]>
Commit | Line | Data |
---|---|---|
0578f103 | 1 | /* CPP Library - lexical analysis. |
806a3d45 | 2 | Copyright (C) 2000-2014 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 | ||
3efb5d22 | 266 | /* Disable on Solaris 2/x86 until the following problem can be properly |
8ab29ece | 267 | autoconfed: |
268 | ||
8ab29ece | 269 | The Solaris 10+ assembler tags objects with the instruction set |
270 | extensions used, so SSE4.2 executables cannot run on machines that | |
271 | don't support that extension. */ | |
272 | ||
273 | #if (GCC_VERSION >= 4005) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__)) | |
2431e8ba | 274 | |
275 | /* Replicated character data to be shared between implementations. | |
276 | Recall that outside of a context with vector support we can't | |
277 | define compatible vector types, therefore these are all defined | |
278 | in terms of raw characters. */ | |
279 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { | |
280 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
281 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, | |
282 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
283 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, | |
284 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
285 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, | |
286 | { '?', '?', '?', '?', '?', '?', '?', '?', | |
287 | '?', '?', '?', '?', '?', '?', '?', '?' }, | |
288 | }; | |
289 | ||
290 | /* A version of the fast scanner using MMX vectorized byte compare insns. | |
291 | ||
292 | This uses the PMOVMSKB instruction which was introduced with "MMX2", | |
cef70b8a | 293 | which was packaged into SSE1; it is also present in the AMD MMX |
2431e8ba | 294 | extension. Mark the function as using "sse" so that we emit a real |
295 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ | |
296 | ||
297 | static const uchar * | |
298 | #ifndef __SSE__ | |
299 | __attribute__((__target__("sse"))) | |
300 | #endif | |
301 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
302 | { | |
303 | typedef char v8qi __attribute__ ((__vector_size__ (8))); | |
304 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); | |
305 | ||
306 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; | |
307 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; | |
308 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; | |
309 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; | |
310 | ||
311 | unsigned int misalign, found, mask; | |
312 | const v8qi *p; | |
313 | v8qi data, t, c; | |
314 | ||
315 | /* Align the source pointer. While MMX doesn't generate unaligned data | |
316 | faults, this allows us to safely scan to the end of the buffer without | |
317 | reading beyond the end of the last page. */ | |
318 | misalign = (uintptr_t)s & 7; | |
319 | p = (const v8qi *)((uintptr_t)s & -8); | |
320 | data = *p; | |
321 | ||
322 | /* Create a mask for the bytes that are valid within the first | |
323 | 16-byte block. The Idea here is that the AND with the mask | |
324 | within the loop is "free", since we need some AND or TEST | |
325 | insn in order to set the flags for the branch anyway. */ | |
326 | mask = -1u << misalign; | |
327 | ||
328 | /* Main loop processing 8 bytes at a time. */ | |
329 | goto start; | |
330 | do | |
331 | { | |
332 | data = *++p; | |
333 | mask = -1; | |
334 | ||
335 | start: | |
336 | t = __builtin_ia32_pcmpeqb(data, repl_nl); | |
337 | c = __builtin_ia32_pcmpeqb(data, repl_cr); | |
338 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
339 | c = __builtin_ia32_pcmpeqb(data, repl_bs); | |
340 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
341 | c = __builtin_ia32_pcmpeqb(data, repl_qm); | |
342 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
343 | found = __builtin_ia32_pmovmskb (t); | |
344 | found &= mask; | |
345 | } | |
346 | while (!found); | |
347 | ||
348 | __builtin_ia32_emms (); | |
349 | ||
350 | /* FOUND contains 1 in bits for which we matched a relevant | |
351 | character. Conversion to the byte index is trivial. */ | |
352 | found = __builtin_ctz(found); | |
353 | return (const uchar *)p + found; | |
354 | } | |
355 | ||
356 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ | |
357 | ||
358 | static const uchar * | |
359 | #ifndef __SSE2__ | |
360 | __attribute__((__target__("sse2"))) | |
361 | #endif | |
362 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
363 | { | |
364 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
365 | ||
366 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; | |
367 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; | |
368 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; | |
369 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; | |
370 | ||
371 | unsigned int misalign, found, mask; | |
372 | const v16qi *p; | |
373 | v16qi data, t; | |
374 | ||
375 | /* Align the source pointer. */ | |
376 | misalign = (uintptr_t)s & 15; | |
377 | p = (const v16qi *)((uintptr_t)s & -16); | |
378 | data = *p; | |
379 | ||
380 | /* Create a mask for the bytes that are valid within the first | |
381 | 16-byte block. The Idea here is that the AND with the mask | |
382 | within the loop is "free", since we need some AND or TEST | |
383 | insn in order to set the flags for the branch anyway. */ | |
384 | mask = -1u << misalign; | |
385 | ||
386 | /* Main loop processing 16 bytes at a time. */ | |
387 | goto start; | |
388 | do | |
389 | { | |
390 | data = *++p; | |
391 | mask = -1; | |
392 | ||
393 | start: | |
394 | t = __builtin_ia32_pcmpeqb128(data, repl_nl); | |
395 | t |= __builtin_ia32_pcmpeqb128(data, repl_cr); | |
396 | t |= __builtin_ia32_pcmpeqb128(data, repl_bs); | |
397 | t |= __builtin_ia32_pcmpeqb128(data, repl_qm); | |
398 | found = __builtin_ia32_pmovmskb128 (t); | |
399 | found &= mask; | |
400 | } | |
401 | while (!found); | |
402 | ||
403 | /* FOUND contains 1 in bits for which we matched a relevant | |
404 | character. Conversion to the byte index is trivial. */ | |
405 | found = __builtin_ctz(found); | |
406 | return (const uchar *)p + found; | |
407 | } | |
408 | ||
b315ae35 | 409 | #ifdef HAVE_SSE4 |
2431e8ba | 410 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
411 | ||
412 | static const uchar * | |
413 | #ifndef __SSE4_2__ | |
414 | __attribute__((__target__("sse4.2"))) | |
415 | #endif | |
416 | search_line_sse42 (const uchar *s, const uchar *end) | |
417 | { | |
418 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
419 | static const v16qi search = { '\n', '\r', '?', '\\' }; | |
420 | ||
421 | uintptr_t si = (uintptr_t)s; | |
422 | uintptr_t index; | |
423 | ||
424 | /* Check for unaligned input. */ | |
425 | if (si & 15) | |
426 | { | |
112f073c | 427 | v16qi sv; |
428 | ||
2431e8ba | 429 | if (__builtin_expect (end - s < 16, 0) |
430 | && __builtin_expect ((si & 0xfff) > 0xff0, 0)) | |
431 | { | |
432 | /* There are less than 16 bytes left in the buffer, and less | |
433 | than 16 bytes left on the page. Reading 16 bytes at this | |
434 | point might generate a spurious page fault. Defer to the | |
435 | SSE2 implementation, which already handles alignment. */ | |
436 | return search_line_sse2 (s, end); | |
437 | } | |
438 | ||
439 | /* ??? The builtin doesn't understand that the PCMPESTRI read from | |
440 | memory need not be aligned. */ | |
112f073c | 441 | sv = __builtin_ia32_loaddqu ((const char *) s); |
442 | index = __builtin_ia32_pcmpestri128 (search, 4, sv, 16, 0); | |
443 | ||
2431e8ba | 444 | if (__builtin_expect (index < 16, 0)) |
445 | goto found; | |
446 | ||
447 | /* Advance the pointer to an aligned address. We will re-scan a | |
448 | few bytes, but we no longer need care for reading past the | |
449 | end of a page, since we're guaranteed a match. */ | |
450 | s = (const uchar *)((si + 16) & -16); | |
451 | } | |
452 | ||
453 | /* Main loop, processing 16 bytes at a time. By doing the whole loop | |
454 | in inline assembly, we can make proper use of the flags set. */ | |
455 | __asm ( "sub $16, %1\n" | |
456 | " .balign 16\n" | |
457 | "0: add $16, %1\n" | |
458 | " %vpcmpestri $0, (%1), %2\n" | |
459 | " jnc 0b" | |
460 | : "=&c"(index), "+r"(s) | |
461 | : "x"(search), "a"(4), "d"(16)); | |
462 | ||
463 | found: | |
464 | return s + index; | |
465 | } | |
466 | ||
b315ae35 | 467 | #else |
468 | /* Work around out-dated assemblers without sse4 support. */ | |
469 | #define search_line_sse42 search_line_sse2 | |
470 | #endif | |
471 | ||
2431e8ba | 472 | /* Check the CPU capabilities. */ |
473 | ||
474 | #include "../gcc/config/i386/cpuid.h" | |
475 | ||
476 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); | |
477 | static search_line_fast_type search_line_fast; | |
478 | ||
b735cc56 | 479 | #define HAVE_init_vectorized_lexer 1 |
480 | static inline void | |
2431e8ba | 481 | init_vectorized_lexer (void) |
482 | { | |
483 | unsigned dummy, ecx = 0, edx = 0; | |
484 | search_line_fast_type impl = search_line_acc_char; | |
485 | int minimum = 0; | |
486 | ||
487 | #if defined(__SSE4_2__) | |
488 | minimum = 3; | |
489 | #elif defined(__SSE2__) | |
490 | minimum = 2; | |
cef70b8a | 491 | #elif defined(__SSE__) |
2431e8ba | 492 | minimum = 1; |
493 | #endif | |
494 | ||
495 | if (minimum == 3) | |
496 | impl = search_line_sse42; | |
497 | else if (__get_cpuid (1, &dummy, &dummy, &ecx, &edx) || minimum == 2) | |
498 | { | |
499 | if (minimum == 3 || (ecx & bit_SSE4_2)) | |
500 | impl = search_line_sse42; | |
501 | else if (minimum == 2 || (edx & bit_SSE2)) | |
502 | impl = search_line_sse2; | |
503 | else if (minimum == 1 || (edx & bit_SSE)) | |
504 | impl = search_line_mmx; | |
505 | } | |
506 | else if (__get_cpuid (0x80000001, &dummy, &dummy, &dummy, &edx)) | |
507 | { | |
85303bd5 | 508 | if (minimum == 1 |
509 | || (edx & (bit_MMXEXT | bit_CMOV)) == (bit_MMXEXT | bit_CMOV)) | |
2431e8ba | 510 | impl = search_line_mmx; |
511 | } | |
512 | ||
513 | search_line_fast = impl; | |
514 | } | |
515 | ||
70ebee13 | 516 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) |
2431e8ba | 517 | |
518 | /* A vection of the fast scanner using AltiVec vectorized byte compares. */ | |
519 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, | |
520 | so we can't compile this function without -maltivec on the command line | |
521 | (or implied by some other switch). */ | |
522 | ||
523 | static const uchar * | |
524 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
525 | { | |
526 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
527 | ||
528 | const vc repl_nl = { | |
529 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
530 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
531 | }; | |
532 | const vc repl_cr = { | |
533 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
534 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
535 | }; | |
536 | const vc repl_bs = { | |
537 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
538 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
539 | }; | |
540 | const vc repl_qm = { | |
541 | '?', '?', '?', '?', '?', '?', '?', '?', | |
542 | '?', '?', '?', '?', '?', '?', '?', '?', | |
543 | }; | |
544 | const vc ones = { | |
545 | -1, -1, -1, -1, -1, -1, -1, -1, | |
546 | -1, -1, -1, -1, -1, -1, -1, -1, | |
547 | }; | |
548 | const vc zero = { 0 }; | |
549 | ||
550 | vc data, mask, t; | |
551 | ||
552 | /* Altivec loads automatically mask addresses with -16. This lets us | |
553 | issue the first load as early as possible. */ | |
554 | data = __builtin_vec_ld(0, (const vc *)s); | |
555 | ||
556 | /* Discard bytes before the beginning of the buffer. Do this by | |
557 | beginning with all ones and shifting in zeros according to the | |
558 | mis-alignment. The LVSR instruction pulls the exact shift we | |
559 | want from the address. */ | |
eb992054 | 560 | #ifdef __BIG_ENDIAN__ |
2431e8ba | 561 | mask = __builtin_vec_lvsr(0, s); |
562 | mask = __builtin_vec_perm(zero, ones, mask); | |
eb992054 | 563 | #else |
564 | mask = __builtin_vec_lvsl(0, s); | |
565 | mask = __builtin_vec_perm(ones, zero, mask); | |
566 | #endif | |
2431e8ba | 567 | data &= mask; |
568 | ||
569 | /* While altivec loads mask addresses, we still need to align S so | |
570 | that the offset we compute at the end is correct. */ | |
571 | s = (const uchar *)((uintptr_t)s & -16); | |
572 | ||
573 | /* Main loop processing 16 bytes at a time. */ | |
574 | goto start; | |
575 | do | |
576 | { | |
577 | vc m_nl, m_cr, m_bs, m_qm; | |
578 | ||
579 | s += 16; | |
580 | data = __builtin_vec_ld(0, (const vc *)s); | |
581 | ||
582 | start: | |
583 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
584 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
585 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
586 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
587 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
588 | ||
589 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
590 | characters. We want to exit the loop if any byte in T is non-zero. | |
591 | Below is the expansion of vec_any_ne(t, zero). */ | |
592 | } | |
593 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
594 | ||
595 | { | |
596 | #define N (sizeof(vc) / sizeof(long)) | |
597 | ||
2431e8ba | 598 | union { |
599 | vc v; | |
5fe44548 | 600 | /* Statically assert that N is 2 or 4. */ |
601 | unsigned long l[(N == 2 || N == 4) ? N : -1]; | |
2431e8ba | 602 | } u; |
603 | unsigned long l, i = 0; | |
604 | ||
605 | u.v = t; | |
606 | ||
607 | /* Find the first word of T that is non-zero. */ | |
608 | switch (N) | |
609 | { | |
610 | case 4: | |
611 | l = u.l[i++]; | |
612 | if (l != 0) | |
613 | break; | |
614 | s += sizeof(unsigned long); | |
615 | l = u.l[i++]; | |
616 | if (l != 0) | |
617 | break; | |
618 | s += sizeof(unsigned long); | |
619 | case 2: | |
620 | l = u.l[i++]; | |
