |
1369 | /* Basic source charset except the above chars. */ | |
1370 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1371 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1372 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1373 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1374 | case 'y': case 'z': | |
1375 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1376 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1377 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1378 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1379 | case 'Y': case 'Z': | |
1380 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1381 | case '6': case '7': case '8': case '9': | |
7fe0ef3a | 1382 | case '_': case '{': case '}': case '#': case '[': case ']': |
538ba11a | 1383 | case '<': case '>': case '%': case ':': case ';': case '.': |
1384 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1385 | case '&': case '|': case '~': case '!': case '=': case ',': | |
7fe0ef3a | 1386 | case '"': case '\'': |
538ba11a | 1387 | raw_prefix_len++; |
1388 | continue; | |
1389 | } | |
1390 | break; | |
1391 | } | |
1392 | ||
7fe0ef3a | 1393 | if (raw_prefix[raw_prefix_len] != '(') |
538ba11a | 1394 | { |
1395 | int col = CPP_BUF_COLUMN (pfile->buffer, raw_prefix + raw_prefix_len) | |
1396 | + 1; | |
1397 | if (raw_prefix_len == 16) | |
1398 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1399 | "raw string delimiter longer than 16 characters"); | |
1400 | else | |
1401 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1402 | "invalid character '%c' in raw string delimiter", | |
1403 | (int) raw_prefix[raw_prefix_len]); | |
1404 | pfile->buffer->cur = raw_prefix - 1; | |
1405 | create_literal (pfile, token, base, raw_prefix - 1 - base, CPP_OTHER); | |
1406 | return; | |
1407 | } | |
1408 | ||
1409 | cur = raw_prefix + raw_prefix_len + 1; | |
1410 | for (;;) | |
1411 | { | |
3a45011c | 1412 | #define BUF_APPEND(STR,LEN) \ |
1413 | do { \ | |
1414 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1415 | &first_buff, &last_buff); \ | |
1416 | total_len += (LEN); \ | |
1417 | } while (0); | |
1418 | ||
1419 | cppchar_t c; | |
1420 | ||
1421 | /* If we previously performed any trigraph or line splicing | |
1422 | transformations, undo them within the body of the raw string. */ | |
1423 | while (note->pos < cur) | |
1424 | ++note; | |
1425 | for (; note->pos == cur; ++note) | |
1426 | { | |
1427 | switch (note->type) | |
1428 | { | |
1429 | case '\\': | |
1430 | case ' ': | |
1431 | /* Restore backslash followed by newline. */ | |
1432 | BUF_APPEND (base, cur - base); | |
1433 | base = cur; | |
1434 | BUF_APPEND ("\\", 1); | |
1435 | after_backslash: | |
1436 | if (note->type == ' ') | |
1437 | { | |
1438 | /* GNU backslash whitespace newline extension. FIXME | |
1439 | could be any sequence of non-vertical space. When we | |
1440 | can properly restore any such sequence, we should mark | |
1441 | this note as handled so _cpp_process_line_notes | |
1442 | doesn't warn. */ | |
1443 | BUF_APPEND (" ", 1); | |
1444 | } | |
1445 | ||
1446 | BUF_APPEND ("\n", 1); | |
1447 | break; | |
1448 | ||
1449 | case 0: | |
1450 | /* Already handled. */ | |
1451 | break; | |
1452 | ||
1453 | default: | |
1454 | if (_cpp_trigraph_map[note->type]) | |
1455 | { | |
1456 | /* Don't warn about this trigraph in | |
1457 | _cpp_process_line_notes, since trigraphs show up as | |
1458 | trigraphs in raw strings. */ | |
56e2ce2d | 1459 | uchar type = note->type; |
3a45011c | 1460 | note->type = 0; |
1461 | ||
1462 | if (!CPP_OPTION (pfile, trigraphs)) | |
1463 | /* If we didn't convert the trigraph in the first | |
1464 | place, don't do anything now either. */ | |
1465 | break; | |
1466 | ||
1467 | BUF_APPEND (base, cur - base); | |
1468 | base = cur; | |
1469 | BUF_APPEND ("??", 2); | |
1470 | ||
1471 | /* ??/ followed by newline gets two line notes, one for | |
1472 | the trigraph and one for the backslash/newline. */ | |
1473 | if (type == '/' && note[1].pos == cur) | |
1474 | { | |
1475 | if (note[1].type != '\\' | |
1476 | && note[1].type != ' ') | |
1477 | abort (); | |
1478 | BUF_APPEND ("/", 1); | |
1479 | ++note; | |
1480 | goto after_backslash; | |
1481 | } | |
1482 | /* The ) from ??) could be part of the suffix. */ | |
1483 | else if (type == ')' | |
1484 | && strncmp ((const char *) cur+1, | |
1485 | (const char *) raw_prefix, | |
1486 | raw_prefix_len) == 0 | |
1487 | && cur[raw_prefix_len+1] == '"') | |
1488 | { | |
0ff0cd04 | 1489 | BUF_APPEND (")", 1); |
1490 | base++; | |
1491 | cur += raw_prefix_len + 2; | |
3a45011c | 1492 | goto break_outer_loop; |
1493 | } | |
1494 | else | |
1495 | { | |
1496 | /* Skip the replacement character. */ | |
1497 | base = ++cur; | |
1498 | BUF_APPEND (&type, 1); | |
1499 | } | |
1500 | } | |
1501 | else | |
1502 | abort (); | |
1503 | break; | |
1504 | } | |
1505 | } | |
1506 | c = *cur++; | |
538ba11a | 1507 | |
7fe0ef3a | 1508 | if (c == ')' |
538ba11a | 1509 | && strncmp ((const char *) cur, (const char *) raw_prefix, |
1510 | raw_prefix_len) == 0 | |
1511 | && cur[raw_prefix_len] == '"') | |
1512 | { | |
1513 | cur += raw_prefix_len + 1; | |
1514 | break; | |
1515 | } | |
1516 | else if (c == '\n') | |
1517 | { | |
1518 | if (pfile->state.in_directive | |
1519 | || pfile->state.parsing_args | |
1520 | || pfile->state.in_deferred_pragma) | |
1521 | { | |
1522 | cur--; | |
1523 | type = CPP_OTHER; | |
1524 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1525 | "unterminated raw string"); | |
1526 | break; | |
1527 | } | |
1528 | ||
3a45011c | 1529 | BUF_APPEND (base, cur - base); |
538ba11a | 1530 | |
1531 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1532 | CPP_INCREMENT_LINE (pfile, 0); | |
1533 | pfile->buffer->need_line = true; | |
1534 | ||
3a45011c | 1535 | pfile->buffer->cur = cur-1; |
1536 | _cpp_process_line_notes (pfile, false); | |
538ba11a | 1537 | if (!_cpp_get_fresh_line (pfile)) |
1538 | { | |
1539 | source_location src_loc = token->src_loc; | |
1540 | token->type = CPP_EOF; | |
1541 | /* Tell the compiler the line number of the EOF token. */ | |
1542 | token->src_loc = pfile->line_table->highest_line; | |
1543 | token->flags = BOL; | |
1544 | if (first_buff != NULL) | |
1545 | _cpp_release_buff (pfile, first_buff); | |
1546 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1547 | "unterminated raw string"); | |
1548 | return; | |
1549 | } | |
1550 | ||
1551 | cur = base = pfile->buffer->cur; | |
3a45011c | 1552 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1553 | } |
538ba11a | 1554 | } |
3a45011c | 1555 | break_outer_loop: |
538ba11a | 1556 | |
244db24d | 1557 | if (CPP_OPTION (pfile, user_literals)) |
1558 | { | |
76d340ac | 1559 | /* According to C++11 [lex.ext]p10, a ud-suffix not starting with an |
1560 | underscore is ill-formed. Since this breaks programs using macros | |
1561 | from inttypes.h, we generate a warning and treat the ud-suffix as a | |
1562 | separate preprocessing token. This approach is under discussion by | |
1563 | the standards committee, and has been adopted as a conforming | |
911b08c6 | 1564 | extension by other front ends such as clang. |
1565 | A special exception is made for the suffix 's' which will be | |
1566 | standardized as a user-defined literal suffix for strings. */ | |
1567 | if (ISALPHA (*cur) && *cur != 's') | |
76d340ac | 1568 | { |
86c82b61 | 1569 | /* Raise a warning, but do not consume subsequent tokens. */ |
76d340ac | 1570 | if (CPP_OPTION (pfile, warn_literal_suffix)) |
1571 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, | |
1572 | token->src_loc, 0, | |
1573 | "invalid suffix on literal; C++11 requires " | |
1574 | "a space between literal and identifier"); | |
1575 | } | |
244db24d | 1576 | /* Grab user defined literal suffix. */ |
911b08c6 | 1577 | else if (ISIDST (*cur)) |
244db24d | 1578 | { |
1579 | type = cpp_userdef_string_add_type (type); | |
1580 | ++cur; | |
76d340ac | 1581 | |
1582 | while (ISIDNUM (*cur)) | |
1583 | ++cur; | |
244db24d | 1584 | } |
244db24d | 1585 | } |
1586 | ||
538ba11a | 1587 | pfile->buffer->cur = cur; |
1588 | if (first_buff == NULL) | |
