]>
Commit | Line | Data |
---|---|---|
0578f103 | 1 | /* CPP Library - lexical analysis. |
2431e8ba | 2 | Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009, 2010 |
6bc9506f | 3 | Free Software Foundation, Inc. |
0578f103 | 4 | Contributed by Per Bothner, 1994-95. |
5 | Based on CCCP program by Paul Rubin, June 1986 | |
6 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
7 | Broken out to separate file, Zack Weinberg, Mar 2000 | |
8 | ||
9 | This program is free software; you can redistribute it and/or modify it | |
10 | under the terms of the GNU General Public License as published by the | |
6bc9506f | 11 | Free Software Foundation; either version 3, or (at your option) any |
0578f103 | 12 | later version. |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
6bc9506f | 20 | along with this program; see the file COPYING3. If not see |
21 | <http://www.gnu.org/licenses/>. */ | |
0578f103 | 22 | |
23 | #include "config.h" | |
24 | #include "system.h" | |
0578f103 | 25 | #include "cpplib.h" |
d856c8a6 | 26 | #include "internal.h" |
0578f103 | 27 | |
79bd622b | 28 | enum spell_type |
241e762e | 29 | { |
79bd622b | 30 | SPELL_OPERATOR = 0, |
79bd622b | 31 | SPELL_IDENT, |
4970d4c2 | 32 | SPELL_LITERAL, |
79bd622b | 33 | SPELL_NONE |
241e762e | 34 | }; |
35 | ||
79bd622b | 36 | struct token_spelling |
241e762e | 37 | { |
79bd622b | 38 | enum spell_type category; |
39 | const unsigned char *name; | |
241e762e | 40 | }; |
41 | ||
0ca849f9 | 42 | static const unsigned char *const digraph_spellings[] = |
924bbf02 | 43 | { UC"%:", UC"%:%:", UC"<:", UC":>", UC"<%", UC"%>" }; |
79bd622b | 44 | |
924bbf02 | 45 | #define OP(e, s) { SPELL_OPERATOR, UC s }, |
46 | #define TK(e, s) { SPELL_ ## s, UC #e }, | |
0ca849f9 | 47 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLE }; |
79bd622b | 48 | #undef OP |
49 | #undef TK | |
50 | ||
51 | #define TOKEN_SPELL(token) (token_spellings[(token)->type].category) | |
52 | #define TOKEN_NAME(token) (token_spellings[(token)->type].name) | |
e2f9a79f | 53 | |
f7fdd7a1 | 54 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); |
55 | static int skip_line_comment (cpp_reader *); | |
56 | static void skip_whitespace (cpp_reader *, cppchar_t); | |
f7fdd7a1 | 57 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); |
58 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); | |
956c6108 | 59 | static void store_comment (cpp_reader *, cpp_token *); |
f7fdd7a1 | 60 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, |
61 | unsigned int, enum cpp_ttype); | |
62 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); | |
63 | static int name_p (cpp_reader *, const cpp_string *); | |
f7fdd7a1 | 64 | static tokenrun *next_tokenrun (tokenrun *); |
65 | ||
f7fdd7a1 | 66 | static _cpp_buff *new_buff (size_t); |
bce8e0c0 | 67 | |
e920deaf | 68 | |
f80e83a9 | 69 | /* Utility routine: |
2c63d6c8 | 70 | |
76faa4c0 | 71 | Compares, the token TOKEN to the NUL-terminated string STRING. |
72 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ | |
f80e83a9 | 73 | int |
f7fdd7a1 | 74 | cpp_ideq (const cpp_token *token, const char *string) |
f80e83a9 | 75 | { |
76faa4c0 | 76 | if (token->type != CPP_NAME) |
f80e83a9 | 77 | return 0; |
76faa4c0 | 78 | |
2ee04baa | 79 | return !ustrcmp (NODE_NAME (token->val.node.node), (const uchar *) string); |
bce8e0c0 | 80 | } |
50fd6b48 | 81 | |
a54e0bf8 | 82 | /* Record a note TYPE at byte POS into the current cleaned logical |
83 | line. */ | |
1e0ef2fd | 84 | static void |
f7fdd7a1 | 85 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) |
338fa5f7 | 86 | { |
a54e0bf8 | 87 | if (buffer->notes_used == buffer->notes_cap) |
88 | { | |
89 | buffer->notes_cap = buffer->notes_cap * 2 + 200; | |
720aca92 | 90 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes, |
91 | buffer->notes_cap); | |
a54e0bf8 | 92 | } |
338fa5f7 | 93 | |
a54e0bf8 | 94 | buffer->notes[buffer->notes_used].pos = pos; |
95 | buffer->notes[buffer->notes_used].type = type; | |
96 | buffer->notes_used++; | |
338fa5f7 | 97 | } |
98 | ||
2431e8ba | 99 | \f |
100 | /* Fast path to find line special characters using optimized character | |
101 | scanning algorithms. Anything complicated falls back to the slow | |
102 | path below. Since this loop is very hot it's worth doing these kinds | |
103 | of optimizations. | |
104 | ||
105 | One of the paths through the ifdefs should provide | |
106 | ||
107 | const uchar *search_line_fast (const uchar *s, const uchar *end); | |
108 | ||
109 | Between S and END, search for \n, \r, \\, ?. Return a pointer to | |
110 | the found character. | |
111 | ||
112 | Note that the last character of the buffer is *always* a newline, | |
113 | as forced by _cpp_convert_input. This fact can be used to avoid | |
114 | explicitly looking for the end of the buffer. */ | |
115 | ||
116 | /* Configure gives us an ifdef test. */ | |
117 | #ifndef WORDS_BIGENDIAN | |
118 | #define WORDS_BIGENDIAN 0 | |
119 | #endif | |
120 | ||
121 | /* We'd like the largest integer that fits into a register. There's nothing | |
122 | in <stdint.h> that gives us that. For most hosts this is unsigned long, | |
123 | but MS decided on an LLP64 model. Thankfully when building with GCC we | |
124 | can get the "real" word size. */ | |
125 | #ifdef __GNUC__ | |
126 | typedef unsigned int word_type __attribute__((__mode__(__word__))); | |
127 | #else | |
128 | typedef unsigned long word_type; | |
129 | #endif | |
130 | ||
131 | /* The code below is only expecting sizes 4 or 8. | |
132 | Die at compile-time if this expectation is violated. */ | |
133 | typedef char check_word_type_size | |
134 | [(sizeof(word_type) == 8 || sizeof(word_type) == 4) * 2 - 1]; | |
135 | ||
136 | /* Return X with the first N bytes forced to values that won't match one | |
137 | of the interesting characters. Note that NUL is not interesting. */ | |
138 | ||
139 | static inline word_type | |
140 | acc_char_mask_misalign (word_type val, unsigned int n) | |
141 | { | |
142 | word_type mask = -1; | |
143 | if (WORDS_BIGENDIAN) | |
144 | mask >>= n * 8; | |
145 | else | |
146 | mask <<= n * 8; | |
147 | return val & mask; | |
148 | } | |
149 | ||
150 | /* Return X replicated to all byte positions within WORD_TYPE. */ | |
151 | ||
152 | static inline word_type | |
153 | acc_char_replicate (uchar x) | |
154 | { | |
155 | word_type ret; | |
156 | ||
157 | ret = (x << 24) | (x << 16) | (x << 8) | x; | |
158 | if (sizeof(word_type) == 8) | |
159 | ret = (ret << 16 << 16) | ret; | |
160 | return ret; | |
161 | } | |
162 | ||
163 | /* Return non-zero if some byte of VAL is (probably) C. */ | |
164 | ||
165 | static inline word_type | |
166 | acc_char_cmp (word_type val, word_type c) | |
167 | { | |
168 | #if defined(__GNUC__) && defined(__alpha__) | |
169 | /* We can get exact results using a compare-bytes instruction. | |
170 | Get (val == c) via (0 >= (val ^ c)). */ | |
171 | return __builtin_alpha_cmpbge (0, val ^ c); | |
172 | #else | |
173 | word_type magic = 0x7efefefeU; | |
174 | if (sizeof(word_type) == 8) | |
175 | magic = (magic << 16 << 16) | 0xfefefefeU; | |
176 | magic |= 1; | |
177 | ||
178 | val ^= c; | |
179 | return ((val + magic) ^ ~val) & ~magic; | |
180 | #endif | |
181 | } | |
182 | ||
183 | /* Given the result of acc_char_cmp is non-zero, return the index of | |
184 | the found character. If this was a false positive, return -1. */ | |
185 | ||
186 | static inline int | |
187 | acc_char_index (word_type cmp ATTRIBUTE_UNUSED, | |
188 | word_type val ATTRIBUTE_UNUSED) | |
189 | { | |
190 | #if defined(__GNUC__) && defined(__alpha__) && !WORDS_BIGENDIAN | |
191 | /* The cmpbge instruction sets *bits* of the result corresponding to | |
192 | matches in the bytes with no false positives. */ | |
193 | return __builtin_ctzl (cmp); | |
194 | #else | |
195 | unsigned int i; | |
196 | ||
197 | /* ??? It would be nice to force unrolling here, | |
198 | and have all of these constants folded. */ | |
199 | for (i = 0; i < sizeof(word_type); ++i) | |
200 | { | |
201 | uchar c; | |
202 | if (WORDS_BIGENDIAN) | |
203 | c = (val >> (sizeof(word_type) - i - 1) * 8) & 0xff; | |
204 | else | |
205 | c = (val >> i * 8) & 0xff; | |
206 | ||
207 | if (c == '\n' || c == '\r' || c == '\\' || c == '?') | |
208 | return i; | |
209 | } | |
210 | ||
211 | return -1; | |
212 | #endif | |
213 | } | |
214 | ||
215 | /* A version of the fast scanner using bit fiddling techniques. | |
216 | ||
217 | For 32-bit words, one would normally perform 16 comparisons and | |
218 | 16 branches. With this algorithm one performs 24 arithmetic | |
219 | operations and one branch. Whether this is faster with a 32-bit | |
220 | word size is going to be somewhat system dependent. | |
221 | ||
222 | For 64-bit words, we eliminate twice the number of comparisons | |
223 | and branches without increasing the number of arithmetic operations. | |
224 | It's almost certainly going to be a win with 64-bit word size. */ | |
225 | ||
226 | static const uchar * search_line_acc_char (const uchar *, const uchar *) | |
227 | ATTRIBUTE_UNUSED; | |
228 | ||
229 | static const uchar * | |
230 | search_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
231 | { | |
232 | const word_type repl_nl = acc_char_replicate ('\n'); | |
233 | const word_type repl_cr = acc_char_replicate ('\r'); | |
234 | const word_type repl_bs = acc_char_replicate ('\\'); | |
235 | const word_type repl_qm = acc_char_replicate ('?'); | |
236 | ||
237 | unsigned int misalign; | |
238 | const word_type *p; | |
239 | word_type val, t; | |
240 | ||
241 | /* Align the buffer. Mask out any bytes from before the beginning. */ | |
242 | p = (word_type *)((uintptr_t)s & -sizeof(word_type)); | |
243 | val = *p; | |
244 | misalign = (uintptr_t)s & (sizeof(word_type) - 1); | |
245 | if (misalign) | |
246 | val = acc_char_mask_misalign (val, misalign); | |
247 | ||
248 | /* Main loop. */ | |
249 | while (1) | |
250 | { | |
251 | t = acc_char_cmp (val, repl_nl); | |
252 | t |= acc_char_cmp (val, repl_cr); | |
253 | t |= acc_char_cmp (val, repl_bs); | |
254 | t |= acc_char_cmp (val, repl_qm); | |
255 | ||
256 | if (__builtin_expect (t != 0, 0)) | |
257 | { | |
258 | int i = acc_char_index (t, val); | |
259 | if (i >= 0) | |
260 | return (const uchar *)p + i; | |
261 | } | |
262 | ||
263 | val = *++p; | |
264 | } | |
265 | } | |
266 | ||
8ab29ece | 267 | /* Disable on Solaris 2/x86 until the following problems can be properly |
268 | autoconfed: | |
269 | ||
270 | The Solaris 8 assembler cannot assemble SSE2/SSE4.2 insns. | |
271 | The Solaris 9 assembler cannot assemble SSE4.2 insns. | |
272 | Before Solaris 9 Update 6, SSE insns cannot be executed. | |
273 | The Solaris 10+ assembler tags objects with the instruction set | |
274 | extensions used, so SSE4.2 executables cannot run on machines that | |
275 | don't support that extension. */ | |
276 | ||
277 | #if (GCC_VERSION >= 4005) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__)) | |
2431e8ba | 278 | |
279 | /* Replicated character data to be shared between implementations. | |
280 | Recall that outside of a context with vector support we can't | |
281 | define compatible vector types, therefore these are all defined | |
282 | in terms of raw characters. */ | |
283 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { | |
284 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
285 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, | |
286 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
287 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, | |
288 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
289 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, | |
290 | { '?', '?', '?', '?', '?', '?', '?', '?', | |
291 | '?', '?', '?', '?', '?', '?', '?', '?' }, | |
292 | }; | |
293 | ||
294 | /* A version of the fast scanner using MMX vectorized byte compare insns. | |
295 | ||
296 | This uses the PMOVMSKB instruction which was introduced with "MMX2", | |
297 | which was packaged into SSE1; it is also present in the AMD 3dNOW-A | |
298 | extension. Mark the function as using "sse" so that we emit a real | |
299 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ | |
300 | ||
301 | static const uchar * | |
302 | #ifndef __SSE__ | |
303 | __attribute__((__target__("sse"))) | |
304 | #endif | |
305 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
306 | { | |
307 | typedef char v8qi __attribute__ ((__vector_size__ (8))); | |
308 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); | |
309 | ||
310 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; | |
311 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; | |
312 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; | |
313 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; | |
314 | ||
315 | unsigned int misalign, found, mask; | |
316 | const v8qi *p; | |
317 | v8qi data, t, c; | |
318 | ||
319 | /* Align the source pointer. While MMX doesn't generate unaligned data | |
320 | faults, this allows us to safely scan to the end of the buffer without | |
321 | reading beyond the end of the last page. */ | |
322 | misalign = (uintptr_t)s & 7; | |
323 | p = (const v8qi *)((uintptr_t)s & -8); | |
324 | data = *p; | |
325 | ||
326 | /* Create a mask for the bytes that are valid within the first | |
327 | 16-byte block. The Idea here is that the AND with the mask | |
328 | within the loop is "free", since we need some AND or TEST | |
329 | insn in order to set the flags for the branch anyway. */ | |
330 | mask = -1u << misalign; | |
331 | ||
332 | /* Main loop processing 8 bytes at a time. */ | |
333 | goto start; | |
