| File: | build/libcpp/lex.cc |
| Warning: | line 5563, column 8 Value stored to 'raw' is never read |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 1 | /* CPP Library - lexical analysis. |
| 2 | Copyright (C) 2000-2023 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 |
| 7 | |
| 8 | This program is free software; you can redistribute it and/or modify it |
| 9 | under the terms of the GNU General Public License as published by the |
| 10 | Free Software Foundation; either version 3, or (at your option) any |
| 11 | later version. |
| 12 | |
| 13 | This program is distributed in the hope that it will be useful, |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | GNU General Public License for more details. |
| 17 | |
| 18 | You should have received a copy of the GNU General Public License |
| 19 | along with this program; see the file COPYING3. If not see |
| 20 | <http://www.gnu.org/licenses/>. */ |
| 21 | |
| 22 | #include "config.h" |
| 23 | #include "system.h" |
| 24 | #include "cpplib.h" |
| 25 | #include "internal.h" |
| 26 | |
| 27 | enum spell_type |
| 28 | { |
| 29 | SPELL_OPERATOR = 0, |
| 30 | SPELL_IDENT, |
| 31 | SPELL_LITERAL, |
| 32 | SPELL_NONE |
| 33 | }; |
| 34 | |
| 35 | struct token_spelling |
| 36 | { |
| 37 | enum spell_type category; |
| 38 | const unsigned char *name; |
| 39 | }; |
| 40 | |
| 41 | static const unsigned char *const digraph_spellings[] = |
| 42 | { UC(const uchar *)"%:", UC(const uchar *)"%:%:", UC(const uchar *)"<:", UC(const uchar *)":>", UC(const uchar *)"<%", UC(const uchar *)"%>" }; |
| 43 | |
| 44 | #define OP(e, s) { SPELL_OPERATOR, UC(const uchar *) s }, |
| 45 | #define TK(e, s) { SPELL_ ## s, UC(const uchar *) #e }, |
| 46 | static const struct token_spelling token_spellings[N_TTYPES] = { TTYPE_TABLEOP(EQ, "=") OP(NOT, "!") OP(GREATER, ">") OP(LESS, "<") OP(PLUS, "+") OP(MINUS, "-") OP(MULT, "*") OP(DIV, "/") OP(MOD , "%") OP(AND, "&") OP(OR, "|") OP(XOR, "^") OP(RSHIFT, ">>" ) OP(LSHIFT, "<<") OP(COMPL, "~") OP(AND_AND, "&&" ) OP(OR_OR, "||") OP(QUERY, "?") OP(COLON, ":") OP(COMMA, "," ) OP(OPEN_PAREN, "(") OP(CLOSE_PAREN, ")") TK((-1), NONE) OP( EQ_EQ, "==") OP(NOT_EQ, "!=") OP(GREATER_EQ, ">=") OP(LESS_EQ , "<=") OP(SPACESHIP, "<=>") OP(PLUS_EQ, "+=") OP(MINUS_EQ , "-=") OP(MULT_EQ, "*=") OP(DIV_EQ, "/=") OP(MOD_EQ, "%=") OP (AND_EQ, "&=") OP(OR_EQ, "|=") OP(XOR_EQ, "^=") OP(RSHIFT_EQ , ">>=") OP(LSHIFT_EQ, "<<=") OP(HASH, "#") OP(PASTE , "##") OP(OPEN_SQUARE, "[") OP(CLOSE_SQUARE, "]") OP(OPEN_BRACE , "{") OP(CLOSE_BRACE, "}") OP(SEMICOLON, ";") OP(ELLIPSIS, "..." ) OP(PLUS_PLUS, "++") OP(MINUS_MINUS, "--") OP(DEREF, "->" ) OP(DOT, ".") OP(SCOPE, "::") OP(DEREF_STAR, "->*") OP(DOT_STAR , ".*") OP(ATSIGN, "@") TK(NAME, IDENT) TK(AT_NAME, IDENT) TK (NUMBER, LITERAL) TK(CHAR, LITERAL) TK(WCHAR, LITERAL) TK(CHAR16 , LITERAL) TK(CHAR32, LITERAL) TK(UTF8CHAR, LITERAL) TK(OTHER , LITERAL) TK(STRING, LITERAL) TK(WSTRING, LITERAL) TK(STRING16 , LITERAL) TK(STRING32, LITERAL) TK(UTF8STRING, LITERAL) TK(OBJC_STRING , LITERAL) TK(HEADER_NAME, LITERAL) TK(CHAR_USERDEF, LITERAL) TK(WCHAR_USERDEF, LITERAL) TK(CHAR16_USERDEF, LITERAL) TK(CHAR32_USERDEF , LITERAL) TK(UTF8CHAR_USERDEF, LITERAL) TK(STRING_USERDEF, LITERAL ) TK(WSTRING_USERDEF, LITERAL) TK(STRING16_USERDEF, LITERAL) TK (STRING32_USERDEF, LITERAL) TK(UTF8STRING_USERDEF,LITERAL) TK (COMMENT, LITERAL) TK(MACRO_ARG, NONE) TK(PRAGMA, NONE) TK(PRAGMA_EOL , NONE) TK(PADDING, NONE) }; |