1589 | create_literal (pfile, token, base, cur - base, type); | |
1590 | else | |
1591 | { | |
1592 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1593 | ||
1594 | token->type = type; | |
1595 | token->val.str.len = total_len + (cur - base); | |
1596 | token->val.str.text = dest; | |
1597 | last_buff = first_buff; | |
1598 | while (last_buff != NULL) | |
1599 | { | |
1600 | memcpy (dest, last_buff->base, | |
1601 | BUFF_FRONT (last_buff) - last_buff->base); | |
1602 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1603 | last_buff = last_buff->next; | |
1604 | } | |
1605 | _cpp_release_buff (pfile, first_buff); | |
1606 | memcpy (dest, base, cur - base); | |
1607 | dest[cur - base] = '\0'; | |
1608 | } | |
1609 | } | |
1610 | ||
5bb46c08 | 1611 | /* Lexes a string, character constant, or angle-bracketed header file |
4970d4c2 | 1612 | name. The stored string contains the spelling, including opening |
538ba11a | 1613 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1614 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1615 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1616 | header name which must be relexed as normal tokens. | |
4970d4c2 | 1617 | |
1618 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1619 | is the first NUL since embedded NULs are preserved. */ | |
f80e83a9 | 1620 | static void |
f7fdd7a1 | 1621 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
0578f103 | 1622 | { |
4970d4c2 | 1623 | bool saw_NUL = false; |
1624 | const uchar *cur; | |
5bb46c08 | 1625 | cppchar_t terminator; |
4970d4c2 | 1626 | enum cpp_ttype type; |
1627 | ||
1628 | cur = base; | |
1629 | terminator = *cur++; | |
538ba11a | 1630 | if (terminator == 'L' || terminator == 'U') |
4970d4c2 | 1631 | terminator = *cur++; |
538ba11a | 1632 | else if (terminator == 'u') |
1633 | { | |
1634 | terminator = *cur++; | |
1635 | if (terminator == '8') | |
1636 | terminator = *cur++; | |
1637 | } | |
1638 | if (terminator == 'R') | |
1639 | { | |
1640 | lex_raw_string (pfile, token, base, cur); | |
1641 | return; | |
1642 | } | |
1643 | if (terminator == '"') | |
924bbf02 | 1644 | type = (*base == 'L' ? CPP_WSTRING : |
1645 | *base == 'U' ? CPP_STRING32 : | |
538ba11a | 1646 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
1647 | : CPP_STRING); | |
4970d4c2 | 1648 | else if (terminator == '\'') |
924bbf02 | 1649 | type = (*base == 'L' ? CPP_WCHAR : |
1650 | *base == 'U' ? CPP_CHAR32 : | |
1651 | *base == 'u' ? CPP_CHAR16 : CPP_CHAR); | |
4970d4c2 | 1652 | else |
1653 | terminator = '>', type = CPP_HEADER_NAME; | |
79bd622b | 1654 | |
338fa5f7 | 1655 | for (;;) |
0578f103 | 1656 | { |
4970d4c2 | 1657 | cppchar_t c = *cur++; |
4b0c16ee | 1658 | |
edaf8cb5 | 1659 | /* In #include-style directives, terminators are not escapable. */ |
4970d4c2 | 1660 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
1661 | cur++; | |
1662 | else if (c == terminator) | |
5bb46c08 | 1663 | break; |
4970d4c2 | 1664 | else if (c == '\n') |
338fa5f7 | 1665 | { |
4970d4c2 | 1666 | cur--; |
7811eab5 | 1667 | /* Unmatched quotes always yield undefined behavior, but |
1668 | greedy lexing means that what appears to be an unterminated | |
1669 | header name may actually be a legitimate sequence of tokens. */ | |
1670 | if (terminator == '>') | |
1671 | { | |
1672 | token->type = CPP_LESS; | |
1673 | return; | |
1674 | } | |
4970d4c2 | 1675 | type = CPP_OTHER; |
1676 | break; | |
0578f103 | 1677 | } |
4970d4c2 | 1678 | else if (c == '\0') |
1679 | saw_NUL = true; | |
0578f103 | 1680 | } |
1681 | ||
4970d4c2 | 1682 | if (saw_NUL && !pfile->state.skipping) |
d80d2074 | 1683 | cpp_error (pfile, CPP_DL_WARNING, |
1684 | "null character(s) preserved in literal"); | |
0578f103 | 1685 | |
0b67f687 | 1686 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
1687 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
1688 | (int) terminator); | |
1689 | ||
244db24d | 1690 | if (CPP_OPTION (pfile, user_literals)) |
1691 | { | |
76d340ac | 1692 | /* According to C++11 [lex.ext]p10, a ud-suffix not starting with an |
1693 | underscore is ill-formed. Since this breaks programs using macros | |
1694 | from inttypes.h, we generate a warning and treat the ud-suffix as a | |
1695 | separate preprocessing token. This approach is under discussion by | |
1696 | the standards committee, and has been adopted as a conforming | |
911b08c6 | 1697 | extension by other front ends such as clang. |
1698 | A special exception is made for the suffix 's' which will be | |
1699 | standardized as a user-defined literal suffix for strings. */ | |
1700 | if (ISALPHA (*cur) && *cur != 's') | |
76d340ac | 1701 | { |
86c82b61 | 1702 | /* Raise a warning, but do not consume subsequent tokens. */ |
76d340ac | 1703 | if (CPP_OPTION (pfile, warn_literal_suffix)) |
1704 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, | |
1705 | token->src_loc, 0, | |
1706 | "invalid suffix on literal; C++11 requires " | |
1707 | "a space between literal and identifier"); | |
1708 | } | |
244db24d | 1709 | /* Grab user defined literal suffix. */ |
911b08c6 | 1710 | else if (ISIDST (*cur)) |
244db24d | 1711 | { |
1712 | type = cpp_userdef_char_add_type (type); | |
1713 | type = cpp_userdef_string_add_type (type); | |
1714 | ++cur; | |
76d340ac | 1715 | |
1716 | while (ISIDNUM (*cur)) | |
1717 | ++cur; | |
244db24d | 1718 | } |
244db24d | 1719 | } |
1720 | ||
4970d4c2 | 1721 | pfile->buffer->cur = cur; |
1722 | create_literal (pfile, token, base, cur - base, type); | |
338fa5f7 | 1723 | } |
f80e83a9 | 1724 | |
956c6108 | 1725 | /* Return the comment table. The client may not make any assumption |
1726 | about the ordering of the table. */ | |
1727 | cpp_comment_table * | |
1728 | cpp_get_comments (cpp_reader *pfile) | |
1729 | { | |
1730 | return &pfile->comments; | |
1731 | } | |
1732 | ||
1733 | /* Append a comment to the end of the comment table. */ | |
1734 | static void | |
1735 | store_comment (cpp_reader *pfile, cpp_token *token) | |
1736 | { | |
1737 | int len; | |
1738 | ||
1739 | if (pfile->comments.allocated == 0) | |
1740 | { | |
1741 | pfile->comments.allocated = 256; | |
1742 | pfile->comments.entries = (cpp_comment *) xmalloc | |
1743 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
1744 | } | |
1745 | ||
1746 | if (pfile->comments.count == pfile->comments.allocated) | |
1747 | { | |
1748 | pfile->comments.allocated *= 2; | |
1749 | pfile->comments.entries = (cpp_comment *) xrealloc | |
1750 | (pfile->comments.entries, | |
1751 | pfile->comments.allocated * sizeof (cpp_comment)); | |
1752 | } | |
1753 | ||
1754 | len = token->val.str.len; | |
1755 | ||
1756 | /* Copy comment. Note, token may not be NULL terminated. */ | |
1757 | pfile->comments.entries[pfile->comments.count].comment = | |
1758 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
1759 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
1760 | token->val.str.text, len); | |
1761 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
1762 | ||
1763 | /* Set source location. */ | |
1764 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
1765 | ||
1766 | /* Increment the count of entries in the comment table. */ | |
1767 | pfile->comments.count++; | |
1768 | } | |
1769 | ||
79bd622b | 1770 | /* The stored comment includes the comment start and any terminator. */ |
2c63d6c8 | 1771 | static void |
f7fdd7a1 | 1772 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
1773 | cppchar_t type) | |
2c63d6c8 | 1774 | { |
f80e83a9 | 1775 | unsigned char *buffer; |
560ab0b2 | 1776 | unsigned int len, clen, i; |
b1a9ff83 | 1777 | |
f0495c2c | 1778 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
1c124f85 | 1779 | |
a543b315 | 1780 | /* C++ comments probably (not definitely) have moved past a new |
1781 | line, which we don't want to save in the comment. */ | |
1c124f85 | 1782 | if (is_vspace (pfile->buffer->cur[-1])) |
a543b315 | 1783 | len--; |
d3f7919d | 1784 | |
560ab0b2 | 1785 | /* If we are currently in a directive or in argument parsing, then |
1786 | we need to store all C++ comments as C comments internally, and | |
1787 | so we need to allocate a little extra space in that case. | |
