1 /* CPP Library - lexical analysis.
2 Copyright (C) 2000-2019 Free Software Foundation, Inc.
3 Contributed by Per Bothner, 1994-95.
4 Based on CCCP program by Paul Rubin, June 1986
5 Adapted to ANSI C, Richard Stallman, Jan 1987
6 Broken out to separate file, Zack Weinberg, Mar 2000
8 This program is free software; you can redistribute it and/or modify it
9 under the terms of the GNU General Public License as published by the
10 Free Software Foundation; either version 3, or (at your option) any
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; see the file COPYING3. If not see
20 <http://www.gnu.org/licenses/>. */
37 enum spell_type category
;
38 const unsigned char *name
;
41 static const unsigned char *const digraph_spellings
[] =
42 { UC
"%:", UC
"%:%:", UC
"<:", UC
":>", UC
"<%", UC
"%>" };
44 #define OP(e, s) { SPELL_OPERATOR, UC s },
45 #define TK(e, s) { SPELL_ ## s, UC #e },
46 static const struct token_spelling token_spellings
[N_TTYPES
] = { TTYPE_TABLE
};
50 #define TOKEN_SPELL(token) (token_spellings[(token)->type].category)
51 #define TOKEN_NAME(token) (token_spellings[(token)->type].name)
53 static void add_line_note (cpp_buffer
*, const uchar
*, unsigned int);
54 static int skip_line_comment (cpp_reader
*);
55 static void skip_whitespace (cpp_reader
*, cppchar_t
);
56 static void lex_string (cpp_reader
*, cpp_token
*, const uchar
*);
57 static void save_comment (cpp_reader
*, cpp_token
*, const uchar
*, cppchar_t
);
58 static void store_comment (cpp_reader
*, cpp_token
*);
59 static void create_literal (cpp_reader
*, cpp_token
*, const uchar
*,
60 unsigned int, enum cpp_ttype
);
61 static bool warn_in_comment (cpp_reader
*, _cpp_line_note
*);
62 static int name_p (cpp_reader
*, const cpp_string
*);
63 static tokenrun
*next_tokenrun (tokenrun
*);
65 static _cpp_buff
*new_buff (size_t);
70 Compares, the token TOKEN to the NUL-terminated string STRING.
71 TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */
73 cpp_ideq (const cpp_token
*token
, const char *string
)
75 if (token
->type
!= CPP_NAME
)
78 return !ustrcmp (NODE_NAME (token
->val
.node
.node
), (const uchar
*) string
);
81 /* Record a note TYPE at byte POS into the current cleaned logical
84 add_line_note (cpp_buffer
*buffer
, const uchar
*pos
, unsigned int type
)
86 if (buffer
->notes_used
== buffer
->notes_cap
)
88 buffer
->notes_cap
= buffer
->notes_cap
* 2 + 200;
89 buffer
->notes
= XRESIZEVEC (_cpp_line_note
, buffer
->notes
,
93 buffer
->notes
[buffer
->notes_used
].pos
= pos
;
94 buffer
->notes
[buffer
->notes_used
].type
= type
;
99 /* Fast path to find line special characters using optimized character
100 scanning algorithms. Anything complicated falls back to the slow
101 path below. Since this loop is very hot it's worth doing these kinds
104 One of the paths through the ifdefs should provide
106 const uchar *search_line_fast (const uchar *s, const uchar *end);
108 Between S and END, search for \n, \r, \\, ?. Return a pointer to
111 Note that the last character of the buffer is *always* a newline,
112 as forced by _cpp_convert_input. This fact can be used to avoid
113 explicitly looking for the end of the buffer. */
115 /* Configure gives us an ifdef test. */
116 #ifndef WORDS_BIGENDIAN
117 #define WORDS_BIGENDIAN 0
120 /* We'd like the largest integer that fits into a register. There's nothing
121 in <stdint.h> that gives us that. For most hosts this is unsigned long,
122 but MS decided on an LLP64 model. Thankfully when building with GCC we
123 can get the "real" word size. */
125 typedef unsigned int word_type
__attribute__((__mode__(__word__
)));
127 typedef unsigned long word_type
;
130 /* The code below is only expecting sizes 4 or 8.
131 Die at compile-time if this expectation is violated. */
132 typedef char check_word_type_size
133 [(sizeof(word_type
) == 8 || sizeof(word_type
) == 4) * 2 - 1];
135 /* Return X with the first N bytes forced to values that won't match one
136 of the interesting characters. Note that NUL is not interesting. */
138 static inline word_type
139 acc_char_mask_misalign (word_type val
, unsigned int n
)
149 /* Return X replicated to all byte positions within WORD_TYPE. */
151 static inline word_type
152 acc_char_replicate (uchar x
)
156 ret
= (x
<< 24) | (x
<< 16) | (x
<< 8) | x
;
157 if (sizeof(word_type
) == 8)
158 ret
= (ret
<< 16 << 16) | ret
;
162 /* Return non-zero if some byte of VAL is (probably) C. */
164 static inline word_type
165 acc_char_cmp (word_type val
, word_type c
)
167 #if defined(__GNUC__) && defined(__alpha__)
168 /* We can get exact results using a compare-bytes instruction.
169 Get (val == c) via (0 >= (val ^ c)). */
170 return __builtin_alpha_cmpbge (0, val
^ c
);
172 word_type magic
= 0x7efefefeU
;
173 if (sizeof(word_type
) == 8)
174 magic
= (magic
<< 16 << 16) | 0xfefefefeU
;
178 return ((val
+ magic
) ^ ~val
) & ~magic
;
182 /* Given the result of acc_char_cmp is non-zero, return the index of
183 the found character. If this was a false positive, return -1. */
186 acc_char_index (word_type cmp ATTRIBUTE_UNUSED
,
187 word_type val ATTRIBUTE_UNUSED
)
189 #if defined(__GNUC__) && defined(__alpha__) && !WORDS_BIGENDIAN
190 /* The cmpbge instruction sets *bits* of the result corresponding to
191 matches in the bytes with no false positives. */
192 return __builtin_ctzl (cmp
);
196 /* ??? It would be nice to force unrolling here,
197 and have all of these constants folded. */
198 for (i
= 0; i
< sizeof(word_type
); ++i
)
202 c
= (val
>> (sizeof(word_type
) - i
- 1) * 8) & 0xff;
204 c
= (val
>> i
* 8) & 0xff;
206 if (c
== '\n' || c
== '\r' || c
== '\\' || c
== '?')
214 /* A version of the fast scanner using bit fiddling techniques.
216 For 32-bit words, one would normally perform 16 comparisons and
217 16 branches. With this algorithm one performs 24 arithmetic
218 operations and one branch. Whether this is faster with a 32-bit
219 word size is going to be somewhat system dependent.
221 For 64-bit words, we eliminate twice the number of comparisons
222 and branches without increasing the number of arithmetic operations.
223 It's almost certainly going to be a win with 64-bit word size. */
225 static const uchar
* search_line_acc_char (const uchar
*, const uchar
*)
229 search_line_acc_char (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
231 const word_type repl_nl
= acc_char_replicate ('\n');
232 const word_type repl_cr
= acc_char_replicate ('\r');
233 const word_type repl_bs
= acc_char_replicate ('\\');
234 const word_type repl_qm
= acc_char_replicate ('?');
236 unsigned int misalign
;
240 /* Align the buffer. Mask out any bytes from before the beginning. */
241 p
= (word_type
*)((uintptr_t)s
& -sizeof(word_type
));
243 misalign
= (uintptr_t)s
& (sizeof(word_type
) - 1);
245 val
= acc_char_mask_misalign (val
, misalign
);
250 t
= acc_char_cmp (val
, repl_nl
);
251 t
|= acc_char_cmp (val
, repl_cr
);
252 t
|= acc_char_cmp (val
, repl_bs
);
253 t
|= acc_char_cmp (val
, repl_qm
);
255 if (__builtin_expect (t
!= 0, 0))
257 int i
= acc_char_index (t
, val
);
259 return (const uchar
*)p
+ i
;
266 /* Disable on Solaris 2/x86 until the following problem can be properly
269 The Solaris 10+ assembler tags objects with the instruction set
270 extensions used, so SSE4.2 executables cannot run on machines that
271 don't support that extension. */
273 #if (GCC_VERSION >= 4005) && (__GNUC__ >= 5 || !defined(__PIC__)) && (defined(__i386__) || defined(__x86_64__)) && !(defined(__sun__) && defined(__svr4__))
275 /* Replicated character data to be shared between implementations.
276 Recall that outside of a context with vector support we can't
277 define compatible vector types, therefore these are all defined
278 in terms of raw characters. */
279 static const char repl_chars
[4][16] __attribute__((aligned(16))) = {
280 { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n',
281 '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' },
282 { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r',
283 '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' },
284 { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\',
285 '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' },
286 { '?', '?', '?', '?', '?', '?', '?', '?',
287 '?', '?', '?', '?', '?', '?', '?', '?' },
290 /* A version of the fast scanner using MMX vectorized byte compare insns.
292 This uses the PMOVMSKB instruction which was introduced with "MMX2",
293 which was packaged into SSE1; it is also present in the AMD MMX
294 extension. Mark the function as using "sse" so that we emit a real
295 "emms" instruction, rather than the 3dNOW "femms" instruction. */
299 __attribute__((__target__("sse")))
301 search_line_mmx (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
303 typedef char v8qi
__attribute__ ((__vector_size__ (8)));
304 typedef int __m64
__attribute__ ((__vector_size__ (8), __may_alias__
));
306 const v8qi repl_nl
= *(const v8qi
*)repl_chars
[0];
307 const v8qi repl_cr
= *(const v8qi
*)repl_chars
[1];
308 const v8qi repl_bs
= *(const v8qi
*)repl_chars
[2];
309 const v8qi repl_qm
= *(const v8qi
*)repl_chars
[3];
311 unsigned int misalign
, found
, mask
;
315 /* Align the source pointer. While MMX doesn't generate unaligned data
316 faults, this allows us to safely scan to the end of the buffer without
317 reading beyond the end of the last page. */
318 misalign
= (uintptr_t)s
& 7;
319 p
= (const v8qi
*)((uintptr_t)s
& -8);
322 /* Create a mask for the bytes that are valid within the first
323 16-byte block. The Idea here is that the AND with the mask
324 within the loop is "free", since we need some AND or TEST
325 insn in order to set the flags for the branch anyway. */
326 mask
= -1u << misalign
;
328 /* Main loop processing 8 bytes at a time. */
336 t
= __builtin_ia32_pcmpeqb(data
, repl_nl
);
337 c
= __builtin_ia32_pcmpeqb(data
, repl_cr
);
338 t
= (v8qi
) __builtin_ia32_por ((__m64
)t
, (__m64
)c
);
339 c
= __builtin_ia32_pcmpeqb(data
, repl_bs
);
340 t
= (v8qi
) __builtin_ia32_por ((__m64
)t
, (__m64
)c
);
341 c
= __builtin_ia32_pcmpeqb(data
, repl_qm
);
342 t
= (v8qi
) __builtin_ia32_por ((__m64
)t
, (__m64
)c
);
343 found
= __builtin_ia32_pmovmskb (t
);
348 __builtin_ia32_emms ();
350 /* FOUND contains 1 in bits for which we matched a relevant
351 character. Conversion to the byte index is trivial. */
352 found
= __builtin_ctz(found
);
353 return (const uchar
*)p
+ found
;
356 /* A version of the fast scanner using SSE2 vectorized byte compare insns. */
360 __attribute__((__target__("sse2")))
362 search_line_sse2 (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
364 typedef char v16qi
__attribute__ ((__vector_size__ (16)));
366 const v16qi repl_nl
= *(const v16qi
*)repl_chars
[0];
367 const v16qi repl_cr
= *(const v16qi
*)repl_chars
[1];
368 const v16qi repl_bs
= *(const v16qi
*)repl_chars
[2];
369 const v16qi repl_qm
= *(const v16qi
*)repl_chars
[3];
371 unsigned int misalign
, found
, mask
;
375 /* Align the source pointer. */
376 misalign
= (uintptr_t)s
& 15;
377 p
= (const v16qi
*)((uintptr_t)s
& -16);
380 /* Create a mask for the bytes that are valid within the first
381 16-byte block. The Idea here is that the AND with the mask
382 within the loop is "free", since we need some AND or TEST
383 insn in order to set the flags for the branch anyway. */
384 mask
= -1u << misalign
;
386 /* Main loop processing 16 bytes at a time. */
394 t
= __builtin_ia32_pcmpeqb128(data
, repl_nl
);
395 t
|= __builtin_ia32_pcmpeqb128(data
, repl_cr
);
396 t
|= __builtin_ia32_pcmpeqb128(data
, repl_bs
);
397 t
|= __builtin_ia32_pcmpeqb128(data
, repl_qm
);
398 found
= __builtin_ia32_pmovmskb128 (t
);
403 /* FOUND contains 1 in bits for which we matched a relevant
404 character. Conversion to the byte index is trivial. */
405 found
= __builtin_ctz(found
);
406 return (const uchar
*)p
+ found
;
410 /* A version of the fast scanner using SSE 4.2 vectorized string insns. */
414 __attribute__((__target__("sse4.2")))
416 search_line_sse42 (const uchar
*s
, const uchar
*end
)
418 typedef char v16qi
__attribute__ ((__vector_size__ (16)));
419 static const v16qi search
= { '\n', '\r', '?', '\\' };
421 uintptr_t si
= (uintptr_t)s
;
424 /* Check for unaligned input. */
429 if (__builtin_expect (end
- s
< 16, 0)
430 && __builtin_expect ((si
& 0xfff) > 0xff0, 0))
432 /* There are less than 16 bytes left in the buffer, and less
433 than 16 bytes left on the page. Reading 16 bytes at this
434 point might generate a spurious page fault. Defer to the
435 SSE2 implementation, which already handles alignment. */
436 return search_line_sse2 (s
, end
);
439 /* ??? The builtin doesn't understand that the PCMPESTRI read from
440 memory need not be aligned. */
441 sv
= __builtin_ia32_loaddqu ((const char *) s
);
442 index
= __builtin_ia32_pcmpestri128 (search
, 4, sv
, 16, 0);
444 if (__builtin_expect (index
< 16, 0))
447 /* Advance the pointer to an aligned address. We will re-scan a
448 few bytes, but we no longer need care for reading past the
449 end of a page, since we're guaranteed a match. */
450 s
= (const uchar
*)((si
+ 15) & -16);
453 /* Main loop, processing 16 bytes at a time. */
454 #ifdef __GCC_ASM_FLAG_OUTPUTS__
459 /* By using inline assembly instead of the builtin,
460 we can use the result, as well as the flags set. */
461 __asm ("%vpcmpestri\t$0, %2, %3"
462 : "=c"(index
), "=@ccc"(f
)
463 : "m"(*s
), "x"(search
), "a"(4), "d"(16));
471 /* By doing the whole loop in inline assembly,
472 we can make proper use of the flags set. */
473 __asm ( ".balign 16\n"
475 " %vpcmpestri\t$0, (%1), %2\n"
477 : "=&c"(index
), "+r"(s
)
478 : "x"(search
), "a"(4), "d"(16));
486 /* Work around out-dated assemblers without sse4 support. */
487 #define search_line_sse42 search_line_sse2
490 /* Check the CPU capabilities. */
492 #include "../gcc/config/i386/cpuid.h"
494 typedef const uchar
* (*search_line_fast_type
) (const uchar
*, const uchar
*);
495 static search_line_fast_type search_line_fast
;
497 #define HAVE_init_vectorized_lexer 1
499 init_vectorized_lexer (void)
501 unsigned dummy
, ecx
= 0, edx
= 0;
502 search_line_fast_type impl
= search_line_acc_char
;
505 #if defined(__SSE4_2__)
507 #elif defined(__SSE2__)
509 #elif defined(__SSE__)
514 impl
= search_line_sse42
;
515 else if (__get_cpuid (1, &dummy
, &dummy
, &ecx
, &edx
) || minimum
== 2)
517 if (minimum
== 3 || (ecx
& bit_SSE4_2
))
518 impl
= search_line_sse42
;
519 else if (minimum
== 2 || (edx
& bit_SSE2
))
520 impl
= search_line_sse2
;
521 else if (minimum
== 1 || (edx
& bit_SSE
))
522 impl
= search_line_mmx
;
524 else if (__get_cpuid (0x80000001, &dummy
, &dummy
, &dummy
, &edx
))
527 || (edx
& (bit_MMXEXT
| bit_CMOV
)) == (bit_MMXEXT
| bit_CMOV
))
528 impl
= search_line_mmx
;
531 search_line_fast
= impl
;
534 #elif defined(_ARCH_PWR8) && defined(__ALTIVEC__)
536 /* A vection of the fast scanner using AltiVec vectorized byte compares
537 and VSX unaligned loads (when VSX is available). This is otherwise
538 the same as the pre-GCC 5 version. */
540 ATTRIBUTE_NO_SANITIZE_UNDEFINED
542 search_line_fast (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
544 typedef __attribute__((altivec(vector
))) unsigned char vc
;
547 '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n',
548 '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n'
551 '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r',
552 '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r'
555 '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\',
556 '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\'
559 '?', '?', '?', '?', '?', '?', '?', '?',
560 '?', '?', '?', '?', '?', '?', '?', '?',
562 const vc zero
= { 0 };
566 /* Main loop processing 16 bytes at a time. */
569 vc m_nl
, m_cr
, m_bs
, m_qm
;
571 data
= __builtin_vec_vsx_ld (0, s
);
574 m_nl
= (vc
) __builtin_vec_cmpeq(data
, repl_nl
);
575 m_cr
= (vc
) __builtin_vec_cmpeq(data
, repl_cr
);
576 m_bs
= (vc
) __builtin_vec_cmpeq(data
, repl_bs
);
577 m_qm
= (vc
) __builtin_vec_cmpeq(data
, repl_qm
);
578 t
= (m_nl
| m_cr
) | (m_bs
| m_qm
);
580 /* T now contains 0xff in bytes for which we matched one of the relevant
581 characters. We want to exit the loop if any byte in T is non-zero.