| 47 | #undef OP |
| 48 | #undef TK |
| 49 | |
| 50 | #define TOKEN_SPELL(token)(token_spellings[(token)->type].category) (token_spellings[(token)->type].category) |
| 51 | #define TOKEN_NAME(token)(token_spellings[(token)->type].name) (token_spellings[(token)->type].name) |
| 52 | |
| 53 | /* ISO 10646 defines the UCS codespace as the range 0-0x10FFFF inclusive. */ |
| 54 | #define UCS_LIMIT0x10FFFF 0x10FFFF |
| 55 | |
| 56 | static void add_line_note (cpp_buffer *, const uchar *, unsigned int); |
| 57 | static int skip_line_comment (cpp_reader *); |
| 58 | static void skip_whitespace (cpp_reader *, cppchar_t); |
| 59 | static void lex_string (cpp_reader *, cpp_token *, const uchar *); |
| 60 | static void save_comment (cpp_reader *, cpp_token *, const uchar *, cppchar_t); |
| 61 | static void store_comment (cpp_reader *, cpp_token *); |
| 62 | static void create_literal (cpp_reader *, cpp_token *, const uchar *, |
| 63 | unsigned int, enum cpp_ttype); |
| 64 | static bool warn_in_comment (cpp_reader *, _cpp_line_note *); |
| 65 | static int name_p (cpp_reader *, const cpp_string *); |
| 66 | static tokenrun *next_tokenrun (tokenrun *); |
| 67 | |
| 68 | static _cpp_buff *new_buff (size_t); |
| 69 | |
| 70 | |
| 71 | /* Utility routine: |
| 72 | |
| 73 | Compares, the token TOKEN to the NUL-terminated string STRING. |
| 74 | TOKEN must be a CPP_NAME. Returns 1 for equal, 0 for unequal. */ |
| 75 | int |
| 76 | cpp_ideq (const cpp_token *token, const char *string) |
| 77 | { |
| 78 | if (token->type != CPP_NAME) |
| 79 | return 0; |
| 80 | |
| 81 | return !ustrcmp (NODE_NAME (token->val.node.node)(((&(token->val.node.node)->ident))->str), (const uchar *) string); |
| 82 | } |
| 83 | |
| 84 | /* Record a note TYPE at byte POS into the current cleaned logical |
| 85 | line. */ |
| 86 | static void |
| 87 | add_line_note (cpp_buffer *buffer, const uchar *pos, unsigned int type) |
| 88 | { |
| 89 | if (buffer->notes_used == buffer->notes_cap) |
| 90 | { |
| 91 | buffer->notes_cap = buffer->notes_cap * 2 + 200; |
| 92 | buffer->notes = XRESIZEVEC (_cpp_line_note, buffer->notes,((_cpp_line_note *) xrealloc ((void *) (buffer->notes), sizeof (_cpp_line_note) * (buffer->notes_cap))) |
| 93 | buffer->notes_cap)((_cpp_line_note *) xrealloc ((void *) (buffer->notes), sizeof (_cpp_line_note) * (buffer->notes_cap))); |
| 94 | } |
| 95 | |
| 96 | buffer->notes[buffer->notes_used].pos = pos; |
| 97 | buffer->notes[buffer->notes_used].type = type; |
| 98 | buffer->notes_used++; |
| 99 | } |
| 100 | |
| 101 | |
| 102 | /* Fast path to find line special characters using optimized character |
| 103 | scanning algorithms. Anything complicated falls back to the slow |
| 104 | path below. Since this loop is very hot it's worth doing these kinds |
| 105 | of optimizations. |
| 106 | |
| 107 | One of the paths through the ifdefs should provide |
| 108 | |
| 109 | const uchar *search_line_fast (const uchar *s, const uchar *end); |
| 110 | |
| 111 | Between S and END, search for \n, \r, \\, ?. Return a pointer to |
| 112 | the found character. |
| 113 | |
| 114 | Note that the last character of the buffer is *always* a newline, |
| 115 | as forced by _cpp_convert_input. This fact can be used to avoid |
| 116 | explicitly looking for the end of the buffer. */ |