d3f7919d | 1788 | |
1789 | Note that the only time we encounter a directive here is | |
1790 | when we are saving comments in a "#define". */ | |
560ab0b2 | 1791 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
1792 | && type == '/') ? len + 2 : len; | |
d3f7919d | 1793 | |
1794 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
b1a9ff83 | 1795 | |
f80e83a9 | 1796 | token->type = CPP_COMMENT; |
d3f7919d | 1797 | token->val.str.len = clen; |
338fa5f7 | 1798 | token->val.str.text = buffer; |
0578f103 | 1799 | |
f0495c2c | 1800 | buffer[0] = '/'; |
1801 | memcpy (buffer + 1, from, len - 1); | |
d3f7919d | 1802 | |
a113df96 | 1803 | /* Finish conversion to a C comment, if necessary. */ |
560ab0b2 | 1804 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
d3f7919d | 1805 | { |
1806 | buffer[1] = '*'; | |
1807 | buffer[clen - 2] = '*'; | |
1808 | buffer[clen - 1] = '/'; | |
560ab0b2 | 1809 | /* As there can be in a C++ comments illegal sequences for C comments |
1810 | we need to filter them out. */ | |
1811 | for (i = 2; i < (clen - 2); i++) | |
1812 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
1813 | buffer[i] = '|'; | |
d3f7919d | 1814 | } |
956c6108 | 1815 | |
1816 | /* Finally store this comment for use by clients of libcpp. */ | |
1817 | store_comment (pfile, token); | |
338fa5f7 | 1818 | } |
0578f103 | 1819 | |
83dcbb5c | 1820 | /* Allocate COUNT tokens for RUN. */ |
1821 | void | |
f7fdd7a1 | 1822 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
83dcbb5c | 1823 | { |
3b298764 | 1824 | run->base = XNEWVEC (cpp_token, count); |
83dcbb5c | 1825 | run->limit = run->base + count; |
1826 | run->next = NULL; | |
1827 | } | |
1828 | ||
1829 | /* Returns the next tokenrun, or creates one if there is none. */ | |
1830 | static tokenrun * | |
f7fdd7a1 | 1831 | next_tokenrun (tokenrun *run) |
83dcbb5c | 1832 | { |
1833 | if (run->next == NULL) | |
1834 | { | |
3b298764 | 1835 | run->next = XNEW (tokenrun); |
fb5ab82c | 1836 | run->next->prev = run; |
83dcbb5c | 1837 | _cpp_init_tokenrun (run->next, 250); |
1838 | } | |
1839 | ||
1840 | return run->next; | |
1841 | } | |
1842 | ||
a2eb22f0 | 1843 | /* Return the number of not yet processed token in a given |
ce70f433 | 1844 | context. */ |
1845 | int | |
a2eb22f0 | 1846 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
ce70f433 | 1847 | { |
ce70f433 | 1848 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
ac6130e2 | 1849 | return (LAST (context).token - FIRST (context).token); |
ce70f433 | 1850 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
1851 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
ac6130e2 | 1852 | return (LAST (context).ptoken - FIRST (context).ptoken); |
ce70f433 | 1853 | else |
1854 | abort (); | |
1855 | } | |
1856 | ||
a2eb22f0 | 1857 | /* Returns the token present at index INDEX in a given context. If |
1858 | INDEX is zero, the next token to be processed is returned. */ | |
ce70f433 | 1859 | static const cpp_token* |
a2eb22f0 | 1860 | _cpp_token_from_context_at (cpp_context *context, int index) |
ce70f433 | 1861 | { |
ce70f433 | 1862 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
1863 | return &(FIRST (context).token[index]); | |
1864 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
1865 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
1866 | return FIRST (context).ptoken[index]; | |
1867 | else | |
1868 | abort (); | |
1869 | } | |
1870 | ||
89768577 | 1871 | /* Look ahead in the input stream. */ |
1872 | const cpp_token * | |
1873 | cpp_peek_token (cpp_reader *pfile, int index) | |
1874 | { | |
1875 | cpp_context *context = pfile->context; | |
1876 | const cpp_token *peektok; | |
1877 | int count; | |
1878 | ||
1879 | /* First, scan through any pending cpp_context objects. */ | |
1880 | while (context->prev) | |
1881 | { | |
a2eb22f0 | 1882 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
89768577 | 1883 | |
1884 | if (index < (int) sz) | |
a2eb22f0 | 1885 | return _cpp_token_from_context_at (context, index); |
89768577 | 1886 | index -= (int) sz; |
1887 | context = context->prev; | |
1888 | } | |
1889 | ||
1890 | /* We will have to read some new tokens after all (and do so | |
1891 | without invalidating preceding tokens). */ | |
1892 | count = index; | |
1893 | pfile->keep_tokens++; | |
1894 | ||
1895 | do | |
1896 | { | |
1897 | peektok = _cpp_lex_token (pfile); | |
1898 | if (peektok->type == CPP_EOF) | |
1899 | return peektok; | |
1900 | } | |
1901 | while (index--); | |
1902 | ||
1903 | _cpp_backup_tokens_direct (pfile, count + 1); | |
1904 | pfile->keep_tokens--; | |
1905 | ||
1906 | return peektok; | |
1907 | } | |
1908 | ||
f9b5f742 | 1909 | /* Allocate a single token that is invalidated at the same time as the |
1910 | rest of the tokens on the line. Has its line and col set to the | |
1911 | same as the last lexed token, so that diagnostics appear in the | |
1912 | right place. */ | |
1913 | cpp_token * | |
f7fdd7a1 | 1914 | _cpp_temp_token (cpp_reader *pfile) |
f9b5f742 | 1915 | { |
1916 | cpp_token *old, *result; | |
89768577 | 1917 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
1918 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
f9b5f742 | 1919 | |
1920 | old = pfile->cur_token - 1; | |
89768577 | 1921 | /* Any pre-existing lookaheads must not be clobbered. */ |
1922 | if (la) | |
1923 | { | |
1924 | if (sz <= la) | |
1925 | { | |
1926 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
1927 | ||
1928 | if (sz < la) | |
1929 | memmove (next->base + 1, next->base, | |
1930 | (la - sz) * sizeof (cpp_token)); | |
1931 | ||
1932 | next->base[0] = pfile->cur_run->limit[-1]; | |
1933 | } | |
1934 | ||
1935 | if (sz > 1) | |
1936 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
1937 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
1938 | } | |
1939 | ||
1940 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
f9b5f742 | 1941 | { |
1942 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
1943 | pfile->cur_token = pfile->cur_run->base; | |
1944 | } | |
1945 | ||
1946 | result = pfile->cur_token++; | |
610625e3 | 1947 | result->src_loc = old->src_loc; |
f9b5f742 | 1948 | return result; |
1949 | } | |
1950 | ||
10b4496a | 1951 | /* Lex a token into RESULT (external interface). Takes care of issues |
1952 | like directive handling, token lookahead, multiple include | |
3fb1e43b | 1953 | optimization and skipping. */ |
c00e481c | 1954 | const cpp_token * |
f7fdd7a1 | 1955 | _cpp_lex_token (cpp_reader *pfile) |
83dcbb5c | 1956 | { |
fb5ab82c | 1957 | cpp_token *result; |
83dcbb5c | 1958 | |
fb5ab82c | 1959 | for (;;) |
83dcbb5c | 1960 | { |
fb5ab82c | 1961 | if (pfile->cur_token == pfile->cur_run->limit) |
83dcbb5c | 1962 | { |
fb5ab82c | 1963 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
1964 | pfile->cur_token = pfile->cur_run->base; | |
83dcbb5c | 1965 | } |
e0ff7935 | 1966 | /* We assume that the current token is somewhere in the current |
1967 | run. */ | |
1968 | if (pfile->cur_token < pfile->cur_run->base | |
1969 | || pfile->cur_token >= pfile->cur_run->limit) | |
1970 | abort (); | |
83dcbb5c | 1971 | |
fb5ab82c | 1972 | if (pfile->lookaheads) |
10b4496a | 1973 | { |
1974 | pfile->lookaheads--; | |
1975 | result = pfile->cur_token++; | |
1976 | } | |
fb5ab82c | 1977 | else |
10b4496a | 1978 | result = _cpp_lex_direct (pfile); |
fb5ab82c | 1979 | |
1980 | if (result->flags & BOL) | |
83dcbb5c | 1981 | { |
fb5ab82c | 1982 | /* Is this a directive. If _cpp_handle_directive returns |
1983 | false, it is an assembler #. */ | |
1984 | if (result->type == CPP_HASH | |
d6af0368 | 1985 | /* 6.10.3 p 11: Directives in a list of macro arguments |
1986 | gives undefined behavior. This implementation | |
1987 | handles the directive as normal. */ | |
b75b98aa | 1988 | && pfile->state.parsing_args != 1) |
d6d3c909 | 1989 | { |
b75b98aa | 1990 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
d6d3c909 | 1991 | { |
b75b98aa | 1992 | if (pfile->directive_result.type == CPP_PADDING) |
1993 | continue; | |
d6d3c909 | 1994 | result = &pfile->directive_result; |
d6d3c909 | 1995 | } |
1996 | } | |
b75b98aa | 1997 | else if (pfile->state.in_deferred_pragma) |