621 | if (l != 0) | |
622 | break; | |
623 | s += sizeof(unsigned long); | |
624 | l = u.l[i]; | |
625 | } | |
626 | ||
627 | /* L now contains 0xff in bytes for which we matched one of the | |
628 | relevant characters. We can find the byte index by finding | |
629 | its bit index and dividing by 8. */ | |
eb992054 | 630 | #ifdef __BIG_ENDIAN__ |
2431e8ba | 631 | l = __builtin_clzl(l) >> 3; |
eb992054 | 632 | #else |
633 | l = __builtin_ctzl(l) >> 3; | |
634 | #endif | |
2431e8ba | 635 | return s + l; |
636 | ||
637 | #undef N | |
638 | } | |
639 | } | |
640 | ||
c434932e | 641 | #elif defined (__ARM_NEON__) |
642 | #include "arm_neon.h" | |
643 | ||
644 | static const uchar * | |
645 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
646 | { | |
647 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); | |
648 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); | |
649 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); | |
650 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); | |
651 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); | |
652 | ||
653 | unsigned int misalign, found, mask; | |
654 | const uint8_t *p; | |
655 | uint8x16_t data; | |
656 | ||
657 | /* Align the source pointer. */ | |
658 | misalign = (uintptr_t)s & 15; | |
659 | p = (const uint8_t *)((uintptr_t)s & -16); | |
660 | data = vld1q_u8 (p); | |
661 | ||
662 | /* Create a mask for the bytes that are valid within the first | |
663 | 16-byte block. The Idea here is that the AND with the mask | |
664 | within the loop is "free", since we need some AND or TEST | |
665 | insn in order to set the flags for the branch anyway. */ | |
666 | mask = (-1u << misalign) & 0xffff; | |
667 | ||
668 | /* Main loop, processing 16 bytes at a time. */ | |
669 | goto start; | |
670 | ||
671 | do | |
672 | { | |
673 | uint8x8_t l; | |
674 | uint16x4_t m; | |
675 | uint32x2_t n; | |
676 | uint8x16_t t, u, v, w; | |
677 | ||
678 | p += 16; | |
679 | data = vld1q_u8 (p); | |
680 | mask = 0xffff; | |
681 | ||
682 | start: | |
683 | t = vceqq_u8 (data, repl_nl); | |
684 | u = vceqq_u8 (data, repl_cr); | |
685 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); | |
686 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); | |
687 | t = vandq_u8 (vorrq_u8 (v, w), xmask); | |
688 | l = vpadd_u8 (vget_low_u8 (t), vget_high_u8 (t)); | |
689 | m = vpaddl_u8 (l); | |
690 | n = vpaddl_u16 (m); | |
691 | ||
692 | found = vget_lane_u32 ((uint32x2_t) vorr_u64 ((uint64x1_t) n, | |
693 | vshr_n_u64 ((uint64x1_t) n, 24)), 0); | |
694 | found &= mask; | |
695 | } | |
696 | while (!found); | |
697 | ||
698 | /* FOUND contains 1 in bits for which we matched a relevant | |
699 | character. Conversion to the byte index is trivial. */ | |
700 | found = __builtin_ctz (found); | |
701 | return (const uchar *)p + found; | |
702 | } | |
703 | ||
2431e8ba | 704 | #else |
705 | ||
706 | /* We only have one accellerated alternative. Use a direct call so that | |
707 | we encourage inlining. */ | |
708 | ||
709 | #define search_line_fast search_line_acc_char | |
710 | ||
711 | #endif | |
712 | ||
b735cc56 | 713 | /* Initialize the lexer if needed. */ |
714 | ||
715 | void | |
716 | _cpp_init_lexer (void) | |
717 | { | |
718 | #ifdef HAVE_init_vectorized_lexer | |
719 | init_vectorized_lexer (); | |
720 | #endif | |
721 | } | |
722 | ||
a54e0bf8 | 723 | /* Returns with a logical line that contains no escaped newlines or |
724 | trigraphs. This is a time-critical inner loop. */ | |
725 | void | |
f7fdd7a1 | 726 | _cpp_clean_line (cpp_reader *pfile) |
0578f103 | 727 | { |
a54e0bf8 | 728 | cpp_buffer *buffer; |
729 | const uchar *s; | |
730 | uchar c, *d, *p; | |
1e0ef2fd | 731 | |
a54e0bf8 | 732 | buffer = pfile->buffer; |
733 | buffer->cur_note = buffer->notes_used = 0; | |
734 | buffer->cur = buffer->line_base = buffer->next_line; | |
735 | buffer->need_line = false; | |
2431e8ba | 736 | s = buffer->next_line; |
1e0ef2fd | 737 | |
a54e0bf8 | 738 | if (!buffer->from_stage3) |
0578f103 | 739 | { |
5008f5c5 | 740 | const uchar *pbackslash = NULL; |
741 | ||
2431e8ba | 742 | /* Fast path. This is the common case of an un-escaped line with |
54d3be91 | 743 | no trigraphs. The primary win here is by not writing any |
744 | data back to memory until we have to. */ | |
2431e8ba | 745 | while (1) |
54d3be91 | 746 | { |
2431e8ba | 747 | /* Perform an optimized search for \n, \r, \\, ?. */ |
748 | s = search_line_fast (s, buffer->rlimit); | |
54d3be91 | 749 | |
2431e8ba | 750 | c = *s; |
751 | if (c == '\\') | |
752 | { | |
753 | /* Record the location of the backslash and continue. */ | |
754 | pbackslash = s++; | |
54d3be91 | 755 | } |
2431e8ba | 756 | else if (__builtin_expect (c == '?', 0)) |
54d3be91 | 757 | { |
2431e8ba | 758 | if (__builtin_expect (s[1] == '?', false) |
759 | && _cpp_trigraph_map[s[2]]) | |
54d3be91 | 760 | { |
2431e8ba | 761 | /* Have a trigraph. We may or may not have to convert |
762 | it. Add a line note regardless, for -Wtrigraphs. */ | |
763 | add_line_note (buffer, s, s[2]); | |
764 | if (CPP_OPTION (pfile, trigraphs)) | |
765 | { | |
766 | /* We do, and that means we have to switch to the | |
767 | slow path. */ | |
768 | d = (uchar *) s; | |
769 | *d = _cpp_trigraph_map[s[2]]; | |
770 | s += 2; | |
771 | goto slow_path; | |
772 | } | |
54d3be91 | 773 | } |
2431e8ba | 774 | /* Not a trigraph. Continue on fast-path. */ |
775 | s++; | |
54d3be91 | 776 | } |
2431e8ba | 777 | else |
778 | break; | |
54d3be91 | 779 | } |
780 | ||
2431e8ba | 781 | /* This must be \r or \n. We're either done, or we'll be forced |
782 | to write back to the buffer and continue on the slow path. */ | |
783 | d = (uchar *) s; | |
784 | ||
785 | if (__builtin_expect (s == buffer->rlimit, false)) | |
786 | goto done; | |
787 | ||
788 | /* DOS line ending? */ | |
789 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
790 | { | |
791 | s++; | |
792 | if (s == buffer->rlimit) | |
793 | goto done; | |
794 | } | |
795 | ||
796 | if (__builtin_expect (pbackslash == NULL, true)) | |
797 | goto done; | |
798 | ||
799 | /* Check for escaped newline. */ | |
800 | p = d; | |
801 | while (is_nvspace (p[-1])) | |
802 | p--; | |
803 | if (p - 1 != pbackslash) | |
804 | goto done; | |
805 | ||
806 | /* Have an escaped newline; process it and proceed to | |
807 | the slow path. */ | |
808 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
809 | d = p - 2; | |
810 | buffer->next_line = p - 1; | |
a54e0bf8 | 811 | |
2431e8ba | 812 | slow_path: |
813 | while (1) | |
4b912310 | 814 | { |
a54e0bf8 | 815 | c = *++s; |
816 | *++d = c; | |
817 | ||
818 | if (c == '\n' || c == '\r') | |
819 | { | |
2431e8ba | 820 | /* Handle DOS line endings. */ |
a54e0bf8 | 821 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
822 | s++; | |
823 | if (s == buffer->rlimit) | |
824 | break; | |
825 | ||
826 | /* Escaped? */ | |
827 | p = d; | |
828 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
829 | p--; | |
830 | if (p == buffer->next_line || p[-1] != '\\') | |
831 | break; | |
832 | ||
aad4a87f | 833 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
a54e0bf8 | 834 | d = p - 2; |
835 | buffer->next_line = p - 1; | |
836 | } | |
837 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
838 | { | |
839 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
aad4a87f | 840 | add_line_note (buffer, d, s[2]); |
a54e0bf8 | 841 | if (CPP_OPTION (pfile, trigraphs)) |
842 | { | |
843 | *d = _cpp_trigraph_map[s[2]]; | |
844 | s += 2; | |
845 | } | |
846 | } | |
4b912310 | 847 | } |
0578f103 | 848 | } |
a54e0bf8 | 849 | else |
850 | { | |
2431e8ba | 851 | while (*s != '\n' && *s != '\r') |
a54e0bf8 | 852 | s++; |
a54e0bf8 | 853 | d = (uchar *) s; |
854 | ||
855 | /* Handle DOS line endings. */ | |
856 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
857 | s++; | |
858 | } | |
338fa5f7 | 859 | |
54d3be91 | 860 | done: |
a54e0bf8 | 861 | *d = '\n'; |
aad4a87f | 862 | /* A sentinel note that should never be processed. */ |
863 | add_line_note (buffer, d + 1, '\n'); | |
a54e0bf8 | 864 | buffer->next_line = s + 1; |
0578f103 | 865 | } |
866 | ||
3078f2b2 | 867 | /* Return true if the trigraph indicated by NOTE should be warned |
868 | about in a comment. */ | |
869 | static bool | |
f7fdd7a1 | 870 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
3078f2b2 | 871 | { |
872 | const uchar *p; | |
873 | ||
874 | /* Within comments we don't warn about trigraphs, unless the | |
875 | trigraph forms an escaped newline, as that may change | |
7ef5b942 | 876 | behavior. */ |
3078f2b2 | 877 | if (note->type != '/') |
878 | return false; | |
879 | ||
880 | /* If -trigraphs, then this was an escaped newline iff the next note | |
881 | is coincident. */ | |
882 | if (CPP_OPTION (pfile, trigraphs)) | |
883 | return note[1].pos == note->pos; | |
884 | ||
885 | /* Otherwise, see if this forms an escaped newline. */ | |
886 | p = note->pos + 3; | |
887 | while (is_nvspace (*p)) | |
888 | p++; | |
889 | ||
890 | /* There might have been escaped newlines between the trigraph and the | |
891 | newline we found. Hence the position test. */ | |
892 | return (*p == '\n' && p < note[1].pos); | |
893 | } | |
894 | ||
a54e0bf8 | 895 | /* Process the notes created by add_line_note as far as the current |
896 | location. */ | |
897 | void | |
f7fdd7a1 | 898 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
0578f103 | 899 | { |
c808d026 | 900 | cpp_buffer *buffer = pfile->buffer; |
901 | ||
a54e0bf8 | 902 | for (;;) |
f80e83a9 | 903 | { |
a54e0bf8 | 904 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
905 | unsigned int col; | |
396ffa86 | 906 | |
a54e0bf8 | 907 | if (note->pos > buffer->cur) |
908 | break; | |
396ffa86 | 909 | |
a54e0bf8 | 910 | buffer->cur_note++; |
911 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
435fb09b | 912 | |
aad4a87f | 913 | if (note->type == '\\' || note->type == ' ') |
a54e0bf8 | 914 | { |
aad4a87f | 915 | if (note->type == ' ' && !in_comment) |
dbddc569 | 916 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
a54e0bf8 | 917 | "backslash and newline separated by space"); |
aad4a87f | 918 | |
a54e0bf8 | 919 | if (buffer->next_line > buffer->rlimit) |
1e0ef2fd | 920 | { |
dbddc569 | 921 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
a54e0bf8 | 922 | "backslash-newline at end of file"); |
923 | /* Prevent "no newline at end of file" warning. */ | |
924 | buffer->next_line = buffer->rlimit; | |
1e0ef2fd | 925 | } |
a54e0bf8 | 926 | |
927 | buffer->line_base = note->pos; | |
610625e3 | 928 | CPP_INCREMENT_LINE (pfile, 0); |
338fa5f7 | 929 | } |
aad4a87f | 930 | else if (_cpp_trigraph_map[note->type]) |
931 | { | |
3078f2b2 | 932 | if (CPP_OPTION (pfile, warn_trigraphs) |
933 | && (!in_comment || warn_in_comment (pfile, note))) | |
aad4a87f | 934 | { |
935 | if (CPP_OPTION (pfile, trigraphs)) | |
3a79f5da | 936 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
937 | pfile->line_table->highest_line, col, | |
938 | "trigraph ??%c converted to %c", | |
939 | note->type, | |
940 | (int) _cpp_trigraph_map[note->type]); | |
aad4a87f | 941 | else |
1542b1ef | 942 | { |
3a79f5da | 943 | cpp_warning_with_line |
944 | (pfile, CPP_W_TRIGRAPHS, | |
945 | pfile->line_table->highest_line, col, | |
1542b1ef | 946 | "trigraph ??%c ignored, use -trigraphs to enable", |
947 | note->type); | |
948 | } | |
aad4a87f | 949 | } |
950 | } | |
3a45011c | 951 | else if (note->type == 0) |
952 | /* Already processed in lex_raw_string. */; | |
aad4a87f | 953 | else |
954 | abort (); | |
f80e83a9 | 955 | } |
0578f103 | 956 | } |
957 | ||
338fa5f7 | 958 | /* Skip a C-style block comment. We find the end of the comment by |
959 | seeing if an asterisk is before every '/' we encounter. Returns | |
edaf8cb5 | 960 | nonzero if comment terminated by EOF, zero otherwise. |
961 | ||
962 | Buffer->cur points to the initial asterisk of the comment. */ | |
a54e0bf8 | 963 | bool |
f7fdd7a1 | 964 | _cpp_skip_block_comment (cpp_reader *pfile) |
0578f103 | 965 | { |
f80e83a9 | 966 | cpp_buffer *buffer = pfile->buffer; |
54d3be91 | 967 | const uchar *cur = buffer->cur; |
968 | uchar c; | |
338fa5f7 | 969 | |
54d3be91 | 970 | cur++; |
971 | if (*cur == '/') | |
972 | cur++; | |
338fa5f7 | 973 | |
a54e0bf8 | 974 | for (;;) |
975 | { | |
338fa5f7 | 976 | /* People like decorating comments with '*', so check for '/' |
977 | instead for efficiency. */ | |
54d3be91 | 978 | c = *cur++; |
979 | ||
f80e83a9 | 980 | if (c == '/') |
0578f103 | 981 | { |
54d3be91 | 982 | if (cur[-2] == '*') |
338fa5f7 | 983 | break; |
f80e83a9 | 984 | |
338fa5f7 | 985 | /* Warn about potential nested comments, but not if the '/' |
3fb1e43b | 986 | comes immediately before the true comment delimiter. |
f80e83a9 | 987 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 988 | if (CPP_OPTION (pfile, warn_comments) |
54d3be91 | 989 | && cur[0] == '*' && cur[1] != '/') |
990 | { | |
991 | buffer->cur = cur; | |
3a79f5da | 992 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
993 | pfile->line_table->highest_line, | |
994 | CPP_BUF_COL (buffer), | |
995 | "\"/*\" within comment"); | |
54d3be91 | 996 | } |
0578f103 | 997 | } |
a54e0bf8 | 998 | else if (c == '\n') |
999 | { | |
610625e3 | 1000 | unsigned int cols; |
54d3be91 | 1001 | buffer->cur = cur - 1; |
a54e0bf8 | 1002 | _cpp_process_line_notes (pfile, true); |
1003 | if (buffer->next_line >= buffer->rlimit) | |
1004 | return true; | |
1005 | _cpp_clean_line (pfile); | |
610625e3 | 1006 | |
1007 | cols = buffer->next_line - buffer->line_base; | |
1008 | CPP_INCREMENT_LINE (pfile, cols); | |
1009 | ||
54d3be91 | 1010 | cur = buffer->cur; |
a54e0bf8 | 1011 | } |
0578f103 | 1012 | } |
f80e83a9 | 1013 | |
54d3be91 | 1014 | buffer->cur = cur; |
3078f2b2 | 1015 | _cpp_process_line_notes (pfile, true); |
a54e0bf8 | 1016 | return false; |
0578f103 | 1017 | } |