| 117 | |
| 118 | /* Configure gives us an ifdef test. */ |
| 119 | #ifndef WORDS_BIGENDIAN0 |
| 120 | #define WORDS_BIGENDIAN0 0 |
| 121 | #endif |
| 122 | |
| 123 | /* We'd like the largest integer that fits into a register. There's nothing |
| 124 | in <stdint.h> that gives us that. For most hosts this is unsigned long, |
| 125 | but MS decided on an LLP64 model. Thankfully when building with GCC we |
| 126 | can get the "real" word size. */ |
| 127 | #ifdef __GNUC__4 |
| 128 | typedef unsigned int word_type __attribute__((__mode__(__word__))); |
| 129 | #else |
| 130 | typedef unsigned long word_type; |
| 131 | #endif |
| 132 | |
| 133 | /* The code below is only expecting sizes 4 or 8. |
| 134 | Die at compile-time if this expectation is violated. */ |
| 135 | typedef char check_word_type_size |
| 136 | [(sizeof(word_type) == 8 || sizeof(word_type) == 4) * 2 - 1]; |
| 137 | |
| 138 | /* Return X with the first N bytes forced to values that won't match one |
| 139 | of the interesting characters. Note that NUL is not interesting. */ |
| 140 | |
| 141 | static inline word_type |
| 142 | acc_char_mask_misalign (word_type val, unsigned int n) |
| 143 | { |
| 144 | word_type mask = -1; |
| 145 | if (WORDS_BIGENDIAN0) |
| 146 | mask >>= n * 8; |
| 147 | else |
| 148 | mask <<= n * 8; |
| 149 | return val & mask; |
| 150 | } |
| 151 | |
| 152 | /* Return X replicated to all byte positions within WORD_TYPE. */ |
| 153 | |
| 154 | static inline word_type |
| 155 | acc_char_replicate (uchar x) |
| 156 | { |
| 157 | word_type ret; |
| 158 | |
| 159 | ret = (x << 24) | (x << 16) | (x << 8) | x; |
| 160 | if (sizeof(word_type) == 8) |
| 161 | ret = (ret << 16 << 16) | ret; |
| 162 | return ret; |
| 163 | } |
| 164 | |
| 165 | /* Return non-zero if some byte of VAL is (probably) C. */ |
| 166 | |
| 167 | static inline word_type |
| 168 | acc_char_cmp (word_type val, word_type c) |
| 169 | { |
| 170 | #if defined(__GNUC__4) && defined(__alpha__) |
| 171 | /* We can get exact results using a compare-bytes instruction. |
| 172 | Get (val == c) via (0 >= (val ^ c)). */ |
| 173 | return __builtin_alpha_cmpbge (0, val ^ c); |
| 174 | #else |
| 175 | word_type magic = 0x7efefefeU; |
| 176 | if (sizeof(word_type) == 8) |
| 177 | magic = (magic << 16 << 16) | 0xfefefefeU; |
| 178 | magic |= 1; |
| 179 | |
| 180 | val ^= c; |
| 181 | return ((val + magic) ^ ~val) & ~magic; |
| 182 | #endif |
| 183 | } |
| 184 | |
| 185 | /* Given the result of acc_char_cmp is non-zero, return the index of |
| 186 | the found character. If this was a false positive, return -1. */ |
| 187 | |
| 188 | static inline int |
| 189 | acc_char_index (word_type cmp ATTRIBUTE_UNUSED__attribute__ ((__unused__)), |
| 190 | word_type val ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 191 | { |
| 192 | #if defined(__GNUC__4) && defined(__alpha__) && !WORDS_BIGENDIAN0 |
| 193 | /* The cmpbge instruction sets *bits* of the result corresponding to |
| 194 | matches in the bytes with no false positives. */ |
| 195 | return __builtin_ctzl (cmp); |
| 196 | #else |
| 197 | unsigned int i; |
| 198 | |
| 199 | /* ??? It would be nice to force unrolling here, |
| 200 | and have all of these constants folded. */ |
| 201 | for (i = 0; i < sizeof(word_type); ++i) |
| 202 | { |
| 203 | uchar c; |
| 204 | if (WORDS_BIGENDIAN0) |