582 Below is the expansion of vec_any_ne(t, zero). */
584 while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t
, zero
));
586 /* Restore s to to point to the 16 bytes we just processed. */
590 #define N (sizeof(vc) / sizeof(long))
594 /* Statically assert that N is 2 or 4. */
595 unsigned long l
[(N
== 2 || N
== 4) ? N
: -1];
597 unsigned long l
, i
= 0;
601 /* Find the first word of T that is non-zero. */
608 s
+= sizeof(unsigned long);
612 s
+= sizeof(unsigned long);
618 s
+= sizeof(unsigned long);
622 /* L now contains 0xff in bytes for which we matched one of the
623 relevant characters. We can find the byte index by finding
624 its bit index and dividing by 8. */
625 #ifdef __BIG_ENDIAN__
626 l
= __builtin_clzl(l
) >> 3;
628 l
= __builtin_ctzl(l
) >> 3;
636 #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__)
638 /* A vection of the fast scanner using AltiVec vectorized byte compares.
639 This cannot be used for little endian because vec_lvsl/lvsr are
640 deprecated for little endian and the code won't work properly. */
641 /* ??? Unfortunately, attribute(target("altivec")) is not yet supported,
642 so we can't compile this function without -maltivec on the command line
643 (or implied by some other switch). */
646 search_line_fast (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
648 typedef __attribute__((altivec(vector
))) unsigned char vc
;
651 '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n',
652 '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n'
655 '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r',
656 '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r'
659 '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\',
660 '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\'
663 '?', '?', '?', '?', '?', '?', '?', '?',
664 '?', '?', '?', '?', '?', '?', '?', '?',
667 -1, -1, -1, -1, -1, -1, -1, -1,
668 -1, -1, -1, -1, -1, -1, -1, -1,
670 const vc zero
= { 0 };
674 /* Altivec loads automatically mask addresses with -16. This lets us
675 issue the first load as early as possible. */
676 data
= __builtin_vec_ld(0, (const vc
*)s
);
678 /* Discard bytes before the beginning of the buffer. Do this by
679 beginning with all ones and shifting in zeros according to the
680 mis-alignment. The LVSR instruction pulls the exact shift we
681 want from the address. */
682 mask
= __builtin_vec_lvsr(0, s
);
683 mask
= __builtin_vec_perm(zero
, ones
, mask
);
686 /* While altivec loads mask addresses, we still need to align S so
687 that the offset we compute at the end is correct. */
688 s
= (const uchar
*)((uintptr_t)s
& -16);
690 /* Main loop processing 16 bytes at a time. */
694 vc m_nl
, m_cr
, m_bs
, m_qm
;
697 data
= __builtin_vec_ld(0, (const vc
*)s
);
700 m_nl
= (vc
) __builtin_vec_cmpeq(data
, repl_nl
);
701 m_cr
= (vc
) __builtin_vec_cmpeq(data
, repl_cr
);
702 m_bs
= (vc
) __builtin_vec_cmpeq(data
, repl_bs
);
703 m_qm
= (vc
) __builtin_vec_cmpeq(data
, repl_qm
);
704 t
= (m_nl
| m_cr
) | (m_bs
| m_qm
);
706 /* T now contains 0xff in bytes for which we matched one of the relevant
707 characters. We want to exit the loop if any byte in T is non-zero.
708 Below is the expansion of vec_any_ne(t, zero). */
710 while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t
, zero
));
713 #define N (sizeof(vc) / sizeof(long))
717 /* Statically assert that N is 2 or 4. */
718 unsigned long l
[(N
== 2 || N
== 4) ? N
: -1];
720 unsigned long l
, i
= 0;
724 /* Find the first word of T that is non-zero. */
731 s
+= sizeof(unsigned long);
735 s
+= sizeof(unsigned long);
741 s
+= sizeof(unsigned long);
745 /* L now contains 0xff in bytes for which we matched one of the
746 relevant characters. We can find the byte index by finding
747 its bit index and dividing by 8. */
748 l
= __builtin_clzl(l
) >> 3;
755 #elif defined (__ARM_NEON) && defined (__ARM_64BIT_STATE)
756 #include "arm_neon.h"
758 /* This doesn't have to be the exact page size, but no system may use
759 a size smaller than this. ARMv8 requires a minimum page size of
760 4k. The impact of being conservative here is a small number of
761 cases will take the slightly slower entry path into the main
764 #define AARCH64_MIN_PAGE_SIZE 4096
767 search_line_fast (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
769 const uint8x16_t repl_nl
= vdupq_n_u8 ('\n');
770 const uint8x16_t repl_cr
= vdupq_n_u8 ('\r');
771 const uint8x16_t repl_bs
= vdupq_n_u8 ('\\');
772 const uint8x16_t repl_qm
= vdupq_n_u8 ('?');
773 const uint8x16_t xmask
= (uint8x16_t
) vdupq_n_u64 (0x8040201008040201ULL
);
775 #ifdef __ARM_BIG_ENDIAN
776 const int16x8_t shift
= {8, 8, 8, 8, 0, 0, 0, 0};
778 const int16x8_t shift
= {0, 0, 0, 0, 8, 8, 8, 8};
788 /* Align the source pointer. */
789 p
= (const uint8_t *)((uintptr_t)s
& -16);
791 /* Assuming random string start positions, with a 4k page size we'll take
792 the slow path about 0.37% of the time. */
793 if (__builtin_expect ((AARCH64_MIN_PAGE_SIZE
794 - (((uintptr_t) s
) & (AARCH64_MIN_PAGE_SIZE
- 1)))
797 /* Slow path: the string starts near a possible page boundary. */
798 uint32_t misalign
, mask
;
800 misalign
= (uintptr_t)s
& 15;
801 mask
= (-1u << misalign
) & 0xffff;
803 t
= vceqq_u8 (data
, repl_nl
);
804 u
= vceqq_u8 (data
, repl_cr
);
805 v
= vorrq_u8 (t
, vceqq_u8 (data
, repl_bs
));
806 w
= vorrq_u8 (u
, vceqq_u8 (data
, repl_qm
));
808 t
= vandq_u8 (t
, xmask
);
810 m
= vshlq_u16 (m
, shift
);
811 found
= vaddvq_u16 (m
);
814 return (const uchar
*)p
+ __builtin_ctz (found
);
818 data
= vld1q_u8 ((const uint8_t *) s
);
819 t
= vceqq_u8 (data
, repl_nl
);
820 u
= vceqq_u8 (data
, repl_cr
);
821 v
= vorrq_u8 (t
, vceqq_u8 (data
, repl_bs
));
822 w
= vorrq_u8 (u
, vceqq_u8 (data
, repl_qm
));
824 if (__builtin_expect (vpaddd_u64 ((uint64x2_t
)t
) != 0, 0))
832 t
= vceqq_u8 (data
, repl_nl
);
833 u
= vceqq_u8 (data
, repl_cr
);
834 v
= vorrq_u8 (t
, vceqq_u8 (data
, repl_bs
));
835 w
= vorrq_u8 (u
, vceqq_u8 (data
, repl_qm
));
837 } while (!vpaddd_u64 ((uint64x2_t
)t
));
840 /* Now that we've found the terminating substring, work out precisely where
842 t
= vandq_u8 (t
, xmask
);
844 m
= vshlq_u16 (m
, shift
);
845 found
= vaddvq_u16 (m
);
846 return (((((uintptr_t) p
) < (uintptr_t) s
) ? s
: (const uchar
*)p
)
847 + __builtin_ctz (found
));
850 #elif defined (__ARM_NEON)
851 #include "arm_neon.h"
854 search_line_fast (const uchar
*s
, const uchar
*end ATTRIBUTE_UNUSED
)
856 const uint8x16_t repl_nl
= vdupq_n_u8 ('\n');
857 const uint8x16_t repl_cr
= vdupq_n_u8 ('\r');
858 const uint8x16_t repl_bs
= vdupq_n_u8 ('\\');
859 const uint8x16_t repl_qm
= vdupq_n_u8 ('?');
860 const uint8x16_t xmask
= (uint8x16_t
) vdupq_n_u64 (0x8040201008040201ULL
);
862 unsigned int misalign
, found
, mask
;
866 /* Align the source pointer. */
867 misalign
= (uintptr_t)s
& 15;
868 p
= (const uint8_t *)((uintptr_t)s
& -16);
871 /* Create a mask for the bytes that are valid within the first
872 16-byte block. The Idea here is that the AND with the mask
873 within the loop is "free", since we need some AND or TEST
874 insn in order to set the flags for the branch anyway. */
875 mask
= (-1u << misalign
) & 0xffff;
877 /* Main loop, processing 16 bytes at a time. */
885 uint8x16_t t
, u
, v
, w
;
892 t
= vceqq_u8 (data
, repl_nl
);
893 u
= vceqq_u8 (data
, repl_cr
);
894 v
= vorrq_u8 (t
, vceqq_u8 (data
, repl_bs
));
895 w
= vorrq_u8 (u
, vceqq_u8 (data
, repl_qm
));
896 t
= vandq_u8 (vorrq_u8 (v
, w
), xmask
);
897 l
= vpadd_u8 (vget_low_u8 (t
), vget_high_u8 (t
));
901 found
= vget_lane_u32 ((uint32x2_t
) vorr_u64 ((uint64x1_t
) n
,
902 vshr_n_u64 ((uint64x1_t
) n
, 24)), 0);
907 /* FOUND contains 1 in bits for which we matched a relevant
908 character. Conversion to the byte index is trivial. */
909 found
= __builtin_ctz (found
);
910 return (const uchar
*)p
+ found
;
915 /* We only have one accelerated alternative. Use a direct call so that
916 we encourage inlining. */
918 #define search_line_fast search_line_acc_char
922 /* Initialize the lexer if needed. */
925 _cpp_init_lexer (void)
927 #ifdef HAVE_init_vectorized_lexer
928 init_vectorized_lexer ();
932 /* Returns with a logical line that contains no escaped newlines or
933 trigraphs. This is a time-critical inner loop. */
935 _cpp_clean_line (cpp_reader
*pfile
)
941 buffer
= pfile
->buffer
;
942 buffer
->cur_note
= buffer
->notes_used
= 0;
943 buffer
->cur
= buffer
->line_base
= buffer
->next_line
;
944 buffer
->need_line
= false;
945 s
= buffer
->next_line
;
947 if (!buffer
->from_stage3
)
949 const uchar
*pbackslash
= NULL
;
951 /* Fast path. This is the common case of an un-escaped line with
952 no trigraphs. The primary win here is by not writing any
953 data back to memory until we have to. */
956 /* Perform an optimized search for \n, \r, \\, ?. */
957 s
= search_line_fast (s
, buffer
->rlimit
);
962 /* Record the location of the backslash and continue. */
965 else if (__builtin_expect (c
== '?', 0))
967 if (__builtin_expect (s
[1] == '?', false)
968 && _cpp_trigraph_map
[s
[2]])
970 /* Have a trigraph. We may or may not have to convert
971 it. Add a line note regardless, for -Wtrigraphs. */
972 add_line_note (buffer
, s
, s
[2]);
973 if (CPP_OPTION (pfile
, trigraphs
))
975 /* We do, and that means we have to switch to the
978 *d
= _cpp_trigraph_map
[s
[2]];
983 /* Not a trigraph. Continue on fast-path. */
990 /* This must be \r or \n. We're either done, or we'll be forced
991 to write back to the buffer and continue on the slow path. */
994 if (__builtin_expect (s
== buffer
->rlimit
, false))
997 /* DOS line ending? */
998 if (__builtin_expect (c
== '\r', false) && s
[1] == '\n')
1001 if (s
== buffer
->rlimit
)
1005 if (__builtin_expect (pbackslash
== NULL
, true))
1008 /* Check for escaped newline. */
1010 while (is_nvspace (p
[-1]))
1012 if (p
- 1 != pbackslash
)
1015 /* Have an escaped newline; process it and proceed to
1017 add_line_note (buffer
, p
- 1, p
!= d
? ' ' : '\\');
1019 buffer
->next_line
= p
- 1;
1027 if (c
== '\n' || c
== '\r')
1029 /* Handle DOS line endings. */
1030 if (c
== '\r' && s
!= buffer
->rlimit
&& s
[1] == '\n')
1032 if (s
== buffer
->rlimit
)
1037 while (p
!= buffer
->next_line
&& is_nvspace (p
[-1]))
1039 if (p
== buffer
->next_line
|| p
[-1] != '\\')
1042 add_line_note (buffer
, p
- 1, p
!= d
? ' ': '\\');
1044 buffer
->next_line
= p
- 1;
1046 else if (c
== '?' && s
[1] == '?' && _cpp_trigraph_map
[s
[2]])
1048 /* Add a note regardless, for the benefit of -Wtrigraphs. */
1049 add_line_note (buffer
, d
, s
[2]);
1050 if (CPP_OPTION (pfile
, trigraphs
))
1052 *d
= _cpp_trigraph_map
[s
[2]];
1060 while (*s
!= '\n' && *s
!= '\r')
1064 /* Handle DOS line endings. */
1065 if (*s
== '\r' && s
!= buffer
->rlimit
&& s
[1] == '\n')
1071 /* A sentinel note that should never be processed. */
1072 add_line_note (buffer
, d
+ 1, '\n');
1073 buffer
->next_line
= s
+ 1;
1076 /* Return true if the trigraph indicated by NOTE should be warned
1077 about in a comment. */
1079 warn_in_comment (cpp_reader
*pfile
, _cpp_line_note
*note
)
1083 /* Within comments we don't warn about trigraphs, unless the
1084 trigraph forms an escaped newline, as that may change
1086 if (note
->type
!= '/')
1089 /* If -trigraphs, then this was an escaped newline iff the next note
1091 if (CPP_OPTION (pfile
, trigraphs
))
1092 return note
[1].pos
== note
->pos
;
1094 /* Otherwise, see if this forms an escaped newline. */
1096 while (is_nvspace (*p
))
1099 /* There might have been escaped newlines between the trigraph and the
1100 newline we found. Hence the position test. */
1101 return (*p
== '\n' && p
< note
[1].pos
);
1104 /* Process the notes created by add_line_note as far as the current
1107 _cpp_process_line_notes (cpp_reader
*pfile
, int in_comment
)
1109 cpp_buffer
*buffer
= pfile
->buffer
;
1113 _cpp_line_note
*note
= &buffer
->notes
[buffer
->cur_note
];
1116 if (note
->pos
> buffer
->cur
)
1120 col
= CPP_BUF_COLUMN (buffer
, note
->pos
+ 1);
1122 if (note
->type
== '\\' || note
->type
== ' ')
1124 if (note
->type
== ' ' && !in_comment
)
1125 cpp_error_with_line (pfile
, CPP_DL_WARNING
, pfile
->line_table
->highest_line
, col
,
1126 "backslash and newline separated by space");
1128 if (buffer
->next_line
> buffer
->rlimit
)
1130 cpp_error_with_line (pfile
, CPP_DL_PEDWARN
, pfile
->line_table
->highest_line
, col
,
1131 "backslash-newline at end of file");
1132 /* Prevent "no newline at end of file" warning. */
1133 buffer
->next_line
= buffer
->rlimit
;
1136 buffer
->line_base
= note
->pos
;
1137 CPP_INCREMENT_LINE (pfile
, 0);
1139 else if (_cpp_trigraph_map
[note
->type
])
1141 if (CPP_OPTION (pfile
, warn_trigraphs
)
1142 && (!in_comment
|| warn_in_comment (pfile
, note
)))
1144 if (CPP_OPTION (pfile
, trigraphs
))
1145 cpp_warning_with_line (pfile
, CPP_W_TRIGRAPHS
,
1146 pfile
->line_table
->highest_line
, col
,
1147 "trigraph ??%c converted to %c",
1149 (int) _cpp_trigraph_map
[note
->type
]);
1152 cpp_warning_with_line
1153 (pfile
, CPP_W_TRIGRAPHS
,
1154 pfile
->line_table
->highest_line
, col
,
1155 "trigraph ??%c ignored, use -trigraphs to enable",
1160 else if (note
->type
== 0)
1161 /* Already processed in lex_raw_string. */;
1167 /* Skip a C-style block comment. We find the end of the comment by
1168 seeing if an asterisk is before every '/' we encounter. Returns
1169 nonzero if comment terminated by EOF, zero otherwise.