334 | do | |
335 | { | |
336 | data = *++p; | |
337 | mask = -1; | |
338 | ||
339 | start: | |
340 | t = __builtin_ia32_pcmpeqb(data, repl_nl); | |
341 | c = __builtin_ia32_pcmpeqb(data, repl_cr); | |
342 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
343 | c = __builtin_ia32_pcmpeqb(data, repl_bs); | |
344 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
345 | c = __builtin_ia32_pcmpeqb(data, repl_qm); | |
346 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); | |
347 | found = __builtin_ia32_pmovmskb (t); | |
348 | found &= mask; | |
349 | } | |
350 | while (!found); | |
351 | ||
352 | __builtin_ia32_emms (); | |
353 | ||
354 | /* FOUND contains 1 in bits for which we matched a relevant | |
355 | character. Conversion to the byte index is trivial. */ | |
356 | found = __builtin_ctz(found); | |
357 | return (const uchar *)p + found; | |
358 | } | |
359 | ||
360 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ | |
361 | ||
362 | static const uchar * | |
363 | #ifndef __SSE2__ | |
364 | __attribute__((__target__("sse2"))) | |
365 | #endif | |
366 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
367 | { | |
368 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
369 | ||
370 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; | |
371 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; | |
372 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; | |
373 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; | |
374 | ||
375 | unsigned int misalign, found, mask; | |
376 | const v16qi *p; | |
377 | v16qi data, t; | |
378 | ||
379 | /* Align the source pointer. */ | |
380 | misalign = (uintptr_t)s & 15; | |
381 | p = (const v16qi *)((uintptr_t)s & -16); | |
382 | data = *p; | |
383 | ||
384 | /* Create a mask for the bytes that are valid within the first | |
385 | 16-byte block. The Idea here is that the AND with the mask | |
386 | within the loop is "free", since we need some AND or TEST | |
387 | insn in order to set the flags for the branch anyway. */ | |
388 | mask = -1u << misalign; | |
389 | ||
390 | /* Main loop processing 16 bytes at a time. */ | |
391 | goto start; | |
392 | do | |
393 | { | |
394 | data = *++p; | |
395 | mask = -1; | |
396 | ||
397 | start: | |
398 | t = __builtin_ia32_pcmpeqb128(data, repl_nl); | |
399 | t |= __builtin_ia32_pcmpeqb128(data, repl_cr); | |
400 | t |= __builtin_ia32_pcmpeqb128(data, repl_bs); | |
401 | t |= __builtin_ia32_pcmpeqb128(data, repl_qm); | |
402 | found = __builtin_ia32_pmovmskb128 (t); | |
403 | found &= mask; | |
404 | } | |
405 | while (!found); | |
406 | ||
407 | /* FOUND contains 1 in bits for which we matched a relevant | |
408 | character. Conversion to the byte index is trivial. */ | |
409 | found = __builtin_ctz(found); | |
410 | return (const uchar *)p + found; | |
411 | } | |
412 | ||
b315ae35 | 413 | #ifdef HAVE_SSE4 |
2431e8ba | 414 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
415 | ||
416 | static const uchar * | |
417 | #ifndef __SSE4_2__ | |
418 | __attribute__((__target__("sse4.2"))) | |
419 | #endif | |
420 | search_line_sse42 (const uchar *s, const uchar *end) | |
421 | { | |
422 | typedef char v16qi __attribute__ ((__vector_size__ (16))); | |
423 | static const v16qi search = { '\n', '\r', '?', '\\' }; | |
424 | ||
425 | uintptr_t si = (uintptr_t)s; | |
426 | uintptr_t index; | |
427 | ||
428 | /* Check for unaligned input. */ | |
429 | if (si & 15) | |
430 | { | |
431 | if (__builtin_expect (end - s < 16, 0) | |
432 | && __builtin_expect ((si & 0xfff) > 0xff0, 0)) | |
433 | { | |
434 | /* There are less than 16 bytes left in the buffer, and less | |
435 | than 16 bytes left on the page. Reading 16 bytes at this | |
436 | point might generate a spurious page fault. Defer to the | |
437 | SSE2 implementation, which already handles alignment. */ | |
438 | return search_line_sse2 (s, end); | |
439 | } | |
440 | ||
441 | /* ??? The builtin doesn't understand that the PCMPESTRI read from | |
442 | memory need not be aligned. */ | |
443 | __asm ("%vpcmpestri $0, (%1), %2" | |
444 | : "=c"(index) : "r"(s), "x"(search), "a"(4), "d"(16)); | |
445 | if (__builtin_expect (index < 16, 0)) | |
446 | goto found; | |
447 | ||
448 | /* Advance the pointer to an aligned address. We will re-scan a | |
449 | few bytes, but we no longer need care for reading past the | |
450 | end of a page, since we're guaranteed a match. */ | |
451 | s = (const uchar *)((si + 16) & -16); | |
452 | } | |
453 | ||
454 | /* Main loop, processing 16 bytes at a time. By doing the whole loop | |
455 | in inline assembly, we can make proper use of the flags set. */ | |
456 | __asm ( "sub $16, %1\n" | |
457 | " .balign 16\n" | |
458 | "0: add $16, %1\n" | |
459 | " %vpcmpestri $0, (%1), %2\n" | |
460 | " jnc 0b" | |
461 | : "=&c"(index), "+r"(s) | |
462 | : "x"(search), "a"(4), "d"(16)); | |
463 | ||
464 | found: | |
465 | return s + index; | |
466 | } | |
467 | ||
b315ae35 | 468 | #else |
469 | /* Work around out-dated assemblers without sse4 support. */ | |
470 | #define search_line_sse42 search_line_sse2 | |
471 | #endif | |
472 | ||
2431e8ba | 473 | /* Check the CPU capabilities. */ |
474 | ||
475 | #include "../gcc/config/i386/cpuid.h" | |
476 | ||
477 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); | |
478 | static search_line_fast_type search_line_fast; | |
479 | ||
480 | static void __attribute__((constructor)) | |
481 | init_vectorized_lexer (void) | |
482 | { | |
483 | unsigned dummy, ecx = 0, edx = 0; | |
484 | search_line_fast_type impl = search_line_acc_char; | |
485 | int minimum = 0; | |
486 | ||
487 | #if defined(__SSE4_2__) | |
488 | minimum = 3; | |
489 | #elif defined(__SSE2__) | |
490 | minimum = 2; | |
491 | #elif defined(__SSE__) || defined(__3dNOW_A__) | |
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 | { | |
508 | if (minimum == 1 || edx & bit_3DNOWP) | |
509 | impl = search_line_mmx; | |
510 | } | |
511 | ||
512 | search_line_fast = impl; | |
513 | } | |
514 | ||
70ebee13 | 515 | #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) |
2431e8ba | 516 | |
517 | /* A vection of the fast scanner using AltiVec vectorized byte compares. */ | |
518 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, | |
519 | so we can't compile this function without -maltivec on the command line | |
520 | (or implied by some other switch). */ | |
521 | ||
522 | static const uchar * | |
523 | search_line_fast (const uchar *s, const uchar *end ATTRIBUTE_UNUSED) | |
524 | { | |
525 | typedef __attribute__((altivec(vector))) unsigned char vc; | |
526 | ||
527 | const vc repl_nl = { | |
528 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', | |
529 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' | |
530 | }; | |
531 | const vc repl_cr = { | |
532 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', | |
533 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' | |
534 | }; | |
535 | const vc repl_bs = { | |
536 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', | |
537 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' | |
538 | }; | |
539 | const vc repl_qm = { | |
540 | '?', '?', '?', '?', '?', '?', '?', '?', | |
541 | '?', '?', '?', '?', '?', '?', '?', '?', | |
542 | }; | |
543 | const vc ones = { | |
544 | -1, -1, -1, -1, -1, -1, -1, -1, | |
545 | -1, -1, -1, -1, -1, -1, -1, -1, | |
546 | }; | |
547 | const vc zero = { 0 }; | |
548 | ||
549 | vc data, mask, t; | |
550 | ||
551 | /* Altivec loads automatically mask addresses with -16. This lets us | |
552 | issue the first load as early as possible. */ | |
553 | data = __builtin_vec_ld(0, (const vc *)s); | |
554 | ||
555 | /* Discard bytes before the beginning of the buffer. Do this by | |
556 | beginning with all ones and shifting in zeros according to the | |
557 | mis-alignment. The LVSR instruction pulls the exact shift we | |
558 | want from the address. */ | |
559 | mask = __builtin_vec_lvsr(0, s); | |
560 | mask = __builtin_vec_perm(zero, ones, mask); | |
561 | data &= mask; | |
562 | ||
563 | /* While altivec loads mask addresses, we still need to align S so | |
564 | that the offset we compute at the end is correct. */ | |
565 | s = (const uchar *)((uintptr_t)s & -16); | |
566 | ||
567 | /* Main loop processing 16 bytes at a time. */ | |
568 | goto start; | |
569 | do | |
570 | { | |
571 | vc m_nl, m_cr, m_bs, m_qm; | |
572 | ||
573 | s += 16; | |
574 | data = __builtin_vec_ld(0, (const vc *)s); | |
575 | ||
576 | start: | |
577 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); | |
578 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); | |
579 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); | |
580 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); | |
581 | t = (m_nl | m_cr) | (m_bs | m_qm); | |
582 | ||
583 | /* T now contains 0xff in bytes for which we matched one of the relevant | |
584 | characters. We want to exit the loop if any byte in T is non-zero. | |
585 | Below is the expansion of vec_any_ne(t, zero). */ | |
586 | } | |
587 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); | |
588 | ||
589 | { | |
590 | #define N (sizeof(vc) / sizeof(long)) | |
591 | ||
592 | typedef char check_count[(N == 2 || N == 4) * 2 - 1]; | |
593 | union { | |
594 | vc v; | |
595 | unsigned long l[N]; | |
596 | } u; | |
597 | unsigned long l, i = 0; | |
598 | ||
599 | u.v = t; | |
600 | ||
601 | /* Find the first word of T that is non-zero. */ | |
602 | switch (N) | |
603 | { | |
604 | case 4: | |
605 | l = u.l[i++]; | |
606 | if (l != 0) | |
607 | break; | |
608 | s += sizeof(unsigned long); | |
609 | l = u.l[i++]; | |
610 | if (l != 0) | |
611 | break; | |
612 | s += sizeof(unsigned long); | |
613 | case 2: | |
614 | l = u.l[i++]; | |
615 | if (l != 0) | |
616 | break; | |
617 | s += sizeof(unsigned long); | |
618 | l = u.l[i]; | |
619 | } | |
620 | ||
621 | /* L now contains 0xff in bytes for which we matched one of the | |
622 | relevant characters. We can find the byte index by finding | |
623 | its bit index and dividing by 8. */ | |
624 | l = __builtin_clzl(l) >> 3; | |
625 | return s + l; | |
626 | ||
627 | #undef N | |
628 | } | |
629 | } | |
630 | ||
631 | #else | |
632 | ||
633 | /* We only have one accellerated alternative. Use a direct call so that | |
634 | we encourage inlining. */ | |
635 | ||
636 | #define search_line_fast search_line_acc_char | |
637 | ||
638 | #endif | |
639 | ||
a54e0bf8 | 640 | /* Returns with a logical line that contains no escaped newlines or |
641 | trigraphs. This is a time-critical inner loop. */ | |
642 | void | |
f7fdd7a1 | 643 | _cpp_clean_line (cpp_reader *pfile) |
0578f103 | 644 | { |
a54e0bf8 | 645 | cpp_buffer *buffer; |
646 | const uchar *s; | |
647 | uchar c, *d, *p; | |
1e0ef2fd | 648 | |
a54e0bf8 | 649 | buffer = pfile->buffer; |
650 | buffer->cur_note = buffer->notes_used = 0; | |
651 | buffer->cur = buffer->line_base = buffer->next_line; | |
652 | buffer->need_line = false; | |
2431e8ba | 653 | s = buffer->next_line; |
1e0ef2fd | 654 | |
a54e0bf8 | 655 | if (!buffer->from_stage3) |
0578f103 | 656 | { |
5008f5c5 | 657 | const uchar *pbackslash = NULL; |
658 | ||
2431e8ba | 659 | /* Fast path. This is the common case of an un-escaped line with |
54d3be91 | 660 | no trigraphs. The primary win here is by not writing any |
661 | data back to memory until we have to. */ | |
2431e8ba | 662 | while (1) |
54d3be91 | 663 | { |
2431e8ba | 664 | /* Perform an optimized search for \n, \r, \\, ?. */ |
665 | s = search_line_fast (s, buffer->rlimit); | |
54d3be91 | 666 | |
2431e8ba | 667 | c = *s; |
668 | if (c == '\\') | |
669 | { | |
670 | /* Record the location of the backslash and continue. */ | |
671 | pbackslash = s++; | |
54d3be91 | 672 | } |
2431e8ba | 673 | else if (__builtin_expect (c == '?', 0)) |
54d3be91 | 674 | { |
2431e8ba | 675 | if (__builtin_expect (s[1] == '?', false) |
676 | && _cpp_trigraph_map[s[2]]) | |
54d3be91 | 677 | { |
2431e8ba | 678 | /* Have a trigraph. We may or may not have to convert |
679 | it. Add a line note regardless, for -Wtrigraphs. */ | |
680 | add_line_note (buffer, s, s[2]); | |
681 | if (CPP_OPTION (pfile, trigraphs)) | |
682 | { | |
683 | /* We do, and that means we have to switch to the | |
684 | slow path. */ | |
685 | d = (uchar *) s; | |
686 | *d = _cpp_trigraph_map[s[2]]; | |
687 | s += 2; | |
688 | goto slow_path; | |
689 | } | |
54d3be91 | 690 | } |
2431e8ba | 691 | /* Not a trigraph. Continue on fast-path. */ |
692 | s++; | |
54d3be91 | 693 | } |
2431e8ba | 694 | else |
695 | break; | |
54d3be91 | 696 | } |
697 | ||
2431e8ba | 698 | /* This must be \r or \n. We're either done, or we'll be forced |
699 | to write back to the buffer and continue on the slow path. */ | |
700 | d = (uchar *) s; | |
701 | ||
702 | if (__builtin_expect (s == buffer->rlimit, false)) | |
703 | goto done; | |
704 | ||
705 | /* DOS line ending? */ | |
706 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') | |
707 | { | |
708 | s++; | |
709 | if (s == buffer->rlimit) | |
710 | goto done; | |
711 | } | |
712 | ||
713 | if (__builtin_expect (pbackslash == NULL, true)) | |
714 | goto done; | |
715 | ||
716 | /* Check for escaped newline. */ | |
717 | p = d; | |
718 | while (is_nvspace (p[-1])) | |
719 | p--; | |
720 | if (p - 1 != pbackslash) | |
721 | goto done; | |
722 | ||
723 | /* Have an escaped newline; process it and proceed to | |
724 | the slow path. */ | |
725 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); | |
726 | d = p - 2; | |
727 | buffer->next_line = p - 1; | |
a54e0bf8 | 728 | |
2431e8ba | 729 | slow_path: |
730 | while (1) | |
4b912310 | 731 | { |
a54e0bf8 | 732 | c = *++s; |