| 205 | c = (val >> (sizeof(word_type) - i - 1) * 8) & 0xff; |
| 206 | else |
| 207 | c = (val >> i * 8) & 0xff; |
| 208 | |
| 209 | if (c == '\n' || c == '\r' || c == '\\' || c == '?') |
| 210 | return i; |
| 211 | } |
| 212 | |
| 213 | return -1; |
| 214 | #endif |
| 215 | } |
| 216 | |
| 217 | /* A version of the fast scanner using bit fiddling techniques. |
| 218 | |
| 219 | For 32-bit words, one would normally perform 16 comparisons and |
| 220 | 16 branches. With this algorithm one performs 24 arithmetic |
| 221 | operations and one branch. Whether this is faster with a 32-bit |
| 222 | word size is going to be somewhat system dependent. |
| 223 | |
| 224 | For 64-bit words, we eliminate twice the number of comparisons |
| 225 | and branches without increasing the number of arithmetic operations. |
| 226 | It's almost certainly going to be a win with 64-bit word size. */ |
| 227 | |
| 228 | static const uchar * search_line_acc_char (const uchar *, const uchar *) |
| 229 | ATTRIBUTE_UNUSED__attribute__ ((__unused__)); |
| 230 | |
| 231 | static const uchar * |
| 232 | search_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 233 | { |
| 234 | const word_type repl_nl = acc_char_replicate ('\n'); |
| 235 | const word_type repl_cr = acc_char_replicate ('\r'); |
| 236 | const word_type repl_bs = acc_char_replicate ('\\'); |
| 237 | const word_type repl_qm = acc_char_replicate ('?'); |
| 238 | |
| 239 | unsigned int misalign; |
| 240 | const word_type *p; |
| 241 | word_type val, t; |
| 242 | |
| 243 | /* Align the buffer. Mask out any bytes from before the beginning. */ |
| 244 | p = (word_type *)((uintptr_t)s & -sizeof(word_type)); |
| 245 | val = *p; |
| 246 | misalign = (uintptr_t)s & (sizeof(word_type) - 1); |
| 247 | if (misalign) |
| 248 | val = acc_char_mask_misalign (val, misalign); |
| 249 | |
| 250 | /* Main loop. */ |
| 251 | while (1) |
| 252 | { |
| 253 | t = acc_char_cmp (val, repl_nl); |
| 254 | t |= acc_char_cmp (val, repl_cr); |
| 255 | t |= acc_char_cmp (val, repl_bs); |
| 256 | t |= acc_char_cmp (val, repl_qm); |
| 257 | |
| 258 | if (__builtin_expect (t != 0, 0)) |
| 259 | { |
| 260 | int i = acc_char_index (t, val); |
| 261 | if (i >= 0) |
| 262 | return (const uchar *)p + i; |
| 263 | } |
| 264 | |
| 265 | val = *++p; |
| 266 | } |
| 267 | } |
| 268 | |
| 269 | /* Disable on Solaris 2/x86 until the following problem can be properly |
| 270 | autoconfed: |
| 271 | |
| 272 | The Solaris 10+ assembler tags objects with the instruction set |
| 273 | extensions used, so SSE4.2 executables cannot run on machines that |
| 274 | don't support that extension. */ |
| 275 | |
| 276 | #if (GCC_VERSION(4 * 1000 + 2) >= 4005) && (__GNUC__4 >= 5 || !defined(__PIC__2)) && (defined(__i386__) || defined(__x86_64__1)) && !(defined(__sun__) && defined(__svr4__)) |
| 277 | |
| 278 | /* Replicated character data to be shared between implementations. |
| 279 | Recall that outside of a context with vector support we can't |
| 280 | define compatible vector types, therefore these are all defined |
| 281 | in terms of raw characters. */ |
| 282 | static const char repl_chars[4][16] __attribute__((aligned(16))) = { |
| 283 | { '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', |
| 284 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' }, |
| 285 | { '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', |