1171 Buffer->cur points to the initial asterisk of the comment. */
1173 _cpp_skip_block_comment (cpp_reader
*pfile
)
1175 cpp_buffer
*buffer
= pfile
->buffer
;
1176 const uchar
*cur
= buffer
->cur
;
1185 /* People like decorating comments with '*', so check for '/'
1186 instead for efficiency. */
1194 /* Warn about potential nested comments, but not if the '/'
1195 comes immediately before the true comment delimiter.
1196 Don't bother to get it right across escaped newlines. */
1197 if (CPP_OPTION (pfile
, warn_comments
)
1198 && cur
[0] == '*' && cur
[1] != '/')
1201 cpp_warning_with_line (pfile
, CPP_W_COMMENTS
,
1202 pfile
->line_table
->highest_line
,
1203 CPP_BUF_COL (buffer
),
1204 "\"/*\" within comment");
1210 buffer
->cur
= cur
- 1;
1211 _cpp_process_line_notes (pfile
, true);
1212 if (buffer
->next_line
>= buffer
->rlimit
)
1214 _cpp_clean_line (pfile
);
1216 cols
= buffer
->next_line
- buffer
->line_base
;
1217 CPP_INCREMENT_LINE (pfile
, cols
);
1224 _cpp_process_line_notes (pfile
, true);
1228 /* Skip a C++ line comment, leaving buffer->cur pointing to the
1229 terminating newline. Handles escaped newlines. Returns nonzero
1230 if a multiline comment. */
1232 skip_line_comment (cpp_reader
*pfile
)
1234 cpp_buffer
*buffer
= pfile
->buffer
;
1235 location_t orig_line
= pfile
->line_table
->highest_line
;
1237 while (*buffer
->cur
!= '\n')
1240 _cpp_process_line_notes (pfile
, true);
1241 return orig_line
!= pfile
->line_table
->highest_line
;
1244 /* Skips whitespace, saving the next non-whitespace character. */
1246 skip_whitespace (cpp_reader
*pfile
, cppchar_t c
)
1248 cpp_buffer
*buffer
= pfile
->buffer
;
1249 bool saw_NUL
= false;
1253 /* Horizontal space always OK. */
1254 if (c
== ' ' || c
== '\t')
1256 /* Just \f \v or \0 left. */
1259 else if (pfile
->state
.in_directive
&& CPP_PEDANTIC (pfile
))
1260 cpp_error_with_line (pfile
, CPP_DL_PEDWARN
, pfile
->line_table
->highest_line
,
1261 CPP_BUF_COL (buffer
),
1262 "%s in preprocessing directive",
1263 c
== '\f' ? "form feed" : "vertical tab");
1267 /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */
1268 while (is_nvspace (c
));
1271 cpp_error (pfile
, CPP_DL_WARNING
, "null character(s) ignored");
1276 /* See if the characters of a number token are valid in a name (no
1277 '.', '+' or '-'). */
1279 name_p (cpp_reader
*pfile
, const cpp_string
*string
)
1283 for (i
= 0; i
< string
->len
; i
++)
1284 if (!is_idchar (string
->text
[i
]))
1290 /* After parsing an identifier or other sequence, produce a warning about
1291 sequences not in NFC/NFKC. */
1293 warn_about_normalization (cpp_reader
*pfile
,
1294 const cpp_token
*token
,
1295 const struct normalize_state
*s
)
1297 if (CPP_OPTION (pfile
, warn_normalize
) < NORMALIZE_STATE_RESULT (s
)
1298 && !pfile
->state
.skipping
)
1300 /* Make sure that the token is printed using UCNs, even
1301 if we'd otherwise happily print UTF-8. */
1302 unsigned char *buf
= XNEWVEC (unsigned char, cpp_token_len (token
));
1305 sz
= cpp_spell_token (pfile
, token
, buf
, false) - buf
;
1306 if (NORMALIZE_STATE_RESULT (s
) == normalized_C
)
1307 cpp_warning_with_line (pfile
, CPP_W_NORMALIZE
, token
->src_loc
, 0,
1308 "`%.*s' is not in NFKC", (int) sz
, buf
);
1310 cpp_warning_with_line (pfile
, CPP_W_NORMALIZE
, token
->src_loc
, 0,
1311 "`%.*s' is not in NFC", (int) sz
, buf
);
1316 static const cppchar_t utf8_signifier
= 0xC0;
1318 /* Returns TRUE if the sequence starting at buffer->cur is valid in
1319 an identifier. FIRST is TRUE if this starts an identifier. */
1321 forms_identifier_p (cpp_reader
*pfile
, int first
,
1322 struct normalize_state
*state
)
1324 cpp_buffer
*buffer
= pfile
->buffer
;
1326 if (*buffer
->cur
== '$')
1328 if (!CPP_OPTION (pfile
, dollars_in_ident
))
1332 if (CPP_OPTION (pfile
, warn_dollars
) && !pfile
->state
.skipping
)
1334 CPP_OPTION (pfile
, warn_dollars
) = 0;
1335 cpp_error (pfile
, CPP_DL_PEDWARN
, "'$' in identifier or number");
1341 /* Is this a syntactically valid UCN or a valid UTF-8 char? */
1342 if (CPP_OPTION (pfile
, extended_identifiers
))
1345 if (*buffer
->cur
>= utf8_signifier
)
1347 if (_cpp_valid_utf8 (pfile
, &buffer
->cur
, buffer
->rlimit
, 1 + !first
,
1351 else if (*buffer
->cur
== '\\'
1352 && (buffer
->cur
[1] == 'u' || buffer
->cur
[1] == 'U'))
1355 if (_cpp_valid_ucn (pfile
, &buffer
->cur
, buffer
->rlimit
, 1 + !first
,
1356 state
, &s
, NULL
, NULL
))
1365 /* Helper function to issue error about improper __VA_OPT__ use. */
1367 maybe_va_opt_error (cpp_reader
*pfile
)
1369 if (CPP_PEDANTIC (pfile
) && !CPP_OPTION (pfile
, va_opt
))
1371 /* __VA_OPT__ should not be accepted at all, but allow it in
1373 if (!cpp_in_system_header (pfile
))
1374 cpp_error (pfile
, CPP_DL_PEDWARN
,
1375 "__VA_OPT__ is not available until C++2a");
1377 else if (!pfile
->state
.va_args_ok
)
1379 /* __VA_OPT__ should only appear in the replacement list of a
1381 cpp_error (pfile
, CPP_DL_PEDWARN
,
1382 "__VA_OPT__ can only appear in the expansion"
1383 " of a C++2a variadic macro");
1387 /* Helper function to get the cpp_hashnode of the identifier BASE. */
1388 static cpp_hashnode
*
1389 lex_identifier_intern (cpp_reader
*pfile
, const uchar
*base
)
1391 cpp_hashnode
*result
;
1394 unsigned int hash
= HT_HASHSTEP (0, *base
);
1397 while (ISIDNUM (*cur
))
1399 hash
= HT_HASHSTEP (hash
, *cur
);
1403 hash
= HT_HASHFINISH (hash
, len
);
1404 result
= CPP_HASHNODE (ht_lookup_with_hash (pfile
->hash_table
,
1405 base
, len
, hash
, HT_ALLOC
));
1407 /* Rarely, identifiers require diagnostics when lexed. */
1408 if (__builtin_expect ((result
->flags
& NODE_DIAGNOSTIC
)
1409 && !pfile
->state
.skipping
, 0))
1411 /* It is allowed to poison the same identifier twice. */
1412 if ((result
->flags
& NODE_POISONED
) && !pfile
->state
.poisoned_ok
)
1413 cpp_error (pfile
, CPP_DL_ERROR
, "attempt to use poisoned \"%s\"",
1414 NODE_NAME (result
));
1416 /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
1417 replacement list of a variadic macro. */
1418 if (result
== pfile
->spec_nodes
.n__VA_ARGS__
1419 && !pfile
->state
.va_args_ok
)
1421 if (CPP_OPTION (pfile
, cplusplus
))
1422 cpp_error (pfile
, CPP_DL_PEDWARN
,
1423 "__VA_ARGS__ can only appear in the expansion"
1424 " of a C++11 variadic macro");
1426 cpp_error (pfile
, CPP_DL_PEDWARN
,
1427 "__VA_ARGS__ can only appear in the expansion"
1428 " of a C99 variadic macro");
1431 if (result
== pfile
->spec_nodes
.n__VA_OPT__
)
1432 maybe_va_opt_error (pfile
);
1434 /* For -Wc++-compat, warn about use of C++ named operators. */
1435 if (result
->flags
& NODE_WARN_OPERATOR
)
1436 cpp_warning (pfile
, CPP_W_CXX_OPERATOR_NAMES
,
1437 "identifier \"%s\" is a special operator name in C++",
1438 NODE_NAME (result
));
1444 /* Get the cpp_hashnode of an identifier specified by NAME in
1445 the current cpp_reader object. If none is found, NULL is returned. */
1447 _cpp_lex_identifier (cpp_reader
*pfile
, const char *name
)
1449 cpp_hashnode
*result
;
1450 result
= lex_identifier_intern (pfile
, (uchar
*) name
);
1454 /* Lex an identifier starting at BUFFER->CUR - 1. */
1455 static cpp_hashnode
*
1456 lex_identifier (cpp_reader
*pfile
, const uchar
*base
, bool starts_ucn
,
1457 struct normalize_state
*nst
, cpp_hashnode
**spelling
)
1459 cpp_hashnode
*result
;
1462 unsigned int hash
= HT_HASHSTEP (0, *base
);
1464 cur
= pfile
->buffer
->cur
;
1467 while (ISIDNUM (*cur
))
1469 hash
= HT_HASHSTEP (hash
, *cur
);
1472 NORMALIZE_STATE_UPDATE_IDNUM (nst
, *(cur
- 1));
1474 pfile
->buffer
->cur
= cur
;
1475 if (starts_ucn
|| forms_identifier_p (pfile
, false, nst
))
1477 /* Slower version for identifiers containing UCNs
1478 or extended chars (including $). */
1480 while (ISIDNUM (*pfile
->buffer
->cur
))
1482 NORMALIZE_STATE_UPDATE_IDNUM (nst
, *pfile
->buffer
->cur
);
1483 pfile
->buffer
->cur
++;
1485 } while (forms_identifier_p (pfile
, false, nst
));
1486 result
= _cpp_interpret_identifier (pfile
, base
,
1487 pfile
->buffer
->cur
- base
);
1488 *spelling
= cpp_lookup (pfile
, base
, pfile
->buffer
->cur
- base
);
1493 hash
= HT_HASHFINISH (hash
, len
);
1495 result
= CPP_HASHNODE (ht_lookup_with_hash (pfile
->hash_table
,
1496 base
, len
, hash
, HT_ALLOC
));
1500 /* Rarely, identifiers require diagnostics when lexed. */
1501 if (__builtin_expect ((result
->flags
& NODE_DIAGNOSTIC
)
1502 && !pfile
->state
.skipping
, 0))
1504 /* It is allowed to poison the same identifier twice. */
1505 if ((result
->flags
& NODE_POISONED
) && !pfile
->state
.poisoned_ok
)
1506 cpp_error (pfile
, CPP_DL_ERROR
, "attempt to use poisoned \"%s\"",
1507 NODE_NAME (result
));
1509 /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the
1510 replacement list of a variadic macro. */
1511 if (result
== pfile
->spec_nodes
.n__VA_ARGS__
1512 && !pfile
->state
.va_args_ok
)
1514 if (CPP_OPTION (pfile
, cplusplus
))
1515 cpp_error (pfile
, CPP_DL_PEDWARN
,
1516 "__VA_ARGS__ can only appear in the expansion"
1517 " of a C++11 variadic macro");
1519 cpp_error (pfile
, CPP_DL_PEDWARN
,
1520 "__VA_ARGS__ can only appear in the expansion"
1521 " of a C99 variadic macro");
1524 /* __VA_OPT__ should only appear in the replacement list of a
1526 if (result
== pfile
->spec_nodes
.n__VA_OPT__
)
1527 maybe_va_opt_error (pfile
);
1529 /* For -Wc++-compat, warn about use of C++ named operators. */
1530 if (result
->flags
& NODE_WARN_OPERATOR
)
1531 cpp_warning (pfile
, CPP_W_CXX_OPERATOR_NAMES
,
1532 "identifier \"%s\" is a special operator name in C++",
1533 NODE_NAME (result
));
1539 /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */
1541 lex_number (cpp_reader
*pfile
, cpp_string
*number
,
1542 struct normalize_state
*nst
)
1548 base
= pfile
->buffer
->cur
- 1;
1551 cur
= pfile
->buffer
->cur
;
1553 /* N.B. ISIDNUM does not include $. */
1554 while (ISIDNUM (*cur
) || *cur
== '.' || DIGIT_SEP (*cur
)
1555 || VALID_SIGN (*cur
, cur
[-1]))
1557 NORMALIZE_STATE_UPDATE_IDNUM (nst
, *cur
);
1560 /* A number can't end with a digit separator. */
1561 while (cur
> pfile
->buffer
->cur
&& DIGIT_SEP (cur
[-1]))
1564 pfile
->buffer
->cur
= cur
;
1566 while (forms_identifier_p (pfile
, false, nst
));
1568 number
->len
= cur
- base
;
1569 dest
= _cpp_unaligned_alloc (pfile
, number
->len
+ 1);
1570 memcpy (dest
, base
, number
->len
);
1571 dest
[number
->len
] = '\0';
1572 number
->text
= dest
;
1575 /* Create a token of type TYPE with a literal spelling. */
1577 create_literal (cpp_reader
*pfile
, cpp_token
*token
, const uchar
*base
,
1578 unsigned int len
, enum cpp_ttype type
)
1580 uchar
*dest
= _cpp_unaligned_alloc (pfile
, len
+ 1);
1582 memcpy (dest
, base
, len
);
1585 token
->val
.str
.len
= len
;
1586 token
->val
.str
.text
= dest
;
1589 /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer
1590 sequence from *FIRST_BUFF_P to LAST_BUFF_P. */
1593 bufring_append (cpp_reader
*pfile
, const uchar
*base
, size_t len
,
1594 _cpp_buff
**first_buff_p
, _cpp_buff
**last_buff_p
)
1596 _cpp_buff
*first_buff
= *first_buff_p
;
1597 _cpp_buff
*last_buff
= *last_buff_p
;
1599 if (first_buff
== NULL
)
1600 first_buff
= last_buff
= _cpp_get_buff (pfile
, len
);
1601 else if (len
> BUFF_ROOM (last_buff
))
1603 size_t room
= BUFF_ROOM (last_buff
);
1604 memcpy (BUFF_FRONT (last_buff
), base
, room
);
1605 BUFF_FRONT (last_buff
) += room
;
1608 last_buff
= _cpp_append_extend_buff (pfile
, last_buff
, len
);
1611 memcpy (BUFF_FRONT (last_buff
), base
, len
);
1612 BUFF_FRONT (last_buff
) += len
;
1614 *first_buff_p
= first_buff
;
1615 *last_buff_p
= last_buff
;
1619 /* Returns true if a macro has been defined.