733 | *++d = c; | |
734 | ||
735 | if (c == '\n' || c == '\r') | |
736 | { | |
2431e8ba | 737 | /* Handle DOS line endings. */ |
a54e0bf8 | 738 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
739 | s++; | |
740 | if (s == buffer->rlimit) | |
741 | break; | |
742 | ||
743 | /* Escaped? */ | |
744 | p = d; | |
745 | while (p != buffer->next_line && is_nvspace (p[-1])) | |
746 | p--; | |
747 | if (p == buffer->next_line || p[-1] != '\\') | |
748 | break; | |
749 | ||
aad4a87f | 750 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
a54e0bf8 | 751 | d = p - 2; |
752 | buffer->next_line = p - 1; | |
753 | } | |
754 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) | |
755 | { | |
756 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ | |
aad4a87f | 757 | add_line_note (buffer, d, s[2]); |
a54e0bf8 | 758 | if (CPP_OPTION (pfile, trigraphs)) |
759 | { | |
760 | *d = _cpp_trigraph_map[s[2]]; | |
761 | s += 2; | |
762 | } | |
763 | } | |
4b912310 | 764 | } |
0578f103 | 765 | } |
a54e0bf8 | 766 | else |
767 | { | |
2431e8ba | 768 | while (*s != '\n' && *s != '\r') |
a54e0bf8 | 769 | s++; |
a54e0bf8 | 770 | d = (uchar *) s; |
771 | ||
772 | /* Handle DOS line endings. */ | |
773 | if (*s == '\r' && s != buffer->rlimit && s[1] == '\n') | |
774 | s++; | |
775 | } | |
338fa5f7 | 776 | |
54d3be91 | 777 | done: |
a54e0bf8 | 778 | *d = '\n'; |
aad4a87f | 779 | /* A sentinel note that should never be processed. */ |
780 | add_line_note (buffer, d + 1, '\n'); | |
a54e0bf8 | 781 | buffer->next_line = s + 1; |
0578f103 | 782 | } |
783 | ||
3078f2b2 | 784 | /* Return true if the trigraph indicated by NOTE should be warned |
785 | about in a comment. */ | |
786 | static bool | |
f7fdd7a1 | 787 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
3078f2b2 | 788 | { |
789 | const uchar *p; | |
790 | ||
791 | /* Within comments we don't warn about trigraphs, unless the | |
792 | trigraph forms an escaped newline, as that may change | |
7ef5b942 | 793 | behavior. */ |
3078f2b2 | 794 | if (note->type != '/') |
795 | return false; | |
796 | ||
797 | /* If -trigraphs, then this was an escaped newline iff the next note | |
798 | is coincident. */ | |
799 | if (CPP_OPTION (pfile, trigraphs)) | |
800 | return note[1].pos == note->pos; | |
801 | ||
802 | /* Otherwise, see if this forms an escaped newline. */ | |
803 | p = note->pos + 3; | |
804 | while (is_nvspace (*p)) | |
805 | p++; | |
806 | ||
807 | /* There might have been escaped newlines between the trigraph and the | |
808 | newline we found. Hence the position test. */ | |
809 | return (*p == '\n' && p < note[1].pos); | |
810 | } | |
811 | ||
a54e0bf8 | 812 | /* Process the notes created by add_line_note as far as the current |
813 | location. */ | |
814 | void | |
f7fdd7a1 | 815 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
0578f103 | 816 | { |
c808d026 | 817 | cpp_buffer *buffer = pfile->buffer; |
818 | ||
a54e0bf8 | 819 | for (;;) |
f80e83a9 | 820 | { |
a54e0bf8 | 821 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
822 | unsigned int col; | |
396ffa86 | 823 | |
a54e0bf8 | 824 | if (note->pos > buffer->cur) |
825 | break; | |
396ffa86 | 826 | |
a54e0bf8 | 827 | buffer->cur_note++; |
828 | col = CPP_BUF_COLUMN (buffer, note->pos + 1); | |
435fb09b | 829 | |
aad4a87f | 830 | if (note->type == '\\' || note->type == ' ') |
a54e0bf8 | 831 | { |
aad4a87f | 832 | if (note->type == ' ' && !in_comment) |
dbddc569 | 833 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
a54e0bf8 | 834 | "backslash and newline separated by space"); |
aad4a87f | 835 | |
a54e0bf8 | 836 | if (buffer->next_line > buffer->rlimit) |
1e0ef2fd | 837 | { |
dbddc569 | 838 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
a54e0bf8 | 839 | "backslash-newline at end of file"); |
840 | /* Prevent "no newline at end of file" warning. */ | |
841 | buffer->next_line = buffer->rlimit; | |
1e0ef2fd | 842 | } |
a54e0bf8 | 843 | |
844 | buffer->line_base = note->pos; | |
610625e3 | 845 | CPP_INCREMENT_LINE (pfile, 0); |
338fa5f7 | 846 | } |
aad4a87f | 847 | else if (_cpp_trigraph_map[note->type]) |
848 | { | |
3078f2b2 | 849 | if (CPP_OPTION (pfile, warn_trigraphs) |
850 | && (!in_comment || warn_in_comment (pfile, note))) | |
aad4a87f | 851 | { |
852 | if (CPP_OPTION (pfile, trigraphs)) | |
3a79f5da | 853 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
854 | pfile->line_table->highest_line, col, | |
855 | "trigraph ??%c converted to %c", | |
856 | note->type, | |
857 | (int) _cpp_trigraph_map[note->type]); | |
aad4a87f | 858 | else |
1542b1ef | 859 | { |
3a79f5da | 860 | cpp_warning_with_line |
861 | (pfile, CPP_W_TRIGRAPHS, | |
862 | pfile->line_table->highest_line, col, | |
1542b1ef | 863 | "trigraph ??%c ignored, use -trigraphs to enable", |
864 | note->type); | |
865 | } | |
aad4a87f | 866 | } |
867 | } | |
3a45011c | 868 | else if (note->type == 0) |
869 | /* Already processed in lex_raw_string. */; | |
aad4a87f | 870 | else |
871 | abort (); | |
f80e83a9 | 872 | } |
0578f103 | 873 | } |
874 | ||
338fa5f7 | 875 | /* Skip a C-style block comment. We find the end of the comment by |
876 | seeing if an asterisk is before every '/' we encounter. Returns | |
edaf8cb5 | 877 | nonzero if comment terminated by EOF, zero otherwise. |
878 | ||
879 | Buffer->cur points to the initial asterisk of the comment. */ | |
a54e0bf8 | 880 | bool |
f7fdd7a1 | 881 | _cpp_skip_block_comment (cpp_reader *pfile) |
0578f103 | 882 | { |
f80e83a9 | 883 | cpp_buffer *buffer = pfile->buffer; |
54d3be91 | 884 | const uchar *cur = buffer->cur; |
885 | uchar c; | |
338fa5f7 | 886 | |
54d3be91 | 887 | cur++; |
888 | if (*cur == '/') | |
889 | cur++; | |
338fa5f7 | 890 | |
a54e0bf8 | 891 | for (;;) |
892 | { | |
338fa5f7 | 893 | /* People like decorating comments with '*', so check for '/' |
894 | instead for efficiency. */ | |
54d3be91 | 895 | c = *cur++; |
896 | ||
f80e83a9 | 897 | if (c == '/') |
0578f103 | 898 | { |
54d3be91 | 899 | if (cur[-2] == '*') |
338fa5f7 | 900 | break; |
f80e83a9 | 901 | |
338fa5f7 | 902 | /* Warn about potential nested comments, but not if the '/' |
3fb1e43b | 903 | comes immediately before the true comment delimiter. |
f80e83a9 | 904 | Don't bother to get it right across escaped newlines. */ |
338fa5f7 | 905 | if (CPP_OPTION (pfile, warn_comments) |
54d3be91 | 906 | && cur[0] == '*' && cur[1] != '/') |
907 | { | |
908 | buffer->cur = cur; | |
3a79f5da | 909 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
910 | pfile->line_table->highest_line, | |
911 | CPP_BUF_COL (buffer), | |
912 | "\"/*\" within comment"); | |
54d3be91 | 913 | } |
0578f103 | 914 | } |
a54e0bf8 | 915 | else if (c == '\n') |
916 | { | |
610625e3 | 917 | unsigned int cols; |
54d3be91 | 918 | buffer->cur = cur - 1; |
a54e0bf8 | 919 | _cpp_process_line_notes (pfile, true); |
920 | if (buffer->next_line >= buffer->rlimit) | |
921 | return true; | |
922 | _cpp_clean_line (pfile); | |
610625e3 | 923 | |
924 | cols = buffer->next_line - buffer->line_base; | |
925 | CPP_INCREMENT_LINE (pfile, cols); | |
926 | ||
54d3be91 | 927 | cur = buffer->cur; |
a54e0bf8 | 928 | } |
0578f103 | 929 | } |
f80e83a9 | 930 | |
54d3be91 | 931 | buffer->cur = cur; |
3078f2b2 | 932 | _cpp_process_line_notes (pfile, true); |
a54e0bf8 | 933 | return false; |
0578f103 | 934 | } |
935 | ||
1c124f85 | 936 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
d10cfa8d | 937 | terminating newline. Handles escaped newlines. Returns nonzero |
1c124f85 | 938 | if a multiline comment. */ |
f80e83a9 | 939 | static int |
f7fdd7a1 | 940 | skip_line_comment (cpp_reader *pfile) |
0578f103 | 941 | { |
f669338a | 942 | cpp_buffer *buffer = pfile->buffer; |
4999c35b | 943 | source_location orig_line = pfile->line_table->highest_line; |
f80e83a9 | 944 | |
a54e0bf8 | 945 | while (*buffer->cur != '\n') |
946 | buffer->cur++; | |
1c124f85 | 947 | |
a54e0bf8 | 948 | _cpp_process_line_notes (pfile, true); |
dbddc569 | 949 | return orig_line != pfile->line_table->highest_line; |
f80e83a9 | 950 | } |
0578f103 | 951 | |
a54e0bf8 | 952 | /* Skips whitespace, saving the next non-whitespace character. */ |
b86584f6 | 953 | static void |
f7fdd7a1 | 954 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
f80e83a9 | 955 | { |
956 | cpp_buffer *buffer = pfile->buffer; | |
fe9eb18b | 957 | bool saw_NUL = false; |
0578f103 | 958 | |
338fa5f7 | 959 | do |
f80e83a9 | 960 | { |
78719282 | 961 | /* Horizontal space always OK. */ |
a54e0bf8 | 962 | if (c == ' ' || c == '\t') |
338fa5f7 | 963 | ; |
338fa5f7 | 964 | /* Just \f \v or \0 left. */ |
78719282 | 965 | else if (c == '\0') |
fe9eb18b | 966 | saw_NUL = true; |
79bd622b | 967 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)) |
dbddc569 | 968 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
73328dce | 969 | CPP_BUF_COL (buffer), |
970 | "%s in preprocessing directive", | |
971 | c == '\f' ? "form feed" : "vertical tab"); | |
338fa5f7 | 972 | |
338fa5f7 | 973 | c = *buffer->cur++; |
0578f103 | 974 | } |
2c0e001b | 975 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
338fa5f7 | 976 | while (is_nvspace (c)); |
977 | ||
fe9eb18b | 978 | if (saw_NUL) |
d80d2074 | 979 | cpp_error (pfile, CPP_DL_WARNING, "null character(s) ignored"); |
fe9eb18b | 980 | |
1c124f85 | 981 | buffer->cur--; |
f80e83a9 | 982 | } |
0578f103 | 983 | |
79bd622b | 984 | /* See if the characters of a number token are valid in a name (no |
985 | '.', '+' or '-'). */ | |
986 | static int | |
f7fdd7a1 | 987 | name_p (cpp_reader *pfile, const cpp_string *string) |
79bd622b | 988 | { |
989 | unsigned int i; | |
990 | ||
991 | for (i = 0; i < string->len; i++) | |
992 | if (!is_idchar (string->text[i])) | |
993 | return 0; | |
994 | ||
b1a9ff83 | 995 | return 1; |
79bd622b | 996 | } |
997 | ||
bce47149 | 998 | /* After parsing an identifier or other sequence, produce a warning about |
999 | sequences not in NFC/NFKC. */ | |
1000 | static void | |
1001 | warn_about_normalization (cpp_reader *pfile, | |
1002 | const cpp_token *token, | |
1003 | const struct normalize_state *s) | |
1004 | { | |
1005 | if (CPP_OPTION (pfile, warn_normalize) < NORMALIZE_STATE_RESULT (s) | |
1006 | && !pfile->state.skipping) | |
1007 | { | |
1008 | /* Make sure that the token is printed using UCNs, even | |
1009 | if we'd otherwise happily print UTF-8. */ | |
720aca92 | 1010 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token)); |
bce47149 | 1011 | size_t sz; |
1012 | ||
1013 | sz = cpp_spell_token (pfile, token, buf, false) - buf; | |
1014 | if (NORMALIZE_STATE_RESULT (s) == normalized_C) | |
3a79f5da | 1015 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1016 | "`%.*s' is not in NFKC", (int) sz, buf); | |
bce47149 | 1017 | else |
3a79f5da | 1018 | cpp_warning_with_line (pfile, CPP_W_NORMALIZE, token->src_loc, 0, |
1019 | "`%.*s' is not in NFC", (int) sz, buf); | |
bce47149 | 1020 | } |
1021 | } | |
1022 | ||
5bb46c08 | 1023 | /* Returns TRUE if the sequence starting at buffer->cur is invalid in |
2cbf1359 | 1024 | an identifier. FIRST is TRUE if this starts an identifier. */ |
5bb46c08 | 1025 | static bool |
bce47149 | 1026 | forms_identifier_p (cpp_reader *pfile, int first, |
1027 | struct normalize_state *state) | |
5bb46c08 | 1028 | { |
2cbf1359 | 1029 | cpp_buffer *buffer = pfile->buffer; |
1030 | ||
1031 | if (*buffer->cur == '$') | |
1032 | { | |
1033 | if (!CPP_OPTION (pfile, dollars_in_ident)) | |
1034 | return false; | |
1035 | ||
1036 | buffer->cur++; | |
f0c2775b | 1037 | if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping) |
2cbf1359 | 1038 | { |
f0c2775b | 1039 | CPP_OPTION (pfile, warn_dollars) = 0; |
d80d2074 | 1040 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
2cbf1359 | 1041 | } |
1042 | ||
1043 | return true; | |
1044 | } | |
5bb46c08 | 1045 | |
2cbf1359 | 1046 | /* Is this a syntactically valid UCN? */ |
865c4e44 | 1047 | if (CPP_OPTION (pfile, extended_identifiers) |
4e9d1e6d | 1048 | && *buffer->cur == '\\' |
2cbf1359 | 1049 | && (buffer->cur[1] == 'u' || buffer->cur[1] == 'U')) |
5bb46c08 | 1050 | { |
2cbf1359 | 1051 | buffer->cur += 2; |
bce47149 | 1052 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
1053 | state)) | |
2cbf1359 | 1054 | return true; |
1055 | buffer->cur -= 2; | |
5bb46c08 | 1056 | } |
5bb46c08 | 1057 | |
2cbf1359 | 1058 | return false; |
5bb46c08 | 1059 | } |
1060 | ||
038c21f1 | 1061 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
1062 | static cpp_hashnode * | |
1063 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) | |
1064 | { | |
1065 | cpp_hashnode *result; | |
1066 | const uchar *cur; | |
1067 | unsigned int len; | |
1068 | unsigned int hash = HT_HASHSTEP (0, *base); | |
1069 | ||
1070 | cur = base + 1; | |
1071 | while (ISIDNUM (*cur)) | |
1072 | { | |
1073 | hash = HT_HASHSTEP (hash, *cur); | |
1074 | cur++; | |
1075 | } | |
1076 | len = cur - base; | |
1077 | hash = HT_HASHFINISH (hash, len); | |
1078 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, | |
1079 | base, len, hash, HT_ALLOC)); | |
1080 | ||
1081 | /* Rarely, identifiers require diagnostics when lexed. */ | |
1082 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) | |
1083 | && !pfile->state.skipping, 0)) | |
1084 | { | |
1085 | /* It is allowed to poison the same identifier twice. */ | |
1086 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
1087 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", | |
1088 | NODE_NAME (result)); | |
1089 | ||
1090 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1091 | replacement list of a variadic macro. */ | |
1092 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1093 | && !pfile->state.va_args_ok) | |