| 286 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' }, |
| 287 | { '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', |
| 288 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' }, |
| 289 | { '?', '?', '?', '?', '?', '?', '?', '?', |
| 290 | '?', '?', '?', '?', '?', '?', '?', '?' }, |
| 291 | }; |
| 292 | |
| 293 | /* A version of the fast scanner using MMX vectorized byte compare insns. |
| 294 | |
| 295 | This uses the PMOVMSKB instruction which was introduced with "MMX2", |
| 296 | which was packaged into SSE1; it is also present in the AMD MMX |
| 297 | extension. Mark the function as using "sse" so that we emit a real |
| 298 | "emms" instruction, rather than the 3dNOW "femms" instruction. */ |
| 299 | |
| 300 | static const uchar * |
| 301 | #ifndef __SSE__1 |
| 302 | __attribute__((__target__("sse"))) |
| 303 | #endif |
| 304 | search_line_mmx (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 305 | { |
| 306 | typedef char v8qi __attribute__ ((__vector_size__ (8))); |
| 307 | typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__)); |
| 308 | |
| 309 | const v8qi repl_nl = *(const v8qi *)repl_chars[0]; |
| 310 | const v8qi repl_cr = *(const v8qi *)repl_chars[1]; |
| 311 | const v8qi repl_bs = *(const v8qi *)repl_chars[2]; |
| 312 | const v8qi repl_qm = *(const v8qi *)repl_chars[3]; |
| 313 | |
| 314 | unsigned int misalign, found, mask; |
| 315 | const v8qi *p; |
| 316 | v8qi data, t, c; |
| 317 | |
| 318 | /* Align the source pointer. While MMX doesn't generate unaligned data |
| 319 | faults, this allows us to safely scan to the end of the buffer without |
| 320 | reading beyond the end of the last page. */ |
| 321 | misalign = (uintptr_t)s & 7; |
| 322 | p = (const v8qi *)((uintptr_t)s & -8); |
| 323 | data = *p; |
| 324 | |
| 325 | /* Create a mask for the bytes that are valid within the first |
| 326 | 16-byte block. The Idea here is that the AND with the mask |
| 327 | within the loop is "free", since we need some AND or TEST |
| 328 | insn in order to set the flags for the branch anyway. */ |
| 329 | mask = -1u << misalign; |
| 330 | |
| 331 | /* Main loop processing 8 bytes at a time. */ |
| 332 | goto start; |
| 333 | do |
| 334 | { |
| 335 | data = *++p; |
| 336 | mask = -1; |
| 337 | |
| 338 | start: |
| 339 | t = __builtin_ia32_pcmpeqb(data, repl_nl); |
| 340 | c = __builtin_ia32_pcmpeqb(data, repl_cr); |
| 341 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); |
| 342 | c = __builtin_ia32_pcmpeqb(data, repl_bs); |
| 343 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); |
| 344 | c = __builtin_ia32_pcmpeqb(data, repl_qm); |
| 345 | t = (v8qi) __builtin_ia32_por ((__m64)t, (__m64)c); |
| 346 | found = __builtin_ia32_pmovmskb (t); |
| 347 | found &= mask; |
| 348 | } |
| 349 | while (!found); |
| 350 | |
| 351 | __builtin_ia32_emms (); |
| 352 | |
| 353 | /* FOUND contains 1 in bits for which we matched a relevant |
| 354 | character. Conversion to the byte index is trivial. */ |
| 355 | found = __builtin_ctz(found); |
| 356 | return (const uchar *)p + found; |
| 357 | } |
| 358 | |
| 359 | /* A version of the fast scanner using SSE2 vectorized byte compare insns. */ |
| 360 | |
| 361 | static const uchar * |
| 362 | #ifndef __SSE2__1 |
| 363 | __attribute__((__target__("sse2"))) |
| 364 | #endif |