1998 | result = &pfile->directive_result; | |
d6d3c909 | 1999 | |
5621a364 | 2000 | if (pfile->cb.line_change && !pfile->state.skipping) |
f7fdd7a1 | 2001 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
83dcbb5c | 2002 | } |
83dcbb5c | 2003 | |
fb5ab82c | 2004 | /* We don't skip tokens in directives. */ |
b75b98aa | 2005 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
fb5ab82c | 2006 | break; |
83dcbb5c | 2007 | |
fb5ab82c | 2008 | /* Outside a directive, invalidate controlling macros. At file |
10b4496a | 2009 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
7ef5b942 | 2010 | get here and MI optimization works. */ |
83dcbb5c | 2011 | pfile->mi_valid = false; |
fb5ab82c | 2012 | |
2013 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
2014 | break; | |
83dcbb5c | 2015 | } |
2016 | ||
c00e481c | 2017 | return result; |
83dcbb5c | 2018 | } |
2019 | ||
a54e0bf8 | 2020 | /* Returns true if a fresh line has been loaded. */ |
2021 | bool | |
f7fdd7a1 | 2022 | _cpp_get_fresh_line (cpp_reader *pfile) |
0bb65704 | 2023 | { |
6e04daf1 | 2024 | int return_at_eof; |
2025 | ||
a54e0bf8 | 2026 | /* We can't get a new line until we leave the current directive. */ |
2027 | if (pfile->state.in_directive) | |
2028 | return false; | |
b1a9ff83 | 2029 | |
a54e0bf8 | 2030 | for (;;) |
fb83e0d6 | 2031 | { |
a54e0bf8 | 2032 | cpp_buffer *buffer = pfile->buffer; |
fb83e0d6 | 2033 | |
a54e0bf8 | 2034 | if (!buffer->need_line) |
2035 | return true; | |
2036 | ||
2037 | if (buffer->next_line < buffer->rlimit) | |
0bb65704 | 2038 | { |
a54e0bf8 | 2039 | _cpp_clean_line (pfile); |
2040 | return true; | |
2041 | } | |
0bb65704 | 2042 | |
a54e0bf8 | 2043 | /* First, get out of parsing arguments state. */ |
2044 | if (pfile->state.parsing_args) | |
2045 | return false; | |
2046 | ||
2047 | /* End of buffer. Non-empty files should end in a newline. */ | |
2048 | if (buffer->buf != buffer->rlimit | |
2049 | && buffer->next_line > buffer->rlimit | |
2050 | && !buffer->from_stage3) | |
2051 | { | |
0448520c | 2052 | /* Clip to buffer size. */ |
a54e0bf8 | 2053 | buffer->next_line = buffer->rlimit; |
a54e0bf8 | 2054 | } |
6e04daf1 | 2055 | |
2056 | return_at_eof = buffer->return_at_eof; | |
a54e0bf8 | 2057 | _cpp_pop_buffer (pfile); |
6e04daf1 | 2058 | if (pfile->buffer == NULL || return_at_eof) |
11b5269c | 2059 | return false; |
a54e0bf8 | 2060 | } |
0bb65704 | 2061 | } |
2062 | ||
edaf8cb5 | 2063 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
2064 | do \ | |
2065 | { \ | |
2066 | result->type = ELSE_TYPE; \ | |
2067 | if (*buffer->cur == CHAR) \ | |
2068 | buffer->cur++, result->type = THEN_TYPE; \ | |
2069 | } \ | |
2070 | while (0) | |
1c124f85 | 2071 | |
10b4496a | 2072 | /* Lex a token into pfile->cur_token, which is also incremented, to |
2073 | get diagnostics pointing to the correct location. | |
2074 | ||
2075 | Does not handle issues such as token lookahead, multiple-include | |
4172d65e | 2076 | optimization, directives, skipping etc. This function is only |
10b4496a | 2077 | suitable for use by _cpp_lex_token, and in special cases like |
2078 | lex_expansion_token which doesn't care for any of these issues. | |
2079 | ||
2080 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
2081 | otherwise returns to the start of the token buffer if permissible. | |
2082 | Returns the location of the lexed token. */ | |
2083 | cpp_token * | |
f7fdd7a1 | 2084 | _cpp_lex_direct (cpp_reader *pfile) |
0578f103 | 2085 | { |
338fa5f7 | 2086 | cppchar_t c; |
230f0943 | 2087 | cpp_buffer *buffer; |
338fa5f7 | 2088 | const unsigned char *comment_start; |
10b4496a | 2089 | cpp_token *result = pfile->cur_token++; |
0653b94e | 2090 | |
83dcbb5c | 2091 | fresh_line: |
a54e0bf8 | 2092 | result->flags = 0; |
82166c5c | 2093 | buffer = pfile->buffer; |
11b5269c | 2094 | if (buffer->need_line) |
a54e0bf8 | 2095 | { |
b75b98aa | 2096 | if (pfile->state.in_deferred_pragma) |
2097 | { | |
2098 | result->type = CPP_PRAGMA_EOL; | |
2099 | pfile->state.in_deferred_pragma = false; | |
2100 | if (!pfile->state.pragma_allow_expansion) | |
2101 | pfile->state.prevent_expansion--; | |
2102 | return result; | |
2103 | } | |
a54e0bf8 | 2104 | if (!_cpp_get_fresh_line (pfile)) |
2105 | { | |
2106 | result->type = CPP_EOF; | |
2908f819 | 2107 | if (!pfile->state.in_directive) |
2108 | { | |
2109 | /* Tell the compiler the line number of the EOF token. */ | |
dbddc569 | 2110 | result->src_loc = pfile->line_table->highest_line; |
2908f819 | 2111 | result->flags = BOL; |
2112 | } | |
a54e0bf8 | 2113 | return result; |
2114 | } | |
2115 | if (!pfile->keep_tokens) | |
2116 | { | |
2117 | pfile->cur_run = &pfile->base_run; | |
2118 | result = pfile->base_run.base; | |
2119 | pfile->cur_token = result + 1; | |
2120 | } | |
2121 | result->flags = BOL; | |
2122 | if (pfile->state.parsing_args == 2) | |
2123 | result->flags |= PREV_WHITE; | |
2124 | } | |
11b5269c | 2125 | buffer = pfile->buffer; |
83dcbb5c | 2126 | update_tokens_line: |
dbddc569 | 2127 | result->src_loc = pfile->line_table->highest_line; |
f80e83a9 | 2128 | |
83dcbb5c | 2129 | skipped_white: |
a54e0bf8 | 2130 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
2131 | && !pfile->overlaid_buffer) | |
2132 | { | |
2133 | _cpp_process_line_notes (pfile, false); | |
dbddc569 | 2134 | result->src_loc = pfile->line_table->highest_line; |
a54e0bf8 | 2135 | } |
1c124f85 | 2136 | c = *buffer->cur++; |
610625e3 | 2137 | |
6ea2c7a3 | 2138 | if (pfile->forced_token_location_p) |
2139 | result->src_loc = *pfile->forced_token_location_p; | |
2140 | else | |
2141 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
2142 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
83dcbb5c | 2143 | |
338fa5f7 | 2144 | switch (c) |
0578f103 | 2145 | { |
435fb09b | 2146 | case ' ': case '\t': case '\f': case '\v': case '\0': |
2147 | result->flags |= PREV_WHITE; | |
a54e0bf8 | 2148 | skip_whitespace (pfile, c); |
2149 | goto skipped_white; | |
338fa5f7 | 2150 | |
a54e0bf8 | 2151 | case '\n': |
610625e3 | 2152 | if (buffer->cur < buffer->rlimit) |
2153 | CPP_INCREMENT_LINE (pfile, 0); | |
a54e0bf8 | 2154 | buffer->need_line = true; |
2155 | goto fresh_line; | |
732cb4c9 | 2156 | |
338fa5f7 | 2157 | case '0': case '1': case '2': case '3': case '4': |
2158 | case '5': case '6': case '7': case '8': case '9': | |
bce47149 | 2159 | { |
2160 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2161 | result->type = CPP_NUMBER; | |
2162 | lex_number (pfile, &result->val.str, &nst); | |
2163 | warn_about_normalization (pfile, result, &nst); | |
2164 | break; | |
2165 | } | |
732cb4c9 | 2166 | |
78c551ad | 2167 | case 'L': |
924bbf02 | 2168 | case 'u': |
2169 | case 'U': | |
538ba11a | 2170 | case 'R': |
2171 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2172 | wide strings or raw strings. */ | |
6f6f3dd7 | 2173 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
2174 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
5bb46c08 | 2175 | { |
538ba11a | 2176 | if ((*buffer->cur == '\'' && c != 'R') |
2177 | || *buffer->cur == '"' | |
2178 | || (*buffer->cur == 'R' | |
2179 | && c != 'R' | |
2180 | && buffer->cur[1] == '"' | |
6f6f3dd7 | 2181 | && CPP_OPTION (pfile, rliterals)) |
538ba11a | 2182 | || (*buffer->cur == '8' |
2183 | && c == 'u' | |
2184 | && (buffer->cur[1] == '"' | |
6f6f3dd7 | 2185 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
2186 | && CPP_OPTION (pfile, rliterals))))) | |
924bbf02 | 2187 | { |
2188 | lex_string (pfile, result, buffer->cur - 1); | |
2189 | break; | |
2190 | } | |
5bb46c08 | 2191 | } |
b1a9ff83 | 2192 | /* Fall through. */ |
78c551ad | 2193 | |
338fa5f7 | 2194 | case '_': |
2195 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2196 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2197 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
924bbf02 | 2198 | case 's': case 't': case 'v': case 'w': case 'x': |
338fa5f7 | 2199 | case 'y': case 'z': |
2200 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
78c551ad | 2201 | case 'G': case 'H': case 'I': case 'J': case 'K': |
538ba11a | 2202 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