1018 | ||
1c124f85 | 1019 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
d10cfa8d | 1020 | terminating newline. Handles escaped newlines. Returns nonzero |
1c124f85 | 1021 | if a multiline comment. */ |
f80e83a9 | 1022 | static int |
f7fdd7a1 | 1023 | skip_line_comment (cpp_reader *pfile) |
0578f103 | 1024 | { |
f669338a | 1025 | cpp_buffer *buffer = pfile->buffer; |
4999c35b | 1026 | source_location orig_line = pfile->line_table->highest_line; |
f80e83a9 | 1027 | |
a54e0bf8 | 1028 | while (*buffer->cur != '\n') |
1029 | buffer->cur++; | |
1c124f85 | 1030 | |
a54e0bf8 | 1031 | _cpp_process_line_notes (pfile, true); |
dbddc569 | 1032 | return orig_line != pfile->line_table->highest_line; |
f80e83a9 | 1033 | } |
0578f103 | 1034 | |
a54e0bf8 | 1035 | /* Skips whitespace, saving the next non-whitespace character. */ |
b86584f6 | 1036 | static void |
f7fdd7a1 | 1037 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
f80e83a9 | 1038 | { |
1039 | cpp_buffer *buffer = pfile->buffer; | |
fe9eb18b | 1040 | bool saw_NUL = false; |
0578f103 | 1041 | |
338fa5f7 | 1042 | do |
f80e83a9 | 1043 | { |
78719282 | 1044 | /* Horizontal space always OK. */ |
a54e0bf8 | 1045 | if (c == ' ' || c == '\t') |
338fa5f7 | 1046 | ; |
338fa5f7 | 1047 | /* Just \f \v or \0 left. */ |
78719282 | 1048 | else if (c == '\0') |
fe9eb18b | 1049 | saw_NUL = true; |
79bd622b | 1050 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
dbddc569 | 1051 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
73328dce | 1052 | CPP_BUF_COL (buffer), |
1053 | "%s in preprocessing directive", | |
1054 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 1055 | |
338fa5f7 | 1056 | c = *buffer->cur++; |
0578f103 | 1057 | } |
2c0e001b | 1058 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
338fa5f7 | 1059 | while (is_nvspace (c)); |
1060 | ||
fe9eb18b | 1061 | if (saw_NUL) |
d80d2074 | 1062 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
fe9eb18b | 1063 | |
1c124f85 | 1064 | buffer->cur--; |
f80e83a9 | 1065 | } |
0578f103 | 1066 | |
79bd622b | 1067 | /* See if the characters of a number token are valid in a name (no |
1068 | '.', '+' or '-'). */ | |
1069 | static int | |
f7fdd7a1 | 1070 | name_p (cpp_reader *pfile, const cpp_string *string) |
79bd622b | 1071 | { |
1072 | unsigned int i; | |
1073 | ||
1074 | for (i = 0; i < string->len; i++) | |
1075 | if (!is_idchar (string->text[i])) | |
1076 | return 0; | |
1077 | ||
b1a9ff83 | 1078 | return 1; |
79bd622b | 1079 | } |
1080 | ||
bce47149 | 1081 | /* After parsing an identifier or other sequence, produce a warning about |
1082 | sequences not in NFC/NFKC. */ | |
1083 | static void | |
1084 | warn_about_normalization (cpp_reader *pfile, | |
1085 | const cpp_token *token, | |
1086 | const struct normalize_state *s) | |
1087 | { | |
1088 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1089 | && !pfile->state.skipping) | |
1090 | { | |
1091 | /* Make sure that the token is printed using UCNs, even | |
1092 | if we'd otherwise happily print UTF-8. */ | |
720aca92 | 1093 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
bce47149 | 1094 | size_t sz; |
1095 | ||
1096 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1097 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
3a79f5da | 1098 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1099 | "`%.*s' is not in NFKC", (int) sz, buf); | |
bce47149 | 1100 | else |
3a79f5da | 1101 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1102 | "`%.*s' is not in NFC", (int) sz, buf); | |
15fc692a | 1103 | free (buf); |
bce47149 | 1104 | } |
1105 | } | |
1106 | ||
5bb46c08 | 1107 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in |
2cbf1359 | 1108 | an identifier. FIRST is TRUE if this starts an identifier. */ |
5bb46c08 | 1109 | static bool |
bce47149 | 1110 | forms_identifier_p (cpp_reader *pfile, int first, |
1111 | struct normalize_state *state) | |
5bb46c08 | 1112 | { |
2cbf1359 | 1113 | cpp_buffer *buffer = pfile->buffer; |
1114 | ||
1115 | if (*buffer->cur == '$') | |
1116 | { | |
1117 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1118 | return false; | |
1119 | ||
1120 | buffer->cur++; | |
f0c2775b | 1121 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
2cbf1359 | 1122 | { |
f0c2775b | 1123 | CPP_OPTION (pfile, warn_dollars) = 0; |
d80d2074 | 1124 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
2cbf1359 | 1125 | } |
1126 | ||
1127 | return true; | |
1128 | } | |
5bb46c08 | 1129 | |
2cbf1359 | 1130 | /* Is this a syntactically valid UCN? */ |
865c4e44 | 1131 | if (CPP_OPTION (pfile, extended_identifiers) |
4e9d1e6d | 1132 | && *buffer->cur == '\\' |
2cbf1359 | 1133 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) |
5bb46c08 | 1134 | { |
2cbf1359 | 1135 | buffer->cur += 2; |
bce47149 | 1136 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1137 | state)) | |
2cbf1359 | 1138 | return true; |
1139 | buffer->cur -= 2; | |
5bb46c08 | 1140 | } |
5bb46c08 | 1141 | |
2cbf1359 | 1142 | return false; |
5bb46c08 | 1143 | } |
1144 | ||
038c21f1 | 1145 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1146 | static cpp_hashnode * | |
1147 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1148 | { | |
1149 | cpp_hashnode *result; | |
1150 | const uchar *cur; | |
1151 | unsigned int len; | |
1152 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1153 | ||
1154 | cur = base + 1; | |
1155 | while (ISIDNUM (*cur)) | |
1156 | { | |
1157 | hash = HT_HASHSTEP (hash, *cur); | |
1158 | cur++; | |
1159 | } | |
1160 | len = cur - base; | |
1161 | hash = HT_HASHFINISH (hash, len); | |
1162 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1163 | base, len, hash, HT_ALLOC)); | |
1164 | ||
1165 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1166 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1167 | && !pfile->state.skipping, 0)) | |
1168 | { | |
1169 | /* It is allowed to poison the same identifier twice. */ | |
1170 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1171 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1172 | NODE_NAME (result)); | |
1173 | ||
1174 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1175 | replacement list of a variadic macro. */ | |
1176 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1177 | && !pfile->state.va_args_ok) | |
49f161b6 | 1178 | { |
1179 | if (CPP_OPTION (pfile, cplusplus)) | |
1180 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1181 | "__VA_ARGS__ can only appear in the expansion" | |
1182 | " of a C++11 variadic macro"); | |
1183 | else | |
1184 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1185 | "__VA_ARGS__ can only appear in the expansion" | |
1186 | " of a C99 variadic macro"); | |
1187 | } | |
038c21f1 | 1188 | |
1189 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1190 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1191 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1192 | "identifier \"%s\" is a special operator name in C++", | |
1193 | NODE_NAME (result)); | |
038c21f1 | 1194 | } |
1195 | ||
1196 | return result; | |
1197 | } | |
1198 | ||
1199 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1200 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1201 | cpp_hashnode * | |
1202 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1203 | { | |
1204 | cpp_hashnode *result; | |
1205 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1206 | return result; | |
1207 | } | |
1208 | ||
5bb46c08 | 1209 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
338fa5f7 | 1210 | static cpp_hashnode * |
bce47149 | 1211 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
1212 | struct normalize_state *nst) | |
0578f103 | 1213 | { |
79bd622b | 1214 | cpp_hashnode *result; |
bb1fa6bb | 1215 | const uchar *cur; |
3eb3f293 | 1216 | unsigned int len; |
1217 | unsigned int hash = HT_HASHSTEP (0, *base); | |
66a5287e | 1218 | |
3eb3f293 | 1219 | cur = pfile->buffer->cur; |
bb1fa6bb | 1220 | if (! starts_ucn) |
460f52aa | 1221 | { |
1222 | while (ISIDNUM (*cur)) | |
1223 | { | |
1224 | hash = HT_HASHSTEP (hash, *cur); | |
1225 | cur++; | |
1226 | } | |
1227 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *(cur - 1)); | |
1228 | } | |
bb1fa6bb | 1229 | pfile->buffer->cur = cur; |
bce47149 | 1230 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
78a11351 | 1231 | { |
bb1fa6bb | 1232 | /* Slower version for identifiers containing UCNs (or $). */ |
1233 | do { | |
1234 | while (ISIDNUM (*pfile->buffer->cur)) | |
bce47149 | 1235 | { |
460f52aa | 1236 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *pfile->buffer->cur); |
bce47149 | 1237 | pfile->buffer->cur++; |
bce47149 | 1238 | } |
1239 | } while (forms_identifier_p (pfile, false, nst)); | |
bb1fa6bb | 1240 | result = _cpp_interpret_identifier (pfile, base, |
1241 | pfile->buffer->cur - base); | |
66a5287e | 1242 | } |
bb1fa6bb | 1243 | else |
1244 | { | |
1245 | len = cur - base; | |
1246 | hash = HT_HASHFINISH (hash, len); | |
5bb46c08 | 1247 | |
e297899b | 1248 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1249 | base, len, hash, HT_ALLOC)); | |
bb1fa6bb | 1250 | } |
66a5287e | 1251 | |
5bb46c08 | 1252 | /* Rarely, identifiers require diagnostics when lexed. */ |
66a5287e | 1253 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1254 | && !pfile->state.skipping, 0)) | |
1255 | { | |
1256 | /* It is allowed to poison the same identifier twice. */ | |
1257 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
d80d2074 | 1258 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
66a5287e | 1259 | NODE_NAME (result)); |
1260 | ||
1261 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1262 | replacement list of a variadic macro. */ | |
1263 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1264 | && !pfile->state.va_args_ok) | |
49f161b6 | 1265 | { |
1266 | if (CPP_OPTION (pfile, cplusplus)) | |
1267 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1268 | "__VA_ARGS__ can only appear in the expansion" | |
1269 | " of a C++11 variadic macro"); | |
1270 | else | |
1271 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1272 | "__VA_ARGS__ can only appear in the expansion" | |
1273 | " of a C99 variadic macro"); | |
1274 | } | |
2a6a6991 | 1275 | |
1276 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1277 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1278 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1279 | "identifier \"%s\" is a special operator name in C++", | |
1280 | NODE_NAME (result)); | |
66a5287e | 1281 | } |
1282 | ||
1283 | return result; | |
1284 | } | |
1285 | ||
5bb46c08 | 1286 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
0578f103 | 1287 | static void |
bce47149 | 1288 | lex_number (cpp_reader *pfile, cpp_string *number, |
1289 | struct normalize_state *nst) | |
0578f103 | 1290 | { |
b6d18b0a | 1291 | const uchar *cur; |
5bb46c08 | 1292 | const uchar *base; |
1293 | uchar *dest; | |
0578f103 | 1294 | |
5bb46c08 | 1295 | base = pfile->buffer->cur - 1; |
1296 | do | |
f80e83a9 | 1297 | { |
5bb46c08 | 1298 | cur = pfile->buffer->cur; |
338fa5f7 | 1299 | |
5bb46c08 | 1300 | /* N.B. ISIDNUM does not include $. */ |
4d6f7dd4 | 1301 | while (ISIDNUM (*cur) || *cur == '.' || DIGIT_SEP (*cur) |
1302 | || VALID_SIGN (*cur, cur[-1])) | |
bce47149 | 1303 | { |
460f52aa | 1304 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *cur); |
bce47149 | 1305 | cur++; |
bce47149 | 1306 | } |
0578f103 | 1307 | |
78a11351 | 1308 | pfile->buffer->cur = cur; |
0578f103 | 1309 | } |
bce47149 | 1310 | while (forms_identifier_p (pfile, false, nst)); |
79bd622b | 1311 | |
5bb46c08 | 1312 | number->len = cur - base; |
1313 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1314 | memcpy (dest, base, number->len); | |
1315 | dest[number->len] = '\0'; | |
1316 | number->text = dest; | |
79bd622b | 1317 | } |
1318 | ||
4970d4c2 | 1319 | /* Create a token of type TYPE with a literal spelling. */ |
1320 | static void | |
f7fdd7a1 | 1321 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1322 | unsigned int len, enum cpp_ttype type) | |
4970d4c2 | 1323 | { |
1324 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1325 | ||
1326 | memcpy (dest, base, len); | |
1327 | dest[len] = '\0'; | |
1328 | token->type = type; | |
1329 | token->val.str.len = len; | |
1330 | token->val.str.text = dest; | |
1331 | } | |
1332 | ||
3a45011c | 1333 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1334 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1335 | ||
1336 | static void | |
1337 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1338 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1339 | { | |
1340 | _cpp_buff *first_buff = *first_buff_p; | |
1341 | _cpp_buff *last_buff = *last_buff_p; | |
1342 | ||
1343 | if (first_buff == NULL) | |
1344 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1345 | else if (len > BUFF_ROOM (last_buff)) | |
1346 | { | |
1347 | size_t room = BUFF_ROOM (last_buff); | |
1348 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1349 | BUFF_FRONT (last_buff) += room; | |
1350 | base += room; | |
1351 | len -= room; | |
1352 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1353 | } | |
1354 | ||
1355 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1356 | BUFF_FRONT (last_buff) += len; | |
1357 | ||
1358 | *first_buff_p = first_buff; | |
1359 | *last_buff_p = last_buff; | |
1360 | } | |
1361 | ||
4e8832f3 | 1362 | |
1363 | /* Returns true if a macro has been defined. | |
1364 | This might not work if compile with -save-temps, | |
1365 | or preprocess separately from compilation. */ | |
1366 | ||
1367 | static bool | |
1368 | is_macro(cpp_reader *pfile, const uchar *base) | |
1369 | { | |
1370 | const uchar *cur = base; | |
1371 | if (! ISIDST (*cur)) | |
1372 | return false; | |
1373 | unsigned int hash = HT_HASHSTEP (0, *cur); | |
1374 | ++cur; | |