1094 | cpp_error (pfile, CPP_DL_PEDWARN, | |
1095 | "__VA_ARGS__ can only appear in the expansion" | |
1096 | " of a C99 variadic macro"); | |
1097 | ||
1098 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1099 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1100 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1101 | "identifier \"%s\" is a special operator name in C++", | |
1102 | NODE_NAME (result)); | |
038c21f1 | 1103 | } |
1104 | ||
1105 | return result; | |
1106 | } | |
1107 | ||
1108 | /* Get the cpp_hashnode of an identifier specified by NAME in | |
1109 | the current cpp_reader object. If none is found, NULL is returned. */ | |
1110 | cpp_hashnode * | |
1111 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) | |
1112 | { | |
1113 | cpp_hashnode *result; | |
1114 | result = lex_identifier_intern (pfile, (uchar *) name); | |
1115 | return result; | |
1116 | } | |
1117 | ||
5bb46c08 | 1118 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
338fa5f7 | 1119 | static cpp_hashnode * |
bce47149 | 1120 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
1121 | struct normalize_state *nst) | |
0578f103 | 1122 | { |
79bd622b | 1123 | cpp_hashnode *result; |
bb1fa6bb | 1124 | const uchar *cur; |
3eb3f293 | 1125 | unsigned int len; |
1126 | unsigned int hash = HT_HASHSTEP (0, *base); | |
66a5287e | 1127 | |
3eb3f293 | 1128 | cur = pfile->buffer->cur; |
bb1fa6bb | 1129 | if (! starts_ucn) |
1130 | while (ISIDNUM (*cur)) | |
1131 | { | |
1132 | hash = HT_HASHSTEP (hash, *cur); | |
1133 | cur++; | |
1134 | } | |
1135 | pfile->buffer->cur = cur; | |
bce47149 | 1136 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
78a11351 | 1137 | { |
bb1fa6bb | 1138 | /* Slower version for identifiers containing UCNs (or $). */ |
1139 | do { | |
1140 | while (ISIDNUM (*pfile->buffer->cur)) | |
bce47149 | 1141 | { |
1142 | pfile->buffer->cur++; | |
1143 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1144 | } | |
1145 | } while (forms_identifier_p (pfile, false, nst)); | |
bb1fa6bb | 1146 | result = _cpp_interpret_identifier (pfile, base, |
1147 | pfile->buffer->cur - base); | |
66a5287e | 1148 | } |
bb1fa6bb | 1149 | else |
1150 | { | |
1151 | len = cur - base; | |
1152 | hash = HT_HASHFINISH (hash, len); | |
5bb46c08 | 1153 | |
e297899b | 1154 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table, |
1155 | base, len, hash, HT_ALLOC)); | |
bb1fa6bb | 1156 | } |
66a5287e | 1157 | |
5bb46c08 | 1158 | /* Rarely, identifiers require diagnostics when lexed. */ |
66a5287e | 1159 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC) |
1160 | && !pfile->state.skipping, 0)) | |
1161 | { | |
1162 | /* It is allowed to poison the same identifier twice. */ | |
1163 | if ((result->flags & NODE_POISONED) && !pfile->state.poisoned_ok) | |
d80d2074 | 1164 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
66a5287e | 1165 | NODE_NAME (result)); |
1166 | ||
1167 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the | |
1168 | replacement list of a variadic macro. */ | |
1169 | if (result == pfile->spec_nodes.n__VA_ARGS__ | |
1170 | && !pfile->state.va_args_ok) | |
d80d2074 | 1171 | cpp_error (pfile, CPP_DL_PEDWARN, |
f7fdd7a1 | 1172 | "__VA_ARGS__ can only appear in the expansion" |
1173 | " of a C99 variadic macro"); | |
2a6a6991 | 1174 | |
1175 | /* For -Wc++-compat, warn about use of C++ named operators. */ | |
1176 | if (result->flags & NODE_WARN_OPERATOR) | |
3a79f5da | 1177 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
1178 | "identifier \"%s\" is a special operator name in C++", | |
1179 | NODE_NAME (result)); | |
66a5287e | 1180 | } |
1181 | ||
1182 | return result; | |
1183 | } | |
1184 | ||
5bb46c08 | 1185 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
0578f103 | 1186 | static void |
bce47149 | 1187 | lex_number (cpp_reader *pfile, cpp_string *number, |
1188 | struct normalize_state *nst) | |
0578f103 | 1189 | { |
b6d18b0a | 1190 | const uchar *cur; |
5bb46c08 | 1191 | const uchar *base; |
1192 | uchar *dest; | |
0578f103 | 1193 | |
5bb46c08 | 1194 | base = pfile->buffer->cur - 1; |
1195 | do | |
f80e83a9 | 1196 | { |
5bb46c08 | 1197 | cur = pfile->buffer->cur; |
338fa5f7 | 1198 | |
5bb46c08 | 1199 | /* N.B. ISIDNUM does not include $. */ |
1200 | while (ISIDNUM (*cur) || *cur == '.' || VALID_SIGN (*cur, cur[-1])) | |
bce47149 | 1201 | { |
1202 | cur++; | |
1203 | NORMALIZE_STATE_UPDATE_IDNUM (nst); | |
1204 | } | |
0578f103 | 1205 | |
78a11351 | 1206 | pfile->buffer->cur = cur; |
0578f103 | 1207 | } |
bce47149 | 1208 | while (forms_identifier_p (pfile, false, nst)); |
79bd622b | 1209 | |
5bb46c08 | 1210 | number->len = cur - base; |
1211 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); | |
1212 | memcpy (dest, base, number->len); | |
1213 | dest[number->len] = '\0'; | |
1214 | number->text = dest; | |
79bd622b | 1215 | } |
1216 | ||
4970d4c2 | 1217 | /* Create a token of type TYPE with a literal spelling. */ |
1218 | static void | |
f7fdd7a1 | 1219 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
1220 | unsigned int len, enum cpp_ttype type) | |
4970d4c2 | 1221 | { |
1222 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); | |
1223 | ||
1224 | memcpy (dest, base, len); | |
1225 | dest[len] = '\0'; | |
1226 | token->type = type; | |
1227 | token->val.str.len = len; | |
1228 | token->val.str.text = dest; | |
1229 | } | |
1230 | ||
3a45011c | 1231 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
1232 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ | |
1233 | ||
1234 | static void | |
1235 | bufring_append (cpp_reader *pfile, const uchar *base, size_t len, | |
1236 | _cpp_buff **first_buff_p, _cpp_buff **last_buff_p) | |
1237 | { | |
1238 | _cpp_buff *first_buff = *first_buff_p; | |
1239 | _cpp_buff *last_buff = *last_buff_p; | |
1240 | ||
1241 | if (first_buff == NULL) | |
1242 | first_buff = last_buff = _cpp_get_buff (pfile, len); | |
1243 | else if (len > BUFF_ROOM (last_buff)) | |
1244 | { | |
1245 | size_t room = BUFF_ROOM (last_buff); | |
1246 | memcpy (BUFF_FRONT (last_buff), base, room); | |
1247 | BUFF_FRONT (last_buff) += room; | |
1248 | base += room; | |
1249 | len -= room; | |
1250 | last_buff = _cpp_append_extend_buff (pfile, last_buff, len); | |
1251 | } | |
1252 | ||
1253 | memcpy (BUFF_FRONT (last_buff), base, len); | |
1254 | BUFF_FRONT (last_buff) += len; | |
1255 | ||
1256 | *first_buff_p = first_buff; | |
1257 | *last_buff_p = last_buff; | |
1258 | } | |
1259 | ||
538ba11a | 1260 | /* Lexes a raw string. The stored string contains the spelling, including |
3a45011c | 1261 | double quotes, delimiter string, '(' and ')', any leading |
538ba11a | 1262 | 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the |
1263 | literal, or CPP_OTHER if it was not properly terminated. | |
1264 | ||
1265 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1266 | is the first NUL since embedded NULs are preserved. */ | |
1267 | ||
1268 | static void | |
1269 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base, | |
1270 | const uchar *cur) | |
1271 | { | |
1272 | source_location saw_NUL = 0; | |
1273 | const uchar *raw_prefix; | |
1274 | unsigned int raw_prefix_len = 0; | |
1275 | enum cpp_ttype type; | |
1276 | size_t total_len = 0; | |
1277 | _cpp_buff *first_buff = NULL, *last_buff = NULL; | |
3a45011c | 1278 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1279 | |
1280 | type = (*base == 'L' ? CPP_WSTRING : | |
1281 | *base == 'U' ? CPP_STRING32 : | |
1282 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) | |
1283 | : CPP_STRING); | |
1284 | ||
1285 | raw_prefix = cur + 1; | |
1286 | while (raw_prefix_len < 16) | |
1287 | { | |
1288 | switch (raw_prefix[raw_prefix_len]) | |
1289 | { | |
7fe0ef3a | 1290 | case ' ': case '(': case ')': case '\\': case '\t': |
538ba11a | 1291 | case '\v': case '\f': case '\n': default: |
1292 | break; | |
1293 | /* Basic source charset except the above chars. */ | |
1294 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1295 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1296 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1297 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1298 | case 'y': case 'z': | |
1299 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1300 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1301 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1302 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1303 | case 'Y': case 'Z': | |
1304 | case '0': case '1': case '2': case '3': case '4': case '5': | |
1305 | case '6': case '7': case '8': case '9': | |
7fe0ef3a | 1306 | case '_': case '{': case '}': case '#': case '[': case ']': |
538ba11a | 1307 | case '<': case '>': case '%': case ':': case ';': case '.': |
1308 | case '?': case '*': case '+': case '-': case '/': case '^': | |
1309 | case '&': case '|': case '~': case '!': case '=': case ',': | |
7fe0ef3a | 1310 | case '"': case '\'': |
538ba11a | 1311 | raw_prefix_len++; |
1312 | continue; | |
1313 | } | |
1314 | break; | |
1315 | } | |
1316 | ||
7fe0ef3a | 1317 | if (raw_prefix[raw_prefix_len] != '(') |
538ba11a | 1318 | { |
1319 | int col = CPP_BUF_COLUMN (pfile->buffer, raw_prefix + raw_prefix_len) | |
1320 | + 1; | |
1321 | if (raw_prefix_len == 16) | |
1322 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1323 | "raw string delimiter longer than 16 characters"); | |
1324 | else | |
1325 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, col, | |
1326 | "invalid character '%c' in raw string delimiter", | |
1327 | (int) raw_prefix[raw_prefix_len]); | |
1328 | pfile->buffer->cur = raw_prefix - 1; | |
1329 | create_literal (pfile, token, base, raw_prefix - 1 - base, CPP_OTHER); | |
1330 | return; | |
1331 | } | |
1332 | ||
1333 | cur = raw_prefix + raw_prefix_len + 1; | |
1334 | for (;;) | |
1335 | { | |
3a45011c | 1336 | #define BUF_APPEND(STR,LEN) \ |
1337 | do { \ | |
1338 | bufring_append (pfile, (const uchar *)(STR), (LEN), \ | |
1339 | &first_buff, &last_buff); \ | |
1340 | total_len += (LEN); \ | |
1341 | } while (0); | |
1342 | ||
1343 | cppchar_t c; | |
1344 | ||
1345 | /* If we previously performed any trigraph or line splicing | |
1346 | transformations, undo them within the body of the raw string. */ | |
1347 | while (note->pos < cur) | |
1348 | ++note; | |
1349 | for (; note->pos == cur; ++note) | |
1350 | { | |
1351 | switch (note->type) | |
1352 | { | |
1353 | case '\\': | |
1354 | case ' ': | |
1355 | /* Restore backslash followed by newline. */ | |
1356 | BUF_APPEND (base, cur - base); | |
1357 | base = cur; | |
1358 | BUF_APPEND ("\\", 1); | |
1359 | after_backslash: | |
1360 | if (note->type == ' ') | |
1361 | { | |
1362 | /* GNU backslash whitespace newline extension. FIXME | |
1363 | could be any sequence of non-vertical space. When we | |
1364 | can properly restore any such sequence, we should mark | |
1365 | this note as handled so _cpp_process_line_notes | |
1366 | doesn't warn. */ | |
1367 | BUF_APPEND (" ", 1); | |
1368 | } | |
1369 | ||
1370 | BUF_APPEND ("\n", 1); | |
1371 | break; | |
1372 | ||
1373 | case 0: | |
1374 | /* Already handled. */ | |
1375 | break; | |
1376 | ||
1377 | default: | |
1378 | if (_cpp_trigraph_map[note->type]) | |
1379 | { | |
1380 | /* Don't warn about this trigraph in | |
1381 | _cpp_process_line_notes, since trigraphs show up as | |
1382 | trigraphs in raw strings. */ | |
56e2ce2d | 1383 | uchar type = note->type; |
3a45011c | 1384 | note->type = 0; |
1385 | ||
1386 | if (!CPP_OPTION (pfile, trigraphs)) | |
1387 | /* If we didn't convert the trigraph in the first | |
1388 | place, don't do anything now either. */ | |
1389 | break; | |
1390 | ||
1391 | BUF_APPEND (base, cur - base); | |
1392 | base = cur; | |
1393 | BUF_APPEND ("??", 2); | |
1394 | ||
1395 | /* ??/ followed by newline gets two line notes, one for | |
1396 | the trigraph and one for the backslash/newline. */ | |
1397 | if (type == '/' && note[1].pos == cur) | |
1398 | { | |
1399 | if (note[1].type != '\\' | |
1400 | && note[1].type != ' ') | |
1401 | abort (); | |
1402 | BUF_APPEND ("/", 1); | |
1403 | ++note; | |
1404 | goto after_backslash; | |
1405 | } | |
1406 | /* The ) from ??) could be part of the suffix. */ | |
1407 | else if (type == ')' | |
1408 | && strncmp ((const char *) cur+1, | |
1409 | (const char *) raw_prefix, | |
1410 | raw_prefix_len) == 0 | |
1411 | && cur[raw_prefix_len+1] == '"') | |
1412 | { | |
0ff0cd04 | 1413 | BUF_APPEND (")", 1); |
1414 | base++; | |
1415 | cur += raw_prefix_len + 2; | |
3a45011c | 1416 | goto break_outer_loop; |
1417 | } | |
1418 | else | |
1419 | { | |
1420 | /* Skip the replacement character. */ | |
1421 | base = ++cur; | |
1422 | BUF_APPEND (&type, 1); | |
1423 | } | |
1424 | } | |
1425 | else | |
1426 | abort (); | |
1427 | break; | |
1428 | } | |
1429 | } | |
1430 | c = *cur++; | |
538ba11a | 1431 | |
7fe0ef3a | 1432 | if (c == ')' |
538ba11a | 1433 | && strncmp ((const char *) cur, (const char *) raw_prefix, |
1434 | raw_prefix_len) == 0 | |
1435 | && cur[raw_prefix_len] == '"') | |
1436 | { | |
1437 | cur += raw_prefix_len + 1; | |
1438 | break; | |
1439 | } | |
1440 | else if (c == '\n') | |
1441 | { | |
1442 | if (pfile->state.in_directive | |
1443 | || pfile->state.parsing_args | |
1444 | || pfile->state.in_deferred_pragma) | |
1445 | { | |
1446 | cur--; | |
1447 | type = CPP_OTHER; | |
1448 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, | |
1449 | "unterminated raw string"); | |
1450 | break; | |
1451 | } | |
1452 | ||
3a45011c | 1453 | BUF_APPEND (base, cur - base); |
538ba11a | 1454 | |
1455 | if (pfile->buffer->cur < pfile->buffer->rlimit) | |