924bbf02 | 2203 | case 'S': case 'T': case 'V': case 'W': case 'X': |
338fa5f7 | 2204 | case 'Y': case 'Z': |
2205 | result->type = CPP_NAME; | |
bce47149 | 2206 | { |
2207 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2ee04baa | 2208 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
2209 | &nst); | |
bce47149 | 2210 | warn_about_normalization (pfile, result, &nst); |
2211 | } | |
338fa5f7 | 2212 | |
338fa5f7 | 2213 | /* Convert named operators to their proper types. */ |
2ee04baa | 2214 | if (result->val.node.node->flags & NODE_OPERATOR) |
338fa5f7 | 2215 | { |
2216 | result->flags |= NAMED_OP; | |
2ee04baa | 2217 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
338fa5f7 | 2218 | } |
2219 | break; | |
2220 | ||
2221 | case '\'': | |
2222 | case '"': | |
4970d4c2 | 2223 | lex_string (pfile, result, buffer->cur - 1); |
338fa5f7 | 2224 | break; |
f80e83a9 | 2225 | |
338fa5f7 | 2226 | case '/': |
f0495c2c | 2227 | /* A potential block or line comment. */ |
2228 | comment_start = buffer->cur; | |
edaf8cb5 | 2229 | c = *buffer->cur; |
2230 | ||
f0495c2c | 2231 | if (c == '*') |
2232 | { | |
a54e0bf8 | 2233 | if (_cpp_skip_block_comment (pfile)) |
d80d2074 | 2234 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
338fa5f7 | 2235 | } |
1c124f85 | 2236 | else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments) |
610625e3 | 2237 | || cpp_in_system_header (pfile))) |
338fa5f7 | 2238 | { |
5db5d057 | 2239 | /* Warn about comments only if pedantically GNUC89, and not |
2240 | in system headers. */ | |
2241 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
66914e49 | 2242 | && ! buffer->warned_cplusplus_comments) |
f80e83a9 | 2243 | { |
d80d2074 | 2244 | cpp_error (pfile, CPP_DL_PEDWARN, |
ba059ac0 | 2245 | "C++ style comments are not allowed in ISO C90"); |
d80d2074 | 2246 | cpp_error (pfile, CPP_DL_PEDWARN, |
73328dce | 2247 | "(this will be reported only once per input file)"); |
f0495c2c | 2248 | buffer->warned_cplusplus_comments = 1; |
2249 | } | |
338fa5f7 | 2250 | |
e1caf668 | 2251 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
3a79f5da | 2252 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
f0495c2c | 2253 | } |
1c124f85 | 2254 | else if (c == '=') |
2255 | { | |
edaf8cb5 | 2256 | buffer->cur++; |
1c124f85 | 2257 | result->type = CPP_DIV_EQ; |
2258 | break; | |
2259 | } | |
2260 | else | |
2261 | { | |
1c124f85 | 2262 | result->type = CPP_DIV; |
2263 | break; | |
2264 | } | |
338fa5f7 | 2265 | |
f0495c2c | 2266 | if (!pfile->state.save_comments) |
2267 | { | |
2268 | result->flags |= PREV_WHITE; | |
83dcbb5c | 2269 | goto update_tokens_line; |
338fa5f7 | 2270 | } |
f0495c2c | 2271 | |
2272 | /* Save the comment as a token in its own right. */ | |
d3f7919d | 2273 | save_comment (pfile, result, comment_start, c); |
fb5ab82c | 2274 | break; |
338fa5f7 | 2275 | |
2276 | case '<': | |
2277 | if (pfile->state.angled_headers) | |
2278 | { | |
4970d4c2 | 2279 | lex_string (pfile, result, buffer->cur - 1); |
7811eab5 | 2280 | if (result->type != CPP_LESS) |
2281 | break; | |
338fa5f7 | 2282 | } |
0578f103 | 2283 | |
edaf8cb5 | 2284 | result->type = CPP_LESS; |
2285 | if (*buffer->cur == '=') | |
2286 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2287 | else if (*buffer->cur == '<') | |
338fa5f7 | 2288 | { |
edaf8cb5 | 2289 | buffer->cur++; |
2290 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
338fa5f7 | 2291 | } |
edaf8cb5 | 2292 | else if (CPP_OPTION (pfile, digraphs)) |
1c124f85 | 2293 | { |
edaf8cb5 | 2294 | if (*buffer->cur == ':') |
2295 | { | |
1aa79d39 | 2296 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
2297 | three characters are <:: and the subsequent character | |
2298 | is neither : nor >, the < is treated as a preprocessor | |
2299 | token by itself". */ | |
2300 | if (CPP_OPTION (pfile, cplusplus) | |
2301 | && (CPP_OPTION (pfile, lang) == CLK_CXX11 | |
2302 | || CPP_OPTION (pfile, lang) == CLK_GNUCXX11) | |
2303 | && buffer->cur[1] == ':' | |
2304 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
2305 | break; | |
2306 | ||
edaf8cb5 | 2307 | buffer->cur++; |
2308 | result->flags |= DIGRAPH; | |
2309 | result->type = CPP_OPEN_SQUARE; | |
2310 | } | |
2311 | else if (*buffer->cur == '%') | |
2312 | { | |
2313 | buffer->cur++; | |
2314 | result->flags |= DIGRAPH; | |
2315 | result->type = CPP_OPEN_BRACE; | |
2316 | } | |
1c124f85 | 2317 | } |
338fa5f7 | 2318 | break; |
2319 | ||
2320 | case '>': | |
edaf8cb5 | 2321 | result->type = CPP_GREATER; |
2322 | if (*buffer->cur == '=') | |
2323 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
2324 | else if (*buffer->cur == '>') | |
338fa5f7 | 2325 | { |
edaf8cb5 | 2326 | buffer->cur++; |
2327 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
2328 | } | |
338fa5f7 | 2329 | break; |
2330 | ||
f669338a | 2331 | case '%': |
edaf8cb5 | 2332 | result->type = CPP_MOD; |
2333 | if (*buffer->cur == '=') | |
2334 | buffer->cur++, result->type = CPP_MOD_EQ; | |
2335 | else if (CPP_OPTION (pfile, digraphs)) | |
1c124f85 | 2336 | { |
edaf8cb5 | 2337 | if (*buffer->cur == ':') |
1c124f85 | 2338 | { |
edaf8cb5 | 2339 | buffer->cur++; |
2340 | result->flags |= DIGRAPH; | |
2341 | result->type = CPP_HASH; | |
2342 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
2ee04baa | 2343 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
edaf8cb5 | 2344 | } |
2345 | else if (*buffer->cur == '>') | |
2346 | { | |
2347 | buffer->cur++; | |
2348 | result->flags |= DIGRAPH; | |
2349 | result->type = CPP_CLOSE_BRACE; | |
1c124f85 | 2350 | } |
1c124f85 | 2351 | } |
338fa5f7 | 2352 | break; |
2353 | ||
f669338a | 2354 | case '.': |
1c124f85 | 2355 | result->type = CPP_DOT; |
edaf8cb5 | 2356 | if (ISDIGIT (*buffer->cur)) |
1c124f85 | 2357 | { |
bce47149 | 2358 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
1c124f85 | 2359 | result->type = CPP_NUMBER; |
bce47149 | 2360 | lex_number (pfile, &result->val.str, &nst); |
2361 | warn_about_normalization (pfile, result, &nst); | |
1c124f85 | 2362 | } |
edaf8cb5 | 2363 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
2364 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
2365 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
2366 | buffer->cur++, result->type = CPP_DOT_STAR; | |
338fa5f7 | 2367 | break; |
0578f103 | 2368 | |
338fa5f7 | 2369 | case '+': |
edaf8cb5 | 2370 | result->type = CPP_PLUS; |
2371 | if (*buffer->cur == '+') | |
2372 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
2373 | else if (*buffer->cur == '=') | |
2374 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
338fa5f7 | 2375 | break; |
ac0749c7 | 2376 | |
338fa5f7 | 2377 | case '-': |
edaf8cb5 | 2378 | result->type = CPP_MINUS; |
2379 | if (*buffer->cur == '>') | |
338fa5f7 | 2380 | { |
edaf8cb5 | 2381 | buffer->cur++; |
1c124f85 | 2382 | result->type = CPP_DEREF; |
edaf8cb5 | 2383 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
2384 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
1c124f85 | 2385 | } |
edaf8cb5 | 2386 | else if (*buffer->cur == '-') |
2387 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
2388 | else if (*buffer->cur == '=') | |
2389 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
338fa5f7 | 2390 | break; |
0578f103 | 2391 | |
338fa5f7 | 2392 | case '&': |
edaf8cb5 | 2393 | result->type = CPP_AND; |
2394 | if (*buffer->cur == '&') | |
2395 | buffer->cur++, result->type = CPP_AND_AND; | |
2396 | else if (*buffer->cur == '=') | |
2397 | buffer->cur++, result->type = CPP_AND_EQ; | |
338fa5f7 | 2398 | break; |
b1a9ff83 | 2399 | |
338fa5f7 | 2400 | case '|': |
edaf8cb5 | 2401 | result->type = CPP_OR; |
2402 | if (*buffer->cur == '|') | |
2403 | buffer->cur++, result->type = CPP_OR_OR; | |
2404 | else if (*buffer->cur == '=') | |
2405 | buffer->cur++, result->type = CPP_OR_EQ; | |
338fa5f7 | 2406 | break; |
0578f103 | 2407 | |
338fa5f7 | 2408 | case ':': |
edaf8cb5 | 2409 | result->type = CPP_COLON; |
2410 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
2411 | buffer->cur++, result->type = CPP_SCOPE; | |
2412 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
338fa5f7 | 2413 | { |
edaf8cb5 | 2414 | buffer->cur++; |