1375 | while (ISIDNUM (*cur)) | |
1376 | { | |
1377 | hash = HT_HASHSTEP (hash, *cur); | |
1378 | ++cur; | |
1379 | } | |
1380 | hash = HT_HASHFINISH (hash, cur - base); | |
1381 | ||
1382 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1383 | base, cur - base, hash, HT_NO_INSERT)); | |
1384 | ||
1385 | return !result ? false : (result->type == NT_MACRO); | |
1386 | } | |
1387 | ||
1388 | ||
538ba11a | 1389 | /* Lexes a raw string. The stored string contains the spelling, including |
3a45011c | 1390 | double quotes, delimiter string, '(' and ')', any leading |
538ba11a | 1391 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1392 | literal, or CPP_OTHER if it was not properly terminated. | |
1393 | ||
1394 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1395 | is the first NUL since embedded NULs are preserved. */ | |
1396 | ||
1397 | static void | |
1398 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1399 | const uchar *cur) | |
1400 | { | |
ffb840b4 | 1401 | uchar raw_prefix[17]; |
1402 | uchar temp_buffer[18]; | |
1403 | const uchar *orig_base; | |
1404 | unsigned int raw_prefix_len = 0, raw_suffix_len = 0; | |
1405 | enum raw_str_phase { RAW_STR_PREFIX, RAW_STR, RAW_STR_SUFFIX }; | |
1406 | raw_str_phase phase = RAW_STR_PREFIX; | |
538ba11a | 1407 | enum cpp_ttype type; |
1408 | size_t total_len = 0; | |
ffb840b4 | 1409 | /* Index into temp_buffer during phases other than RAW_STR, |
1410 | during RAW_STR phase 17 to tell BUF_APPEND that nothing should | |
1411 | be appended to temp_buffer. */ | |
1412 | size_t temp_buffer_len = 0; | |
538ba11a | 1413 | _cpp_buff *first_buff = NULL, *last_buff = NULL; |
ffb840b4 | 1414 | size_t raw_prefix_start; |
3a45011c | 1415 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1416 | |
1417 | type = (*base == 'L' ? CPP_WSTRING : | |
1418 | *base == 'U' ? CPP_STRING32 : | |
1419 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1420 | : CPP_STRING); | |
1421 | ||
3a45011c | 1422 | #define BUF_APPEND(STR,LEN) \ |
1423 | do { \ | |
1424 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1425 | &first_buff, &last_buff); \ | |
1426 | total_len += (LEN); \ | |
ffb840b4 | 1427 | if (__builtin_expect (temp_buffer_len < 17, 0) \ |
1428 | && (const uchar *)(STR) != base \ | |
1429 | && (LEN) <= 2) \ | |
1430 | { \ | |
1431 | memcpy (temp_buffer + temp_buffer_len, \ | |
1432 | (const uchar *)(STR), (LEN)); \ | |
1433 | temp_buffer_len += (LEN); \ | |
1434 | } \ | |
3a45011c | 1435 | } while (0); |
1436 | ||
ffb840b4 | 1437 | orig_base = base; |
1438 | ++cur; | |
1439 | raw_prefix_start = cur - base; | |
1440 | for (;;) | |
1441 | { | |
3a45011c | 1442 | cppchar_t c; |
1443 | ||
1444 | /* If we previously performed any trigraph or line splicing | |
ffb840b4 | 1445 | transformations, undo them in between the opening and closing |
1446 | double quote. */ | |
3a45011c | 1447 | while (note->pos < cur) |
1448 | ++note; | |
1449 | for (; note->pos == cur; ++note) | |
1450 | { | |
1451 | switch (note->type) | |
1452 | { | |
1453 | case '\\': | |
1454 | case ' ': | |
1455 | /* Restore backslash followed by newline. */ | |
1456 | BUF_APPEND (base, cur - base); | |
1457 | base = cur; | |
1458 | BUF_APPEND ("\\", 1); | |
1459 | after_backslash: | |
1460 | if (note->type == ' ') | |
1461 | { | |
1462 | /* GNU backslash whitespace newline extension. FIXME | |
1463 | could be any sequence of non-vertical space. When we | |
1464 | can properly restore any such sequence, we should mark | |
1465 | this note as handled so _cpp_process_line_notes | |
1466 | doesn't warn. */ | |
1467 | BUF_APPEND (" ", 1); | |
1468 | } | |
1469 | ||
1470 | BUF_APPEND ("\n", 1); | |
1471 | break; | |
1472 | ||
1473 | case 0: | |
1474 | /* Already handled. */ | |
1475 | break; | |
1476 | ||
1477 | default: | |
1478 | if (_cpp_trigraph_map[note->type]) | |
1479 | { | |
1480 | /* Don't warn about this trigraph in | |
1481 | _cpp_process_line_notes, since trigraphs show up as | |
1482 | trigraphs in raw strings. */ | |
56e2ce2d | 1483 | uchar type = note->type; |
3a45011c | 1484 | note->type = 0; |
1485 | ||
1486 | if (!CPP_OPTION (pfile, trigraphs)) | |
1487 | /* If we didn't convert the trigraph in the first | |
1488 | place, don't do anything now either. */ | |
1489 | break; | |
1490 | ||
1491 | BUF_APPEND (base, cur - base); | |
1492 | base = cur; | |
1493 | BUF_APPEND ("??", 2); | |
1494 | ||
1495 | /* ??/ followed by newline gets two line notes, one for | |
1496 | the trigraph and one for the backslash/newline. */ | |
1497 | if (type == '/' && note[1].pos == cur) | |
1498 | { | |
1499 | if (note[1].type != '\\' | |
1500 | && note[1].type != ' ') | |
1501 | abort (); | |
1502 | BUF_APPEND ("/", 1); | |
1503 | ++note; | |
1504 | goto after_backslash; | |
1505 | } | |
3a45011c | 1506 | else |
1507 | { | |
1508 | /* Skip the replacement character. */ | |
1509 | base = ++cur; | |
1510 | BUF_APPEND (&type, 1); | |
ffb840b4 | 1511 | c = type; |
1512 | goto check_c; | |
3a45011c | 1513 | } |
1514 | } | |
1515 | else | |
1516 | abort (); | |
1517 | break; | |
1518 | } | |
1519 | } | |
1520 | c = *cur++; | |
ffb840b4 | 1521 | if (__builtin_expect (temp_buffer_len < 17, 0)) |
1522 | temp_buffer[temp_buffer_len++] = c; | |
538ba11a | 1523 | |
ffb840b4 | 1524 | check_c: |
1525 | if (phase == RAW_STR_PREFIX) | |
538ba11a | 1526 | { |
ffb840b4 | 1527 | while (raw_prefix_len < temp_buffer_len) |
1528 | { | |
1529 | raw_prefix[raw_prefix_len] = temp_buffer[raw_prefix_len]; | |
1530 | switch (raw_prefix[raw_prefix_len]) | |
1531 | { | |
1532 | case ' ': case '(': case ')': case '\\': case '\t': | |
1533 | case '\v': case '\f': case '\n': default: | |
1534 | break; | |
1535 | /* Basic source charset except the above chars. */ | |
1536 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1537 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1538 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1539 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1540 | case 'y': case 'z': | |
1541 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1542 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1543 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1544 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1545 | case 'Y': case 'Z': | |
1546 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1547 | case '6': case '7': case '8': case '9': | |
1548 | case '_': case '{': case '}': case '#': case '[': case ']': | |
1549 | case '<': case '>': case '%': case ':': case ';': case '.': | |
1550 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1551 | case '&': case '|': case '~': case '!': case '=': case ',': | |
1552 | case '"': case '\'': | |
1553 | if (raw_prefix_len < 16) | |
1554 | { | |
1555 | raw_prefix_len++; | |
1556 | continue; | |
1557 | } | |
1558 | break; | |
1559 | } | |
1560 | ||
1561 | if (raw_prefix[raw_prefix_len] != '(') | |
1562 | { | |
1563 | int col = CPP_BUF_COLUMN (pfile->buffer, cur) + 1; | |
1564 | if (raw_prefix_len == 16) | |
1565 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1566 | col, "raw string delimiter longer " | |
1567 | "than 16 characters"); | |
1568 | else if (raw_prefix[raw_prefix_len] == '\n') | |
1569 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1570 | col, "invalid new-line in raw " | |
1571 | "string delimiter"); | |
1572 | else | |
1573 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, | |
1574 | col, "invalid character '%c' in " | |
1575 | "raw string delimiter", | |
1576 | (int) raw_prefix[raw_prefix_len]); | |
1577 | pfile->buffer->cur = orig_base + raw_prefix_start - 1; | |
1578 | create_literal (pfile, token, orig_base, | |
1579 | raw_prefix_start - 1, CPP_OTHER); | |
1580 | if (first_buff) | |
1581 | _cpp_release_buff (pfile, first_buff); | |
1582 | return; | |
1583 | } | |
1584 | raw_prefix[raw_prefix_len] = '"'; | |
1585 | phase = RAW_STR; | |
1586 | /* Nothing should be appended to temp_buffer during | |
1587 | RAW_STR phase. */ | |
1588 | temp_buffer_len = 17; | |
1589 | break; | |
1590 | } | |
1591 | continue; | |
1592 | } | |
1593 | else if (phase == RAW_STR_SUFFIX) | |
1594 | { | |
1595 | while (raw_suffix_len <= raw_prefix_len | |
1596 | && raw_suffix_len < temp_buffer_len | |
1597 | && temp_buffer[raw_suffix_len] == raw_prefix[raw_suffix_len]) | |
1598 | raw_suffix_len++; | |
1599 | if (raw_suffix_len > raw_prefix_len) | |
1600 | break; | |
1601 | if (raw_suffix_len == temp_buffer_len) | |
1602 | continue; | |
1603 | phase = RAW_STR; | |
1604 | /* Nothing should be appended to temp_buffer during | |
1605 | RAW_STR phase. */ | |
1606 | temp_buffer_len = 17; | |
1607 | } | |
1608 | if (c == ')') | |
1609 | { | |
1610 | phase = RAW_STR_SUFFIX; | |
1611 | raw_suffix_len = 0; | |
1612 | temp_buffer_len = 0; | |
538ba11a | 1613 | } |
1614 | else if (c == '\n') | |
1615 | { | |
1616 | if (pfile->state.in_directive | |
c7691e08 | 1617 | || (pfile->state.parsing_args |
1618 | && pfile->buffer->next_line >= pfile->buffer->rlimit)) | |
538ba11a | 1619 | { |
1620 | cur--; | |
1621 | type = CPP_OTHER; | |
1622 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1623 | "unterminated raw string"); | |
1624 | break; | |
1625 | } | |
1626 | ||
3a45011c | 1627 | BUF_APPEND (base, cur - base); |
538ba11a | 1628 | |
1629 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1630 | CPP_INCREMENT_LINE (pfile, 0); | |
1631 | pfile->buffer->need_line = true; | |
1632 | ||
3a45011c | 1633 | pfile->buffer->cur = cur-1; |
1634 | _cpp_process_line_notes (pfile, false); | |
538ba11a | 1635 | if (!_cpp_get_fresh_line (pfile)) |
1636 | { | |
1637 | source_location src_loc = token->src_loc; | |
1638 | token->type = CPP_EOF; | |
1639 | /* Tell the compiler the line number of the EOF token. */ | |
1640 | token->src_loc = pfile->line_table->highest_line; | |
1641 | token->flags = BOL; | |
1642 | if (first_buff != NULL) | |
1643 | _cpp_release_buff (pfile, first_buff); | |
1644 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1645 | "unterminated raw string"); | |
1646 | return; | |
1647 | } | |
1648 | ||
1649 | cur = base = pfile->buffer->cur; | |
3a45011c | 1650 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1651 | } |
538ba11a | 1652 | } |
1653 | ||
244db24d | 1654 | if (CPP_OPTION (pfile, user_literals)) |
1655 | { | |
4e8832f3 | 1656 | /* If a string format macro, say from inttypes.h, is placed touching |
1657 | a string literal it could be parsed as a C++11 user-defined string | |
1658 | literal thus breaking the program. | |
1659 | Try to identify macros with is_macro. A warning is issued. */ | |
1660 | if (is_macro (pfile, cur)) | |
76d340ac | 1661 | { |
86c82b61 | 1662 | /* Raise a warning, but do not consume subsequent tokens. */ |
8947e5dc | 1663 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
76d340ac | 1664 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1665 | token->src_loc, 0, | |
1666 | "invalid suffix on literal; C++11 requires " | |
4e8832f3 | 1667 | "a space between literal and string macro"); |
76d340ac | 1668 | } |
244db24d | 1669 | /* Grab user defined literal suffix. */ |
911b08c6 | 1670 | else if (ISIDST (*cur)) |
244db24d | 1671 | { |
1672 | type = cpp_userdef_string_add_type (type); | |
1673 | ++cur; | |
76d340ac | 1674 | |
1675 | while (ISIDNUM (*cur)) | |
1676 | ++cur; | |
244db24d | 1677 | } |
244db24d | 1678 | } |
1679 | ||
538ba11a | 1680 | pfile->buffer->cur = cur; |
1681 | if (first_buff == NULL) | |
1682 | create_literal (pfile, token, base, cur - base, type); | |
1683 | else | |
1684 | { | |
1685 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1686 | ||
1687 | token->type = type; | |
1688 | token->val.str.len = total_len + (cur - base); | |
1689 | token->val.str.text = dest; | |
1690 | last_buff = first_buff; | |
1691 | while (last_buff != NULL) | |
1692 | { | |
1693 | memcpy (dest, last_buff->base, | |
1694 | BUFF_FRONT (last_buff) - last_buff->base); | |
1695 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1696 | last_buff = last_buff->next; | |
1697 | } | |
1698 | _cpp_release_buff (pfile, first_buff); | |
1699 | memcpy (dest, base, cur - base); | |
1700 | dest[cur - base] = '\0'; | |
1701 | } | |
1702 | } | |
1703 | ||
5bb46c08 | 1704 | /* Lexes a string, character constant, or angle-bracketed header file |
4970d4c2 | 1705 | name. The stored string contains the spelling, including opening |
538ba11a | 1706 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1707 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1708 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1709 | header name which must be relexed as normal tokens. | |
4970d4c2 | 1710 | |
1711 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1712 | is the first NUL since embedded NULs are preserved. */ | |
f80e83a9 | 1713 | static void |
f7fdd7a1 | 1714 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
0578f103 | 1715 | { |
4970d4c2 | 1716 | bool saw_NUL = false; |
1717 | const uchar *cur; | |
5bb46c08 | 1718 | cppchar_t terminator; |
4970d4c2 | 1719 | enum cpp_ttype type; |
1720 | ||
1721 | cur = base; | |
1722 | terminator = *cur++; | |
538ba11a | 1723 | if (terminator == 'L' || terminator == 'U') |
4970d4c2 | 1724 | terminator = *cur++; |
538ba11a | 1725 | else if (terminator == 'u') |
1726 | { | |
1727 | terminator = *cur++; | |
1728 | if (terminator == '8') | |
1729 | terminator = *cur++; | |
1730 | } | |
1731 | if (terminator == 'R') | |
1732 | { | |
1733 | lex_raw_string (pfile, token, base, cur); | |
1734 | return; | |
1735 | } | |
1736 | if (terminator == '"') | |
924bbf02 | 1737 | type = (*base == 'L' ? CPP_WSTRING : |
1738 | *base == 'U' ? CPP_STRING32 : | |
538ba11a | 1739 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
1740 | : CPP_STRING); | |
4970d4c2 | 1741 | else if (terminator == '\'') |
924bbf02 | 1742 | type = (*base == 'L' ? CPP_WCHAR : |