1456 | CPP_INCREMENT_LINE (pfile, 0); | |
1457 | pfile->buffer->need_line = true; | |
1458 | ||
3a45011c | 1459 | pfile->buffer->cur = cur-1; |
1460 | _cpp_process_line_notes (pfile, false); | |
538ba11a | 1461 | if (!_cpp_get_fresh_line (pfile)) |
1462 | { | |
1463 | source_location src_loc = token->src_loc; | |
1464 | token->type = CPP_EOF; | |
1465 | /* Tell the compiler the line number of the EOF token. */ | |
1466 | token->src_loc = pfile->line_table->highest_line; | |
1467 | token->flags = BOL; | |
1468 | if (first_buff != NULL) | |
1469 | _cpp_release_buff (pfile, first_buff); | |
1470 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, | |
1471 | "unterminated raw string"); | |
1472 | return; | |
1473 | } | |
1474 | ||
1475 | cur = base = pfile->buffer->cur; | |
3a45011c | 1476 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
538ba11a | 1477 | } |
1478 | else if (c == '\0' && !saw_NUL) | |
1479 | LINEMAP_POSITION_FOR_COLUMN (saw_NUL, pfile->line_table, | |
1480 | CPP_BUF_COLUMN (pfile->buffer, cur)); | |
1481 | } | |
3a45011c | 1482 | break_outer_loop: |
538ba11a | 1483 | |
1484 | if (saw_NUL && !pfile->state.skipping) | |
1485 | cpp_error_with_line (pfile, CPP_DL_WARNING, saw_NUL, 0, | |
1486 | "null character(s) preserved in literal"); | |
1487 | ||
1488 | pfile->buffer->cur = cur; | |
1489 | if (first_buff == NULL) | |
1490 | create_literal (pfile, token, base, cur - base, type); | |
1491 | else | |
1492 | { | |
1493 | uchar *dest = _cpp_unaligned_alloc (pfile, total_len + (cur - base) + 1); | |
1494 | ||
1495 | token->type = type; | |
1496 | token->val.str.len = total_len + (cur - base); | |
1497 | token->val.str.text = dest; | |
1498 | last_buff = first_buff; | |
1499 | while (last_buff != NULL) | |
1500 | { | |
1501 | memcpy (dest, last_buff->base, | |
1502 | BUFF_FRONT (last_buff) - last_buff->base); | |
1503 | dest += BUFF_FRONT (last_buff) - last_buff->base; | |
1504 | last_buff = last_buff->next; | |
1505 | } | |
1506 | _cpp_release_buff (pfile, first_buff); | |
1507 | memcpy (dest, base, cur - base); | |
1508 | dest[cur - base] = '\0'; | |
1509 | } | |
1510 | } | |
1511 | ||
5bb46c08 | 1512 | /* Lexes a string, character constant, or angle-bracketed header file |
4970d4c2 | 1513 | name. The stored string contains the spelling, including opening |
538ba11a | 1514 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
1515 | 'R' modifier. It returns the type of the literal, or CPP_OTHER | |
1516 | if it was not properly terminated, or CPP_LESS for an unterminated | |
1517 | header name which must be relexed as normal tokens. | |
4970d4c2 | 1518 | |
1519 | The spelling is NUL-terminated, but it is not guaranteed that this | |
1520 | is the first NUL since embedded NULs are preserved. */ | |
f80e83a9 | 1521 | static void |
f7fdd7a1 | 1522 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
0578f103 | 1523 | { |
4970d4c2 | 1524 | bool saw_NUL = false; |
1525 | const uchar *cur; | |
5bb46c08 | 1526 | cppchar_t terminator; |
4970d4c2 | 1527 | enum cpp_ttype type; |
1528 | ||
1529 | cur = base; | |
1530 | terminator = *cur++; | |
538ba11a | 1531 | if (terminator == 'L' || terminator == 'U') |
4970d4c2 | 1532 | terminator = *cur++; |
538ba11a | 1533 | else if (terminator == 'u') |
1534 | { | |
1535 | terminator = *cur++; | |
1536 | if (terminator == '8') | |
1537 | terminator = *cur++; | |
1538 | } | |
1539 | if (terminator == 'R') | |
1540 | { | |
1541 | lex_raw_string (pfile, token, base, cur); | |
1542 | return; | |
1543 | } | |
1544 | if (terminator == '"') | |
924bbf02 | 1545 | type = (*base == 'L' ? CPP_WSTRING : |
1546 | *base == 'U' ? CPP_STRING32 : | |
538ba11a | 1547 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
1548 | : CPP_STRING); | |
4970d4c2 | 1549 | else if (terminator == '\'') |
924bbf02 | 1550 | type = (*base == 'L' ? CPP_WCHAR : |
1551 | *base == 'U' ? CPP_CHAR32 : | |
1552 | *base == 'u' ? CPP_CHAR16 : CPP_CHAR); | |
4970d4c2 | 1553 | else |
1554 | terminator = '>', type = CPP_HEADER_NAME; | |
79bd622b | 1555 | |
338fa5f7 | 1556 | for (;;) |
0578f103 | 1557 | { |
4970d4c2 | 1558 | cppchar_t c = *cur++; |
4b0c16ee | 1559 | |
edaf8cb5 | 1560 | /* In #include-style directives, terminators are not escapable. */ |
4970d4c2 | 1561 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
1562 | cur++; | |
1563 | else if (c == terminator) | |
5bb46c08 | 1564 | break; |
4970d4c2 | 1565 | else if (c == '\n') |
338fa5f7 | 1566 | { |
4970d4c2 | 1567 | cur--; |
7811eab5 | 1568 | /* Unmatched quotes always yield undefined behavior, but |
1569 | greedy lexing means that what appears to be an unterminated | |
1570 | header name may actually be a legitimate sequence of tokens. */ | |
1571 | if (terminator == '>') | |
1572 | { | |
1573 | token->type = CPP_LESS; | |
1574 | return; | |
1575 | } | |
4970d4c2 | 1576 | type = CPP_OTHER; |
1577 | break; | |
0578f103 | 1578 | } |
4970d4c2 | 1579 | else if (c == '\0') |
1580 | saw_NUL = true; | |
0578f103 | 1581 | } |
1582 | ||
4970d4c2 | 1583 | if (saw_NUL && !pfile->state.skipping) |
d80d2074 | 1584 | cpp_error (pfile, CPP_DL_WARNING, |
1585 | "null character(s) preserved in literal"); | |
0578f103 | 1586 | |
0b67f687 | 1587 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang) != CLK_ASM) |
1588 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", | |
1589 | (int) terminator); | |
1590 | ||
4970d4c2 | 1591 | pfile->buffer->cur = cur; |
1592 | create_literal (pfile, token, base, cur - base, type); | |
338fa5f7 | 1593 | } |
f80e83a9 | 1594 | |
956c6108 | 1595 | /* Return the comment table. The client may not make any assumption |
1596 | about the ordering of the table. */ | |
1597 | cpp_comment_table * | |
1598 | cpp_get_comments (cpp_reader *pfile) | |
1599 | { | |
1600 | return &pfile->comments; | |
1601 | } | |
1602 | ||
1603 | /* Append a comment to the end of the comment table. */ | |
1604 | static void | |
1605 | store_comment (cpp_reader *pfile, cpp_token *token) | |
1606 | { | |
1607 | int len; | |
1608 | ||
1609 | if (pfile->comments.allocated == 0) | |
1610 | { | |
1611 | pfile->comments.allocated = 256; | |
1612 | pfile->comments.entries = (cpp_comment *) xmalloc | |
1613 | (pfile->comments.allocated * sizeof (cpp_comment)); | |
1614 | } | |
1615 | ||
1616 | if (pfile->comments.count == pfile->comments.allocated) | |
1617 | { | |
1618 | pfile->comments.allocated *= 2; | |
1619 | pfile->comments.entries = (cpp_comment *) xrealloc | |
1620 | (pfile->comments.entries, | |
1621 | pfile->comments.allocated * sizeof (cpp_comment)); | |
1622 | } | |
1623 | ||
1624 | len = token->val.str.len; | |
1625 | ||
1626 | /* Copy comment. Note, token may not be NULL terminated. */ | |
1627 | pfile->comments.entries[pfile->comments.count].comment = | |
1628 | (char *) xmalloc (sizeof (char) * (len + 1)); | |
1629 | memcpy (pfile->comments.entries[pfile->comments.count].comment, | |
1630 | token->val.str.text, len); | |
1631 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; | |
1632 | ||
1633 | /* Set source location. */ | |
1634 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; | |
1635 | ||
1636 | /* Increment the count of entries in the comment table. */ | |
1637 | pfile->comments.count++; | |
1638 | } | |
1639 | ||
79bd622b | 1640 | /* The stored comment includes the comment start and any terminator. */ |
2c63d6c8 | 1641 | static void |
f7fdd7a1 | 1642 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
1643 | cppchar_t type) | |
2c63d6c8 | 1644 | { |
f80e83a9 | 1645 | unsigned char *buffer; |
560ab0b2 | 1646 | unsigned int len, clen, i; |
b1a9ff83 | 1647 | |
f0495c2c | 1648 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
1c124f85 | 1649 | |
a543b315 | 1650 | /* C++ comments probably (not definitely) have moved past a new |
1651 | line, which we don't want to save in the comment. */ | |
1c124f85 | 1652 | if (is_vspace (pfile->buffer->cur[-1])) |
a543b315 | 1653 | len--; |
d3f7919d | 1654 | |
560ab0b2 | 1655 | /* If we are currently in a directive or in argument parsing, then |
1656 | we need to store all C++ comments as C comments internally, and | |
1657 | so we need to allocate a little extra space in that case. | |
d3f7919d | 1658 | |
1659 | Note that the only time we encounter a directive here is | |
1660 | when we are saving comments in a "#define". */ | |
560ab0b2 | 1661 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
1662 | && type == '/') ? len + 2 : len; | |
d3f7919d | 1663 | |
1664 | buffer = _cpp_unaligned_alloc (pfile, clen); | |
b1a9ff83 | 1665 | |
f80e83a9 | 1666 | token->type = CPP_COMMENT; |
d3f7919d | 1667 | token->val.str.len = clen; |
338fa5f7 | 1668 | token->val.str.text = buffer; |
0578f103 | 1669 | |
f0495c2c | 1670 | buffer[0] = '/'; |
1671 | memcpy (buffer + 1, from, len - 1); | |
d3f7919d | 1672 | |
a113df96 | 1673 | /* Finish conversion to a C comment, if necessary. */ |
560ab0b2 | 1674 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
d3f7919d | 1675 | { |
1676 | buffer[1] = '*'; | |
1677 | buffer[clen - 2] = '*'; | |
1678 | buffer[clen - 1] = '/'; | |
560ab0b2 | 1679 | /* As there can be in a C++ comments illegal sequences for C comments |
1680 | we need to filter them out. */ | |
1681 | for (i = 2; i < (clen - 2); i++) | |
1682 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) | |
1683 | buffer[i] = '|'; | |
d3f7919d | 1684 | } |
956c6108 | 1685 | |
1686 | /* Finally store this comment for use by clients of libcpp. */ | |
1687 | store_comment (pfile, token); | |
338fa5f7 | 1688 | } |
0578f103 | 1689 | |
83dcbb5c | 1690 | /* Allocate COUNT tokens for RUN. */ |
1691 | void | |
f7fdd7a1 | 1692 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
83dcbb5c | 1693 | { |
3b298764 | 1694 | run->base = XNEWVEC (cpp_token, count); |
83dcbb5c | 1695 | run->limit = run->base + count; |
1696 | run->next = NULL; | |
1697 | } | |
1698 | ||
1699 | /* Returns the next tokenrun, or creates one if there is none. */ | |
1700 | static tokenrun * | |
f7fdd7a1 | 1701 | next_tokenrun (tokenrun *run) |
83dcbb5c | 1702 | { |
1703 | if (run->next == NULL) | |
1704 | { | |
3b298764 | 1705 | run->next = XNEW (tokenrun); |
fb5ab82c | 1706 | run->next->prev = run; |
83dcbb5c | 1707 | _cpp_init_tokenrun (run->next, 250); |
1708 | } | |
1709 | ||
1710 | return run->next; | |
1711 | } | |
1712 | ||
89768577 | 1713 | /* Look ahead in the input stream. */ |
1714 | const cpp_token * | |
1715 | cpp_peek_token (cpp_reader *pfile, int index) | |
1716 | { | |
1717 | cpp_context *context = pfile->context; | |
1718 | const cpp_token *peektok; | |
1719 | int count; | |
1720 | ||
1721 | /* First, scan through any pending cpp_context objects. */ | |
1722 | while (context->prev) | |
1723 | { | |
1724 | ptrdiff_t sz = (context->direct_p | |
1725 | ? LAST (context).token - FIRST (context).token | |
1726 | : LAST (context).ptoken - FIRST (context).ptoken); | |
1727 | ||
1728 | if (index < (int) sz) | |
1729 | return (context->direct_p | |
1730 | ? FIRST (context).token + index | |
1731 | : *(FIRST (context).ptoken + index)); | |
1732 | ||
1733 | index -= (int) sz; | |
1734 | context = context->prev; | |
1735 | } | |
1736 | ||
1737 | /* We will have to read some new tokens after all (and do so | |
1738 | without invalidating preceding tokens). */ | |
1739 | count = index; | |
1740 | pfile->keep_tokens++; | |
1741 | ||
1742 | do | |
1743 | { | |
1744 | peektok = _cpp_lex_token (pfile); | |
1745 | if (peektok->type == CPP_EOF) | |
1746 | return peektok; | |
1747 | } | |
1748 | while (index--); | |
1749 | ||
1750 | _cpp_backup_tokens_direct (pfile, count + 1); | |
1751 | pfile->keep_tokens--; | |
1752 | ||
1753 | return peektok; | |
1754 | } | |
1755 | ||
f9b5f742 | 1756 | /* Allocate a single token that is invalidated at the same time as the |
1757 | rest of the tokens on the line. Has its line and col set to the | |
1758 | same as the last lexed token, so that diagnostics appear in the | |
1759 | right place. */ | |
1760 | cpp_token * | |
f7fdd7a1 | 1761 | _cpp_temp_token (cpp_reader *pfile) |
f9b5f742 | 1762 | { |
1763 | cpp_token *old, *result; | |
89768577 | 1764 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
1765 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; | |
f9b5f742 | 1766 | |
1767 | old = pfile->cur_token - 1; | |
89768577 | 1768 | /* Any pre-existing lookaheads must not be clobbered. */ |
1769 | if (la) | |
1770 | { | |
1771 | if (sz <= la) | |
1772 | { | |
1773 | tokenrun *next = next_tokenrun (pfile->cur_run); | |
1774 | ||
1775 | if (sz < la) | |
1776 | memmove (next->base + 1, next->base, | |
1777 | (la - sz) * sizeof (cpp_token)); | |
1778 | ||
1779 | next->base[0] = pfile->cur_run->limit[-1]; | |
1780 | } | |
1781 | ||
1782 | if (sz > 1) | |
1783 | memmove (pfile->cur_token + 1, pfile->cur_token, | |
1784 | MIN (la, sz - 1) * sizeof (cpp_token)); | |
1785 | } | |
1786 | ||
1787 | if (!sz && pfile->cur_token == pfile->cur_run->limit) | |
f9b5f742 | 1788 | { |
1789 | pfile->cur_run = next_tokenrun (pfile->cur_run); | |
1790 | pfile->cur_token = pfile->cur_run->base; | |
1791 | } | |
1792 | ||
1793 | result = pfile->cur_token++; | |
610625e3 | 1794 | result->src_loc = old->src_loc; |
f9b5f742 | 1795 | return result; |
1796 | } | |
1797 | ||
10b4496a | 1798 | /* Lex a token into RESULT (external interface). Takes care of issues |
1799 | like directive handling, token lookahead, multiple include | |
3fb1e43b | 1800 | optimization and skipping. */ |
c00e481c | 1801 | const cpp_token * |
f7fdd7a1 | 1802 | _cpp_lex_token (cpp_reader *pfile) |
83dcbb5c | 1803 | { |
fb5ab82c | 1804 | cpp_token *result; |
83dcbb5c | 1805 | |
fb5ab82c | 1806 | for (;;) |
83dcbb5c | 1807 | { |