1620 This might not work if compile with -save-temps,
1621 or preprocess separately from compilation. */
1624 is_macro(cpp_reader
*pfile
, const uchar
*base
)
1626 const uchar
*cur
= base
;
1627 if (! ISIDST (*cur
))
1629 unsigned int hash
= HT_HASHSTEP (0, *cur
);
1631 while (ISIDNUM (*cur
))
1633 hash
= HT_HASHSTEP (hash
, *cur
);
1636 hash
= HT_HASHFINISH (hash
, cur
- base
);
1638 cpp_hashnode
*result
= CPP_HASHNODE (ht_lookup_with_hash (pfile
->hash_table
,
1639 base
, cur
- base
, hash
, HT_NO_INSERT
));
1641 return result
&& cpp_macro_p (result
);
1644 /* Returns true if a literal suffix does not have the expected form
1645 and is defined as a macro. */
1648 is_macro_not_literal_suffix(cpp_reader
*pfile
, const uchar
*base
)
1650 /* User-defined literals outside of namespace std must start with a single
1651 underscore, so assume anything of that form really is a UDL suffix.
1652 We don't need to worry about UDLs defined inside namespace std because
1653 their names are reserved, so cannot be used as macro names in valid
1655 if (base
[0] == '_' && base
[1] != '_')
1657 return is_macro (pfile
, base
);
1660 /* Lexes a raw string. The stored string contains the spelling, including
1661 double quotes, delimiter string, '(' and ')', any leading
1662 'L', 'u', 'U' or 'u8' and 'R' modifier. It returns the type of the
1663 literal, or CPP_OTHER if it was not properly terminated.
1665 The spelling is NUL-terminated, but it is not guaranteed that this
1666 is the first NUL since embedded NULs are preserved. */
1669 lex_raw_string (cpp_reader
*pfile
, cpp_token
*token
, const uchar
*base
,
1672 uchar raw_prefix
[17];
1673 uchar temp_buffer
[18];
1674 const uchar
*orig_base
;
1675 unsigned int raw_prefix_len
= 0, raw_suffix_len
= 0;
1676 enum raw_str_phase
{ RAW_STR_PREFIX
, RAW_STR
, RAW_STR_SUFFIX
};
1677 raw_str_phase phase
= RAW_STR_PREFIX
;
1678 enum cpp_ttype type
;
1679 size_t total_len
= 0;
1680 /* Index into temp_buffer during phases other than RAW_STR,
1681 during RAW_STR phase 17 to tell BUF_APPEND that nothing should
1682 be appended to temp_buffer. */
1683 size_t temp_buffer_len
= 0;
1684 _cpp_buff
*first_buff
= NULL
, *last_buff
= NULL
;
1685 size_t raw_prefix_start
;
1686 _cpp_line_note
*note
= &pfile
->buffer
->notes
[pfile
->buffer
->cur_note
];
1688 type
= (*base
== 'L' ? CPP_WSTRING
:
1689 *base
== 'U' ? CPP_STRING32
:
1690 *base
== 'u' ? (base
[1] == '8' ? CPP_UTF8STRING
: CPP_STRING16
)
1693 #define BUF_APPEND(STR,LEN) \
1695 bufring_append (pfile, (const uchar *)(STR), (LEN), \
1696 &first_buff, &last_buff); \
1697 total_len += (LEN); \
1698 if (__builtin_expect (temp_buffer_len < 17, 0) \
1699 && (const uchar *)(STR) != base \
1702 memcpy (temp_buffer + temp_buffer_len, \
1703 (const uchar *)(STR), (LEN)); \
1704 temp_buffer_len += (LEN); \
1710 raw_prefix_start
= cur
- base
;
1715 /* If we previously performed any trigraph or line splicing
1716 transformations, undo them in between the opening and closing
1718 while (note
->pos
< cur
)
1720 for (; note
->pos
== cur
; ++note
)
1726 /* Restore backslash followed by newline. */
1727 BUF_APPEND (base
, cur
- base
);
1729 BUF_APPEND ("\\", 1);
1731 if (note
->type
== ' ')
1733 /* GNU backslash whitespace newline extension. FIXME
1734 could be any sequence of non-vertical space. When we
1735 can properly restore any such sequence, we should mark
1736 this note as handled so _cpp_process_line_notes
1738 BUF_APPEND (" ", 1);
1741 BUF_APPEND ("\n", 1);
1745 /* Already handled. */
1749 if (_cpp_trigraph_map
[note
->type
])
1751 /* Don't warn about this trigraph in
1752 _cpp_process_line_notes, since trigraphs show up as
1753 trigraphs in raw strings. */
1754 uchar type
= note
->type
;
1757 if (!CPP_OPTION (pfile
, trigraphs
))
1758 /* If we didn't convert the trigraph in the first
1759 place, don't do anything now either. */
1762 BUF_APPEND (base
, cur
- base
);
1764 BUF_APPEND ("??", 2);
1766 /* ??/ followed by newline gets two line notes, one for
1767 the trigraph and one for the backslash/newline. */
1768 if (type
== '/' && note
[1].pos
== cur
)
1770 if (note
[1].type
!= '\\'
1771 && note
[1].type
!= ' ')
1773 BUF_APPEND ("/", 1);
1775 goto after_backslash
;
1779 /* Skip the replacement character. */
1781 BUF_APPEND (&type
, 1);
1792 if (__builtin_expect (temp_buffer_len
< 17, 0))
1793 temp_buffer
[temp_buffer_len
++] = c
;
1796 if (phase
== RAW_STR_PREFIX
)
1798 while (raw_prefix_len
< temp_buffer_len
)
1800 raw_prefix
[raw_prefix_len
] = temp_buffer
[raw_prefix_len
];
1801 switch (raw_prefix
[raw_prefix_len
])
1803 case ' ': case '(': case ')': case '\\': case '\t':
1804 case '\v': case '\f': case '\n': default:
1806 /* Basic source charset except the above chars. */
1807 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
1808 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
1809 case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
1810 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
1812 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
1813 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
1814 case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
1815 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
1817 case '0': case '1': case '2': case '3': case '4': case '5':
1818 case '6': case '7': case '8': case '9':
1819 case '_': case '{': case '}': case '#': case '[': case ']':
1820 case '<': case '>': case '%': case ':': case ';': case '.':
1821 case '?': case '*': case '+': case '-': case '/': case '^':
1822 case '&': case '|': case '~': case '!': case '=': case ',':
1823 case '"': case '\'':
1824 if (raw_prefix_len
< 16)
1832 if (raw_prefix
[raw_prefix_len
] != '(')
1834 int col
= CPP_BUF_COLUMN (pfile
->buffer
, cur
) + 1;
1835 if (raw_prefix_len
== 16)
1836 cpp_error_with_line (pfile
, CPP_DL_ERROR
, token
->src_loc
,
1837 col
, "raw string delimiter longer "
1838 "than 16 characters");
1839 else if (raw_prefix
[raw_prefix_len
] == '\n')
1840 cpp_error_with_line (pfile
, CPP_DL_ERROR
, token
->src_loc
,
1841 col
, "invalid new-line in raw "
1842 "string delimiter");
1844 cpp_error_with_line (pfile
, CPP_DL_ERROR
, token
->src_loc
,
1845 col
, "invalid character '%c' in "
1846 "raw string delimiter",
1847 (int) raw_prefix
[raw_prefix_len
]);
1848 pfile
->buffer
->cur
= orig_base
+ raw_prefix_start
- 1;
1849 create_literal (pfile
, token
, orig_base
,
1850 raw_prefix_start
- 1, CPP_OTHER
);
1852 _cpp_release_buff (pfile
, first_buff
);
1855 raw_prefix
[raw_prefix_len
] = '"';
1857 /* Nothing should be appended to temp_buffer during
1859 temp_buffer_len
= 17;
1864 else if (phase
== RAW_STR_SUFFIX
)
1866 while (raw_suffix_len
<= raw_prefix_len
1867 && raw_suffix_len
< temp_buffer_len
1868 && temp_buffer
[raw_suffix_len
] == raw_prefix
[raw_suffix_len
])
1870 if (raw_suffix_len
> raw_prefix_len
)
1872 if (raw_suffix_len
== temp_buffer_len
)
1875 /* Nothing should be appended to temp_buffer during
1877 temp_buffer_len
= 17;
1881 phase
= RAW_STR_SUFFIX
;
1883 temp_buffer_len
= 0;
1887 if (pfile
->state
.in_directive
1888 || (pfile
->state
.parsing_args
1889 && pfile
->buffer
->next_line
>= pfile
->buffer
->rlimit
))
1893 cpp_error_with_line (pfile
, CPP_DL_ERROR
, token
->src_loc
, 0,
1894 "unterminated raw string");
1898 BUF_APPEND (base
, cur
- base
);
1900 if (pfile
->buffer
->cur
< pfile
->buffer
->rlimit
)
1901 CPP_INCREMENT_LINE (pfile
, 0);
1902 pfile
->buffer
->need_line
= true;
1904 pfile
->buffer
->cur
= cur
-1;
1905 _cpp_process_line_notes (pfile
, false);
1906 if (!_cpp_get_fresh_line (pfile
))
1908 location_t src_loc
= token
->src_loc
;
1909 token
->type
= CPP_EOF
;
1910 /* Tell the compiler the line number of the EOF token. */
1911 token
->src_loc
= pfile
->line_table
->highest_line
;
1913 if (first_buff
!= NULL
)
1914 _cpp_release_buff (pfile
, first_buff
);
1915 cpp_error_with_line (pfile
, CPP_DL_ERROR
, src_loc
, 0,
1916 "unterminated raw string");
1920 cur
= base
= pfile
->buffer
->cur
;
1921 note
= &pfile
->buffer
->notes
[pfile
->buffer
->cur_note
];
1925 if (CPP_OPTION (pfile
, user_literals
))
1927 /* If a string format macro, say from inttypes.h, is placed touching
1928 a string literal it could be parsed as a C++11 user-defined string
1929 literal thus breaking the program. */
1930 if (is_macro_not_literal_suffix (pfile
, cur
))
1932 /* Raise a warning, but do not consume subsequent tokens. */
1933 if (CPP_OPTION (pfile
, warn_literal_suffix
) && !pfile
->state
.skipping
)
1934 cpp_warning_with_line (pfile
, CPP_W_LITERAL_SUFFIX
,
1936 "invalid suffix on literal; C++11 requires "
1937 "a space between literal and string macro");
1939 /* Grab user defined literal suffix. */
1940 else if (ISIDST (*cur
))
1942 type
= cpp_userdef_string_add_type (type
);
1945 while (ISIDNUM (*cur
))
1950 pfile
->buffer
->cur
= cur
;
1951 if (first_buff
== NULL
)
1952 create_literal (pfile
, token
, base
, cur
- base
, type
);
1955 uchar
*dest
= _cpp_unaligned_alloc (pfile
, total_len
+ (cur
- base
) + 1);
1958 token
->val
.str
.len
= total_len
+ (cur
- base
);
1959 token
->val
.str
.text
= dest
;
1960 last_buff
= first_buff
;
1961 while (last_buff
!= NULL
)
1963 memcpy (dest
, last_buff
->base
,
1964 BUFF_FRONT (last_buff
) - last_buff
->base
);
1965 dest
+= BUFF_FRONT (last_buff
) - last_buff
->base
;
1966 last_buff
= last_buff
->next
;
1968 _cpp_release_buff (pfile
, first_buff
);
1969 memcpy (dest
, base
, cur
- base
);
1970 dest
[cur
- base
] = '\0';
1974 /* Lexes a string, character constant, or angle-bracketed header file
1975 name. The stored string contains the spelling, including opening
1976 quote and any leading 'L', 'u', 'U' or 'u8' and optional
1977 'R' modifier. It returns the type of the literal, or CPP_OTHER
1978 if it was not properly terminated, or CPP_LESS for an unterminated
1979 header name which must be relexed as normal tokens.