338fa5f7 | 2415 | result->flags |= DIGRAPH; |
1c124f85 | 2416 | result->type = CPP_CLOSE_SQUARE; |
2417 | } | |
338fa5f7 | 2418 | break; |
0578f103 | 2419 | |
1c124f85 | 2420 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
2421 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
2422 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
2423 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
2ee04baa | 2424 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
1c124f85 | 2425 | |
a54e0bf8 | 2426 | case '?': result->type = CPP_QUERY; break; |
338fa5f7 | 2427 | case '~': result->type = CPP_COMPL; break; |
2428 | case ',': result->type = CPP_COMMA; break; | |
2429 | case '(': result->type = CPP_OPEN_PAREN; break; | |
2430 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
2431 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
2432 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
2433 | case '{': result->type = CPP_OPEN_BRACE; break; | |
2434 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
2435 | case ';': result->type = CPP_SEMICOLON; break; | |
2436 | ||
7fd957fe | 2437 | /* @ is a punctuator in Objective-C. */ |
9ee99ac6 | 2438 | case '@': result->type = CPP_ATSIGN; break; |
338fa5f7 | 2439 | |
78c551ad | 2440 | case '$': |
2cbf1359 | 2441 | case '\\': |
2442 | { | |
2443 | const uchar *base = --buffer->cur; | |
bce47149 | 2444 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
78c551ad | 2445 | |
bce47149 | 2446 | if (forms_identifier_p (pfile, true, &nst)) |
2cbf1359 | 2447 | { |
2448 | result->type = CPP_NAME; | |
2ee04baa | 2449 | result->val.node.node = lex_identifier (pfile, base, true, &nst); |
bce47149 | 2450 | warn_about_normalization (pfile, result, &nst); |
2cbf1359 | 2451 | break; |
2452 | } | |
2453 | buffer->cur++; | |
bc205914 | 2454 | } |
2cbf1359 | 2455 | |
bc205914 | 2456 | default: |
4970d4c2 | 2457 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); |
2458 | break; | |
338fa5f7 | 2459 | } |
fb5ab82c | 2460 | |
2461 | return result; | |
338fa5f7 | 2462 | } |
2463 | ||
b1280514 | 2464 | /* An upper bound on the number of bytes needed to spell TOKEN. |
2465 | Does not include preceding whitespace. */ | |
79bd622b | 2466 | unsigned int |
f7fdd7a1 | 2467 | cpp_token_len (const cpp_token *token) |
338fa5f7 | 2468 | { |
79bd622b | 2469 | unsigned int len; |
cfad5579 | 2470 | |
79bd622b | 2471 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2472 | { |
cd740bd5 | 2473 | default: len = 6; break; |
4970d4c2 | 2474 | case SPELL_LITERAL: len = token->val.str.len; break; |
2ee04baa | 2475 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
f80e83a9 | 2476 | } |
b1280514 | 2477 | |
2478 | return len; | |
cfad5579 | 2479 | } |
2480 | ||
bb1fa6bb | 2481 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
2482 | Return the number of bytes read out of NAME. (There are always | |
2483 | 10 bytes written to BUFFER.) */ | |
2484 | ||
2485 | static size_t | |
2486 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
2487 | { | |
2488 | int j; | |
2489 | int ucn_len = 0; | |
2490 | int ucn_len_c; | |
2491 | unsigned t; | |
2492 | unsigned long utf32; | |
2493 | ||
2494 | /* Compute the length of the UTF-8 sequence. */ | |
2495 | for (t = *name; t & 0x80; t <<= 1) | |
2496 | ucn_len++; | |
2497 | ||
2498 | utf32 = *name & (0x7F >> ucn_len); | |
2499 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
2500 | { | |
2501 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
2502 | ||
2503 | /* Ill-formed UTF-8. */ | |
2504 | if ((*name & ~0x3F) != 0x80) | |
2505 | abort (); | |
2506 | } | |
2507 | ||
2508 | *buffer++ = '\\'; | |
2509 | *buffer++ = 'U'; | |
2510 | for (j = 7; j >= 0; j--) | |
2511 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
2512 | return ucn_len; | |
2513 | } | |
2514 | ||
ba99525e | 2515 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
2516 | the spelling of the digraph. */ | |
2517 | static const unsigned char * | |
2518 | cpp_digraph2name (enum cpp_ttype type) | |
2519 | { | |
2520 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
2521 | } | |
bb1fa6bb | 2522 | |
f80e83a9 | 2523 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 2524 | already contain the enough space to hold the token's spelling. |
f7fdd7a1 | 2525 | Returns a pointer to the character after the last character written. |
bb1fa6bb | 2526 | FORSTRING is true if this is to be the spelling after translation |
2527 | phase 1 (this is different for UCNs). | |
f7fdd7a1 | 2528 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
79bd622b | 2529 | unsigned char * |
f7fdd7a1 | 2530 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
bb1fa6bb | 2531 | unsigned char *buffer, bool forstring) |
f80e83a9 | 2532 | { |
7e842f95 | 2533 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2534 | { |
2535 | case SPELL_OPERATOR: | |
2536 | { | |
2537 | const unsigned char *spelling; | |
2538 | unsigned char c; | |
ab12a39c | 2539 | |
f80e83a9 | 2540 | if (token->flags & DIGRAPH) |
ba99525e | 2541 | spelling = cpp_digraph2name (token->type); |
31674461 | 2542 | else if (token->flags & NAMED_OP) |
2543 | goto spell_ident; | |
f80e83a9 | 2544 | else |
7e842f95 | 2545 | spelling = TOKEN_NAME (token); |
b1a9ff83 | 2546 | |
f80e83a9 | 2547 | while ((c = *spelling++) != '\0') |
2548 | *buffer++ = c; | |
2549 | } | |
2550 | break; | |
ab12a39c | 2551 | |
8d27e472 | 2552 | spell_ident: |
f80e83a9 | 2553 | case SPELL_IDENT: |
bb1fa6bb | 2554 | if (forstring) |
2555 | { | |
2ee04baa | 2556 | memcpy (buffer, NODE_NAME (token->val.node.node), |
2557 | NODE_LEN (token->val.node.node)); | |
2558 | buffer += NODE_LEN (token->val.node.node); | |
bb1fa6bb | 2559 | } |
2560 | else | |
2561 | { | |
2562 | size_t i; | |
2ee04baa | 2563 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2564 | |
2ee04baa | 2565 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2566 | if (name[i] & ~0x7F) |
2567 | { | |
2568 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2569 | buffer += 10; | |
2570 | } | |
2571 | else | |
2ee04baa | 2572 | *buffer++ = NODE_NAME (token->val.node.node)[i]; |
bb1fa6bb | 2573 | } |
f80e83a9 | 2574 | break; |
ab12a39c | 2575 | |
4970d4c2 | 2576 | case SPELL_LITERAL: |
8d27e472 | 2577 | memcpy (buffer, token->val.str.text, token->val.str.len); |
2578 | buffer += token->val.str.len; | |
2579 | break; | |
2580 | ||
f80e83a9 | 2581 | case SPELL_NONE: |
d80d2074 | 2582 | cpp_error (pfile, CPP_DL_ICE, |
2583 | "unspellable token %s", TOKEN_NAME (token)); | |
f80e83a9 | 2584 | break; |
2585 | } | |
ab12a39c | 2586 | |
f80e83a9 | 2587 | return buffer; |
2588 | } | |
ab12a39c | 2589 | |
e484a1cc | 2590 | /* Returns TOKEN spelt as a null-terminated string. The string is |
2591 | freed when the reader is destroyed. Useful for diagnostics. */ | |
79bd622b | 2592 | unsigned char * |
f7fdd7a1 | 2593 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
b1280514 | 2594 | { |
2595 | unsigned int len = cpp_token_len (token) + 1; | |
1fdf6039 | 2596 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
6060326b | 2597 | |
bb1fa6bb | 2598 | end = cpp_spell_token (pfile, token, start, false); |
79bd622b | 2599 | end[0] = '\0'; |
6060326b | 2600 | |
79bd622b | 2601 | return start; |
2602 | } | |
6060326b | 2603 | |
ba99525e | 2604 | /* Returns a pointer to a string which spells the token defined by |
2605 | TYPE and FLAGS. Used by C front ends, which really should move to | |
2606 | using cpp_token_as_text. */ | |
79bd622b | 2607 | const char * |
ba99525e | 2608 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
79bd622b | 2609 | { |
ba99525e | 2610 | if (flags & DIGRAPH) |
2611 | return (const char *) cpp_digraph2name (type); | |
2612 | else if (flags & NAMED_OP) | |
2613 | return cpp_named_operator2name (type); | |
2614 | ||
79bd622b | 2615 | return (const char *) token_spellings[type].name; |
2616 | } | |
6060326b | 2617 | |
f9b5f742 | 2618 | /* Writes the spelling of token to FP, without any preceding space. |
2619 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
2620 | double-buffering. */ | |
79bd622b | 2621 | void |
f7fdd7a1 | 2622 | cpp_output_token (const cpp_token *token, FILE *fp) |