1743 | *base == 'U' ? CPP_CHAR32 : | |
1744 | *base == 'u' ? CPP_CHAR16 : CPP_CHAR); | |
4970d4c2 | 1745 | else |
1746 | terminator = '>', type = CPP_HEADER_NAME; | |
79bd622b | 1747 | |
338fa5f7 | 1748 | for (;;) |
0578f103 | 1749 | { |
4970d4c2 | 1750 | cppchar_t c = *cur++; |
4b0c16ee | 1751 | |
edaf8cb5 | 1752 | /* In #include-style directives, terminators are not escapable. */ |
4970d4c2 | 1753 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
1754 | cur++; | |
1755 | else if (c == terminator) | |
5bb46c08 | 1756 | break; |
4970d4c2 | 1757 | else if (c == '\n') |
338fa5f7 | 1758 | { |
4970d4c2 | 1759 | cur--; |
7811eab5 | 1760 | /* Unmatched quotes always yield undefined behavior, but |
1761 | greedy lexing means that what appears to be an unterminated | |
1762 | header name may actually be a legitimate sequence of tokens. */ | |
1763 | if (terminator == '>') | |
1764 | { | |
1765 | token->type = CPP_LESS; | |
1766 | return; | |
1767 | } | |
4970d4c2 | 1768 | type = CPP_OTHER; |
1769 | break; | |
0578f103 | 1770 | } |
4970d4c2 | 1771 | else if (c == '\0') |
1772 | saw_NUL = true; | |
0578f103 | 1773 | } |
1774 | ||
4970d4c2 | 1775 | if (saw_NUL && !pfile->state.skipping) |
d80d2074 | 1776 | cpp_error (pfile, CPP_DL_WARNING, |
1777 | "null character(s) preserved in literal"); | |
0578f103 | 1778 | |
0b67f687 | 1779 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
1780 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
1781 | (int) terminator); | |
1782 | ||
244db24d | 1783 | if (CPP_OPTION (pfile, user_literals)) |
1784 | { | |
4e8832f3 | 1785 | /* If a string format macro, say from inttypes.h, is placed touching |
1786 | a string literal it could be parsed as a C++11 user-defined string | |
1787 | literal thus breaking the program. | |
1788 | Try to identify macros with is_macro. A warning is issued. */ | |
1789 | if (is_macro (pfile, cur)) | |
76d340ac | 1790 | { |
86c82b61 | 1791 | /* Raise a warning, but do not consume subsequent tokens. */ |
8947e5dc | 1792 | if (CPP_OPTION (pfile, warn_literal_suffix) && !pfile->state.skipping) |
76d340ac | 1793 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
1794 | token->src_loc, 0, | |
1795 | "invalid suffix on literal; C++11 requires " | |
4e8832f3 | 1796 | "a space between literal and string macro"); |
76d340ac | 1797 | } |
244db24d | 1798 | /* Grab user defined literal suffix. */ |
911b08c6 | 1799 | else if (ISIDST (*cur)) |
244db24d | 1800 | { |
1801 | type = cpp_userdef_char_add_type (type); | |
1802 | type = cpp_userdef_string_add_type (type); | |
1803 | ++cur; | |
76d340ac | 1804 | |
1805 | while (ISIDNUM (*cur)) | |
1806 | ++cur; | |
244db24d | 1807 | } |
244db24d | 1808 | } |
1809 | ||
4970d4c2 | 1810 | pfile->buffer->cur = cur; |
1811 | create_literal (pfile, token, base, cur - base, type); | |
338fa5f7 | 1812 | } |
f80e83a9 | 1813 | |
956c6108 | 1814 | /* Return the comment table. The client may not make any assumption |
1815 | about the ordering of the table. */ | |
1816 | cpp_comment_table * | |
1817 | cpp_get_comments (cpp_reader *pfile) | |
1818 | { | |
1819 | return &pfile->comments; | |
1820 | } | |
1821 | ||
1822 | /* Append a comment to the end of the comment table. */ | |
1823 | static void | |
1824 | store_comment (cpp_reader *pfile, cpp_token *token) | |
1825 | { | |
1826 | int len; | |
1827 | ||
1828 | if (pfile->comments.allocated == 0) | |
1829 | { | |
1830 | pfile->comments.allocated = 256; | |
1831 | pfile->comments.entries = (cpp_comment *) xmalloc | |
1832 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
1833 | } | |
1834 | ||
1835 | if (pfile->comments.count == pfile->comments.allocated) | |
1836 | { | |
1837 | pfile->comments.allocated *= 2; | |
1838 | pfile->comments.entries = (cpp_comment *) xrealloc | |
1839 | (pfile->comments.entries, | |
1840 | pfile->comments.allocated * sizeof (cpp_comment)); | |
1841 | } | |
1842 | ||
1843 | len = token->val.str.len; | |
1844 | ||
1845 | /* Copy comment. Note, token may not be NULL terminated. */ | |
1846 | pfile->comments.entries[pfile->comments.count].comment = | |
1847 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
1848 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
1849 | token->val.str.text, len); | |
1850 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
1851 | ||
1852 | /* Set source location. */ | |
1853 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
1854 | ||
1855 | /* Increment the count of entries in the comment table. */ | |
1856 | pfile->comments.count++; | |
1857 | } | |
1858 | ||
79bd622b | 1859 | /* The stored comment includes the comment start and any terminator. */ |
2c63d6c8 | 1860 | static void |
f7fdd7a1 | 1861 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
1862 | cppchar_t type) | |
2c63d6c8 | 1863 | { |
f80e83a9 | 1864 | unsigned char *buffer; |
560ab0b2 | 1865 | unsigned int len, clen, i; |
b1a9ff83 | 1866 | |
f0495c2c | 1867 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
1c124f85 | 1868 | |
a543b315 | 1869 | /* C++ comments probably (not definitely) have moved past a new |
1870 | line, which we don't want to save in the comment. */ | |
1c124f85 | 1871 | if (is_vspace (pfile->buffer->cur[-1])) |
a543b315 | 1872 | len--; |
d3f7919d | 1873 | |
560ab0b2 | 1874 | /* If we are currently in a directive or in argument parsing, then |
1875 | we need to store all C++ comments as C comments internally, and | |
1876 | so we need to allocate a little extra space in that case. | |
d3f7919d | 1877 | |
1878 | Note that the only time we encounter a directive here is | |
1879 | when we are saving comments in a "#define". */ | |
560ab0b2 | 1880 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
1881 | && type == '/') ? len + 2 : len; | |
d3f7919d | 1882 | |
1883 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
b1a9ff83 | 1884 | |
f80e83a9 | 1885 | token->type = CPP_COMMENT; |
d3f7919d | 1886 | token->val.str.len = clen; |
338fa5f7 | 1887 | token->val.str.text = buffer; |
0578f103 | 1888 | |
f0495c2c | 1889 | buffer[0] = '/'; |
1890 | memcpy (buffer + 1, from, len - 1); | |
d3f7919d | 1891 | |
a113df96 | 1892 | /* Finish conversion to a C comment, if necessary. */ |
560ab0b2 | 1893 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
d3f7919d | 1894 | { |
1895 | buffer[1] = '*'; | |
1896 | buffer[clen - 2] = '*'; | |
1897 | buffer[clen - 1] = '/'; | |
560ab0b2 | 1898 | /* As there can be in a C++ comments illegal sequences for C comments |
1899 | we need to filter them out. */ | |
1900 | for (i = 2; i < (clen - 2); i++) | |
1901 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
1902 | buffer[i] = '|'; | |
d3f7919d | 1903 | } |
956c6108 | 1904 | |
1905 | /* Finally store this comment for use by clients of libcpp. */ | |
1906 | store_comment (pfile, token); | |
338fa5f7 | 1907 | } |
0578f103 | 1908 | |
83dcbb5c | 1909 | /* Allocate COUNT tokens for RUN. */ |
1910 | void | |
f7fdd7a1 | 1911 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
83dcbb5c | 1912 | { |
3b298764 | 1913 | run->base = XNEWVEC (cpp_token, count); |
83dcbb5c | 1914 | run->limit = run->base + count; |
1915 | run->next = NULL; | |
1916 | } | |
1917 | ||
1918 | /* Returns the next tokenrun, or creates one if there is none. */ | |
1919 | static tokenrun * | |
f7fdd7a1 | 1920 | next_tokenrun (tokenrun *run) |
83dcbb5c | 1921 | { |
1922 | if (run->next == NULL) | |
1923 | { | |
3b298764 | 1924 | run->next = XNEW (tokenrun); |
fb5ab82c | 1925 | run->next->prev = run; |
83dcbb5c | 1926 | _cpp_init_tokenrun (run->next, 250); |
1927 | } | |
1928 | ||
1929 | return run->next; | |
1930 | } | |
1931 | ||
a2eb22f0 | 1932 | /* Return the number of not yet processed token in a given |
ce70f433 | 1933 | context. */ |
1934 | int | |
a2eb22f0 | 1935 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
ce70f433 | 1936 | { |
ce70f433 | 1937 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
ac6130e2 | 1938 | return (LAST (context).token - FIRST (context).token); |
ce70f433 | 1939 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
1940 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
ac6130e2 | 1941 | return (LAST (context).ptoken - FIRST (context).ptoken); |
ce70f433 | 1942 | else |
1943 | abort (); | |
1944 | } | |
1945 | ||
a2eb22f0 | 1946 | /* Returns the token present at index INDEX in a given context. If |
1947 | INDEX is zero, the next token to be processed is returned. */ | |
ce70f433 | 1948 | static const cpp_token* |
a2eb22f0 | 1949 | _cpp_token_from_context_at (cpp_context *context, int index) |
ce70f433 | 1950 | { |
ce70f433 | 1951 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
1952 | return &(FIRST (context).token[index]); | |
1953 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT | |
1954 | || context->tokens_kind == TOKENS_KIND_EXTENDED) | |
1955 | return FIRST (context).ptoken[index]; | |
1956 | else | |
1957 | abort (); | |
1958 | } | |
1959 | ||
89768577 | 1960 | /* Look ahead in the input stream. */ |
1961 | const cpp_token * | |
1962 | cpp_peek_token (cpp_reader *pfile, int index) | |
1963 | { | |
1964 | cpp_context *context = pfile->context; | |
1965 | const cpp_token *peektok; | |
1966 | int count; | |
1967 | ||
1968 | /* First, scan through any pending cpp_context objects. */ | |
1969 | while (context->prev) | |
1970 | { | |
a2eb22f0 | 1971 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
89768577 | 1972 | |
1973 | if (index < (int) sz) | |
a2eb22f0 | 1974 | return _cpp_token_from_context_at (context, index); |
89768577 | 1975 | index -= (int) sz; |
1976 | context = context->prev; | |
1977 | } | |
1978 | ||
1979 | /* We will have to read some new tokens after all (and do so | |
1980 | without invalidating preceding tokens). */ | |
1981 | count = index; | |
1982 | pfile->keep_tokens++; | |
1983 | ||
1984 | do | |
1985 | { | |
1986 | peektok = _cpp_lex_token (pfile); | |
1987 | if (peektok->type == CPP_EOF) | |
1988 | return peektok; | |
1989 | } | |
1990 | while (index--); | |
1991 | ||
1992 | _cpp_backup_tokens_direct (pfile, count + 1); | |
1993 | pfile->keep_tokens--; | |
1994 | ||
1995 | return peektok; | |
1996 | } | |
1997 | ||
f9b5f742 | 1998 | /* Allocate a single token that is invalidated at the same time as the |
1999 | rest of the tokens on the line. Has its line and col set to the | |
2000 | same as the last lexed token, so that diagnostics appear in the | |
2001 | right place. */ | |
2002 | cpp_token * | |
f7fdd7a1 | 2003 | _cpp_temp_token (cpp_reader *pfile) |
f9b5f742 | 2004 | { |
2005 | cpp_token *old, *result; | |
89768577 | 2006 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
2007 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
f9b5f742 | 2008 | |
2009 | old = pfile->cur_token - 1; | |
89768577 | 2010 | /* Any pre-existing lookaheads must not be clobbered. */ |
2011 | if (la) | |
2012 | { | |
2013 | if (sz <= la) | |
2014 | { | |
2015 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
2016 | ||
2017 | if (sz < la) | |
2018 | memmove (next->base + 1, next->base, | |
2019 | (la - sz) * sizeof (cpp_token)); | |
2020 | ||
2021 | next->base[0] = pfile->cur_run->limit[-1]; | |
2022 | } | |
2023 | ||
2024 | if (sz > 1) | |
2025 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
2026 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
2027 | } | |
2028 | ||
2029 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
f9b5f742 | 2030 | { |
2031 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
2032 | pfile->cur_token = pfile->cur_run->base; | |
2033 | } | |
2034 | ||
2035 | result = pfile->cur_token++; | |
610625e3 | 2036 | result->src_loc = old->src_loc; |
f9b5f742 | 2037 | return result; |
2038 | } | |
2039 | ||
10b4496a | 2040 | /* Lex a token into RESULT (external interface). Takes care of issues |
2041 | like directive handling, token lookahead, multiple include | |
3fb1e43b | 2042 | optimization and skipping. */ |
c00e481c | 2043 | const cpp_token * |
f7fdd7a1 | 2044 | _cpp_lex_token (cpp_reader *pfile) |
83dcbb5c | 2045 | { |
fb5ab82c | 2046 | cpp_token *result; |
83dcbb5c | 2047 | |
fb5ab82c | 2048 | for (;;) |
83dcbb5c | 2049 | { |
fb5ab82c | 2050 | if (pfile->cur_token == pfile->cur_run->limit) |
83dcbb5c | 2051 | { |
fb5ab82c | 2052 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
2053 | pfile->cur_token = pfile->cur_run->base; | |
83dcbb5c | 2054 | } |
e0ff7935 | 2055 | /* We assume that the current token is somewhere in the current |
2056 | run. */ | |
2057 | if (pfile->cur_token < pfile->cur_run->base | |
2058 | || pfile->cur_token >= pfile->cur_run->limit) | |
2059 | abort (); | |
83dcbb5c | 2060 | |
fb5ab82c | 2061 | if (pfile->lookaheads) |
10b4496a | 2062 | { |
2063 | pfile->lookaheads--; | |
2064 | result = pfile->cur_token++; | |
2065 | } | |
fb5ab82c | 2066 | else |
10b4496a | 2067 | result = _cpp_lex_direct (pfile); |
fb5ab82c | 2068 | |
2069 | if (result->flags & BOL) | |
83dcbb5c | 2070 | { |
fb5ab82c | 2071 | /* Is this a directive. If _cpp_handle_directive returns |
2072 | false, it is an assembler #. */ | |
2073 | if (result->type == CPP_HASH | |
d6af0368 | 2074 | /* 6.10.3 p 11: Directives in a list of macro arguments |
2075 | gives undefined behavior. This implementation | |
2076 | handles the directive as normal. */ | |
b75b98aa | 2077 | && pfile->state.parsing_args != 1) |
d6d3c909 | 2078 | { |
b75b98aa | 2079 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
d6d3c909 | 2080 | { |
b75b98aa | 2081 | if (pfile->directive_result.type == CPP_PADDING) |
2082 | continue; | |
d6d3c909 | 2083 | result = &pfile->directive_result; |
d6d3c909 | 2084 | } |
2085 | } | |
b75b98aa | 2086 | else if (pfile->state.in_deferred_pragma) |
2087 | result = &pfile->directive_result; | |
d6d3c909 | 2088 | |
5621a364 | 2089 | if (pfile->cb.line_change && !pfile->state.skipping) |
f7fdd7a1 | 2090 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