fb5ab82c | 1808 | if (pfile->cur_token == pfile->cur_run->limit) |
83dcbb5c | 1809 | { |
fb5ab82c | 1810 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
1811 | pfile->cur_token = pfile->cur_run->base; | |
83dcbb5c | 1812 | } |
e0ff7935 | 1813 | /* We assume that the current token is somewhere in the current |
1814 | run. */ | |
1815 | if (pfile->cur_token < pfile->cur_run->base | |
1816 | || pfile->cur_token >= pfile->cur_run->limit) | |
1817 | abort (); | |
83dcbb5c | 1818 | |
fb5ab82c | 1819 | if (pfile->lookaheads) |
10b4496a | 1820 | { |
1821 | pfile->lookaheads--; | |
1822 | result = pfile->cur_token++; | |
1823 | } | |
fb5ab82c | 1824 | else |
10b4496a | 1825 | result = _cpp_lex_direct (pfile); |
fb5ab82c | 1826 | |
1827 | if (result->flags & BOL) | |
83dcbb5c | 1828 | { |
fb5ab82c | 1829 | /* Is this a directive. If _cpp_handle_directive returns |
1830 | false, it is an assembler #. */ | |
1831 | if (result->type == CPP_HASH | |
d6af0368 | 1832 | /* 6.10.3 p 11: Directives in a list of macro arguments |
1833 | gives undefined behavior. This implementation | |
1834 | handles the directive as normal. */ | |
b75b98aa | 1835 | && pfile->state.parsing_args != 1) |
d6d3c909 | 1836 | { |
b75b98aa | 1837 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE)) |
d6d3c909 | 1838 | { |
b75b98aa | 1839 | if (pfile->directive_result.type == CPP_PADDING) |
1840 | continue; | |
d6d3c909 | 1841 | result = &pfile->directive_result; |
d6d3c909 | 1842 | } |
1843 | } | |
b75b98aa | 1844 | else if (pfile->state.in_deferred_pragma) |
1845 | result = &pfile->directive_result; | |
d6d3c909 | 1846 | |
5621a364 | 1847 | if (pfile->cb.line_change && !pfile->state.skipping) |
f7fdd7a1 | 1848 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
83dcbb5c | 1849 | } |
83dcbb5c | 1850 | |
fb5ab82c | 1851 | /* We don't skip tokens in directives. */ |
b75b98aa | 1852 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
fb5ab82c | 1853 | break; |
83dcbb5c | 1854 | |
fb5ab82c | 1855 | /* Outside a directive, invalidate controlling macros. At file |
10b4496a | 1856 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
7ef5b942 | 1857 | get here and MI optimization works. */ |
83dcbb5c | 1858 | pfile->mi_valid = false; |
fb5ab82c | 1859 | |
1860 | if (!pfile->state.skipping || result->type == CPP_EOF) | |
1861 | break; | |
83dcbb5c | 1862 | } |
1863 | ||
c00e481c | 1864 | return result; |
83dcbb5c | 1865 | } |
1866 | ||
a54e0bf8 | 1867 | /* Returns true if a fresh line has been loaded. */ |
1868 | bool | |
f7fdd7a1 | 1869 | _cpp_get_fresh_line (cpp_reader *pfile) |
0bb65704 | 1870 | { |
6e04daf1 | 1871 | int return_at_eof; |
1872 | ||
a54e0bf8 | 1873 | /* We can't get a new line until we leave the current directive. */ |
1874 | if (pfile->state.in_directive) | |
1875 | return false; | |
b1a9ff83 | 1876 | |
a54e0bf8 | 1877 | for (;;) |
fb83e0d6 | 1878 | { |
a54e0bf8 | 1879 | cpp_buffer *buffer = pfile->buffer; |
fb83e0d6 | 1880 | |
a54e0bf8 | 1881 | if (!buffer->need_line) |
1882 | return true; | |
1883 | ||
1884 | if (buffer->next_line < buffer->rlimit) | |
0bb65704 | 1885 | { |
a54e0bf8 | 1886 | _cpp_clean_line (pfile); |
1887 | return true; | |
1888 | } | |
0bb65704 | 1889 | |
a54e0bf8 | 1890 | /* First, get out of parsing arguments state. */ |
1891 | if (pfile->state.parsing_args) | |
1892 | return false; | |
1893 | ||
1894 | /* End of buffer. Non-empty files should end in a newline. */ | |
1895 | if (buffer->buf != buffer->rlimit | |
1896 | && buffer->next_line > buffer->rlimit | |
1897 | && !buffer->from_stage3) | |
1898 | { | |
0448520c | 1899 | /* Clip to buffer size. */ |
a54e0bf8 | 1900 | buffer->next_line = buffer->rlimit; |
a54e0bf8 | 1901 | } |
6e04daf1 | 1902 | |
1903 | return_at_eof = buffer->return_at_eof; | |
a54e0bf8 | 1904 | _cpp_pop_buffer (pfile); |
6e04daf1 | 1905 | if (pfile->buffer == NULL || return_at_eof) |
11b5269c | 1906 | return false; |
a54e0bf8 | 1907 | } |
0bb65704 | 1908 | } |
1909 | ||
edaf8cb5 | 1910 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \ |
1911 | do \ | |
1912 | { \ | |
1913 | result->type = ELSE_TYPE; \ | |
1914 | if (*buffer->cur == CHAR) \ | |
1915 | buffer->cur++, result->type = THEN_TYPE; \ | |
1916 | } \ | |
1917 | while (0) | |
1c124f85 | 1918 | |
10b4496a | 1919 | /* Lex a token into pfile->cur_token, which is also incremented, to |
1920 | get diagnostics pointing to the correct location. | |
1921 | ||
1922 | Does not handle issues such as token lookahead, multiple-include | |
4172d65e | 1923 | optimization, directives, skipping etc. This function is only |
10b4496a | 1924 | suitable for use by _cpp_lex_token, and in special cases like |
1925 | lex_expansion_token which doesn't care for any of these issues. | |
1926 | ||
1927 | When meeting a newline, returns CPP_EOF if parsing a directive, | |
1928 | otherwise returns to the start of the token buffer if permissible. | |
1929 | Returns the location of the lexed token. */ | |
1930 | cpp_token * | |
f7fdd7a1 | 1931 | _cpp_lex_direct (cpp_reader *pfile) |
0578f103 | 1932 | { |
338fa5f7 | 1933 | cppchar_t c; |
230f0943 | 1934 | cpp_buffer *buffer; |
338fa5f7 | 1935 | const unsigned char *comment_start; |
10b4496a | 1936 | cpp_token *result = pfile->cur_token++; |
0653b94e | 1937 | |
83dcbb5c | 1938 | fresh_line: |
a54e0bf8 | 1939 | result->flags = 0; |
82166c5c | 1940 | buffer = pfile->buffer; |
11b5269c | 1941 | if (buffer->need_line) |
a54e0bf8 | 1942 | { |
b75b98aa | 1943 | if (pfile->state.in_deferred_pragma) |
1944 | { | |
1945 | result->type = CPP_PRAGMA_EOL; | |
1946 | pfile->state.in_deferred_pragma = false; | |
1947 | if (!pfile->state.pragma_allow_expansion) | |
1948 | pfile->state.prevent_expansion--; | |
1949 | return result; | |
1950 | } | |
a54e0bf8 | 1951 | if (!_cpp_get_fresh_line (pfile)) |
1952 | { | |
1953 | result->type = CPP_EOF; | |
2908f819 | 1954 | if (!pfile->state.in_directive) |
1955 | { | |
1956 | /* Tell the compiler the line number of the EOF token. */ | |
dbddc569 | 1957 | result->src_loc = pfile->line_table->highest_line; |
2908f819 | 1958 | result->flags = BOL; |
1959 | } | |
a54e0bf8 | 1960 | return result; |
1961 | } | |
1962 | if (!pfile->keep_tokens) | |
1963 | { | |
1964 | pfile->cur_run = &pfile->base_run; | |
1965 | result = pfile->base_run.base; | |
1966 | pfile->cur_token = result + 1; | |
1967 | } | |
1968 | result->flags = BOL; | |
1969 | if (pfile->state.parsing_args == 2) | |
1970 | result->flags |= PREV_WHITE; | |
1971 | } | |
11b5269c | 1972 | buffer = pfile->buffer; |
83dcbb5c | 1973 | update_tokens_line: |
dbddc569 | 1974 | result->src_loc = pfile->line_table->highest_line; |
f80e83a9 | 1975 | |
83dcbb5c | 1976 | skipped_white: |
a54e0bf8 | 1977 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
1978 | && !pfile->overlaid_buffer) | |
1979 | { | |
1980 | _cpp_process_line_notes (pfile, false); | |
dbddc569 | 1981 | result->src_loc = pfile->line_table->highest_line; |
a54e0bf8 | 1982 | } |
1c124f85 | 1983 | c = *buffer->cur++; |
610625e3 | 1984 | |
dbddc569 | 1985 | LINEMAP_POSITION_FOR_COLUMN (result->src_loc, pfile->line_table, |
1986 | CPP_BUF_COLUMN (buffer, buffer->cur)); | |
83dcbb5c | 1987 | |
338fa5f7 | 1988 | switch (c) |
0578f103 | 1989 | { |
435fb09b | 1990 | case ' ': case '\t': case '\f': case '\v': case '\0': |
1991 | result->flags |= PREV_WHITE; | |
a54e0bf8 | 1992 | skip_whitespace (pfile, c); |
1993 | goto skipped_white; | |
338fa5f7 | 1994 | |
a54e0bf8 | 1995 | case '\n': |
610625e3 | 1996 | if (buffer->cur < buffer->rlimit) |
1997 | CPP_INCREMENT_LINE (pfile, 0); | |
a54e0bf8 | 1998 | buffer->need_line = true; |
1999 | goto fresh_line; | |
732cb4c9 | 2000 | |
338fa5f7 | 2001 | case '0': case '1': case '2': case '3': case '4': |
2002 | case '5': case '6': case '7': case '8': case '9': | |
bce47149 | 2003 | { |
2004 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2005 | result->type = CPP_NUMBER; | |
2006 | lex_number (pfile, &result->val.str, &nst); | |
2007 | warn_about_normalization (pfile, result, &nst); | |
2008 | break; | |
2009 | } | |
732cb4c9 | 2010 | |
78c551ad | 2011 | case 'L': |
924bbf02 | 2012 | case 'u': |
2013 | case 'U': | |
538ba11a | 2014 | case 'R': |
2015 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, | |
2016 | wide strings or raw strings. */ | |
924bbf02 | 2017 | if (c == 'L' || CPP_OPTION (pfile, uliterals)) |
5bb46c08 | 2018 | { |
538ba11a | 2019 | if ((*buffer->cur == '\'' && c != 'R') |
2020 | || *buffer->cur == '"' | |
2021 | || (*buffer->cur == 'R' | |
2022 | && c != 'R' | |
2023 | && buffer->cur[1] == '"' | |
2024 | && CPP_OPTION (pfile, uliterals)) | |
2025 | || (*buffer->cur == '8' | |
2026 | && c == 'u' | |
2027 | && (buffer->cur[1] == '"' | |
2028 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"')))) | |
924bbf02 | 2029 | { |
2030 | lex_string (pfile, result, buffer->cur - 1); | |
2031 | break; | |
2032 | } | |
5bb46c08 | 2033 | } |
b1a9ff83 | 2034 | /* Fall through. */ |
78c551ad | 2035 | |
338fa5f7 | 2036 | case '_': |
2037 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
2038 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
2039 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
924bbf02 | 2040 | case 's': case 't': case 'v': case 'w': case 'x': |
338fa5f7 | 2041 | case 'y': case 'z': |
2042 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
78c551ad | 2043 | case 'G': case 'H': case 'I': case 'J': case 'K': |
538ba11a | 2044 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
924bbf02 | 2045 | case 'S': case 'T': case 'V': case 'W': case 'X': |
338fa5f7 | 2046 | case 'Y': case 'Z': |
2047 | result->type = CPP_NAME; | |
bce47149 | 2048 | { |
2049 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; | |
2ee04baa | 2050 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
2051 | &nst); | |
bce47149 | 2052 | warn_about_normalization (pfile, result, &nst); |
2053 | } | |
338fa5f7 | 2054 | |
338fa5f7 | 2055 | /* Convert named operators to their proper types. */ |
2ee04baa | 2056 | if (result->val.node.node->flags & NODE_OPERATOR) |
338fa5f7 | 2057 | { |
2058 | result->flags |= NAMED_OP; | |
2ee04baa | 2059 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
338fa5f7 | 2060 | } |
2061 | break; | |
2062 | ||
2063 | case '\'': | |
2064 | case '"': | |
4970d4c2 | 2065 | lex_string (pfile, result, buffer->cur - 1); |
338fa5f7 | 2066 | break; |
f80e83a9 | 2067 | |
338fa5f7 | 2068 | case '/': |
f0495c2c | 2069 | /* A potential block or line comment. */ |
2070 | comment_start = buffer->cur; | |
edaf8cb5 | 2071 | c = *buffer->cur; |
2072 | ||
f0495c2c | 2073 | if (c == '*') |
2074 | { | |
a54e0bf8 | 2075 | if (_cpp_skip_block_comment (pfile)) |
d80d2074 | 2076 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
338fa5f7 | 2077 | } |
1c124f85 | 2078 | else if (c == '/' && (CPP_OPTION (pfile, cplusplus_comments) |
610625e3 | 2079 | || cpp_in_system_header (pfile))) |
338fa5f7 | 2080 | { |
5db5d057 | 2081 | /* Warn about comments only if pedantically GNUC89, and not |
2082 | in system headers. */ | |
2083 | if (CPP_OPTION (pfile, lang) == CLK_GNUC89 && CPP_PEDANTIC (pfile) | |
66914e49 | 2084 | && ! buffer->warned_cplusplus_comments) |
f80e83a9 | 2085 | { |
d80d2074 | 2086 | cpp_error (pfile, CPP_DL_PEDWARN, |
ba059ac0 | 2087 | "C++ style comments are not allowed in ISO C90"); |
d80d2074 | 2088 | cpp_error (pfile, CPP_DL_PEDWARN, |
73328dce | 2089 | "(this will be reported only once per input file)"); |
f0495c2c | 2090 | buffer->warned_cplusplus_comments = 1; |
2091 | } | |
338fa5f7 | 2092 | |
e1caf668 | 2093 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)) |
3a79f5da | 2094 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
f0495c2c | 2095 | } |
1c124f85 | 2096 | else if (c == '=') |
2097 | { | |
edaf8cb5 | 2098 | buffer->cur++; |
1c124f85 | 2099 | result->type = CPP_DIV_EQ; |
2100 | break; | |
2101 | } | |
2102 | else | |
2103 | { | |
1c124f85 | 2104 | result->type = CPP_DIV; |
2105 | break; | |
2106 | } | |
338fa5f7 | 2107 | |
f0495c2c | 2108 | if (!pfile->state.save_comments) |
2109 | { | |
2110 | result->flags |= PREV_WHITE; | |
83dcbb5c | 2111 | goto update_tokens_line; |
338fa5f7 | 2112 | } |
f0495c2c | 2113 | |
2114 | /* Save the comment as a token in its own right. */ | |
d3f7919d | 2115 | save_comment (pfile, result, comment_start, c); |
fb5ab82c | 2116 | break; |
338fa5f7 | 2117 | |
2118 | case '<': | |
2119 | if (pfile->state.angled_headers) | |
2120 | { | |
4970d4c2 | 2121 | lex_string (pfile, result, buffer->cur - 1); |
7811eab5 | 2122 | if (result->type != CPP_LESS) |
2123 | break; | |
338fa5f7 | 2124 | } |
0578f103 | 2125 | |
edaf8cb5 | 2126 | result->type = CPP_LESS; |
2127 | if (*buffer->cur == '=') | |
2128 | buffer->cur++, result->type = CPP_LESS_EQ; | |
2129 | else if (*buffer->cur == '<') | |
338fa5f7 | 2130 | { |
edaf8cb5 | 2131 | buffer->cur++; |
2132 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT); | |
338fa5f7 | 2133 | } |
edaf8cb5 | 2134 | else if (CPP_OPTION (pfile, digraphs)) |
1c124f85 | 2135 | { |
edaf8cb5 | 2136 | if (*buffer->cur == ':') |
2137 | { | |
2138 | buffer->cur++; | |
2139 | result->flags |= DIGRAPH; | |
2140 | result->type = CPP_OPEN_SQUARE; | |
2141 | } | |
2142 | else if (*buffer->cur == '%') | |
2143 | { | |
2144 | buffer->cur++; | |
2145 | result->flags |= DIGRAPH; | |
2146 | result->type = CPP_OPEN_BRACE; | |
2147 | } | |
1c124f85 | 2148 | } |
338fa5f7 | 2149 | break; |
2150 | ||
2151 | case '>': | |
edaf8cb5 | 2152 | result->type = CPP_GREATER; |
2153 | if (*buffer->cur == '=') | |
2154 | buffer->cur++, result->type = CPP_GREATER_EQ; | |
2155 | else if (*buffer->cur == '>') | |