1981 The spelling is NUL-terminated, but it is not guaranteed that this
1982 is the first NUL since embedded NULs are preserved. */
1984 lex_string (cpp_reader
*pfile
, cpp_token
*token
, const uchar
*base
)
1986 bool saw_NUL
= false;
1988 cppchar_t terminator
;
1989 enum cpp_ttype type
;
1992 terminator
= *cur
++;
1993 if (terminator
== 'L' || terminator
== 'U')
1994 terminator
= *cur
++;
1995 else if (terminator
== 'u')
1997 terminator
= *cur
++;
1998 if (terminator
== '8')
1999 terminator
= *cur
++;
2001 if (terminator
== 'R')
2003 lex_raw_string (pfile
, token
, base
, cur
);
2006 if (terminator
== '"')
2007 type
= (*base
== 'L' ? CPP_WSTRING
:
2008 *base
== 'U' ? CPP_STRING32
:
2009 *base
== 'u' ? (base
[1] == '8' ? CPP_UTF8STRING
: CPP_STRING16
)
2011 else if (terminator
== '\'')
2012 type
= (*base
== 'L' ? CPP_WCHAR
:
2013 *base
== 'U' ? CPP_CHAR32
:
2014 *base
== 'u' ? (base
[1] == '8' ? CPP_UTF8CHAR
: CPP_CHAR16
)
2017 terminator
= '>', type
= CPP_HEADER_NAME
;
2021 cppchar_t c
= *cur
++;
2023 /* In #include-style directives, terminators are not escapable. */
2024 if (c
== '\\' && !pfile
->state
.angled_headers
&& *cur
!= '\n')
2026 else if (c
== terminator
)
2031 /* Unmatched quotes always yield undefined behavior, but
2032 greedy lexing means that what appears to be an unterminated
2033 header name may actually be a legitimate sequence of tokens. */
2034 if (terminator
== '>')
2036 token
->type
= CPP_LESS
;
2046 if (saw_NUL
&& !pfile
->state
.skipping
)
2047 cpp_error (pfile
, CPP_DL_WARNING
,
2048 "null character(s) preserved in literal");
2050 if (type
== CPP_OTHER
&& CPP_OPTION (pfile
, lang
) != CLK_ASM
)
2051 cpp_error (pfile
, CPP_DL_PEDWARN
, "missing terminating %c character",
2054 if (CPP_OPTION (pfile
, user_literals
))
2056 /* If a string format macro, say from inttypes.h, is placed touching
2057 a string literal it could be parsed as a C++11 user-defined string
2058 literal thus breaking the program. */
2059 if (is_macro_not_literal_suffix (pfile
, cur
))
2061 /* Raise a warning, but do not consume subsequent tokens. */
2062 if (CPP_OPTION (pfile
, warn_literal_suffix
) && !pfile
->state
.skipping
)
2063 cpp_warning_with_line (pfile
, CPP_W_LITERAL_SUFFIX
,
2065 "invalid suffix on literal; C++11 requires "
2066 "a space between literal and string macro");
2068 /* Grab user defined literal suffix. */
2069 else if (ISIDST (*cur
))
2071 type
= cpp_userdef_char_add_type (type
);
2072 type
= cpp_userdef_string_add_type (type
);
2075 while (ISIDNUM (*cur
))
2079 else if (CPP_OPTION (pfile
, cpp_warn_cxx11_compat
)
2080 && is_macro (pfile
, cur
)
2081 && !pfile
->state
.skipping
)
2082 cpp_warning_with_line (pfile
, CPP_W_CXX11_COMPAT
,
2083 token
->src_loc
, 0, "C++11 requires a space "
2084 "between string literal and macro");
2086 pfile
->buffer
->cur
= cur
;
2087 create_literal (pfile
, token
, base
, cur
- base
, type
);
2090 /* Return the comment table. The client may not make any assumption
2091 about the ordering of the table. */
2093 cpp_get_comments (cpp_reader
*pfile
)
2095 return &pfile
->comments
;
2098 /* Append a comment to the end of the comment table. */
2100 store_comment (cpp_reader
*pfile
, cpp_token
*token
)
2104 if (pfile
->comments
.allocated
== 0)
2106 pfile
->comments
.allocated
= 256;
2107 pfile
->comments
.entries
= (cpp_comment
*) xmalloc
2108 (pfile
->comments
.allocated
* sizeof (cpp_comment
));
2111 if (pfile
->comments
.count
== pfile
->comments
.allocated
)
2113 pfile
->comments
.allocated
*= 2;
2114 pfile
->comments
.entries
= (cpp_comment
*) xrealloc
2115 (pfile
->comments
.entries
,
2116 pfile
->comments
.allocated
* sizeof (cpp_comment
));
2119 len
= token
->val
.str
.len
;
2121 /* Copy comment. Note, token may not be NULL terminated. */
2122 pfile
->comments
.entries
[pfile
->comments
.count
].comment
=
2123 (char *) xmalloc (sizeof (char) * (len
+ 1));
2124 memcpy (pfile
->comments
.entries
[pfile
->comments
.count
].comment
,
2125 token
->val
.str
.text
, len
);
2126 pfile
->comments
.entries
[pfile
->comments
.count
].comment
[len
] = '\0';
2128 /* Set source location. */
2129 pfile
->comments
.entries
[pfile
->comments
.count
].sloc
= token
->src_loc
;
2131 /* Increment the count of entries in the comment table. */
2132 pfile
->comments
.count
++;
2135 /* The stored comment includes the comment start and any terminator. */
2137 save_comment (cpp_reader
*pfile
, cpp_token
*token
, const unsigned char *from
,
2140 unsigned char *buffer
;
2141 unsigned int len
, clen
, i
;
2143 len
= pfile
->buffer
->cur
- from
+ 1; /* + 1 for the initial '/'. */
2145 /* C++ comments probably (not definitely) have moved past a new
2146 line, which we don't want to save in the comment. */
2147 if (is_vspace (pfile
->buffer
->cur
[-1]))
2150 /* If we are currently in a directive or in argument parsing, then
2151 we need to store all C++ comments as C comments internally, and
2152 so we need to allocate a little extra space in that case.
2154 Note that the only time we encounter a directive here is
2155 when we are saving comments in a "#define". */
2156 clen
= ((pfile
->state
.in_directive
|| pfile
->state
.parsing_args
)
2157 && type
== '/') ? len
+ 2 : len
;
2159 buffer
= _cpp_unaligned_alloc (pfile
, clen
);
2161 token
->type
= CPP_COMMENT
;
2162 token
->val
.str
.len
= clen
;
2163 token
->val
.str
.text
= buffer
;
2166 memcpy (buffer
+ 1, from
, len
- 1);
2168 /* Finish conversion to a C comment, if necessary. */
2169 if ((pfile
->state
.in_directive
|| pfile
->state
.parsing_args
) && type
== '/')
2172 buffer
[clen
- 2] = '*';
2173 buffer
[clen
- 1] = '/';
2174 /* As there can be in a C++ comments illegal sequences for C comments
2175 we need to filter them out. */
2176 for (i
= 2; i
< (clen
- 2); i
++)
2177 if (buffer
[i
] == '/' && (buffer
[i
- 1] == '*' || buffer
[i
+ 1] == '*'))
2181 /* Finally store this comment for use by clients of libcpp. */
2182 store_comment (pfile
, token
);
2185 /* Returns true if comment at COMMENT_START is a recognized FALLTHROUGH
2189 fallthrough_comment_p (cpp_reader
*pfile
, const unsigned char *comment_start
)
2191 const unsigned char *from
= comment_start
+ 1;
2193 switch (CPP_OPTION (pfile
, cpp_warn_implicit_fallthrough
))
2195 /* For both -Wimplicit-fallthrough=0 and -Wimplicit-fallthrough=5 we
2196 don't recognize any comments. The latter only checks attributes,
2197 the former doesn't warn. */
2201 /* -Wimplicit-fallthrough=1 considers any comment, no matter what
2206 /* -Wimplicit-fallthrough=2 looks for (case insensitive)
2207 .*falls?[ \t-]*thr(u|ough).* regex. */
2208 for (; (size_t) (pfile
->buffer
->cur
- from
) >= sizeof "fallthru" - 1;
2211 /* Is there anything like strpbrk with upper boundary, or
2212 memchr looking for 2 characters rather than just one? */
2213 if (from
[0] != 'f' && from
[0] != 'F')
2215 if (from
[1] != 'a' && from
[1] != 'A')
2217 if (from
[2] != 'l' && from
[2] != 'L')
2219 if (from
[3] != 'l' && from
[3] != 'L')
2221 from
+= sizeof "fall" - 1;
2222 if (from
[0] == 's' || from
[0] == 'S')
2224 while (*from
== ' ' || *from
== '\t' || *from
== '-')
2226 if (from
[0] != 't' && from
[0] != 'T')
2228 if (from
[1] != 'h' && from
[1] != 'H')
2230 if (from
[2] != 'r' && from
[2] != 'R')
2232 if (from
[3] == 'u' || from
[3] == 'U')
2234 if (from
[3] != 'o' && from
[3] != 'O')
2236 if (from
[4] != 'u' && from
[4] != 'U')
2238 if (from
[5] != 'g' && from
[5] != 'G')
2240 if (from
[6] != 'h' && from
[6] != 'H')
2250 /* Whole comment contents:
2254 if (*from
== '-' || *from
== '@')
2256 size_t len
= sizeof "fallthrough" - 1;
2257 if ((size_t) (pfile
->buffer
->cur
- from
- 1) < len
)
2259 if (memcmp (from
+ 1, "fallthrough", len
))
2263 if (from
[len
+ 1] != '@')
2269 /* Whole comment contents (regex):
2270 lint -fallthrough[ \t]*
2272 else if (*from
== 'l')
2274 size_t len
= sizeof "int -fallthrough" - 1;
2275 if ((size_t) (pfile
->buffer
->cur
- from
- 1) < len
)
2277 if (memcmp (from
+ 1, "int -fallthrough", len
))
2280 while (*from
== ' ' || *from
== '\t')
2283 /* Whole comment contents (regex):
2284 [ \t]*FALLTHR(U|OUGH)[ \t]*
2286 else if (CPP_OPTION (pfile
, cpp_warn_implicit_fallthrough
) == 4)
2288 while (*from
== ' ' || *from
== '\t')
2290 if ((size_t) (pfile
->buffer
->cur
- from
) < sizeof "FALLTHRU" - 1)
2292 if (memcmp (from
, "FALLTHR", sizeof "FALLTHR" - 1))
2294 from
+= sizeof "FALLTHR" - 1;
2297 else if ((size_t) (pfile
->buffer
->cur
- from
) < sizeof "OUGH" - 1)
2299 else if (memcmp (from
, "OUGH", sizeof "OUGH" - 1))
2302 from
+= sizeof "OUGH" - 1;
2303 while (*from
== ' ' || *from
== '\t')
2306 /* Whole comment contents (regex):
2307 [ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?
2308 [ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?
2309 [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?
2313 while (*from
== ' ' || *from
== '\t' || *from
== '.' || *from
== '!')
2315 unsigned char f
= *from
;
2316 bool all_upper
= false;
2317 if (f
== 'E' || f
== 'e')
2319 if ((size_t) (pfile
->buffer
->cur
- from
)
2320 < sizeof "else fallthru" - 1)
2322 if (f
== 'E' && memcmp (from
+ 1, "LSE", sizeof "LSE" - 1) == 0)
2324 else if (memcmp (from
+ 1, "lse", sizeof "lse" - 1))
2326 from
+= sizeof "else" - 1;
2332 if (all_upper
&& *from
== 'f')
2334 if (f
== 'e' && *from
== 'F')
2338 else if (f
== 'I' || f
== 'i')
2340 if ((size_t) (pfile
->buffer
->cur
- from
)
2341 < sizeof "intentional fallthru" - 1)
2343 if (f
== 'I' && memcmp (from
+ 1, "NTENTIONAL",
2344 sizeof "NTENTIONAL" - 1) == 0)
2346 else if (memcmp (from
+ 1, "ntentional",
2347 sizeof "ntentional" - 1))
2349 from
+= sizeof "intentional" - 1;
2353 if (all_upper
&& *from
== 'f')
2358 if (memcmp (from
, "LY F", sizeof "LY F" - 1))
2360 from
+= sizeof "LY " - 1;
2364 if (memcmp (from
, "ly ", sizeof "ly " - 1))
2366 from
+= sizeof "ly " - 1;
2368 if (f
== 'i' && *from
== 'F')
2372 if (f
!= 'F' && f
!= 'f')
2374 if ((size_t) (pfile
->buffer
->cur
- from
) < sizeof "fallthru" - 1)
2376 if (f
== 'F' && memcmp (from
+ 1, "ALL", sizeof "ALL" - 1) == 0)
2380 else if (memcmp (from
+ 1, "all", sizeof "all" - 1))
2382 from
+= sizeof "fall" - 1;
2383 if (*from
== (all_upper
? 'S' : 's') && from
[1] == ' ')
2385 else if (*from
== ' ' || *from
== '-')
2387 else if (*from
!= (all_upper
? 'T' : 't'))
2389 if ((f
== 'f' || *from
!= 'T') && (all_upper
|| *from
!= 't'))
2391 if ((size_t) (pfile
->buffer
->cur
- from
) < sizeof "thru" - 1)
2393 if (memcmp (from
+ 1, all_upper
? "HRU" : "hru", sizeof "hru" - 1))
2395 if ((size_t) (pfile
->buffer
->cur
- from
) < sizeof "through" - 1)
2397 if (memcmp (from
+ 1, all_upper
? "HROUGH" : "hrough",
2398 sizeof "hrough" - 1))
2400 from
+= sizeof "through" - 1;
2403 from
+= sizeof "thru" - 1;
2404 while (*from
== ' ' || *from
== '\t' || *from
== '.' || *from
== '!')