79bd622b | 2623 | { |
79bd622b | 2624 | switch (TOKEN_SPELL (token)) |
6060326b | 2625 | { |
79bd622b | 2626 | case SPELL_OPERATOR: |
2627 | { | |
2628 | const unsigned char *spelling; | |
28874558 | 2629 | int c; |
6060326b | 2630 | |
79bd622b | 2631 | if (token->flags & DIGRAPH) |
ba99525e | 2632 | spelling = cpp_digraph2name (token->type); |
79bd622b | 2633 | else if (token->flags & NAMED_OP) |
2634 | goto spell_ident; | |
2635 | else | |
2636 | spelling = TOKEN_NAME (token); | |
f80e83a9 | 2637 | |
28874558 | 2638 | c = *spelling; |
2639 | do | |
2640 | putc (c, fp); | |
2641 | while ((c = *++spelling) != '\0'); | |
79bd622b | 2642 | } |
2643 | break; | |
f80e83a9 | 2644 | |
79bd622b | 2645 | spell_ident: |
2646 | case SPELL_IDENT: | |
bb1fa6bb | 2647 | { |
2648 | size_t i; | |
2ee04baa | 2649 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2650 | |
2ee04baa | 2651 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2652 | if (name[i] & ~0x7F) |
2653 | { | |
2654 | unsigned char buffer[10]; | |
2655 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2656 | fwrite (buffer, 1, 10, fp); | |
2657 | } | |
2658 | else | |
2ee04baa | 2659 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
bb1fa6bb | 2660 | } |
2661 | break; | |
f80e83a9 | 2662 | |
4970d4c2 | 2663 | case SPELL_LITERAL: |
8d27e472 | 2664 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
2665 | break; | |
2666 | ||
79bd622b | 2667 | case SPELL_NONE: |
2668 | /* An error, most probably. */ | |
2669 | break; | |
f80e83a9 | 2670 | } |
6060326b | 2671 | } |
2672 | ||
79bd622b | 2673 | /* Compare two tokens. */ |
2674 | int | |
f7fdd7a1 | 2675 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
6060326b | 2676 | { |
79bd622b | 2677 | if (a->type == b->type && a->flags == b->flags) |
2678 | switch (TOKEN_SPELL (a)) | |
2679 | { | |
2680 | default: /* Keep compiler happy. */ | |
2681 | case SPELL_OPERATOR: | |
2ee04baa | 2682 | /* token_no is used to track where multiple consecutive ## |
941f2388 | 2683 | tokens were originally located. */ |
2ee04baa | 2684 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
79bd622b | 2685 | case SPELL_NONE: |
2ee04baa | 2686 | return (a->type != CPP_MACRO_ARG |
2687 | || a->val.macro_arg.arg_no == b->val.macro_arg.arg_no); | |
79bd622b | 2688 | case SPELL_IDENT: |
2ee04baa | 2689 | return a->val.node.node == b->val.node.node; |
4970d4c2 | 2690 | case SPELL_LITERAL: |
79bd622b | 2691 | return (a->val.str.len == b->val.str.len |
2692 | && !memcmp (a->val.str.text, b->val.str.text, | |
2693 | a->val.str.len)); | |
2694 | } | |
6060326b | 2695 | |
f80e83a9 | 2696 | return 0; |
2697 | } | |
2698 | ||
79bd622b | 2699 | /* Returns nonzero if a space should be inserted to avoid an |
2700 | accidental token paste for output. For simplicity, it is | |
2701 | conservative, and occasionally advises a space where one is not | |
2702 | needed, e.g. "." and ".2". */ | |
79bd622b | 2703 | int |
f7fdd7a1 | 2704 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
2705 | const cpp_token *token2) | |
6060326b | 2706 | { |
79bd622b | 2707 | enum cpp_ttype a = token1->type, b = token2->type; |
2708 | cppchar_t c; | |
6060326b | 2709 | |
79bd622b | 2710 | if (token1->flags & NAMED_OP) |
2711 | a = CPP_NAME; | |
2712 | if (token2->flags & NAMED_OP) | |
2713 | b = CPP_NAME; | |
6060326b | 2714 | |
79bd622b | 2715 | c = EOF; |
2716 | if (token2->flags & DIGRAPH) | |
ee6c4e4b | 2717 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
79bd622b | 2718 | else if (token_spellings[b].category == SPELL_OPERATOR) |
2719 | c = token_spellings[b].name[0]; | |
6060326b | 2720 | |
79bd622b | 2721 | /* Quickly get everything that can paste with an '='. */ |
ee6c4e4b | 2722 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
79bd622b | 2723 | return 1; |
6060326b | 2724 | |
79bd622b | 2725 | switch (a) |
6060326b | 2726 | { |
e58c07f7 | 2727 | case CPP_GREATER: return c == '>'; |
2728 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
79bd622b | 2729 | case CPP_PLUS: return c == '+'; |
2730 | case CPP_MINUS: return c == '-' || c == '>'; | |
2731 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
2732 | case CPP_MOD: return c == ':' || c == '>'; | |
2733 | case CPP_AND: return c == '&'; | |
2734 | case CPP_OR: return c == '|'; | |
2735 | case CPP_COLON: return c == ':' || c == '>'; | |
2736 | case CPP_DEREF: return c == '*'; | |
efdcc728 | 2737 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
79bd622b | 2738 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
2739 | case CPP_NAME: return ((b == CPP_NUMBER | |
2740 | && name_p (pfile, &token2->val.str)) | |
2741 | || b == CPP_NAME | |
2742 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
2743 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
2744 | || c == '.' || c == '+' || c == '-'); | |
2cbf1359 | 2745 | /* UCNs */ |
bc205914 | 2746 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
2747 | && b == CPP_NAME) | |
2cbf1359 | 2748 | || (CPP_OPTION (pfile, objc) |
bc205914 | 2749 | && token1->val.str.text[0] == '@' |
2cbf1359 | 2750 | && (b == CPP_NAME || b == CPP_STRING))); |
79bd622b | 2751 | default: break; |
6060326b | 2752 | } |
6060326b | 2753 | |
deb356cf | 2754 | return 0; |
6060326b | 2755 | } |
2756 | ||
79bd622b | 2757 | /* Output all the remaining tokens on the current line, and a newline |
f9b5f742 | 2758 | character, to FP. Leading whitespace is removed. If there are |
2759 | macros, special token padding is not performed. */ | |
6060326b | 2760 | void |
f7fdd7a1 | 2761 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
6060326b | 2762 | { |
f9b5f742 | 2763 | const cpp_token *token; |
7e842f95 | 2764 | |
f9b5f742 | 2765 | token = cpp_get_token (pfile); |
2766 | while (token->type != CPP_EOF) | |
7e842f95 | 2767 | { |
f9b5f742 | 2768 | cpp_output_token (token, fp); |
2769 | token = cpp_get_token (pfile); | |
2770 | if (token->flags & PREV_WHITE) | |
2771 | putc (' ', fp); | |
7e842f95 | 2772 | } |
2773 | ||
79bd622b | 2774 | putc ('\n', fp); |
f80e83a9 | 2775 | } |
6060326b | 2776 | |
c0770282 | 2777 | /* Return a string representation of all the remaining tokens on the |
2778 | current line. The result is allocated using xmalloc and must be | |
2779 | freed by the caller. */ | |
2780 | unsigned char * | |
2781 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
2782 | { | |
2783 | const cpp_token *token; | |
2784 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
2785 | unsigned int alloced = 120 + out; | |
2786 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
2787 | ||
2788 | /* If DIR_NAME is empty, there are no initial contents. */ | |
2789 | if (dir_name) | |
2790 | { | |
2791 | sprintf ((char *) result, "#%s ", dir_name); | |
2792 | out += 2; | |
2793 | } | |
2794 | ||
2795 | token = cpp_get_token (pfile); | |
2796 | while (token->type != CPP_EOF) | |
2797 | { | |
2798 | unsigned char *last; | |
2799 | /* Include room for a possible space and the terminating nul. */ | |
2800 | unsigned int len = cpp_token_len (token) + 2; | |
2801 | ||
2802 | if (out + len > alloced) | |
2803 | { | |
2804 | alloced *= 2; | |
2805 | if (out + len > alloced) | |
2806 | alloced = out + len; | |
2807 | result = (unsigned char *) xrealloc (result, alloced); | |
2808 | } | |
2809 | ||
2810 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
2811 | out = last - result; | |
2812 | ||
2813 | token = cpp_get_token (pfile); | |
2814 | if (token->flags & PREV_WHITE) | |
2815 | result[out++] = ' '; | |
2816 | } | |
2817 | ||
2818 | result[out] = '\0'; | |
2819 | return result; | |
2820 | } | |
2821 | ||
084163dc | 2822 | /* Memory buffers. Changing these three constants can have a dramatic |
2823 | effect on performance. The values here are reasonable defaults, | |
2824 | but might be tuned. If you adjust them, be sure to test across a | |
2825 | range of uses of cpplib, including heavy nested function-like macro | |
2826 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2827 | good tool for this). */ | |
2828 | #define MIN_BUFF_SIZE 8000 | |
1e0ef2fd | 2829 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
084163dc | 2830 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