338fa5f7 | 2156 | { |
edaf8cb5 | 2157 | buffer->cur++; |
2158 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT); | |
2159 | } | |
338fa5f7 | 2160 | break; |
2161 | ||
f669338a | 2162 | case '%': |
edaf8cb5 | 2163 | result->type = CPP_MOD; |
2164 | if (*buffer->cur == '=') | |
2165 | buffer->cur++, result->type = CPP_MOD_EQ; | |
2166 | else if (CPP_OPTION (pfile, digraphs)) | |
1c124f85 | 2167 | { |
edaf8cb5 | 2168 | if (*buffer->cur == ':') |
1c124f85 | 2169 | { |
edaf8cb5 | 2170 | buffer->cur++; |
2171 | result->flags |= DIGRAPH; | |
2172 | result->type = CPP_HASH; | |
2173 | if (*buffer->cur == '%' && buffer->cur[1] == ':') | |
2ee04baa | 2174 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
edaf8cb5 | 2175 | } |
2176 | else if (*buffer->cur == '>') | |
2177 | { | |
2178 | buffer->cur++; | |
2179 | result->flags |= DIGRAPH; | |
2180 | result->type = CPP_CLOSE_BRACE; | |
1c124f85 | 2181 | } |
1c124f85 | 2182 | } |
338fa5f7 | 2183 | break; |
2184 | ||
f669338a | 2185 | case '.': |
1c124f85 | 2186 | result->type = CPP_DOT; |
edaf8cb5 | 2187 | if (ISDIGIT (*buffer->cur)) |
1c124f85 | 2188 | { |
bce47149 | 2189 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
1c124f85 | 2190 | result->type = CPP_NUMBER; |
bce47149 | 2191 | lex_number (pfile, &result->val.str, &nst); |
2192 | warn_about_normalization (pfile, result, &nst); | |
1c124f85 | 2193 | } |
edaf8cb5 | 2194 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
2195 | buffer->cur += 2, result->type = CPP_ELLIPSIS; | |
2196 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) | |
2197 | buffer->cur++, result->type = CPP_DOT_STAR; | |
338fa5f7 | 2198 | break; |
0578f103 | 2199 | |
338fa5f7 | 2200 | case '+': |
edaf8cb5 | 2201 | result->type = CPP_PLUS; |
2202 | if (*buffer->cur == '+') | |
2203 | buffer->cur++, result->type = CPP_PLUS_PLUS; | |
2204 | else if (*buffer->cur == '=') | |
2205 | buffer->cur++, result->type = CPP_PLUS_EQ; | |
338fa5f7 | 2206 | break; |
ac0749c7 | 2207 | |
338fa5f7 | 2208 | case '-': |
edaf8cb5 | 2209 | result->type = CPP_MINUS; |
2210 | if (*buffer->cur == '>') | |
338fa5f7 | 2211 | { |
edaf8cb5 | 2212 | buffer->cur++; |
1c124f85 | 2213 | result->type = CPP_DEREF; |
edaf8cb5 | 2214 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)) |
2215 | buffer->cur++, result->type = CPP_DEREF_STAR; | |
1c124f85 | 2216 | } |
edaf8cb5 | 2217 | else if (*buffer->cur == '-') |
2218 | buffer->cur++, result->type = CPP_MINUS_MINUS; | |
2219 | else if (*buffer->cur == '=') | |
2220 | buffer->cur++, result->type = CPP_MINUS_EQ; | |
338fa5f7 | 2221 | break; |
0578f103 | 2222 | |
338fa5f7 | 2223 | case '&': |
edaf8cb5 | 2224 | result->type = CPP_AND; |
2225 | if (*buffer->cur == '&') | |
2226 | buffer->cur++, result->type = CPP_AND_AND; | |
2227 | else if (*buffer->cur == '=') | |
2228 | buffer->cur++, result->type = CPP_AND_EQ; | |
338fa5f7 | 2229 | break; |
b1a9ff83 | 2230 | |
338fa5f7 | 2231 | case '|': |
edaf8cb5 | 2232 | result->type = CPP_OR; |
2233 | if (*buffer->cur == '|') | |
2234 | buffer->cur++, result->type = CPP_OR_OR; | |
2235 | else if (*buffer->cur == '=') | |
2236 | buffer->cur++, result->type = CPP_OR_EQ; | |
338fa5f7 | 2237 | break; |
0578f103 | 2238 | |
338fa5f7 | 2239 | case ':': |
edaf8cb5 | 2240 | result->type = CPP_COLON; |
2241 | if (*buffer->cur == ':' && CPP_OPTION (pfile, cplusplus)) | |
2242 | buffer->cur++, result->type = CPP_SCOPE; | |
2243 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)) | |
338fa5f7 | 2244 | { |
edaf8cb5 | 2245 | buffer->cur++; |
338fa5f7 | 2246 | result->flags |= DIGRAPH; |
1c124f85 | 2247 | result->type = CPP_CLOSE_SQUARE; |
2248 | } | |
338fa5f7 | 2249 | break; |
0578f103 | 2250 | |
1c124f85 | 2251 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT); break; |
2252 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ); break; | |
2253 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT); break; | |
2254 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR); break; | |
2ee04baa | 2255 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH); result->val.token_no = 0; break; |
1c124f85 | 2256 | |
a54e0bf8 | 2257 | case '?': result->type = CPP_QUERY; break; |
338fa5f7 | 2258 | case '~': result->type = CPP_COMPL; break; |
2259 | case ',': result->type = CPP_COMMA; break; | |
2260 | case '(': result->type = CPP_OPEN_PAREN; break; | |
2261 | case ')': result->type = CPP_CLOSE_PAREN; break; | |
2262 | case '[': result->type = CPP_OPEN_SQUARE; break; | |
2263 | case ']': result->type = CPP_CLOSE_SQUARE; break; | |
2264 | case '{': result->type = CPP_OPEN_BRACE; break; | |
2265 | case '}': result->type = CPP_CLOSE_BRACE; break; | |
2266 | case ';': result->type = CPP_SEMICOLON; break; | |
2267 | ||
7fd957fe | 2268 | /* @ is a punctuator in Objective-C. */ |
9ee99ac6 | 2269 | case '@': result->type = CPP_ATSIGN; break; |
338fa5f7 | 2270 | |
78c551ad | 2271 | case '$': |
2cbf1359 | 2272 | case '\\': |
2273 | { | |
2274 | const uchar *base = --buffer->cur; | |
bce47149 | 2275 | struct normalize_state nst = INITIAL_NORMALIZE_STATE; |
78c551ad | 2276 | |
bce47149 | 2277 | if (forms_identifier_p (pfile, true, &nst)) |
2cbf1359 | 2278 | { |
2279 | result->type = CPP_NAME; | |
2ee04baa | 2280 | result->val.node.node = lex_identifier (pfile, base, true, &nst); |
bce47149 | 2281 | warn_about_normalization (pfile, result, &nst); |
2cbf1359 | 2282 | break; |
2283 | } | |
2284 | buffer->cur++; | |
bc205914 | 2285 | } |
2cbf1359 | 2286 | |
bc205914 | 2287 | default: |
4970d4c2 | 2288 | create_literal (pfile, result, buffer->cur - 1, 1, CPP_OTHER); |
2289 | break; | |
338fa5f7 | 2290 | } |
fb5ab82c | 2291 | |
2292 | return result; | |
338fa5f7 | 2293 | } |
2294 | ||
b1280514 | 2295 | /* An upper bound on the number of bytes needed to spell TOKEN. |
2296 | Does not include preceding whitespace. */ | |
79bd622b | 2297 | unsigned int |
f7fdd7a1 | 2298 | cpp_token_len (const cpp_token *token) |
338fa5f7 | 2299 | { |
79bd622b | 2300 | unsigned int len; |
cfad5579 | 2301 | |
79bd622b | 2302 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2303 | { |
cd740bd5 | 2304 | default: len = 6; break; |
4970d4c2 | 2305 | case SPELL_LITERAL: len = token->val.str.len; break; |
2ee04baa | 2306 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node) * 10; break; |
f80e83a9 | 2307 | } |
b1280514 | 2308 | |
2309 | return len; | |
cfad5579 | 2310 | } |
2311 | ||
bb1fa6bb | 2312 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
2313 | Return the number of bytes read out of NAME. (There are always | |
2314 | 10 bytes written to BUFFER.) */ | |
2315 | ||
2316 | static size_t | |
2317 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) | |
2318 | { | |
2319 | int j; | |
2320 | int ucn_len = 0; | |
2321 | int ucn_len_c; | |
2322 | unsigned t; | |
2323 | unsigned long utf32; | |
2324 | ||
2325 | /* Compute the length of the UTF-8 sequence. */ | |
2326 | for (t = *name; t & 0x80; t <<= 1) | |
2327 | ucn_len++; | |
2328 | ||
2329 | utf32 = *name & (0x7F >> ucn_len); | |
2330 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) | |
2331 | { | |
2332 | utf32 = (utf32 << 6) | (*++name & 0x3F); | |
2333 | ||
2334 | /* Ill-formed UTF-8. */ | |
2335 | if ((*name & ~0x3F) != 0x80) | |
2336 | abort (); | |
2337 | } | |
2338 | ||
2339 | *buffer++ = '\\'; | |
2340 | *buffer++ = 'U'; | |
2341 | for (j = 7; j >= 0; j--) | |
2342 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; | |
2343 | return ucn_len; | |
2344 | } | |
2345 | ||
ba99525e | 2346 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
2347 | the spelling of the digraph. */ | |
2348 | static const unsigned char * | |
2349 | cpp_digraph2name (enum cpp_ttype type) | |
2350 | { | |
2351 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; | |
2352 | } | |
bb1fa6bb | 2353 | |
f80e83a9 | 2354 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
c5ea33a8 | 2355 | already contain the enough space to hold the token's spelling. |
f7fdd7a1 | 2356 | Returns a pointer to the character after the last character written. |
bb1fa6bb | 2357 | FORSTRING is true if this is to be the spelling after translation |
2358 | phase 1 (this is different for UCNs). | |
f7fdd7a1 | 2359 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
79bd622b | 2360 | unsigned char * |
f7fdd7a1 | 2361 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
bb1fa6bb | 2362 | unsigned char *buffer, bool forstring) |
f80e83a9 | 2363 | { |
7e842f95 | 2364 | switch (TOKEN_SPELL (token)) |
f80e83a9 | 2365 | { |
2366 | case SPELL_OPERATOR: | |
2367 | { | |
2368 | const unsigned char *spelling; | |
2369 | unsigned char c; | |
ab12a39c | 2370 | |
f80e83a9 | 2371 | if (token->flags & DIGRAPH) |
ba99525e | 2372 | spelling = cpp_digraph2name (token->type); |
31674461 | 2373 | else if (token->flags & NAMED_OP) |
2374 | goto spell_ident; | |
f80e83a9 | 2375 | else |
7e842f95 | 2376 | spelling = TOKEN_NAME (token); |
b1a9ff83 | 2377 | |
f80e83a9 | 2378 | while ((c = *spelling++) != '\0') |
2379 | *buffer++ = c; | |
2380 | } | |
2381 | break; | |
ab12a39c | 2382 | |
8d27e472 | 2383 | spell_ident: |
f80e83a9 | 2384 | case SPELL_IDENT: |
bb1fa6bb | 2385 | if (forstring) |
2386 | { | |
2ee04baa | 2387 | memcpy (buffer, NODE_NAME (token->val.node.node), |
2388 | NODE_LEN (token->val.node.node)); | |
2389 | buffer += NODE_LEN (token->val.node.node); | |
bb1fa6bb | 2390 | } |
2391 | else | |
2392 | { | |
2393 | size_t i; | |
2ee04baa | 2394 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2395 | |
2ee04baa | 2396 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2397 | if (name[i] & ~0x7F) |
2398 | { | |
2399 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2400 | buffer += 10; | |
2401 | } | |
2402 | else | |
2ee04baa | 2403 | *buffer++ = NODE_NAME (token->val.node.node)[i]; |
bb1fa6bb | 2404 | } |
f80e83a9 | 2405 | break; |
ab12a39c | 2406 | |
4970d4c2 | 2407 | case SPELL_LITERAL: |
8d27e472 | 2408 | memcpy (buffer, token->val.str.text, token->val.str.len); |
2409 | buffer += token->val.str.len; | |
2410 | break; | |
2411 | ||
f80e83a9 | 2412 | case SPELL_NONE: |
d80d2074 | 2413 | cpp_error (pfile, CPP_DL_ICE, |
2414 | "unspellable token %s", TOKEN_NAME (token)); | |
f80e83a9 | 2415 | break; |
2416 | } | |
ab12a39c | 2417 | |
f80e83a9 | 2418 | return buffer; |
2419 | } | |
ab12a39c | 2420 | |
e484a1cc | 2421 | /* Returns TOKEN spelt as a null-terminated string. The string is |
2422 | freed when the reader is destroyed. Useful for diagnostics. */ | |
79bd622b | 2423 | unsigned char * |
f7fdd7a1 | 2424 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
b1280514 | 2425 | { |
2426 | unsigned int len = cpp_token_len (token) + 1; | |
1fdf6039 | 2427 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
6060326b | 2428 | |
bb1fa6bb | 2429 | end = cpp_spell_token (pfile, token, start, false); |
79bd622b | 2430 | end[0] = '\0'; |
6060326b | 2431 | |
79bd622b | 2432 | return start; |
2433 | } | |
6060326b | 2434 | |
ba99525e | 2435 | /* Returns a pointer to a string which spells the token defined by |
2436 | TYPE and FLAGS. Used by C front ends, which really should move to | |
2437 | using cpp_token_as_text. */ | |
79bd622b | 2438 | const char * |
ba99525e | 2439 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
79bd622b | 2440 | { |
ba99525e | 2441 | if (flags & DIGRAPH) |
2442 | return (const char *) cpp_digraph2name (type); | |
2443 | else if (flags & NAMED_OP) | |
2444 | return cpp_named_operator2name (type); | |
2445 | ||
79bd622b | 2446 | return (const char *) token_spellings[type].name; |
2447 | } | |
6060326b | 2448 | |
f9b5f742 | 2449 | /* Writes the spelling of token to FP, without any preceding space. |
2450 | Separated from cpp_spell_token for efficiency - to avoid stdio | |
2451 | double-buffering. */ | |
79bd622b | 2452 | void |
f7fdd7a1 | 2453 | cpp_output_token (const cpp_token *token, FILE *fp) |
79bd622b | 2454 | { |
79bd622b | 2455 | switch (TOKEN_SPELL (token)) |
6060326b | 2456 | { |
79bd622b | 2457 | case SPELL_OPERATOR: |
2458 | { | |
2459 | const unsigned char *spelling; | |
28874558 | 2460 | int c; |
6060326b | 2461 | |
79bd622b | 2462 | if (token->flags & DIGRAPH) |
ba99525e | 2463 | spelling = cpp_digraph2name (token->type); |
79bd622b | 2464 | else if (token->flags & NAMED_OP) |
2465 | goto spell_ident; | |
2466 | else | |
2467 | spelling = TOKEN_NAME (token); | |
f80e83a9 | 2468 | |
28874558 | 2469 | c = *spelling; |
2470 | do | |
2471 | putc (c, fp); | |
2472 | while ((c = *++spelling) != '\0'); | |
79bd622b | 2473 | } |
2474 | break; | |
f80e83a9 | 2475 | |
79bd622b | 2476 | spell_ident: |
2477 | case SPELL_IDENT: | |
bb1fa6bb | 2478 | { |
2479 | size_t i; | |
2ee04baa | 2480 | const unsigned char * name = NODE_NAME (token->val.node.node); |
bb1fa6bb | 2481 | |
2ee04baa | 2482 | for (i = 0; i < NODE_LEN (token->val.node.node); i++) |
bb1fa6bb | 2483 | if (name[i] & ~0x7F) |
2484 | { | |
2485 | unsigned char buffer[10]; | |
2486 | i += utf8_to_ucn (buffer, name + i) - 1; | |
2487 | fwrite (buffer, 1, 10, fp); | |
2488 | } | |
2489 | else | |
2ee04baa | 2490 | fputc (NODE_NAME (token->val.node.node)[i], fp); |
bb1fa6bb | 2491 | } |
2492 | break; | |
f80e83a9 | 2493 | |
4970d4c2 | 2494 | case SPELL_LITERAL: |
8d27e472 | 2495 | fwrite (token->val.str.text, 1, token->val.str.len, fp); |
2496 | break; | |
2497 | ||
79bd622b | 2498 | case SPELL_NONE: |
2499 | /* An error, most probably. */ | |
2500 | break; | |
f80e83a9 | 2501 | } |
6060326b | 2502 | } |
2503 | ||
79bd622b | 2504 | /* Compare two tokens. */ |
2505 | int | |
f7fdd7a1 | 2506 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
6060326b | 2507 | { |