2409 if (*comment_start
== '*')
2413 while (*from
&& *from
!= '*'
2414 && *from
!= '\n' && *from
!= '\r')
2416 if (*from
!= '*' || from
[1] == '/')
2423 while (*from
&& *from
!= '\n' && *from
!= '\r')
2427 /* C block comment. */
2428 if (*comment_start
== '*')
2430 if (*from
!= '*' || from
[1] != '/')
2433 /* C++ line comment. */
2434 else if (*from
!= '\n')
2440 /* Allocate COUNT tokens for RUN. */
2442 _cpp_init_tokenrun (tokenrun
*run
, unsigned int count
)
2444 run
->base
= XNEWVEC (cpp_token
, count
);
2445 run
->limit
= run
->base
+ count
;
2449 /* Returns the next tokenrun, or creates one if there is none. */
2451 next_tokenrun (tokenrun
*run
)
2453 if (run
->next
== NULL
)
2455 run
->next
= XNEW (tokenrun
);
2456 run
->next
->prev
= run
;
2457 _cpp_init_tokenrun (run
->next
, 250);
2463 /* Return the number of not yet processed token in a given
2466 _cpp_remaining_tokens_num_in_context (cpp_context
*context
)
2468 if (context
->tokens_kind
== TOKENS_KIND_DIRECT
)
2469 return (LAST (context
).token
- FIRST (context
).token
);
2470 else if (context
->tokens_kind
== TOKENS_KIND_INDIRECT
2471 || context
->tokens_kind
== TOKENS_KIND_EXTENDED
)
2472 return (LAST (context
).ptoken
- FIRST (context
).ptoken
);
2477 /* Returns the token present at index INDEX in a given context. If
2478 INDEX is zero, the next token to be processed is returned. */
2479 static const cpp_token
*
2480 _cpp_token_from_context_at (cpp_context
*context
, int index
)
2482 if (context
->tokens_kind
== TOKENS_KIND_DIRECT
)
2483 return &(FIRST (context
).token
[index
]);
2484 else if (context
->tokens_kind
== TOKENS_KIND_INDIRECT
2485 || context
->tokens_kind
== TOKENS_KIND_EXTENDED
)
2486 return FIRST (context
).ptoken
[index
];
2491 /* Look ahead in the input stream. */
2493 cpp_peek_token (cpp_reader
*pfile
, int index
)
2495 cpp_context
*context
= pfile
->context
;
2496 const cpp_token
*peektok
;
2499 /* First, scan through any pending cpp_context objects. */
2500 while (context
->prev
)
2502 ptrdiff_t sz
= _cpp_remaining_tokens_num_in_context (context
);
2504 if (index
< (int) sz
)
2505 return _cpp_token_from_context_at (context
, index
);
2507 context
= context
->prev
;
2510 /* We will have to read some new tokens after all (and do so
2511 without invalidating preceding tokens). */
2513 pfile
->keep_tokens
++;
2515 /* For peeked tokens temporarily disable line_change reporting,
2516 until the tokens are parsed for real. */
2517 void (*line_change
) (cpp_reader
*, const cpp_token
*, int)
2518 = pfile
->cb
.line_change
;
2519 pfile
->cb
.line_change
= NULL
;
2523 peektok
= _cpp_lex_token (pfile
);
2524 if (peektok
->type
== CPP_EOF
)
2532 _cpp_backup_tokens_direct (pfile
, count
- index
);
2533 pfile
->keep_tokens
--;
2534 pfile
->cb
.line_change
= line_change
;
2539 /* Allocate a single token that is invalidated at the same time as the
2540 rest of the tokens on the line. Has its line and col set to the
2541 same as the last lexed token, so that diagnostics appear in the
2544 _cpp_temp_token (cpp_reader
*pfile
)
2546 cpp_token
*old
, *result
;
2547 ptrdiff_t sz
= pfile
->cur_run
->limit
- pfile
->cur_token
;
2548 ptrdiff_t la
= (ptrdiff_t) pfile
->lookaheads
;
2550 old
= pfile
->cur_token
- 1;
2551 /* Any pre-existing lookaheads must not be clobbered. */
2556 tokenrun
*next
= next_tokenrun (pfile
->cur_run
);
2559 memmove (next
->base
+ 1, next
->base
,
2560 (la
- sz
) * sizeof (cpp_token
));
2562 next
->base
[0] = pfile
->cur_run
->limit
[-1];
2566 memmove (pfile
->cur_token
+ 1, pfile
->cur_token
,
2567 MIN (la
, sz
- 1) * sizeof (cpp_token
));
2570 if (!sz
&& pfile
->cur_token
== pfile
->cur_run
->limit
)
2572 pfile
->cur_run
= next_tokenrun (pfile
->cur_run
);
2573 pfile
->cur_token
= pfile
->cur_run
->base
;
2576 result
= pfile
->cur_token
++;
2577 result
->src_loc
= old
->src_loc
;
2581 /* Lex a token into RESULT (external interface). Takes care of issues
2582 like directive handling, token lookahead, multiple include
2583 optimization and skipping. */
2585 _cpp_lex_token (cpp_reader
*pfile
)
2591 if (pfile
->cur_token
== pfile
->cur_run
->limit
)
2593 pfile
->cur_run
= next_tokenrun (pfile
->cur_run
);
2594 pfile
->cur_token
= pfile
->cur_run
->base
;
2596 /* We assume that the current token is somewhere in the current
2598 if (pfile
->cur_token
< pfile
->cur_run
->base
2599 || pfile
->cur_token
>= pfile
->cur_run
->limit
)
2602 if (pfile
->lookaheads
)
2604 pfile
->lookaheads
--;
2605 result
= pfile
->cur_token
++;
2608 result
= _cpp_lex_direct (pfile
);
2610 if (result
->flags
& BOL
)
2612 /* Is this a directive. If _cpp_handle_directive returns
2613 false, it is an assembler #. */
2614 if (result
->type
== CPP_HASH
2615 /* 6.10.3 p 11: Directives in a list of macro arguments
2616 gives undefined behavior. This implementation
2617 handles the directive as normal. */
2618 && pfile
->state
.parsing_args
!= 1)
2620 if (_cpp_handle_directive (pfile
, result
->flags
& PREV_WHITE
))
2622 if (pfile
->directive_result
.type
== CPP_PADDING
)
2624 result
= &pfile
->directive_result
;
2627 else if (pfile
->state
.in_deferred_pragma
)
2628 result
= &pfile
->directive_result
;
2630 if (pfile
->cb
.line_change
&& !pfile
->state
.skipping
)
2631 pfile
->cb
.line_change (pfile
, result
, pfile
->state
.parsing_args
);
2634 /* We don't skip tokens in directives. */
2635 if (pfile
->state
.in_directive
|| pfile
->state
.in_deferred_pragma
)
2638 /* Outside a directive, invalidate controlling macros. At file
2639 EOF, _cpp_lex_direct takes care of popping the buffer, so we never
2640 get here and MI optimization works. */
2641 pfile
->mi_valid
= false;
2643 if (!pfile
->state
.skipping
|| result
->type
== CPP_EOF
)
2650 /* Returns true if a fresh line has been loaded. */
2652 _cpp_get_fresh_line (cpp_reader
*pfile
)
2656 /* We can't get a new line until we leave the current directive. */
2657 if (pfile
->state
.in_directive
)
2662 cpp_buffer
*buffer
= pfile
->buffer
;
2664 if (!buffer
->need_line
)
2667 if (buffer
->next_line
< buffer
->rlimit
)
2669 _cpp_clean_line (pfile
);
2673 /* First, get out of parsing arguments state. */
2674 if (pfile
->state
.parsing_args
)
2677 /* End of buffer. Non-empty files should end in a newline. */
2678 if (buffer
->buf
!= buffer
->rlimit
2679 && buffer
->next_line
> buffer
->rlimit
2680 && !buffer
->from_stage3
)
2682 /* Clip to buffer size. */
2683 buffer
->next_line
= buffer
->rlimit
;
2686 return_at_eof
= buffer
->return_at_eof
;
2687 _cpp_pop_buffer (pfile
);
2688 if (pfile
->buffer
== NULL
|| return_at_eof
)
2693 #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE) \
2696 result->type = ELSE_TYPE; \
2697 if (*buffer->cur == CHAR) \
2698 buffer->cur++, result->type = THEN_TYPE; \
2702 /* Lex a token into pfile->cur_token, which is also incremented, to
2703 get diagnostics pointing to the correct location.
2705 Does not handle issues such as token lookahead, multiple-include
2706 optimization, directives, skipping etc. This function is only
2707 suitable for use by _cpp_lex_token, and in special cases like
2708 lex_expansion_token which doesn't care for any of these issues.
2710 When meeting a newline, returns CPP_EOF if parsing a directive,
2711 otherwise returns to the start of the token buffer if permissible.
2712 Returns the location of the lexed token. */
2714 _cpp_lex_direct (cpp_reader
*pfile
)
2718 const unsigned char *comment_start
;
2719 bool fallthrough_comment
= false;
2720 cpp_token
*result
= pfile
->cur_token
++;
2724 buffer
= pfile
->buffer
;
2725 if (buffer
->need_line
)
2727 if (pfile
->state
.in_deferred_pragma
)
2729 result
->type
= CPP_PRAGMA_EOL
;
2730 pfile
->state
.in_deferred_pragma
= false;
2731 if (!pfile
->state
.pragma_allow_expansion
)
2732 pfile
->state
.prevent_expansion
--;
2735 if (!_cpp_get_fresh_line (pfile
))
2737 result
->type
= CPP_EOF
;
2738 if (!pfile
->state
.in_directive
)
2740 /* Tell the compiler the line number of the EOF token. */
2741 result
->src_loc
= pfile
->line_table
->highest_line
;
2742 result
->flags
= BOL
;
2746 if (buffer
!= pfile
->buffer
)
2747 fallthrough_comment
= false;
2748 if (!pfile
->keep_tokens
)
2750 pfile
->cur_run
= &pfile
->base_run
;
2751 result
= pfile
->base_run
.base
;
2752 pfile
->cur_token
= result
+ 1;
2754 result
->flags
= BOL
;
2755 if (pfile
->state
.parsing_args
== 2)
2756 result
->flags
|= PREV_WHITE
;
2758 buffer
= pfile
->buffer
;
2760 result
->src_loc
= pfile
->line_table
->highest_line
;
2763 if (buffer
->cur
>= buffer
->notes
[buffer
->cur_note
].pos
2764 && !pfile
->overlaid_buffer
)
2766 _cpp_process_line_notes (pfile
, false);
2767 result
->src_loc
= pfile
->line_table
->highest_line
;
2771 if (pfile
->forced_token_location
)
2772 result
->src_loc
= pfile
->forced_token_location
;
2774 result
->src_loc
= linemap_position_for_column (pfile
->line_table
,
2775 CPP_BUF_COLUMN (buffer
, buffer
->cur
));
2779 case ' ': case '\t': case '\f': case '\v': case '\0':
2780 result
->flags
|= PREV_WHITE
;
2781 skip_whitespace (pfile
, c
);
2785 /* Increment the line, unless this is the last line ... */
2786 if (buffer
->cur
< buffer
->rlimit
2787 /* ... or this is a #include, (where _cpp_stack_file needs to
2788 unwind by one line) ... */
2789 || (pfile
->state
.in_directive
> 1
2790 /* ... except traditional-cpp increments this elsewhere. */
2791 && !CPP_OPTION (pfile
, traditional
)))
2792 CPP_INCREMENT_LINE (pfile
, 0);
2793 buffer
->need_line
= true;
2796 case '0': case '1': case '2': case '3': case '4':
2797 case '5': case '6': case '7': case '8': case '9':
2799 struct normalize_state nst
= INITIAL_NORMALIZE_STATE
;
2800 result
->type
= CPP_NUMBER
;
2801 lex_number (pfile
, &result
->val
.str
, &nst
);
2802 warn_about_normalization (pfile
, result
, &nst
);
2810 /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters,
2811 wide strings or raw strings. */
2812 if (c
== 'L' || CPP_OPTION (pfile
, rliterals
)
2813 || (c
!= 'R' && CPP_OPTION (pfile
, uliterals
)))
2815 if ((*buffer
->cur
== '\'' && c
!= 'R')
2816 || *buffer
->cur
== '"'
2817 || (*buffer
->cur
== 'R'
2819 && buffer
->cur
[1] == '"'
2820 && CPP_OPTION (pfile
, rliterals
))
2821 || (*buffer
->cur
== '8'
2823 && ((buffer
->cur
[1] == '"' || (buffer
->cur
[1] == '\''
2824 && CPP_OPTION (pfile
, utf8_char_literals
)))
2825 || (buffer
->cur
[1] == 'R' && buffer
->cur
[2] == '"'
2826 && CPP_OPTION (pfile
, rliterals
)))))
2828 lex_string (pfile
, result
, buffer
->cur
- 1);
2835 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
2836 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
2837 case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
2838 case 's': case 't': case 'v': case 'w': case 'x':
2840 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
2841 case 'G': case 'H': case 'I': case 'J': case 'K':
2842 case 'M': case 'N': case 'O': case 'P': case 'Q':
2843 case 'S': case 'T': case 'V': case 'W': case 'X':
2845 result
->type
= CPP_NAME
;
2847 struct normalize_state nst
= INITIAL_NORMALIZE_STATE
;
2848 result
->val
.node
.node
= lex_identifier (pfile
, buffer
->cur
- 1, false,
2850 &result
->val
.node
.spelling
);
2851 warn_about_normalization (pfile
, result
, &nst
);
2854 /* Convert named operators to their proper types. */
2855 if (result
->val
.node
.node
->flags
& NODE_OPERATOR
)
2857 result
->flags
|= NAMED_OP
;
2858 result
->type
= (enum cpp_ttype
) result
->val
.node
.node
->directive_index
;
2861 /* Signal FALLTHROUGH comment followed by another token. */
2862 if (fallthrough_comment
)
2863 result
->flags
|= PREV_FALLTHROUGH
;
2868 lex_string (pfile
, result
, buffer
->cur
- 1);
2872 /* A potential block or line comment. */
2873 comment_start
= buffer
->cur
;
2878 if (_cpp_skip_block_comment (pfile
))
2879 cpp_error (pfile
, CPP_DL_ERROR
, "unterminated comment");
2881 else if (c
== '/' && ! CPP_OPTION (pfile
, traditional
))
2883 /* Don't warn for system headers. */
2884 if (cpp_in_system_header (pfile
))
2886 /* Warn about comments if pedantically GNUC89, and not
2887 in system headers. */
2888 else if (CPP_OPTION (pfile
, lang
) == CLK_GNUC89
2889 && CPP_PEDANTIC (pfile
)
2890 && ! buffer
->warned_cplusplus_comments
)
2892 if (cpp_error (pfile
, CPP_DL_PEDWARN
,
2893 "C++ style comments are not allowed in ISO C90"))
2894 cpp_error (pfile
, CPP_DL_NOTE
,
2895 "(this will be reported only once per input file)");