83dcbb5c | 2091 | } |
83dcbb5c | 2092 | |
fb5ab82c | 2093 | /* We don't skip tokens in directives. */ |
b75b98aa | 2094 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
fb5ab82c | 2095 | break; |
83dcbb5c | 2096 | |
fb5ab82c | 2097 | /* Outside a directive, invalidate controlling macros. At file |
10b4496a | 2098 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
7ef5b942 | 2099 | get here and MI optimization works. */ |
83dcbb5c | 2100 | pfile->mi_valid = false; |
fb5ab82c | 2101 | |
2102 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
2103 | break; | |
83dcbb5c | 2104 | } |
2105 | ||
c00e481c | 2106 | return result; |
83dcbb5c | 2107 | } |
2108 | ||
a54e0bf8 | 2109 | /* Returns true if a fresh line has been loaded. */ |
2110 | bool | |
f7fdd7a1 | 2111 | _cpp_get_fresh_line (cpp_reader *pfile) |
0bb65704 | 2112 | { |
6e04daf1 | 2113 | int return_at_eof; |
2114 | ||
a54e0bf8 | 2115 | /* We can't get a new line until we leave the current directive. */ |
2116 | if (pfile->state.in_directive) | |
2117 | return false; | |
b1a9ff83 | 2118 | |
a54e0bf8 | 2119 | for (;;) |
fb83e0d6 | 2120 | { |
a54e0bf8 | 2121 | cpp_buffer *buffer = pfile->buffer; |
fb83e0d6 | 2122 | |
a54e0bf8 | 2123 | if (!buffer->need_line) |
2124 | return true; | |
2125 | ||
2126 | if (buffer->next_line < buffer->rlimit) | |
0bb65704 | 2127 | { |
a54e0bf8 | 2128 | _cpp_clean_line (pfile); |
2129 | return true; | |
2130 | } | |
0bb65704 | 2131 | |
a54e0bf8 | 2132 | /* First, get out of parsing arguments state. */ |
2133 | if (pfile->state.parsing_args) | |
2134 | return false; | |
2135 | ||
2136 | /* End of buffer. Non-empty files should end in a newline. */ | |
2137 | if (buffer->buf != buffer->rlimit | |
2138 | && buffer->next_line > buffer->rlimit | |
2139 | && !buffer->from_stage3) | |
2140 | { | |
0448520c | 2141 | /* Clip to buffer size. */ |
a54e0bf8 | 2142 | buffer->next_line = buffer->rlimit; |
a54e0bf8 | 2143 | } |
6e04daf1 | 2144 | |
2145 | return_at_eof = buffer->return_at_eof; | |
a54e0bf8 | 2146 | _cpp_pop_buffer (pfile); |
6e04daf1 | 2147 | if (pfile->buffer == NULL || return_at_eof) |
11b5269c | 2148 | return false; |
a54e0bf8 | 2149 | } |
0bb65704 | 2150 | } |
2151 | ||
edaf8cb5 | 2152 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
2153 | do \ | |
2154 | { \ | |
2155 | result->type = ELSE_TYPE; \ | |
2156 | if (*buffer->cur == CHAR) \ | |
2157 | buffer->cur++, result->type = THEN_TYPE; \ | |
2158 | } \ | |
2159 | while (0) | |
1c124f85 | 2160 | |
10b4496a | 2161 | /* Lex a token into pfile->cur_token, which is also incremented, to |
2162 | get diagnostics pointing to the correct location. | |
2163 | ||
2164 | Does not handle issues such as token lookahead, multiple-include | |
4172d65e | 2165 | optimization, directives, skipping etc. This function is only |
10b4496a | 2166 | suitable for use by _cpp_lex_token, and in special cases like |
2167 | lex_expansion_token which doesn't care for any of these issues. | |
2168 | ||
2169 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
2170 | otherwise returns to the start of the token buffer if permissible. | |
2171 | Returns the location of the lexed token. */ | |
2172 | cpp_token * | |
f7fdd7a1 | 2173 | _cpp_lex_direct (cpp_reader *pfile) |
0578f103 | 2174 | { |
338fa5f7 | 2175 | cppchar_t c; |
230f0943 | 2176 | cpp_buffer *buffer; |
338fa5f7 | 2177 | const unsigned char *comment_start; |
10b4496a | 2178 | cpp_token *result = pfile->cur_token++; |
0653b94e | 2179 | |
83dcbb5c | 2180 | fresh_line: |
a54e0bf8 | 2181 | result->flags = 0; |
82166c5c | 2182 | buffer = pfile->buffer; |
11b5269c | 2183 | if (buffer->need_line) |
a54e0bf8 | 2184 | { |
b75b98aa | 2185 | if (pfile->state.in_deferred_pragma) |
2186 | { | |
2187 | result->type = CPP_PRAGMA_EOL; | |
2188 | pfile->state.in_deferred_pragma = false; | |
2189 | if (!pfile->state.pragma_allow_expansion) | |
2190 | pfile->state.prevent_expansion--; | |
2191 | return result; | |
2192 | } | |
a54e0bf8 | 2193 | if (!_cpp_get_fresh_line (pfile)) |
2194 | { | |
2195 | result->type = CPP_EOF; | |
2908f819 | 2196 | if (!pfile->state.in_directive) |
2197 | { | |
2198 | /* Tell the compiler the line number of the EOF token. */ | |
dbddc569 | 2199 | result->src_loc = pfile->line_table->highest_line; |
2908f819 | 2200 | result->flags = BOL; |
2201 | } | |
a54e0bf8 | 2202 | return result; |
2203 | } | |
2204 | if (!pfile->keep_tokens) | |
2205 | { | |
2206 | pfile->cur_run = &pfile->base_run; | |
2207 | result = pfile->base_run.base; | |
2208 | pfile->cur_token = result + 1; | |
2209 | } | |
2210 | result->flags = BOL; | |
2211 | if (pfile->state.parsing_args == 2) | |
2212 | result->flags |= PREV_WHITE; | |
2213 | } | |
11b5269c | 2214 | buffer = pfile->buffer; |
83dcbb5c | 2215 | update_tokens_line: |
dbddc569 | 2216 | result->src_loc = pfile->line_table->highest_line; |
f80e83a9 | 2217 | |
83dcbb5c | 2218 | skipped_white: |
a54e0bf8 | 2219 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
2220 | && !pfile->overlaid_buffer) | |
2221 | { | |
2222 | _cpp_process_line_notes (pfile, false); | |
dbddc569 | 2223 | result->src_loc = pfile->line_table->highest_line; |
a54e0bf8 | 2224 | } |
1c124f85 | 2225 | c = *buffer->cur++; |
610625e3 | 2226 | |
6ea2c7a3 | 2227 | if (pfile->forced_token_location_p) |
2228 | result->src_loc = *pfile->forced_token_location_p; | |
2229 | else | |
2230 | result->src_loc = linemap_position_for_column (pfile->line_table, | |
2231 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
83dcbb5c | 2232 | |
338fa5f7 | 2233 | switch (c) |
0578f103 | 2234 | { |
435fb09b | 2235 | case ' ': case '\t': case '\f': case '\v': case '\0': |
2236 | result->flags |= PREV_WHITE; | |
a54e0bf8 | 2237 | skip_whitespace (pfile, c); |
2238 | goto skipped_white; | |
338fa5f7 | 2239 | |
a54e0bf8 | 2240 | case '\n': |
610625e3 | 2241 | if (buffer->cur < buffer->rlimit) |
2242 | CPP_INCREMENT_LINE (pfile, 0); | |
a54e0bf8 | 2243 | buffer->need_line = true; |
2244 | goto fresh_line; | |
732cb4c9 | 2245 | |
338fa5f7 | 2246 | case '0': case '1': case '2': case '3': case '4': |
2247 | case '5': case '6': case '7': case '8': case '9': | |
bce47149 | 2248 | { |
2249 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2250 | result->type = CPP_NUMBER; | |
2251 | lex_number (pfile, &result->val.str, &nst); | |
2252 | warn_about_normalization (pfile, result, &nst); | |
2253 | break; | |
2254 | } | |
732cb4c9 | 2255 | |
78c551ad | 2256 | case 'L': |
924bbf02 | 2257 | case 'u': |
2258 | case 'U': | |
538ba11a | 2259 | case 'R': |
2260 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2261 | wide strings or raw strings. */ | |
6f6f3dd7 | 2262 | if (c == 'L' || CPP_OPTION (pfile, rliterals) |
2263 | || (c != 'R' && CPP_OPTION (pfile, uliterals))) | |
5bb46c08 | 2264 | { |
538ba11a | 2265 | if ((*buffer->cur == '\'' && c != 'R') |
2266 | || *buffer->cur == '"' | |
2267 | || (*buffer->cur == 'R' | |
2268 | && c != 'R' | |
2269 | && buffer->cur[1] == '"' | |
6f6f3dd7 | 2270 | && CPP_OPTION (pfile, rliterals)) |
538ba11a | 2271 | || (*buffer->cur == '8' |
2272 | && c == 'u' | |
2273 | && (buffer->cur[1] == '"' | |
6f6f3dd7 | 2274 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
2275 | && CPP_OPTION (pfile, rliterals))))) | |
924bbf02 | 2276 | { |
2277 | lex_string (pfile, result, buffer->cur - 1); | |
2278 | break; | |
2279 | } | |
5bb46c08 | 2280 | } |
b1a9ff83 | 2281 | /* Fall through. */ |
78c551ad | 2282 | |
338fa5f7 | 2283 | case '_': |
2284 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2285 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2286 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
924bbf02 | 2287 | case 's': case 't': case 'v': case 'w': case 'x': |
338fa5f7 | 2288 | case 'y': case 'z': |
2289 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
78c551ad | 2290 | case 'G': case 'H': case 'I': case 'J': case 'K': |
538ba11a | 2291 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
924bbf02 | 2292 | case 'S': case 'T': case 'V': case 'W': case 'X': |
338fa5f7 | 2293 | case 'Y': case 'Z': |
2294 | result->type = CPP_NAME; | |
bce47149 | 2295 | { |
2296 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2ee04baa | 2297 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
2298 | &nst); | |
bce47149 | 2299 | warn_about_normalization (pfile, result, &nst); |
2300 | } | |
338fa5f7 | 2301 | |
338fa5f7 | 2302 | /* Convert named operators to their proper types. */ |
2ee04baa | 2303 | if (result->val.node.node->flags & NODE_OPERATOR) |
338fa5f7 | 2304 | { |
2305 | result->flags |= NAMED_OP; | |
2ee04baa | 2306 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
338fa5f7 | 2307 | } |
2308 | break; | |
2309 | ||
2310 | case '\'': | |
2311 | case '"': | |
4970d4c2 | 2312 | lex_string (pfile, result, buffer->cur - 1); |
338fa5f7 | 2313 | break; |
f80e83a9 | 2314 | |
338fa5f7 | 2315 | case '/': |
f0495c2c | 2316 | /* A potential block or line comment. */ |
2317 | comment_start = buffer->cur; | |
edaf8cb5 | 2318 | c = *buffer->cur; |
2319 | ||
f0495c2c | 2320 | if (c == '*') |
2321 | { | |
a54e0bf8 | 2322 | if (_cpp_skip_block_comment (pfile)) |
d80d2074 | 2323 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
338fa5f7 | 2324 | } |
1c124f85 | 2325 | else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments) |
610625e3 | 2326 | || cpp_in_system_header (pfile))) |
338fa5f7 | 2327 | { |
890c2e2f | 2328 | /* Warn about comments if pedantically GNUC89, and not |
5db5d057 | 2329 | in system headers. */ |
2330 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
66914e49 | 2331 | && ! buffer->warned_cplusplus_comments) |
f80e83a9 | 2332 | { |
d80d2074 | 2333 | cpp_error (pfile, CPP_DL_PEDWARN, |
ba059ac0 | 2334 | "C++ style comments are not allowed in ISO C90"); |
d80d2074 | 2335 | cpp_error (pfile, CPP_DL_PEDWARN, |
73328dce | 2336 | "(this will be reported only once per input file)"); |
f0495c2c | 2337 | buffer->warned_cplusplus_comments = 1; |
2338 | } | |
890c2e2f | 2339 | /* Or if specifically desired via -Wc90-c99-compat. */ |
806fe15e | 2340 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat) > 0 |
890c2e2f | 2341 | && ! buffer->warned_cplusplus_comments) |
2342 | { | |
2343 | cpp_error (pfile, CPP_DL_WARNING, | |
2344 | "C++ style comments are are incompatible with C90"); | |
2345 | cpp_error (pfile, CPP_DL_WARNING, | |
2346 | "(this will be reported only once per input file)"); | |
2347 | buffer->warned_cplusplus_comments = 1; | |
2348 | } | |
338fa5f7 | 2349 | |
e1caf668 | 2350 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
3a79f5da | 2351 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
f0495c2c | 2352 | } |
1c124f85 | 2353 | else if (c == '=') |
2354 | { | |
edaf8cb5 | 2355 | buffer->cur++; |
1c124f85 | 2356 | result->type = CPP_DIV_EQ; |
2357 | break; | |
2358 | } | |
2359 | else | |
2360 | { | |
1c124f85 | 2361 | result->type = CPP_DIV; |
2362 | break; | |
2363 | } | |
338fa5f7 | 2364 | |
f0495c2c | 2365 | if (!pfile->state.save_comments) |
2366 | { | |
2367 | result->flags |= PREV_WHITE; | |
83dcbb5c | 2368 | goto update_tokens_line; |
338fa5f7 | 2369 | } |
f0495c2c | 2370 | |
2371 | /* Save the comment as a token in its own right. */ | |
d3f7919d | 2372 | save_comment (pfile, result, comment_start, c); |
fb5ab82c | 2373 | break; |
338fa5f7 | 2374 | |
2375 | case '<': | |
2376 | if (pfile->state.angled_headers) | |
2377 | { | |
4970d4c2 | 2378 | lex_string (pfile, result, buffer->cur - 1); |
7811eab5 | 2379 | if (result->type != CPP_LESS) |
2380 | break; | |
338fa5f7 | 2381 | } |
0578f103 | 2382 | |
edaf8cb5 | 2383 | result->type = CPP_LESS; |
2384 | if (*buffer->cur == '=') | |
2385 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2386 | else if (*buffer->cur == '<') | |
338fa5f7 | 2387 | { |
edaf8cb5 | 2388 | buffer->cur++; |
2389 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
338fa5f7 | 2390 | } |
edaf8cb5 | 2391 | else if (CPP_OPTION (pfile, digraphs)) |
1c124f85 | 2392 | { |
edaf8cb5 | 2393 | if (*buffer->cur == ':') |
2394 | { | |
1aa79d39 | 2395 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
2396 | three characters are <:: and the subsequent character | |
2397 | is neither : nor >, the < is treated as a preprocessor | |
2398 | token by itself". */ | |
2399 | if (CPP_OPTION (pfile, cplusplus) | |
1638c736 | 2400 | && CPP_OPTION (pfile, lang) != CLK_CXX98 |
2401 | && CPP_OPTION (pfile, lang) != CLK_GNUCXX | |
1aa79d39 | 2402 | && buffer->cur[1] == ':' |
2403 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') | |
2404 | break; | |
2405 | ||
edaf8cb5 | 2406 | buffer->cur++; |
2407 | result->flags |= DIGRAPH; | |
2408 | result->type = CPP_OPEN_SQUARE; | |
2409 | } | |
2410 | else if (*buffer->cur == '%') | |
2411 | { | |
2412 | buffer->cur++; | |
2413 | result->flags |= DIGRAPH; | |
2414 | result->type = CPP_OPEN_BRACE; | |
2415 | } | |
1c124f85 | 2416 | } |
338fa5f7 | 2417 | break; |
2418 | ||
2419 | case '>': | |
edaf8cb5 | 2420 | result->type = CPP_GREATER; |
2421 | if (*buffer->cur == '=') | |
2422 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
2423 | else if (*buffer->cur == '>') | |
338fa5f7 | 2424 | { |
edaf8cb5 | 2425 | buffer->cur++; |
2426 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
2427 | } | |
338fa5f7 | 2428 | break; |
2429 | ||
f669338a | 2430 | case '%': |
edaf8cb5 | 2431 | result->type = CPP_MOD; |
2432 | if (*buffer->cur == '=') | |
2433 | buffer->cur++, result->type = CPP_MOD_EQ; | |
2434 | else if (CPP_OPTION (pfile, digraphs)) | |
1c124f85 | 2435 | { |
edaf8cb5 | 2436 | if (*buffer->cur == ':') |
1c124f85 | 2437 | { |
edaf8cb5 | 2438 | buffer->cur++; |
2439 | result->flags |= DIGRAPH; | |
2440 | result->type = CPP_HASH; | |
2441 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
2ee04baa | 2442 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