79bd622b | 2508 | if (a->type == b->type && a->flags == b->flags) |
2509 | switch (TOKEN_SPELL (a)) | |
2510 | { | |
2511 | default: /* Keep compiler happy. */ | |
2512 | case SPELL_OPERATOR: | |
2ee04baa | 2513 | /* token_no is used to track where multiple consecutive ## |
941f2388 | 2514 | tokens were originally located. */ |
2ee04baa | 2515 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
79bd622b | 2516 | case SPELL_NONE: |
2ee04baa | 2517 | return (a->type != CPP_MACRO_ARG |
2518 | || a->val.macro_arg.arg_no == b->val.macro_arg.arg_no); | |
79bd622b | 2519 | case SPELL_IDENT: |
2ee04baa | 2520 | return a->val.node.node == b->val.node.node; |
4970d4c2 | 2521 | case SPELL_LITERAL: |
79bd622b | 2522 | return (a->val.str.len == b->val.str.len |
2523 | && !memcmp (a->val.str.text, b->val.str.text, | |
2524 | a->val.str.len)); | |
2525 | } | |
6060326b | 2526 | |
f80e83a9 | 2527 | return 0; |
2528 | } | |
2529 | ||
79bd622b | 2530 | /* Returns nonzero if a space should be inserted to avoid an |
2531 | accidental token paste for output. For simplicity, it is | |
2532 | conservative, and occasionally advises a space where one is not | |
2533 | needed, e.g. "." and ".2". */ | |
79bd622b | 2534 | int |
f7fdd7a1 | 2535 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
2536 | const cpp_token *token2) | |
6060326b | 2537 | { |
79bd622b | 2538 | enum cpp_ttype a = token1->type, b = token2->type; |
2539 | cppchar_t c; | |
6060326b | 2540 | |
79bd622b | 2541 | if (token1->flags & NAMED_OP) |
2542 | a = CPP_NAME; | |
2543 | if (token2->flags & NAMED_OP) | |
2544 | b = CPP_NAME; | |
6060326b | 2545 | |
79bd622b | 2546 | c = EOF; |
2547 | if (token2->flags & DIGRAPH) | |
ee6c4e4b | 2548 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
79bd622b | 2549 | else if (token_spellings[b].category == SPELL_OPERATOR) |
2550 | c = token_spellings[b].name[0]; | |
6060326b | 2551 | |
79bd622b | 2552 | /* Quickly get everything that can paste with an '='. */ |
ee6c4e4b | 2553 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
79bd622b | 2554 | return 1; |
6060326b | 2555 | |
79bd622b | 2556 | switch (a) |
6060326b | 2557 | { |
e58c07f7 | 2558 | case CPP_GREATER: return c == '>'; |
2559 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; | |
79bd622b | 2560 | case CPP_PLUS: return c == '+'; |
2561 | case CPP_MINUS: return c == '-' || c == '>'; | |
2562 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ | |
2563 | case CPP_MOD: return c == ':' || c == '>'; | |
2564 | case CPP_AND: return c == '&'; | |
2565 | case CPP_OR: return c == '|'; | |
2566 | case CPP_COLON: return c == ':' || c == '>'; | |
2567 | case CPP_DEREF: return c == '*'; | |
efdcc728 | 2568 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
79bd622b | 2569 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
2570 | case CPP_NAME: return ((b == CPP_NUMBER | |
2571 | && name_p (pfile, &token2->val.str)) | |
2572 | || b == CPP_NAME | |
2573 | || b == CPP_CHAR || b == CPP_STRING); /* L */ | |
2574 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME | |
2575 | || c == '.' || c == '+' || c == '-'); | |
2cbf1359 | 2576 | /* UCNs */ |
bc205914 | 2577 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
2578 | && b == CPP_NAME) | |
2cbf1359 | 2579 | || (CPP_OPTION (pfile, objc) |
bc205914 | 2580 | && token1->val.str.text[0] == '@' |
2cbf1359 | 2581 | && (b == CPP_NAME || b == CPP_STRING))); |
79bd622b | 2582 | default: break; |
6060326b | 2583 | } |
6060326b | 2584 | |
deb356cf | 2585 | return 0; |
6060326b | 2586 | } |
2587 | ||
79bd622b | 2588 | /* Output all the remaining tokens on the current line, and a newline |
f9b5f742 | 2589 | character, to FP. Leading whitespace is removed. If there are |
2590 | macros, special token padding is not performed. */ | |
6060326b | 2591 | void |
f7fdd7a1 | 2592 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
6060326b | 2593 | { |
f9b5f742 | 2594 | const cpp_token *token; |
7e842f95 | 2595 | |
f9b5f742 | 2596 | token = cpp_get_token (pfile); |
2597 | while (token->type != CPP_EOF) | |
7e842f95 | 2598 | { |
f9b5f742 | 2599 | cpp_output_token (token, fp); |
2600 | token = cpp_get_token (pfile); | |
2601 | if (token->flags & PREV_WHITE) | |
2602 | putc (' ', fp); | |
7e842f95 | 2603 | } |
2604 | ||
79bd622b | 2605 | putc ('\n', fp); |
f80e83a9 | 2606 | } |
6060326b | 2607 | |
c0770282 | 2608 | /* Return a string representation of all the remaining tokens on the |
2609 | current line. The result is allocated using xmalloc and must be | |
2610 | freed by the caller. */ | |
2611 | unsigned char * | |
2612 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) | |
2613 | { | |
2614 | const cpp_token *token; | |
2615 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; | |
2616 | unsigned int alloced = 120 + out; | |
2617 | unsigned char *result = (unsigned char *) xmalloc (alloced); | |
2618 | ||
2619 | /* If DIR_NAME is empty, there are no initial contents. */ | |
2620 | if (dir_name) | |
2621 | { | |
2622 | sprintf ((char *) result, "#%s ", dir_name); | |
2623 | out += 2; | |
2624 | } | |
2625 | ||
2626 | token = cpp_get_token (pfile); | |
2627 | while (token->type != CPP_EOF) | |
2628 | { | |
2629 | unsigned char *last; | |
2630 | /* Include room for a possible space and the terminating nul. */ | |
2631 | unsigned int len = cpp_token_len (token) + 2; | |
2632 | ||
2633 | if (out + len > alloced) | |
2634 | { | |
2635 | alloced *= 2; | |
2636 | if (out + len > alloced) | |
2637 | alloced = out + len; | |
2638 | result = (unsigned char *) xrealloc (result, alloced); | |
2639 | } | |
2640 | ||
2641 | last = cpp_spell_token (pfile, token, &result[out], 0); | |
2642 | out = last - result; | |
2643 | ||
2644 | token = cpp_get_token (pfile); | |
2645 | if (token->flags & PREV_WHITE) | |
2646 | result[out++] = ' '; | |
2647 | } | |
2648 | ||
2649 | result[out] = '\0'; | |
2650 | return result; | |
2651 | } | |
2652 | ||
084163dc | 2653 | /* Memory buffers. Changing these three constants can have a dramatic |
2654 | effect on performance. The values here are reasonable defaults, | |
2655 | but might be tuned. If you adjust them, be sure to test across a | |
2656 | range of uses of cpplib, including heavy nested function-like macro | |
2657 | expansion. Also check the change in peak memory usage (NJAMD is a | |
2658 | good tool for this). */ | |
2659 | #define MIN_BUFF_SIZE 8000 | |
1e0ef2fd | 2660 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2) |
084163dc | 2661 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \ |
2662 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) | |
deb356cf | 2663 | |
1e0ef2fd | 2664 | #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0) |
2665 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE! | |
2666 | #endif | |
2667 | ||
1785b647 | 2668 | /* Create a new allocation buffer. Place the control block at the end |
2669 | of the buffer, so that buffer overflows will cause immediate chaos. */ | |
06c92cbc | 2670 | static _cpp_buff * |
f7fdd7a1 | 2671 | new_buff (size_t len) |
06c92cbc | 2672 | { |
2673 | _cpp_buff *result; | |
1fdf6039 | 2674 | unsigned char *base; |
06c92cbc | 2675 | |
084163dc | 2676 | if (len < MIN_BUFF_SIZE) |
2677 | len = MIN_BUFF_SIZE; | |
198b48a0 | 2678 | len = CPP_ALIGN (len); |
06c92cbc | 2679 | |
720aca92 | 2680 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff)); |
06c92cbc | 2681 | result = (_cpp_buff *) (base + len); |
2682 | result->base = base; | |
2683 | result->cur = base; | |
2684 | result->limit = base + len; | |
2685 | result->next = NULL; | |
2686 | return result; | |
2687 | } | |
2688 | ||
2689 | /* Place a chain of unwanted allocation buffers on the free list. */ | |
2690 | void | |
f7fdd7a1 | 2691 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
06c92cbc | 2692 | { |
2693 | _cpp_buff *end = buff; | |
2694 | ||
2695 | while (end->next) | |
2696 | end = end->next; | |
2697 | end->next = pfile->free_buffs; | |
2698 | pfile->free_buffs = buff; | |
2699 | } | |
2700 | ||
2701 | /* Return a free buffer of size at least MIN_SIZE. */ | |
2702 | _cpp_buff * | |
f7fdd7a1 | 2703 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
06c92cbc | 2704 | { |
2705 | _cpp_buff *result, **p; | |
2706 | ||
2707 | for (p = &pfile->free_buffs;; p = &(*p)->next) | |
2708 | { | |
4b31a107 | 2709 | size_t size; |
084163dc | 2710 | |
2711 | if (*p == NULL) | |
06c92cbc | 2712 | return new_buff (min_size); |
084163dc | 2713 | result = *p; |
2714 | size = result->limit - result->base; | |
2715 | /* Return a buffer that's big enough, but don't waste one that's | |
2716 | way too big. */ | |
4085c149 | 2717 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)) |
06c92cbc | 2718 | break; |
2719 | } | |
2720 | ||
2721 | *p = result->next; | |
2722 | result->next = NULL; | |
2723 | result->cur = result->base; | |
2724 | return result; | |
2725 | } | |
2726 | ||
20dd417a | 2727 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 2728 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
2729 | the excess bytes to the new buffer. Chains the new buffer after | |
2730 | BUFF, and returns the new buffer. */ | |
06c92cbc | 2731 | _cpp_buff * |
f7fdd7a1 | 2732 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
06c92cbc | 2733 | { |
4b31a107 | 2734 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra); |
e6a5f963 | 2735 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
06c92cbc | 2736 | |
e6a5f963 | 2737 | buff->next = new_buff; |
2738 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)); | |
2739 | return new_buff; | |
2740 | } | |
2741 | ||
20dd417a | 2742 | /* Creates a new buffer with enough space to hold the uncommitted |
e6a5f963 | 2743 | remaining bytes of the buffer pointed to by BUFF, and at least |
2744 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. | |
2745 | Chains the new buffer before the buffer pointed to by BUFF, and | |
2746 | updates the pointer to point to the new buffer. */ | |
2747 | void | |
f7fdd7a1 | 2748 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
e6a5f963 | 2749 | { |
2750 | _cpp_buff *new_buff, *old_buff = *pbuff; | |
2751 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra); | |
2752 | ||
2753 | new_buff = _cpp_get_buff (pfile, size); | |
2754 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)); | |
2755 | new_buff->next = old_buff; | |
2756 | *pbuff = new_buff; | |
06c92cbc | 2757 | } |
2758 | ||
2759 | /* Free a chain of buffers starting at BUFF. */ | |
2760 | void | |
f82b06e0 | 2761 | _cpp_free_buff (_cpp_buff *buff) |
06c92cbc | 2762 | { |
2763 | _cpp_buff *next; | |
2764 | ||
2765 | for (; buff; buff = next) | |
2766 | { | |
2767 | next = buff->next; | |
2768 | free (buff->base); | |
2769 | } | |
2770 | } | |
deb356cf | 2771 | |
1fdf6039 | 2772 | /* Allocate permanent, unaligned storage of length LEN. */ |
2773 | unsigned char * | |
f7fdd7a1 | 2774 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
1fdf6039 | 2775 | { |
2776 | _cpp_buff *buff = pfile->u_buff; | |
2777 | unsigned char *result = buff->cur; | |
2778 | ||
2779 | if (len > (size_t) (buff->limit - result)) | |
2780 | { | |
2781 | buff = _cpp_get_buff (pfile, len); | |
2782 | buff->next = pfile->u_buff; | |
2783 | pfile->u_buff = buff; | |
2784 | result = buff->cur; | |
2785 | } | |
2786 | ||
2787 | buff->cur = result + len; | |
2788 | return result; | |
2789 | } | |
2790 | ||
1e0ef2fd | 2791 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
2792 | That buffer is used for growing allocations when saving macro | |
2793 | replacement lists in a #define, and when parsing an answer to an | |
2794 | assertion in #assert, #unassert or #if (and therefore possibly | |
2795 | whilst expanding macros). It therefore must not be used by any | |
2796 | code that they might call: specifically the lexer and the guts of | |
2797 | the macro expander. | |
2798 | ||
2799 | All existing other uses clearly fit this restriction: storing | |
2800 | registered pragmas during initialization. */ | |
79bd622b | 2801 | unsigned char * |
f7fdd7a1 | 2802 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
89b05ef6 | 2803 | { |
e6a5f963 | 2804 | _cpp_buff *buff = pfile->a_buff; |
2805 | unsigned char *result = buff->cur; | |
89b05ef6 | 2806 | |
e6a5f963 | 2807 | if (len > (size_t) (buff->limit - result)) |
89b05ef6 | 2808 | { |
e6a5f963 | 2809 | buff = _cpp_get_buff (pfile, len); |
2810 | buff->next = pfile->a_buff; | |
2811 | pfile->a_buff = buff; | |
2812 | result = buff->cur; | |
89b05ef6 | 2813 | } |
f80e83a9 | 2814 | |
e6a5f963 | 2815 | buff->cur = result + len; |
79bd622b | 2816 | return result; |
f80e83a9 | 2817 | } |
c39ed964 | 2818 | |
2819 | /* Say which field of TOK is in use. */ | |
2820 | ||
2821 | enum cpp_token_fld_kind | |
2822 | cpp_token_val_index (cpp_token *tok) | |
2823 | { | |
2824 | switch (TOKEN_SPELL (tok)) | |
2825 | { | |
2826 | case SPELL_IDENT: | |
2827 | return CPP_TOKEN_FLD_NODE; | |
2828 | case SPELL_LITERAL: | |
2829 | return CPP_TOKEN_FLD_STR; | |
941f2388 | 2830 | case SPELL_OPERATOR: |
2831 | if (tok->type == CPP_PASTE) | |
2ee04baa | 2832 | return CPP_TOKEN_FLD_TOKEN_NO; |
941f2388 | 2833 | else |
2834 | return CPP_TOKEN_FLD_NONE; | |
c39ed964 | 2835 | case SPELL_NONE: |
2836 | if (tok->type == CPP_MACRO_ARG) | |
2837 | return CPP_TOKEN_FLD_ARG_NO; | |
2838 | else if (tok->type == CPP_PADDING) | |
2839 | return CPP_TOKEN_FLD_SOURCE; | |
d6d3c909 | 2840 | else if (tok->type == CPP_PRAGMA) |
b75b98aa | 2841 | return CPP_TOKEN_FLD_PRAGMA; |
c39ed964 | 2842 | /* else fall through */ |
2843 | default: | |
2844 | return CPP_TOKEN_FLD_NONE; | |
2845 | } | |
2846 | } |