2896 buffer
->warned_cplusplus_comments
= 1;
2898 /* Or if specifically desired via -Wc90-c99-compat. */
2899 else if (CPP_OPTION (pfile
, cpp_warn_c90_c99_compat
) > 0
2900 && ! CPP_OPTION (pfile
, cplusplus
)
2901 && ! buffer
->warned_cplusplus_comments
)
2903 if (cpp_error (pfile
, CPP_DL_WARNING
,
2904 "C++ style comments are incompatible with C90"))
2905 cpp_error (pfile
, CPP_DL_NOTE
,
2906 "(this will be reported only once per input file)");
2907 buffer
->warned_cplusplus_comments
= 1;
2909 /* In C89/C94, C++ style comments are forbidden. */
2910 else if ((CPP_OPTION (pfile
, lang
) == CLK_STDC89
2911 || CPP_OPTION (pfile
, lang
) == CLK_STDC94
))
2913 /* But don't be confused about valid code such as
2914 - // immediately followed by *,
2915 - // in a preprocessing directive,
2916 - // in an #if 0 block. */
2917 if (buffer
->cur
[1] == '*'
2918 || pfile
->state
.in_directive
2919 || pfile
->state
.skipping
)
2921 result
->type
= CPP_DIV
;
2924 else if (! buffer
->warned_cplusplus_comments
)
2926 if (cpp_error (pfile
, CPP_DL_ERROR
,
2927 "C++ style comments are not allowed in "
2929 cpp_error (pfile
, CPP_DL_NOTE
,
2930 "(this will be reported only once per input "
2932 buffer
->warned_cplusplus_comments
= 1;
2935 if (skip_line_comment (pfile
) && CPP_OPTION (pfile
, warn_comments
))
2936 cpp_warning (pfile
, CPP_W_COMMENTS
, "multi-line comment");
2941 result
->type
= CPP_DIV_EQ
;
2946 result
->type
= CPP_DIV
;
2950 if (fallthrough_comment_p (pfile
, comment_start
))
2951 fallthrough_comment
= true;
2953 if (pfile
->cb
.comment
)
2955 size_t len
= pfile
->buffer
->cur
- comment_start
;
2956 pfile
->cb
.comment (pfile
, result
->src_loc
, comment_start
- 1,
2960 if (!pfile
->state
.save_comments
)
2962 result
->flags
|= PREV_WHITE
;
2963 goto update_tokens_line
;
2966 if (fallthrough_comment
)
2967 result
->flags
|= PREV_FALLTHROUGH
;
2969 /* Save the comment as a token in its own right. */
2970 save_comment (pfile
, result
, comment_start
, c
);
2974 if (pfile
->state
.angled_headers
)
2976 lex_string (pfile
, result
, buffer
->cur
- 1);
2977 if (result
->type
!= CPP_LESS
)
2981 result
->type
= CPP_LESS
;
2982 if (*buffer
->cur
== '=')
2984 buffer
->cur
++, result
->type
= CPP_LESS_EQ
;
2985 if (*buffer
->cur
== '>'
2986 && CPP_OPTION (pfile
, cplusplus
)
2987 && CPP_OPTION (pfile
, lang
) >= CLK_GNUCXX2A
)
2988 buffer
->cur
++, result
->type
= CPP_SPACESHIP
;
2990 else if (*buffer
->cur
== '<')
2993 IF_NEXT_IS ('=', CPP_LSHIFT_EQ
, CPP_LSHIFT
);
2995 else if (CPP_OPTION (pfile
, digraphs
))
2997 if (*buffer
->cur
== ':')
2999 /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next
3000 three characters are <:: and the subsequent character
3001 is neither : nor >, the < is treated as a preprocessor
3002 token by itself". */
3003 if (CPP_OPTION (pfile
, cplusplus
)
3004 && CPP_OPTION (pfile
, lang
) != CLK_CXX98
3005 && CPP_OPTION (pfile
, lang
) != CLK_GNUCXX
3006 && buffer
->cur
[1] == ':'
3007 && buffer
->cur
[2] != ':' && buffer
->cur
[2] != '>')
3011 result
->flags
|= DIGRAPH
;
3012 result
->type
= CPP_OPEN_SQUARE
;
3014 else if (*buffer
->cur
== '%')
3017 result
->flags
|= DIGRAPH
;
3018 result
->type
= CPP_OPEN_BRACE
;
3024 result
->type
= CPP_GREATER
;
3025 if (*buffer
->cur
== '=')
3026 buffer
->cur
++, result
->type
= CPP_GREATER_EQ
;
3027 else if (*buffer
->cur
== '>')
3030 IF_NEXT_IS ('=', CPP_RSHIFT_EQ
, CPP_RSHIFT
);
3035 result
->type
= CPP_MOD
;
3036 if (*buffer
->cur
== '=')
3037 buffer
->cur
++, result
->type
= CPP_MOD_EQ
;
3038 else if (CPP_OPTION (pfile
, digraphs
))
3040 if (*buffer
->cur
== ':')
3043 result
->flags
|= DIGRAPH
;
3044 result
->type
= CPP_HASH
;
3045 if (*buffer
->cur
== '%' && buffer
->cur
[1] == ':')
3046 buffer
->cur
+= 2, result
->type
= CPP_PASTE
, result
->val
.token_no
= 0;
3048 else if (*buffer
->cur
== '>')
3051 result
->flags
|= DIGRAPH
;
3052 result
->type
= CPP_CLOSE_BRACE
;
3058 result
->type
= CPP_DOT
;
3059 if (ISDIGIT (*buffer
->cur
))
3061 struct normalize_state nst
= INITIAL_NORMALIZE_STATE
;
3062 result
->type
= CPP_NUMBER
;
3063 lex_number (pfile
, &result
->val
.str
, &nst
);
3064 warn_about_normalization (pfile
, result
, &nst
);
3066 else if (*buffer
->cur
== '.' && buffer
->cur
[1] == '.')
3067 buffer
->cur
+= 2, result
->type
= CPP_ELLIPSIS
;
3068 else if (*buffer
->cur
== '*' && CPP_OPTION (pfile
, cplusplus
))
3069 buffer
->cur
++, result
->type
= CPP_DOT_STAR
;
3073 result
->type
= CPP_PLUS
;
3074 if (*buffer
->cur
== '+')
3075 buffer
->cur
++, result
->type
= CPP_PLUS_PLUS
;
3076 else if (*buffer
->cur
== '=')
3077 buffer
->cur
++, result
->type
= CPP_PLUS_EQ
;
3081 result
->type
= CPP_MINUS
;
3082 if (*buffer
->cur
== '>')
3085 result
->type
= CPP_DEREF
;
3086 if (*buffer
->cur
== '*' && CPP_OPTION (pfile
, cplusplus
))
3087 buffer
->cur
++, result
->type
= CPP_DEREF_STAR
;
3089 else if (*buffer
->cur
== '-')
3090 buffer
->cur
++, result
->type
= CPP_MINUS_MINUS
;
3091 else if (*buffer
->cur
== '=')
3092 buffer
->cur
++, result
->type
= CPP_MINUS_EQ
;
3096 result
->type
= CPP_AND
;
3097 if (*buffer
->cur
== '&')
3098 buffer
->cur
++, result
->type
= CPP_AND_AND
;
3099 else if (*buffer
->cur
== '=')
3100 buffer
->cur
++, result
->type
= CPP_AND_EQ
;
3104 result
->type
= CPP_OR
;
3105 if (*buffer
->cur
== '|')
3106 buffer
->cur
++, result
->type
= CPP_OR_OR
;
3107 else if (*buffer
->cur
== '=')
3108 buffer
->cur
++, result
->type
= CPP_OR_EQ
;
3112 result
->type
= CPP_COLON
;
3113 if (*buffer
->cur
== ':' && CPP_OPTION (pfile
, scope
))
3114 buffer
->cur
++, result
->type
= CPP_SCOPE
;
3115 else if (*buffer
->cur
== '>' && CPP_OPTION (pfile
, digraphs
))
3118 result
->flags
|= DIGRAPH
;
3119 result
->type
= CPP_CLOSE_SQUARE
;
3123 case '*': IF_NEXT_IS ('=', CPP_MULT_EQ
, CPP_MULT
); break;
3124 case '=': IF_NEXT_IS ('=', CPP_EQ_EQ
, CPP_EQ
); break;
3125 case '!': IF_NEXT_IS ('=', CPP_NOT_EQ
, CPP_NOT
); break;
3126 case '^': IF_NEXT_IS ('=', CPP_XOR_EQ
, CPP_XOR
); break;
3127 case '#': IF_NEXT_IS ('#', CPP_PASTE
, CPP_HASH
); result
->val
.token_no
= 0; break;
3129 case '?': result
->type
= CPP_QUERY
; break;
3130 case '~': result
->type
= CPP_COMPL
; break;
3131 case ',': result
->type
= CPP_COMMA
; break;
3132 case '(': result
->type
= CPP_OPEN_PAREN
; break;
3133 case ')': result
->type
= CPP_CLOSE_PAREN
; break;
3134 case '[': result
->type
= CPP_OPEN_SQUARE
; break;
3135 case ']': result
->type
= CPP_CLOSE_SQUARE
; break;
3136 case '{': result
->type
= CPP_OPEN_BRACE
; break;
3137 case '}': result
->type
= CPP_CLOSE_BRACE
; break;
3138 case ';': result
->type
= CPP_SEMICOLON
; break;
3140 /* @ is a punctuator in Objective-C. */
3141 case '@': result
->type
= CPP_ATSIGN
; break;
3145 const uchar
*base
= --buffer
->cur
;
3147 /* Check for an extended identifier ($ or UCN or UTF-8). */
3148 struct normalize_state nst
= INITIAL_NORMALIZE_STATE
;
3149 if (forms_identifier_p (pfile
, true, &nst
))
3151 result
->type
= CPP_NAME
;
3152 result
->val
.node
.node
= lex_identifier (pfile
, base
, true, &nst
,
3153 &result
->val
.node
.spelling
);
3154 warn_about_normalization (pfile
, result
, &nst
);
3158 /* Otherwise this will form a CPP_OTHER token. Parse valid UTF-8 as a
3161 if (c
>= utf8_signifier
)
3163 const uchar
*pstr
= base
;
3165 if (_cpp_valid_utf8 (pfile
, &pstr
, buffer
->rlimit
, 0, NULL
, &s
))
3168 create_literal (pfile
, result
, base
, buffer
->cur
- base
, CPP_OTHER
);
3174 /* Potentially convert the location of the token to a range. */
3175 if (result
->src_loc
>= RESERVED_LOCATION_COUNT
3176 && result
->type
!= CPP_EOF
)
3178 /* Ensure that any line notes are processed, so that we have the
3179 correct physical line/column for the end-point of the token even
3180 when a logical line is split via one or more backslashes. */
3181 if (buffer
->cur
>= buffer
->notes
[buffer
->cur_note
].pos
3182 && !pfile
->overlaid_buffer
)
3183 _cpp_process_line_notes (pfile
, false);
3185 source_range tok_range
;
3186 tok_range
.m_start
= result
->src_loc
;
3188 = linemap_position_for_column (pfile
->line_table
,
3189 CPP_BUF_COLUMN (buffer
, buffer
->cur
));
3191 result
->src_loc
= COMBINE_LOCATION_DATA (pfile
->line_table
,
3199 /* An upper bound on the number of bytes needed to spell TOKEN.
3200 Does not include preceding whitespace. */
3202 cpp_token_len (const cpp_token
*token
)
3206 switch (TOKEN_SPELL (token
))
3208 default: len
= 6; break;
3209 case SPELL_LITERAL
: len
= token
->val
.str
.len
; break;
3210 case SPELL_IDENT
: len
= NODE_LEN (token
->val
.node
.node
) * 10; break;
3216 /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER.
3217 Return the number of bytes read out of NAME. (There are always
3218 10 bytes written to BUFFER.) */
3221 utf8_to_ucn (unsigned char *buffer
, const unsigned char *name
)
3227 unsigned long utf32
;
3229 /* Compute the length of the UTF-8 sequence. */
3230 for (t
= *name
; t
& 0x80; t
<<= 1)
3233 utf32
= *name
& (0x7F >> ucn_len
);
3234 for (ucn_len_c
= 1; ucn_len_c
< ucn_len
; ucn_len_c
++)
3236 utf32
= (utf32
<< 6) | (*++name
& 0x3F);
3238 /* Ill-formed UTF-8. */
3239 if ((*name
& ~0x3F) != 0x80)
3245 for (j
= 7; j
>= 0; j
--)
3246 *buffer
++ = "0123456789abcdef"[(utf32
>> (4 * j
)) & 0xF];
3250 /* Given a token TYPE corresponding to a digraph, return a pointer to
3251 the spelling of the digraph. */
3252 static const unsigned char *
3253 cpp_digraph2name (enum cpp_ttype type
)
3255 return digraph_spellings
[(int) type
- (int) CPP_FIRST_DIGRAPH
];
3258 /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER.
3259 The buffer must already contain the enough space to hold the
3260 token's spelling. Returns a pointer to the character after the
3261 last character written. */
3263 _cpp_spell_ident_ucns (unsigned char *buffer
, cpp_hashnode
*ident
)
3266 const unsigned char *name
= NODE_NAME (ident
);
3268 for (i
= 0; i
< NODE_LEN (ident
); i
++)
3269 if (name
[i
] & ~0x7F)
3271 i
+= utf8_to_ucn (buffer
, name
+ i
) - 1;
3275 *buffer
++ = name
[i
];
3280 /* Write the spelling of a token TOKEN to BUFFER. The buffer must
3281 already contain the enough space to hold the token's spelling.
3282 Returns a pointer to the character after the last character written.
3283 FORSTRING is true if this is to be the spelling after translation
3284 phase 1 (with the original spelling of extended identifiers), false
3285 if extended identifiers should always be written using UCNs (there is
3286 no option for always writing them in the internal UTF-8 form).
3287 FIXME: Would be nice if we didn't need the PFILE argument. */
3289 cpp_spell_token (cpp_reader
*pfile
, const cpp_token
*token
,
3290 unsigned char *buffer
, bool forstring
)
3292 switch (TOKEN_SPELL (token
))
3294 case SPELL_OPERATOR
:
3296 const unsigned char *spelling
;
3299 if (token
->flags
& DIGRAPH
)
3300 spelling
= cpp_digraph2name (token
->type
);
3301 else if (token
->flags
& NAMED_OP
)
3304 spelling
= TOKEN_NAME (token
);
3306 while ((c
= *spelling
++) != '\0')
3315 memcpy (buffer
, NODE_NAME (token
->val
.node
.spelling
),
3316 NODE_LEN (token
->val
.node
.spelling
));
3317 buffer
+= NODE_LEN (token
->val
.node
.spelling
);
3320 buffer
= _cpp_spell_ident_ucns (buffer
, token
->val
.node
.node
);
3324 memcpy (buffer
, token
->val
.str
.text
, token
->val
.str
.len
);
3325 buffer
+= token
->val
.str
.len
;
3329 cpp_error (pfile
, CPP_DL_ICE
,
3330 "unspellable token %s", TOKEN_NAME (token
));
3337 /* Returns TOKEN spelt as a null-terminated string. The string is
3338 freed when the reader is destroyed. Useful for diagnostics. */
3340 cpp_token_as_text (cpp_reader
*pfile
, const cpp_token
*token
)
3342 unsigned int len
= cpp_token_len (token
) + 1;
3343 unsigned char *start
= _cpp_unaligned_alloc (pfile
, len
), *end
;
3345 end
= cpp_spell_token (pfile
, token
, start
, false);
3351 /* Returns a pointer to a string which spells the token defined by
3352 TYPE and FLAGS. Used by C front ends, which really should move to
3353 using cpp_token_as_text. */
3355 cpp_type2name (enum cpp_ttype type
, unsigned char flags
)
3357 if (flags
& DIGRAPH
)
3358 return (const char *) cpp_digraph2name (type
);
3359 else if (flags
& NAMED_OP
)
3360 return cpp_named_operator2name (type
);
3362 return (const char *) token_spellings
[type
].name
;
3365 /* Writes the spelling of token to FP, without any preceding space.