edaf8cb5 | 2443 | } |
2444 | else if (*buffer->cur == '>') | |
2445 | { | |
2446 | buffer->cur++; | |
2447 | result->flags |= DIGRAPH; | |
2448 | result->type = CPP_CLOSE_BRACE; | |
1c124f85 | 2449 | } |
1c124f85 | 2450 | } |
338fa5f7 | 2451 | break; |
2452 | ||
f669338a | 2453 | case '.': |
1c124f85 | 2454 | result->type = CPP_DOT; |
edaf8cb5 | 2455 | if (ISDIGIT (*buffer->cur)) |
1c124f85 | 2456 | { |
bce47149 | 2457 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
1c124f85 | 2458 | result->type = CPP_NUMBER; |
bce47149 | 2459 | lex_number (pfile, &result->val.str, &nst); |
2460 | warn_about_normalization (pfile, result, &nst); | |
1c124f85 | 2461 | } |
edaf8cb5 | 2462 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
2463 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
2464 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
2465 | buffer->cur++, result->type = CPP_DOT_STAR; | |
338fa5f7 | 2466 | break; |
0578f103 | 2467 | |
338fa5f7 | 2468 | case '+': |
edaf8cb5 | 2469 | result->type = CPP_PLUS; |
2470 | if (*buffer->cur == '+') | |
2471 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
2472 | else if (*buffer->cur == '=') | |
2473 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
338fa5f7 | 2474 | break; |
ac0749c7 | 2475 | |
338fa5f7 | 2476 | case '-': |
edaf8cb5 | 2477 | result->type = CPP_MINUS; |
2478 | if (*buffer->cur == '>') | |
338fa5f7 | 2479 | { |
edaf8cb5 | 2480 | buffer->cur++; |
1c124f85 | 2481 | result->type = CPP_DEREF; |
edaf8cb5 | 2482 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
2483 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
1c124f85 | 2484 | } |
edaf8cb5 | 2485 | else if (*buffer->cur == '-') |
2486 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
2487 | else if (*buffer->cur == '=') | |
2488 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
338fa5f7 | 2489 | break; |
0578f103 | 2490 | |
338fa5f7 | 2491 | case '&': |
edaf8cb5 | 2492 | result->type = CPP_AND; |
2493 | if (*buffer->cur == '&') | |
2494 | buffer->cur++, result->type = CPP_AND_AND; | |
2495 | else if (*buffer->cur == '=') | |
2496 | buffer->cur++, result->type = CPP_AND_EQ; | |
338fa5f7 | 2497 | break; |
b1a9ff83 | 2498 | |
338fa5f7 | 2499 | case '|': |
edaf8cb5 | 2500 | result->type = CPP_OR; |
2501 | if (*buffer->cur == '|') | |
2502 | buffer->cur++, result->type = CPP_OR_OR; | |
2503 | else if (*buffer->cur == '=') | |
2504 | buffer->cur++, result->type = CPP_OR_EQ; | |
338fa5f7 | 2505 | break; |
0578f103 | 2506 | |
338fa5f7 | 2507 | case ':': |
edaf8cb5 | 2508 | result->type = CPP_COLON; |
2509 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
2510 | buffer->cur++, result->type = CPP_SCOPE; | |
2511 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
338fa5f7 | 2512 | { |
edaf8cb5 | 2513 | buffer->cur++; |
338fa5f7 | 2514 | result->flags |= DIGRAPH; |
1c124f85 | 2515 | result->type = CPP_CLOSE_SQUARE; |
2516 | } | |
338fa5f7 | 2517 | break; |
0578f103 | 2518 | |
1c124f85 | 2519 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
2520 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
2521 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
2522 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
2ee04baa | 2523 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
1c124f85 | 2524 | |
a54e0bf8 | 2525 | case '?': result->type = CPP_QUERY; break; |
338fa5f7 | 2526 | case '~': result->type = CPP_COMPL; break; |
2527 | case ',': result->type = CPP_COMMA; break; | |
2528 | case '(': result->type = CPP_OPEN_PAREN; break; | |
2529 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
2530 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
2531 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
2532 | case '{': result->type = CPP_OPEN_BRACE; break; | |
2533 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
2534 | case ';': result->type = CPP_SEMICOLON; break; | |
2535 | ||
7fd957fe | 2536 | /* @ is a punctuator in Objective-C. */ |
9ee99ac6 | 2537 | case '@': result->type = CPP_ATSIGN; break; |
338fa5f7 | 2538 | |
78c551ad | 2539 | case '$': |
2cbf1359 | 2540 | case '\\': |
2541 | { | |
2542 | const uchar *base = --buffer->cur; | |
bce47149 | 2543 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
78c551ad | 2544 | |
bce47149 | 2545 | if (forms_identifier_p (pfile, true, &nst)) |
2cbf1359 | 2546 | { |
2547 | result->type = CPP_NAME; | |
2ee04baa | 2548 | result->val.node.node = lex_identifier (pfile, base, true, &nst); |
bce47149 | 2549 | warn_about_normalization (pfile, result, &nst); |
2cbf1359 | 2550 | break; |
2551 | } | |
2552 | buffer->cur++; | |
bc205914 | 2553 | } |
2cbf1359 | 2554 | |
bc205914 | 2555 | default: |
4970d4c2 | 2556 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); |
2557 | break; | |
338fa5f7 | 2558 | } |
fb5ab82c | 2559 | |
2560 | return result; | |
338fa5f7 | 2561 | } |
2562 | ||
b1280514 | 2563 | /* An upper bound on the number of bytes needed to spell TOKEN. |
2564 | Does not include preceding whitespace. */ | |
79bd622b | 2565 | unsigned int |
f7fdd7a1 | 2566 | cpp_token_len (const cpp_token *token) |
338fa5f7 | 2567 | { |
79bd622b | 2568 | unsigned int len; |
cfad5579 | 2569 | |
79bd622b | 2570 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2571 | { |
cd740bd5 | 2572 | default: len = 6; break; |
4970d4c2 | 2573 | case SPELL_LITERAL: len = token->val.str.len; break; |
2ee04baa | 2574 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
f80e83a9 | 2575 | } |
b1280514 | 2576 | |
2577 | return len; | |
cfad5579 | 2578 | } |
2579 | ||
bb1fa6bb | 2580 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
2581 | Return the number of bytes read out of NAME. (There are always | |
2582 | 10 bytes written to BUFFER.) */ | |
2583 | ||
2584 | static size_t | |
2585 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
2586 | { | |
2587 | int j; | |
2588 | int ucn_len = 0; | |
2589 | int ucn_len_c; | |
2590 | unsigned t; | |
2591 | unsigned long utf32; | |
2592 | ||
2593 | /* Compute the length of the UTF-8 sequence. */ | |
2594 | for (t = *name; t & 0x80; t <<= 1) | |
2595 | ucn_len++; | |
2596 | ||
2597 | utf32 = *name & (0x7F >> ucn_len); | |
2598 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
2599 | { | |
2600 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
2601 | ||
2602 | /* Ill-formed UTF-8. */ | |
2603 | if ((*name & ~0x3F) != 0x80) | |
2604 | abort (); | |
2605 | } | |
2606 | ||
2607 | *buffer++ = '\\'; | |
2608 | *buffer++ = 'U'; | |
2609 | for (j = 7; j >= 0; j--) | |
2610 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
2611 | return ucn_len; | |
2612 | } | |
2613 | ||
ba99525e | 2614 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
2615 | the spelling of the digraph. */ | |
2616 | static const unsigned char * | |
2617 | cpp_digraph2name (enum cpp_ttype type) | |
2618 | { | |
2619 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
2620 | } | |
bb1fa6bb | 2621 | |
f80e83a9 | 2622 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 2623 | already contain the enough space to hold the token's spelling. |
f7fdd7a1 | 2624 | Returns a pointer to the character after the last character written. |
bb1fa6bb | 2625 | FORSTRING is true if this is to be the spelling after translation |
2626 | phase 1 (this is different for UCNs). | |
f7fdd7a1 | 2627 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
79bd622b | 2628 | unsigned char * |
f7fdd7a1 | 2629 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
bb1fa6bb | 2630 | unsigned char *buffer, bool forstring) |
f80e83a9 | 2631 | { |
7e842f95 | 2632 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2633 | { |
2634 | case SPELL_OPERATOR: | |
2635 | { | |
2636 | const unsigned char *spelling; | |
2637 | unsigned char c; | |
ab12a39c | 2638 | |
f80e83a9 | 2639 | if (token->flags & DIGRAPH) |
ba99525e | 2640 | spelling = cpp_digraph2name (token->type); |
31674461 | 2641 | else if (token->flags & NAMED_OP) |
2642 | goto spell_ident; | |
f80e83a9 | 2643 | else |
7e842f95 | 2644 | spelling = TOKEN_NAME (token); |
b1a9ff83 | 2645 | |
f80e83a9 | 2646 | while ((c = *spelling++) != '\0') |
2647 | *buffer++ = c; | |
2648 | } | |
2649 | break; | |
ab12a39c | 2650 | |
8d27e472 | 2651 | spell_ident: |
f80e83a9 | 2652 | case SPELL_IDENT: |
bb1fa6bb | 2653 | if (forstring) |
2654 | { | |
2ee04baa | 2655 | memcpy (buffer, NODE_NAME (token->val.node.node), |
2656 | NODE_LEN (token->val.node.node)); | |
2657 | buffer += NODE_LEN (token->val.node.node); | |
bb1fa6bb | 2658 | } |
2659 | else | |
2660 | { | |
2661 | size_t i; | |
2ee04baa | 2662 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2663 | |
2ee04baa | 2664 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2665 | if (name[i] & ~0x7F) |
2666 | { | |
2667 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2668 | buffer += 10; | |
2669 | } | |
2670 | else | |
2ee04baa | 2671 | *buffer++ = NODE_NAME (token->val.node.node)[i]; |
bb1fa6bb | 2672 | } |
f80e83a9 | 2673 | break; |
ab12a39c | 2674 | |
4970d4c2 | 2675 | case SPELL_LITERAL: |
8d27e472 | 2676 | memcpy (buffer, token->val.str.text, token->val.str.len); |
2677 | buffer += token->val.str.len; | |
2678 | break; | |
2679 | ||
f80e83a9 | 2680 | case SPELL_NONE: |
d80d2074 | 2681 | cpp_error (pfile, CPP_DL_ICE, |
2682 | "unspellable token %s", TOKEN_NAME (token)); | |
f80e83a9 | 2683 | break; |
2684 | } | |
ab12a39c | 2685 | |
f80e83a9 | 2686 | return buffer; |
2687 | } | |
ab12a39c | 2688 | |
e484a1cc | 2689 | /* Returns TOKEN spelt as a null-terminated string. The string is |
2690 | freed when the reader is destroyed. Useful for diagnostics. */ | |
79bd622b | 2691 | unsigned char * |
f7fdd7a1 | 2692 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
b1280514 | 2693 | { |
2694 | unsigned int len = cpp_token_len (token) + 1; | |
1fdf6039 | 2695 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
6060326b | 2696 | |
bb1fa6bb | 2697 | end = cpp_spell_token (pfile, token, start, false); |
79bd622b | 2698 | end[0] = '\0'; |
6060326b | 2699 | |
79bd622b | 2700 | return start; |
2701 | } | |
6060326b | 2702 | |
ba99525e | 2703 | /* Returns a pointer to a string which spells the token defined by |
2704 | TYPE and FLAGS. Used by C front ends, which really should move to | |
2705 | using cpp_token_as_text. */ | |
79bd622b | 2706 | const char * |
ba99525e | 2707 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
79bd622b | 2708 | { |
ba99525e | 2709 | if (flags & DIGRAPH) |
2710 | return (const char *) cpp_digraph2name (type); | |
2711 | else if (flags & NAMED_OP) | |
2712 | return cpp_named_operator2name (type); | |
2713 | ||
79bd622b | 2714 | return (const char *) token_spellings[type].name; |
2715 | } | |
6060326b | 2716 | |
f9b5f742 | 2717 | /* Writes the spelling of token to FP, without any preceding space. |
2718 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
2719 | double-buffering. */ | |
79bd622b | 2720 | void |
f7fdd7a1 | 2721 | cpp_output_token (const cpp_token *token, FILE *fp) |
79bd622b | 2722 | { |
79bd622b | 2723 | switch (TOKEN_SPELL (token)) |
6060326b | 2724 | { |
79bd622b | 2725 | case SPELL_OPERATOR: |
2726 | { | |
2727 | const unsigned char *spelling; | |
28874558 | 2728 | int c; |
6060326b | 2729 | |
79bd622b | 2730 | if (token->flags & DIGRAPH) |
ba99525e | 2731 | spelling = cpp_digraph2name (token->type); |
79bd622b | 2732 | else if (token->flags & NAMED_OP) |
2733 | goto spell_ident; | |
2734 | else | |
2735 | spelling = TOKEN_NAME (token); | |
f80e83a9 | 2736 | |
28874558 | 2737 | c = *spelling; |
2738 | do | |
2739 | putc (c, fp); | |
2740 | while ((c = *++spelling) != '\0'); | |
79bd622b | 2741 | } |
2742 | break; | |
f80e83a9 | 2743 | |
79bd622b | 2744 | spell_ident: |
2745 | case SPELL_IDENT: | |
bb1fa6bb | 2746 | { |
2747 | size_t i; | |
2ee04baa | 2748 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2749 | |
2ee04baa | 2750 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2751 | if (name[i] & ~0x7F) |
2752 | { | |
2753 | unsigned char buffer[10]; | |
2754 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2755 | fwrite (buffer, 1, 10, fp); | |
2756 | } | |
2757 | else | |
2ee04baa | 2758 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
bb1fa6bb | 2759 | } |
2760 | break; | |
f80e83a9 | 2761 | |
4970d4c2 | 2762 | case SPELL_LITERAL: |
8d27e472 | 2763 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
2764 | break; | |
2765 | ||
79bd622b | 2766 | case SPELL_NONE: |
2767 | /* An error, most probably. */ | |
2768 | break; | |
f80e83a9 | 2769 | } |
6060326b | 2770 | } |
2771 | ||
79bd622b | 2772 | /* Compare two tokens. */ |
2773 | int | |
f7fdd7a1 | 2774 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
6060326b | 2775 | { |
79bd622b | 2776 | if (a->type == b->type && a->flags == b->flags) |
2777 | switch (TOKEN_SPELL (a)) | |
2778 | { | |
2779 | default: /* Keep compiler happy. */ | |
2780 | case SPELL_OPERATOR: | |
2ee04baa | 2781 | /* token_no is used to track where multiple consecutive ## |
941f2388 | 2782 | tokens were originally located. */ |
2ee04baa | 2783 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
79bd622b | 2784 | case SPELL_NONE: |
2ee04baa | 2785 | return (a->type != CPP_MACRO_ARG |
2786 | || a->val.macro_arg.arg_no == b->val.macro_arg.arg_no); | |
79bd622b | 2787 | case SPELL_IDENT: |
2ee04baa | 2788 | return a->val.node.node == b->val.node.node; |
4970d4c2 | 2789 | case SPELL_LITERAL: |
79bd622b | 2790 | return (a->val.str.len == b->val.str.len |
2791 | && !memcmp (a->val.str.text, b->val.str.text, | |
2792 | a->val.str.len)); | |
2793 | } | |
6060326b | 2794 | |
f80e83a9 | 2795 | return 0; |
2796 | } | |
2797 | ||
79bd622b | 2798 | /* Returns nonzero if a space should be inserted to avoid an |
2799 | accidental token paste for output. For simplicity, it is | |
2800 | conservative, and occasionally advises a space where one is not | |