2831 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
deb356cf | 2832 | |
1e0ef2fd | 2833 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
2834 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
2835 | #endif | |
2836 | ||
1785b647 | 2837 | /* Create a new allocation buffer. Place the control block at the end |
2838 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
06c92cbc | 2839 | static _cpp_buff * |
f7fdd7a1 | 2840 | new_buff (size_t len) |
06c92cbc | 2841 | { |
2842 | _cpp_buff *result; | |
1fdf6039 | 2843 | unsigned char *base; |
06c92cbc | 2844 | |
084163dc | 2845 | if (len < MIN_BUFF_SIZE) |
2846 | len = MIN_BUFF_SIZE; | |
198b48a0 | 2847 | len = CPP_ALIGN (len); |
06c92cbc | 2848 | |
61ed1f10 | 2849 | #ifdef ENABLE_VALGRIND_CHECKING |
2850 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff | |
2851 | struct first. */ | |
2852 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
2853 | base = XNEWVEC (unsigned char, len + slen); | |
2854 | result = (_cpp_buff *) base; | |
2855 | base += slen; | |
2856 | #else | |
720aca92 | 2857 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
06c92cbc | 2858 | result = (_cpp_buff *) (base + len); |
61ed1f10 | 2859 | #endif |
06c92cbc | 2860 | result->base = base; |
2861 | result->cur = base; | |
2862 | result->limit = base + len; | |
2863 | result->next = NULL; | |
2864 | return result; | |
2865 | } | |
2866 | ||
2867 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2868 | void | |
f7fdd7a1 | 2869 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
06c92cbc | 2870 | { |
2871 | _cpp_buff *end = buff; | |
2872 | ||
2873 | while (end->next) | |
2874 | end = end->next; | |
2875 | end->next = pfile->free_buffs; | |
2876 | pfile->free_buffs = buff; | |
2877 | } | |
2878 | ||
2879 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2880 | _cpp_buff * | |
f7fdd7a1 | 2881 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
06c92cbc | 2882 | { |
2883 | _cpp_buff *result, **p; | |
2884 | ||
2885 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2886 | { | |
4b31a107 | 2887 | size_t size; |
084163dc | 2888 | |
2889 | if (*p == NULL) | |
06c92cbc | 2890 | return new_buff (min_size); |
084163dc | 2891 | result = *p; |
2892 | size = result->limit - result->base; | |
2893 | /* Return a buffer that's big enough, but don't waste one that's | |
2894 | way too big. */ | |
4085c149 | 2895 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
06c92cbc | 2896 | break; |
2897 | } | |
2898 | ||
2899 | *p = result->next; | |
2900 | result->next = NULL; | |
2901 | result->cur = result->base; | |
2902 | return result; | |
2903 | } | |
2904 | ||
20dd417a | 2905 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 2906 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
2907 | the excess bytes to the new buffer. Chains the new buffer after | |
2908 | BUFF, and returns the new buffer. */ | |
06c92cbc | 2909 | _cpp_buff * |
f7fdd7a1 | 2910 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
06c92cbc | 2911 | { |
4b31a107 | 2912 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
e6a5f963 | 2913 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
06c92cbc | 2914 | |
e6a5f963 | 2915 | buff->next = new_buff; |
2916 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
2917 | return new_buff; | |
2918 | } | |
2919 | ||
20dd417a | 2920 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 2921 | remaining bytes of the buffer pointed to by BUFF, and at least |
2922 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
2923 | Chains the new buffer before the buffer pointed to by BUFF, and | |
2924 | updates the pointer to point to the new buffer. */ | |
2925 | void | |
f7fdd7a1 | 2926 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
e6a5f963 | 2927 | { |
2928 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
2929 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
2930 | ||
2931 | new_buff = _cpp_get_buff (pfile, size); | |
2932 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
2933 | new_buff->next = old_buff; | |
2934 | *pbuff = new_buff; | |
06c92cbc | 2935 | } |
2936 | ||
2937 | /* Free a chain of buffers starting at BUFF. */ | |
2938 | void | |
f82b06e0 | 2939 | _cpp_free_buff (_cpp_buff *buff) |
06c92cbc | 2940 | { |
2941 | _cpp_buff *next; | |
2942 | ||
2943 | for (; buff; buff = next) | |
2944 | { | |
2945 | next = buff->next; | |
61ed1f10 | 2946 | #ifdef ENABLE_VALGRIND_CHECKING |
2947 | free (buff); | |
2948 | #else | |
06c92cbc | 2949 | free (buff->base); |
61ed1f10 | 2950 | #endif |
06c92cbc | 2951 | } |
2952 | } | |
deb356cf | 2953 | |
1fdf6039 | 2954 | /* Allocate permanent, unaligned storage of length LEN. */ |
2955 | unsigned char * | |
f7fdd7a1 | 2956 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
1fdf6039 | 2957 | { |
2958 | _cpp_buff *buff = pfile->u_buff; | |
2959 | unsigned char *result = buff->cur; | |
2960 | ||
2961 | if (len > (size_t) (buff->limit - result)) | |
2962 | { | |
2963 | buff = _cpp_get_buff (pfile, len); | |
2964 | buff->next = pfile->u_buff; | |
2965 | pfile->u_buff = buff; | |
2966 | result = buff->cur; | |
2967 | } | |
2968 | ||
2969 | buff->cur = result + len; | |
2970 | return result; | |
2971 | } | |
2972 | ||
1e0ef2fd | 2973 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
2974 | That buffer is used for growing allocations when saving macro | |
2975 | replacement lists in a #define, and when parsing an answer to an | |
2976 | assertion in #assert, #unassert or #if (and therefore possibly | |
2977 | whilst expanding macros). It therefore must not be used by any | |
2978 | code that they might call: specifically the lexer and the guts of | |
2979 | the macro expander. | |
2980 | ||
2981 | All existing other uses clearly fit this restriction: storing | |
2982 | registered pragmas during initialization. */ | |
79bd622b | 2983 | unsigned char * |
f7fdd7a1 | 2984 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
89b05ef6 | 2985 | { |
e6a5f963 | 2986 | _cpp_buff *buff = pfile->a_buff; |
2987 | unsigned char *result = buff->cur; | |
89b05ef6 | 2988 | |
e6a5f963 | 2989 | if (len > (size_t) (buff->limit - result)) |
89b05ef6 | 2990 | { |
e6a5f963 | 2991 | buff = _cpp_get_buff (pfile, len); |
2992 | buff->next = pfile->a_buff; | |
2993 | pfile->a_buff = buff; | |
2994 | result = buff->cur; | |
89b05ef6 | 2995 | } |
f80e83a9 | 2996 | |
e6a5f963 | 2997 | buff->cur = result + len; |
79bd622b | 2998 | return result; |
f80e83a9 | 2999 | } |
c39ed964 | 3000 | |
3001 | /* Say which field of TOK is in use. */ | |
3002 | ||
3003 | enum cpp_token_fld_kind | |
3004 | cpp_token_val_index (cpp_token *tok) | |
3005 | { | |
3006 | switch (TOKEN_SPELL (tok)) | |
3007 | { | |
3008 | case SPELL_IDENT: | |
3009 | return CPP_TOKEN_FLD_NODE; | |
3010 | case SPELL_LITERAL: | |
3011 | return CPP_TOKEN_FLD_STR; | |
941f2388 | 3012 | case SPELL_OPERATOR: |
3013 | if (tok->type == CPP_PASTE) | |
2ee04baa | 3014 | return CPP_TOKEN_FLD_TOKEN_NO; |
941f2388 | 3015 | else |
3016 | return CPP_TOKEN_FLD_NONE; | |
c39ed964 | 3017 | case SPELL_NONE: |
3018 | if (tok->type == CPP_MACRO_ARG) | |
3019 | return CPP_TOKEN_FLD_ARG_NO; | |
3020 | else if (tok->type == CPP_PADDING) | |
3021 | return CPP_TOKEN_FLD_SOURCE; | |
d6d3c909 | 3022 | else if (tok->type == CPP_PRAGMA) |
b75b98aa | 3023 | return CPP_TOKEN_FLD_PRAGMA; |
c39ed964 | 3024 | /* else fall through */ |
3025 | default: | |
3026 | return CPP_TOKEN_FLD_NONE; | |
3027 | } | |
3028 | } | |
6ea2c7a3 | 3029 | |
3030 | /* All tokens lexed in R after calling this function will be forced to have | |
3031 | their source_location the same as the location referenced by P, until | |
3032 | cpp_stop_forcing_token_locations is called for R. */ | |
3033 | ||
3034 | void | |
3035 | cpp_force_token_locations (cpp_reader *r, source_location *p) | |
3036 | { | |
3037 | r->forced_token_location_p = p; | |
3038 | } | |
3039 | ||
3040 | /* Go back to assigning locations naturally for lexed tokens. */ | |
3041 | ||
3042 | void | |
3043 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
3044 | { | |
3045 | r->forced_token_location_p = NULL; | |
3046 | } |