3366 Separated from cpp_spell_token for efficiency - to avoid stdio
3367 double-buffering. */
3369 cpp_output_token (const cpp_token
*token
, FILE *fp
)
3371 switch (TOKEN_SPELL (token
))
3373 case SPELL_OPERATOR
:
3375 const unsigned char *spelling
;
3378 if (token
->flags
& DIGRAPH
)
3379 spelling
= cpp_digraph2name (token
->type
);
3380 else if (token
->flags
& NAMED_OP
)
3383 spelling
= TOKEN_NAME (token
);
3388 while ((c
= *++spelling
) != '\0');
3396 const unsigned char * name
= NODE_NAME (token
->val
.node
.node
);
3398 for (i
= 0; i
< NODE_LEN (token
->val
.node
.node
); i
++)
3399 if (name
[i
] & ~0x7F)
3401 unsigned char buffer
[10];
3402 i
+= utf8_to_ucn (buffer
, name
+ i
) - 1;
3403 fwrite (buffer
, 1, 10, fp
);
3406 fputc (NODE_NAME (token
->val
.node
.node
)[i
], fp
);
3411 fwrite (token
->val
.str
.text
, 1, token
->val
.str
.len
, fp
);
3415 /* An error, most probably. */
3420 /* Compare two tokens. */
3422 _cpp_equiv_tokens (const cpp_token
*a
, const cpp_token
*b
)
3424 if (a
->type
== b
->type
&& a
->flags
== b
->flags
)
3425 switch (TOKEN_SPELL (a
))
3427 default: /* Keep compiler happy. */
3428 case SPELL_OPERATOR
:
3429 /* token_no is used to track where multiple consecutive ##
3430 tokens were originally located. */
3431 return (a
->type
!= CPP_PASTE
|| a
->val
.token_no
== b
->val
.token_no
);
3433 return (a
->type
!= CPP_MACRO_ARG
3434 || (a
->val
.macro_arg
.arg_no
== b
->val
.macro_arg
.arg_no
3435 && a
->val
.macro_arg
.spelling
== b
->val
.macro_arg
.spelling
));
3437 return (a
->val
.node
.node
== b
->val
.node
.node
3438 && a
->val
.node
.spelling
== b
->val
.node
.spelling
);
3440 return (a
->val
.str
.len
== b
->val
.str
.len
3441 && !memcmp (a
->val
.str
.text
, b
->val
.str
.text
,
3448 /* Returns nonzero if a space should be inserted to avoid an
3449 accidental token paste for output. For simplicity, it is
3450 conservative, and occasionally advises a space where one is not
3451 needed, e.g. "." and ".2". */
3453 cpp_avoid_paste (cpp_reader
*pfile
, const cpp_token
*token1
,
3454 const cpp_token
*token2
)
3456 enum cpp_ttype a
= token1
->type
, b
= token2
->type
;
3459 if (token1
->flags
& NAMED_OP
)
3461 if (token2
->flags
& NAMED_OP
)
3465 if (token2
->flags
& DIGRAPH
)
3466 c
= digraph_spellings
[(int) b
- (int) CPP_FIRST_DIGRAPH
][0];
3467 else if (token_spellings
[b
].category
== SPELL_OPERATOR
)
3468 c
= token_spellings
[b
].name
[0];
3470 /* Quickly get everything that can paste with an '='. */
3471 if ((int) a
<= (int) CPP_LAST_EQ
&& c
== '=')
3476 case CPP_GREATER
: return c
== '>';
3477 case CPP_LESS
: return c
== '<' || c
== '%' || c
== ':';
3478 case CPP_PLUS
: return c
== '+';
3479 case CPP_MINUS
: return c
== '-' || c
== '>';
3480 case CPP_DIV
: return c
== '/' || c
== '*'; /* Comments. */
3481 case CPP_MOD
: return c
== ':' || c
== '>';
3482 case CPP_AND
: return c
== '&';
3483 case CPP_OR
: return c
== '|';
3484 case CPP_COLON
: return c
== ':' || c
== '>';
3485 case CPP_DEREF
: return c
== '*';
3486 case CPP_DOT
: return c
== '.' || c
== '%' || b
== CPP_NUMBER
;
3487 case CPP_HASH
: return c
== '#' || c
== '%'; /* Digraph form. */
3488 case CPP_NAME
: return ((b
== CPP_NUMBER
3489 && name_p (pfile
, &token2
->val
.str
))
3491 || b
== CPP_CHAR
|| b
== CPP_STRING
); /* L */
3492 case CPP_NUMBER
: return (b
== CPP_NUMBER
|| b
== CPP_NAME
3493 || c
== '.' || c
== '+' || c
== '-');
3495 case CPP_OTHER
: return ((token1
->val
.str
.text
[0] == '\\'
3497 || (CPP_OPTION (pfile
, objc
)
3498 && token1
->val
.str
.text
[0] == '@'
3499 && (b
== CPP_NAME
|| b
== CPP_STRING
)));
3500 case CPP_LESS_EQ
: return c
== '>';
3503 case CPP_UTF8STRING
:
3505 case CPP_STRING32
: return (CPP_OPTION (pfile
, user_literals
)
3507 || (TOKEN_SPELL (token2
) == SPELL_LITERAL
3508 && ISIDST (token2
->val
.str
.text
[0]))));
3516 /* Output all the remaining tokens on the current line, and a newline
3517 character, to FP. Leading whitespace is removed. If there are
3518 macros, special token padding is not performed. */
3520 cpp_output_line (cpp_reader
*pfile
, FILE *fp
)
3522 const cpp_token
*token
;
3524 token
= cpp_get_token (pfile
);
3525 while (token
->type
!= CPP_EOF
)
3527 cpp_output_token (token
, fp
);
3528 token
= cpp_get_token (pfile
);
3529 if (token
->flags
& PREV_WHITE
)
3536 /* Return a string representation of all the remaining tokens on the
3537 current line. The result is allocated using xmalloc and must be
3538 freed by the caller. */
3540 cpp_output_line_to_string (cpp_reader
*pfile
, const unsigned char *dir_name
)
3542 const cpp_token
*token
;
3543 unsigned int out
= dir_name
? ustrlen (dir_name
) : 0;
3544 unsigned int alloced
= 120 + out
;
3545 unsigned char *result
= (unsigned char *) xmalloc (alloced
);
3547 /* If DIR_NAME is empty, there are no initial contents. */
3550 sprintf ((char *) result
, "#%s ", dir_name
);
3554 token
= cpp_get_token (pfile
);
3555 while (token
->type
!= CPP_EOF
)
3557 unsigned char *last
;
3558 /* Include room for a possible space and the terminating nul. */
3559 unsigned int len
= cpp_token_len (token
) + 2;
3561 if (out
+ len
> alloced
)
3564 if (out
+ len
> alloced
)
3565 alloced
= out
+ len
;
3566 result
= (unsigned char *) xrealloc (result
, alloced
);
3569 last
= cpp_spell_token (pfile
, token
, &result
[out
], 0);
3570 out
= last
- result
;
3572 token
= cpp_get_token (pfile
);
3573 if (token
->flags
& PREV_WHITE
)
3574 result
[out
++] = ' ';
3581 /* Memory buffers. Changing these three constants can have a dramatic
3582 effect on performance. The values here are reasonable defaults,
3583 but might be tuned. If you adjust them, be sure to test across a
3584 range of uses of cpplib, including heavy nested function-like macro
3585 expansion. Also check the change in peak memory usage (NJAMD is a
3586 good tool for this). */
3587 #define MIN_BUFF_SIZE 8000
3588 #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE) (MIN_BUFF_SIZE + (MIN_SIZE) * 3 / 2)
3589 #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA) \
3590 (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2)
3592 #if MIN_BUFF_SIZE > BUFF_SIZE_UPPER_BOUND (0)
3593 #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE!
3596 /* Create a new allocation buffer. Place the control block at the end
3597 of the buffer, so that buffer overflows will cause immediate chaos. */
3599 new_buff (size_t len
)
3602 unsigned char *base
;
3604 if (len
< MIN_BUFF_SIZE
)
3605 len
= MIN_BUFF_SIZE
;
3606 len
= CPP_ALIGN (len
);
3608 #ifdef ENABLE_VALGRIND_ANNOTATIONS
3609 /* Valgrind warns about uses of interior pointers, so put _cpp_buff
3611 size_t slen
= CPP_ALIGN2 (sizeof (_cpp_buff
), 2 * DEFAULT_ALIGNMENT
);
3612 base
= XNEWVEC (unsigned char, len
+ slen
);
3613 result
= (_cpp_buff
*) base
;
3616 base
= XNEWVEC (unsigned char, len
+ sizeof (_cpp_buff
));
3617 result
= (_cpp_buff
*) (base
+ len
);
3619 result
->base
= base
;
3621 result
->limit
= base
+ len
;
3622 result
->next
= NULL
;
3626 /* Place a chain of unwanted allocation buffers on the free list. */
3628 _cpp_release_buff (cpp_reader
*pfile
, _cpp_buff
*buff
)
3630 _cpp_buff
*end
= buff
;
3634 end
->next
= pfile
->free_buffs
;
3635 pfile
->free_buffs
= buff
;
3638 /* Return a free buffer of size at least MIN_SIZE. */
3640 _cpp_get_buff (cpp_reader
*pfile
, size_t min_size
)
3642 _cpp_buff
*result
, **p
;
3644 for (p
= &pfile
->free_buffs
;; p
= &(*p
)->next
)
3649 return new_buff (min_size
);
3651 size
= result
->limit
- result
->base
;
3652 /* Return a buffer that's big enough, but don't waste one that's
3654 if (size
>= min_size
&& size
<= BUFF_SIZE_UPPER_BOUND (min_size
))
3659 result
->next
= NULL
;
3660 result
->cur
= result
->base
;
3664 /* Creates a new buffer with enough space to hold the uncommitted
3665 remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies
3666 the excess bytes to the new buffer. Chains the new buffer after
3667 BUFF, and returns the new buffer. */
3669 _cpp_append_extend_buff (cpp_reader
*pfile
, _cpp_buff
*buff
, size_t min_extra
)
3671 size_t size
= EXTENDED_BUFF_SIZE (buff
, min_extra
);
3672 _cpp_buff
*new_buff
= _cpp_get_buff (pfile
, size
);
3674 buff
->next
= new_buff
;
3675 memcpy (new_buff
->base
, buff
->cur
, BUFF_ROOM (buff
));
3679 /* Creates a new buffer with enough space to hold the uncommitted
3680 remaining bytes of the buffer pointed to by BUFF, and at least
3681 MIN_EXTRA more bytes. Copies the excess bytes to the new buffer.
3682 Chains the new buffer before the buffer pointed to by BUFF, and
3683 updates the pointer to point to the new buffer. */
3685 _cpp_extend_buff (cpp_reader
*pfile
, _cpp_buff
**pbuff
, size_t min_extra
)
3687 _cpp_buff
*new_buff
, *old_buff
= *pbuff
;
3688 size_t size
= EXTENDED_BUFF_SIZE (old_buff
, min_extra
);
3690 new_buff
= _cpp_get_buff (pfile
, size
);
3691 memcpy (new_buff
->base
, old_buff
->cur
, BUFF_ROOM (old_buff
));
3692 new_buff
->next
= old_buff
;
3696 /* Free a chain of buffers starting at BUFF. */
3698 _cpp_free_buff (_cpp_buff
*buff
)
3702 for (; buff
; buff
= next
)
3705 #ifdef ENABLE_VALGRIND_ANNOTATIONS
3713 /* Allocate permanent, unaligned storage of length LEN. */
3715 _cpp_unaligned_alloc (cpp_reader
*pfile
, size_t len
)
3717 _cpp_buff
*buff
= pfile
->u_buff
;
3718 unsigned char *result
= buff
->cur
;
3720 if (len
> (size_t) (buff
->limit
- result
))
3722 buff
= _cpp_get_buff (pfile
, len
);
3723 buff
->next
= pfile
->u_buff
;
3724 pfile
->u_buff
= buff
;
3728 buff
->cur
= result
+ len
;
3732 /* Allocate permanent, unaligned storage of length LEN from a_buff.
3733 That buffer is used for growing allocations when saving macro
3734 replacement lists in a #define, and when parsing an answer to an
3735 assertion in #assert, #unassert or #if (and therefore possibly
3736 whilst expanding macros). It therefore must not be used by any
3737 code that they might call: specifically the lexer and the guts of
3740 All existing other uses clearly fit this restriction: storing
3741 registered pragmas during initialization. */
3743 _cpp_aligned_alloc (cpp_reader
*pfile
, size_t len
)
3745 _cpp_buff
*buff
= pfile
->a_buff
;
3746 unsigned char *result
= buff
->cur
;
3748 if (len
> (size_t) (buff
->limit
- result
))
3750 buff
= _cpp_get_buff (pfile
, len
);
3751 buff
->next
= pfile
->a_buff
;
3752 pfile
->a_buff
= buff
;
3756 buff
->cur
= result
+ len
;
3760 /* Commit or allocate storage from a buffer. */
3763 _cpp_commit_buff (cpp_reader
*pfile
, size_t size
)
3765 void *ptr
= BUFF_FRONT (pfile
->a_buff
);
3767 if (pfile
->hash_table
->alloc_subobject
)
3769 void *copy
= pfile
->hash_table
->alloc_subobject (size
);
3770 memcpy (copy
, ptr
, size
);
3774 BUFF_FRONT (pfile
->a_buff
) += size
;
3779 /* Say which field of TOK is in use. */
3781 enum cpp_token_fld_kind
3782 cpp_token_val_index (const cpp_token
*tok
)
3784 switch (TOKEN_SPELL (tok
))
3787 return CPP_TOKEN_FLD_NODE
;
3789 return CPP_TOKEN_FLD_STR
;
3790 case SPELL_OPERATOR
:
3791 /* Operands which were originally spelled as ident keep around
3792 the node for the exact spelling. */
3793 if (tok
->flags
& NAMED_OP
)
3794 return CPP_TOKEN_FLD_NODE
;
3795 else if (tok
->type
== CPP_PASTE
)
3796 return CPP_TOKEN_FLD_TOKEN_NO
;
3798 return CPP_TOKEN_FLD_NONE
;
3800 if (tok
->type
== CPP_MACRO_ARG
)
3801 return CPP_TOKEN_FLD_ARG_NO
;
3802 else if (tok
->type
== CPP_PADDING
)
3803 return CPP_TOKEN_FLD_SOURCE
;
3804 else if (tok
->type
== CPP_PRAGMA
)
3805 return CPP_TOKEN_FLD_PRAGMA
;
3808 return CPP_TOKEN_FLD_NONE
;
3812 /* All tokens lexed in R after calling this function will be forced to
3813 have their location_t to be P, until
3814 cpp_stop_forcing_token_locations is called for R. */
3817 cpp_force_token_locations (cpp_reader
*r
, location_t loc
)
3819 r
->forced_token_location
= loc
;
3822 /* Go back to assigning locations naturally for lexed tokens. */
3825 cpp_stop_forcing_token_locations (cpp_reader
*r
)
3827 r
->forced_token_location
= 0;