2801 | needed, e.g. "." and ".2". */ | |
79bd622b | 2802 | int |
f7fdd7a1 | 2803 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
2804 | const cpp_token *token2) | |
6060326b | 2805 | { |
79bd622b | 2806 | enum cpp_ttype a = token1->type, b = token2->type; |
2807 | cppchar_t c; | |
6060326b | 2808 | |
79bd622b | 2809 | if (token1->flags & NAMED_OP) |
2810 | a = CPP_NAME; | |
2811 | if (token2->flags & NAMED_OP) | |
2812 | b = CPP_NAME; | |
6060326b | 2813 | |
79bd622b | 2814 | c = EOF; |
2815 | if (token2->flags & DIGRAPH) | |
ee6c4e4b | 2816 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
79bd622b | 2817 | else if (token_spellings[b].category == SPELL_OPERATOR) |
2818 | c = token_spellings[b].name[0]; | |
6060326b | 2819 | |
79bd622b | 2820 | /* Quickly get everything that can paste with an '='. */ |
ee6c4e4b | 2821 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
79bd622b | 2822 | return 1; |
6060326b | 2823 | |
79bd622b | 2824 | switch (a) |
6060326b | 2825 | { |
e58c07f7 | 2826 | case CPP_GREATER: return c == '>'; |
2827 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
79bd622b | 2828 | case CPP_PLUS: return c == '+'; |
2829 | case CPP_MINUS: return c == '-' || c == '>'; | |
2830 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
2831 | case CPP_MOD: return c == ':' || c == '>'; | |
2832 | case CPP_AND: return c == '&'; | |
2833 | case CPP_OR: return c == '|'; | |
2834 | case CPP_COLON: return c == ':' || c == '>'; | |
2835 | case CPP_DEREF: return c == '*'; | |
efdcc728 | 2836 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
79bd622b | 2837 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
2838 | case CPP_NAME: return ((b == CPP_NUMBER | |
2839 | && name_p (pfile, &token2->val.str)) | |
2840 | || b == CPP_NAME | |
2841 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
2842 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
2843 | || c == '.' || c == '+' || c == '-'); | |
2cbf1359 | 2844 | /* UCNs */ |
bc205914 | 2845 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
2846 | && b == CPP_NAME) | |
2cbf1359 | 2847 | || (CPP_OPTION (pfile, objc) |
bc205914 | 2848 | && token1->val.str.text[0] == '@' |
2cbf1359 | 2849 | && (b == CPP_NAME || b == CPP_STRING))); |
bd285415 | 2850 | case CPP_STRING: |
2851 | case CPP_WSTRING: | |
2852 | case CPP_UTF8STRING: | |
2853 | case CPP_STRING16: | |
2854 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals) | |
2855 | && (b == CPP_NAME | |
2856 | || (TOKEN_SPELL (token2) == SPELL_LITERAL | |
2857 | && ISIDST (token2->val.str.text[0])))); | |
2858 | ||
79bd622b | 2859 | default: break; |
6060326b | 2860 | } |
6060326b | 2861 | |
deb356cf | 2862 | return 0; |
6060326b | 2863 | } |
2864 | ||
79bd622b | 2865 | /* Output all the remaining tokens on the current line, and a newline |
f9b5f742 | 2866 | character, to FP. Leading whitespace is removed. If there are |
2867 | macros, special token padding is not performed. */ | |
6060326b | 2868 | void |
f7fdd7a1 | 2869 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
6060326b | 2870 | { |
f9b5f742 | 2871 | const cpp_token *token; |
7e842f95 | 2872 | |
f9b5f742 | 2873 | token = cpp_get_token (pfile); |
2874 | while (token->type != CPP_EOF) | |
7e842f95 | 2875 | { |
f9b5f742 | 2876 | cpp_output_token (token, fp); |
2877 | token = cpp_get_token (pfile); | |
2878 | if (token->flags & PREV_WHITE) | |
2879 | putc (' ', fp); | |
7e842f95 | 2880 | } |
2881 | ||
79bd622b | 2882 | putc ('\n', fp); |
f80e83a9 | 2883 | } |
6060326b | 2884 | |
c0770282 | 2885 | /* Return a string representation of all the remaining tokens on the |
2886 | current line. The result is allocated using xmalloc and must be | |
2887 | freed by the caller. */ | |
2888 | unsigned char * | |
2889 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
2890 | { | |
2891 | const cpp_token *token; | |
2892 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
2893 | unsigned int alloced = 120 + out; | |
2894 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
2895 | ||
2896 | /* If DIR_NAME is empty, there are no initial contents. */ | |
2897 | if (dir_name) | |
2898 | { | |
2899 | sprintf ((char *) result, "#%s ", dir_name); | |
2900 | out += 2; | |
2901 | } | |
2902 | ||
2903 | token = cpp_get_token (pfile); | |
2904 | while (token->type != CPP_EOF) | |
2905 | { | |
2906 | unsigned char *last; | |
2907 | /* Include room for a possible space and the terminating nul. */ | |
2908 | unsigned int len = cpp_token_len (token) + 2; | |
2909 | ||
2910 | if (out + len > alloced) | |
2911 | { | |
2912 | alloced *= 2; | |
2913 | if (out + len > alloced) | |
2914 | alloced = out + len; | |
2915 | result = (unsigned char *) xrealloc (result, alloced); | |
2916 | } | |
2917 | ||
2918 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
2919 | out = last - result; | |
2920 | ||
2921 | token = cpp_get_token (pfile); | |
2922 | if (token->flags & PREV_WHITE) | |
2923 | result[out++] = ' '; | |
2924 | } | |
2925 | ||
2926 | result[out] = '\0'; | |
2927 | return result; | |
2928 | } | |
2929 | ||
084163dc | 2930 | /* Memory buffers. Changing these three constants can have a dramatic |
2931 | effect on performance. The values here are reasonable defaults, | |
2932 | but might be tuned. If you adjust them, be sure to test across a | |
2933 | range of uses of cpplib, including heavy nested function-like macro | |
2934 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2935 | good tool for this). */ | |
2936 | #define MIN_BUFF_SIZE 8000 | |
1e0ef2fd | 2937 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
084163dc | 2938 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
2939 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
deb356cf | 2940 | |
1e0ef2fd | 2941 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
2942 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
2943 | #endif | |
2944 | ||
1785b647 | 2945 | /* Create a new allocation buffer. Place the control block at the end |
2946 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
06c92cbc | 2947 | static _cpp_buff * |
f7fdd7a1 | 2948 | new_buff (size_t len) |
06c92cbc | 2949 | { |
2950 | _cpp_buff *result; | |
1fdf6039 | 2951 | unsigned char *base; |
06c92cbc | 2952 | |
084163dc | 2953 | if (len < MIN_BUFF_SIZE) |
2954 | len = MIN_BUFF_SIZE; | |
198b48a0 | 2955 | len = CPP_ALIGN (len); |
06c92cbc | 2956 | |
61ed1f10 | 2957 | #ifdef ENABLE_VALGRIND_CHECKING |
2958 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff | |
2959 | struct first. */ | |
2960 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT); | |
2961 | base = XNEWVEC (unsigned char, len + slen); | |
2962 | result = (_cpp_buff *) base; | |
2963 | base += slen; | |
2964 | #else | |
720aca92 | 2965 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
06c92cbc | 2966 | result = (_cpp_buff *) (base + len); |
61ed1f10 | 2967 | #endif |
06c92cbc | 2968 | result->base = base; |
2969 | result->cur = base; | |
2970 | result->limit = base + len; | |
2971 | result->next = NULL; | |
2972 | return result; | |
2973 | } | |
2974 | ||
2975 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2976 | void | |
f7fdd7a1 | 2977 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
06c92cbc | 2978 | { |
2979 | _cpp_buff *end = buff; | |
2980 | ||
2981 | while (end->next) | |
2982 | end = end->next; | |
2983 | end->next = pfile->free_buffs; | |
2984 | pfile->free_buffs = buff; | |
2985 | } | |
2986 | ||
2987 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2988 | _cpp_buff * | |
f7fdd7a1 | 2989 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
06c92cbc | 2990 | { |
2991 | _cpp_buff *result, **p; | |
2992 | ||
2993 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2994 | { | |
4b31a107 | 2995 | size_t size; |
084163dc | 2996 | |
2997 | if (*p == NULL) | |
06c92cbc | 2998 | return new_buff (min_size); |
084163dc | 2999 | result = *p; |
3000 | size = result->limit - result->base; | |
3001 | /* Return a buffer that's big enough, but don't waste one that's | |
3002 | way too big. */ | |
4085c149 | 3003 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
06c92cbc | 3004 | break; |
3005 | } | |
3006 | ||
3007 | *p = result->next; | |
3008 | result->next = NULL; | |
3009 | result->cur = result->base; | |
3010 | return result; | |
3011 | } | |
3012 | ||
20dd417a | 3013 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 3014 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
3015 | the excess bytes to the new buffer. Chains the new buffer after | |
3016 | BUFF, and returns the new buffer. */ | |
06c92cbc | 3017 | _cpp_buff * |
f7fdd7a1 | 3018 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
06c92cbc | 3019 | { |
4b31a107 | 3020 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
e6a5f963 | 3021 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
06c92cbc | 3022 | |
e6a5f963 | 3023 | buff->next = new_buff; |
3024 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
3025 | return new_buff; | |
3026 | } | |
3027 | ||
20dd417a | 3028 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 3029 | remaining bytes of the buffer pointed to by BUFF, and at least |
3030 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
3031 | Chains the new buffer before the buffer pointed to by BUFF, and | |
3032 | updates the pointer to point to the new buffer. */ | |
3033 | void | |
f7fdd7a1 | 3034 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
e6a5f963 | 3035 | { |
3036 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
3037 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
3038 | ||
3039 | new_buff = _cpp_get_buff (pfile, size); | |
3040 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
3041 | new_buff->next = old_buff; | |
3042 | *pbuff = new_buff; | |
06c92cbc | 3043 | } |
3044 | ||
3045 | /* Free a chain of buffers starting at BUFF. */ | |
3046 | void | |
f82b06e0 | 3047 | _cpp_free_buff (_cpp_buff *buff) |
06c92cbc | 3048 | { |
3049 | _cpp_buff *next; | |
3050 | ||
3051 | for (; buff; buff = next) | |
3052 | { | |
3053 | next = buff->next; | |
61ed1f10 | 3054 | #ifdef ENABLE_VALGRIND_CHECKING |
3055 | free (buff); | |
3056 | #else | |
06c92cbc | 3057 | free (buff->base); |
61ed1f10 | 3058 | #endif |
06c92cbc | 3059 | } |
3060 | } | |
deb356cf | 3061 | |
1fdf6039 | 3062 | /* Allocate permanent, unaligned storage of length LEN. */ |
3063 | unsigned char * | |
f7fdd7a1 | 3064 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
1fdf6039 | 3065 | { |
3066 | _cpp_buff *buff = pfile->u_buff; | |
3067 | unsigned char *result = buff->cur; | |
3068 | ||
3069 | if (len > (size_t) (buff->limit - result)) | |
3070 | { | |
3071 | buff = _cpp_get_buff (pfile, len); | |
3072 | buff->next = pfile->u_buff; | |
3073 | pfile->u_buff = buff; | |
3074 | result = buff->cur; | |
3075 | } | |
3076 | ||
3077 | buff->cur = result + len; | |
3078 | return result; | |
3079 | } | |
3080 | ||
1e0ef2fd | 3081 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
3082 | That buffer is used for growing allocations when saving macro | |
3083 | replacement lists in a #define, and when parsing an answer to an | |
3084 | assertion in #assert, #unassert or #if (and therefore possibly | |
3085 | whilst expanding macros). It therefore must not be used by any | |
3086 | code that they might call: specifically the lexer and the guts of | |
3087 | the macro expander. | |
3088 | ||
3089 | All existing other uses clearly fit this restriction: storing | |
3090 | registered pragmas during initialization. */ | |
79bd622b | 3091 | unsigned char * |
f7fdd7a1 | 3092 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
89b05ef6 | 3093 | { |
e6a5f963 | 3094 | _cpp_buff *buff = pfile->a_buff; |
3095 | unsigned char *result = buff->cur; | |
89b05ef6 | 3096 | |
e6a5f963 | 3097 | if (len > (size_t) (buff->limit - result)) |
89b05ef6 | 3098 | { |
e6a5f963 | 3099 | buff = _cpp_get_buff (pfile, len); |
3100 | buff->next = pfile->a_buff; | |
3101 | pfile->a_buff = buff; | |
3102 | result = buff->cur; | |
89b05ef6 | 3103 | } |
f80e83a9 | 3104 | |
e6a5f963 | 3105 | buff->cur = result + len; |
79bd622b | 3106 | return result; |
f80e83a9 | 3107 | } |
c39ed964 | 3108 | |
3109 | /* Say which field of TOK is in use. */ | |
3110 | ||
3111 | enum cpp_token_fld_kind | |
da31536d | 3112 | cpp_token_val_index (const cpp_token *tok) |
c39ed964 | 3113 | { |
3114 | switch (TOKEN_SPELL (tok)) | |
3115 | { | |
3116 | case SPELL_IDENT: | |
3117 | return CPP_TOKEN_FLD_NODE; | |
3118 | case SPELL_LITERAL: | |
3119 | return CPP_TOKEN_FLD_STR; | |
941f2388 | 3120 | case SPELL_OPERATOR: |
3121 | if (tok->type == CPP_PASTE) | |
2ee04baa | 3122 | return CPP_TOKEN_FLD_TOKEN_NO; |
941f2388 | 3123 | else |
3124 | return CPP_TOKEN_FLD_NONE; | |
c39ed964 | 3125 | case SPELL_NONE: |
3126 | if (tok->type == CPP_MACRO_ARG) | |
3127 | return CPP_TOKEN_FLD_ARG_NO; | |
3128 | else if (tok->type == CPP_PADDING) | |
3129 | return CPP_TOKEN_FLD_SOURCE; | |
d6d3c909 | 3130 | else if (tok->type == CPP_PRAGMA) |
b75b98aa | 3131 | return CPP_TOKEN_FLD_PRAGMA; |
c39ed964 | 3132 | /* else fall through */ |
3133 | default: | |
3134 | return CPP_TOKEN_FLD_NONE; | |
3135 | } | |
3136 | } | |
6ea2c7a3 | 3137 | |
3138 | /* All tokens lexed in R after calling this function will be forced to have | |
3139 | their source_location the same as the location referenced by P, until | |
3140 | cpp_stop_forcing_token_locations is called for R. */ | |
3141 | ||
3142 | void | |
3143 | cpp_force_token_locations (cpp_reader *r, source_location *p) | |
3144 | { | |
3145 | r->forced_token_location_p = p; | |
3146 | } | |
3147 | ||
3148 | /* Go back to assigning locations naturally for lexed tokens. */ | |
3149 | ||
3150 | void | |
3151 | cpp_stop_forcing_token_locations (cpp_reader *r) | |
3152 | { | |
3153 | r->forced_token_location_p = NULL; | |
3154 | } |