| 365 | search_line_sse2 (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 366 | { |
| 367 | typedef char v16qi __attribute__ ((__vector_size__ (16))); |
| 368 | |
| 369 | const v16qi repl_nl = *(const v16qi *)repl_chars[0]; |
| 370 | const v16qi repl_cr = *(const v16qi *)repl_chars[1]; |
| 371 | const v16qi repl_bs = *(const v16qi *)repl_chars[2]; |
| 372 | const v16qi repl_qm = *(const v16qi *)repl_chars[3]; |
| 373 | |
| 374 | unsigned int misalign, found, mask; |
| 375 | const v16qi *p; |
| 376 | v16qi data, t; |
| 377 | |
| 378 | /* Align the source pointer. */ |
| 379 | misalign = (uintptr_t)s & 15; |
| 380 | p = (const v16qi *)((uintptr_t)s & -16); |
| 381 | data = *p; |
| 382 | |
| 383 | /* Create a mask for the bytes that are valid within the first |
| 384 | 16-byte block. The Idea here is that the AND with the mask |
| 385 | within the loop is "free", since we need some AND or TEST |
| 386 | insn in order to set the flags for the branch anyway. */ |
| 387 | mask = -1u << misalign; |
| 388 | |
| 389 | /* Main loop processing 16 bytes at a time. */ |
| 390 | goto start; |
| 391 | do |
| 392 | { |
| 393 | data = *++p; |
| 394 | mask = -1; |
| 395 | |
| 396 | start: |
| 397 | t = data == repl_nl; |
| 398 | t |= data == repl_cr; |
| 399 | t |= data == repl_bs; |
| 400 | t |= data == repl_qm; |
| 401 | found = __builtin_ia32_pmovmskb128 (t); |
| 402 | found &= mask; |
| 403 | } |
| 404 | while (!found); |
| 405 | |
| 406 | /* FOUND contains 1 in bits for which we matched a relevant |
| 407 | character. Conversion to the byte index is trivial. */ |
| 408 | found = __builtin_ctz(found); |
| 409 | return (const uchar *)p + found; |
| 410 | } |
| 411 | |
| 412 | #ifdef HAVE_SSE41 |
| 413 | /* A version of the fast scanner using SSE 4.2 vectorized string insns. */ |
| 414 | |
| 415 | static const uchar * |
| 416 | #ifndef __SSE4_2__ |
| 417 | __attribute__((__target__("sse4.2"))) |
| 418 | #endif |
| 419 | search_line_sse42 (const uchar *s, const uchar *end) |
| 420 | { |
| 421 | typedef char v16qi __attribute__ ((__vector_size__ (16))); |
| 422 | static const v16qi search = { '\n', '\r', '?', '\\' }; |
| 423 | |
| 424 | uintptr_t si = (uintptr_t)s; |
| 425 | uintptr_t index; |
| 426 | |
| 427 | /* Check for unaligned input. */ |
| 428 | if (si & 15) |
| 429 | { |
| 430 | v16qi sv; |
| 431 | |
| 432 | if (__builtin_expect (end - s < 16, 0) |
| 433 | && __builtin_expect ((si & 0xfff) > 0xff0, 0)) |
| 434 | { |
| 435 | /* There are less than 16 bytes left in the buffer, and less |
| 436 | than 16 bytes left on the page. Reading 16 bytes at this |
| 437 | point might generate a spurious page fault. Defer to the |
| 438 | SSE2 implementation, which already handles alignment. */ |
| 439 | return search_line_sse2 (s, end); |
| 440 | } |
| 441 | |
| 442 | /* ??? The builtin doesn't understand that the PCMPESTRI read from |
| 443 | memory need not be aligned. */ |
| 444 | sv = __builtin_ia32_loaddqu ((const char *) s); |
| 445 | index = __builtin_ia32_pcmpestri128 (search, 4, sv, 16, 0); |
| 446 | |
| 447 | if (__builtin_expect (index < 16, 0)) |
| 448 | goto found; |
| 449 | |
| 450 | /* Advance the pointer to an aligned address. We will re-scan a |
| 451 | few bytes, but we no longer need care for reading past the |
| 452 | end of a page, since we're guaranteed a match. */ |
| 453 | s = (const uchar *)((si + 15) & -16); |
| 454 | } |
| 455 | |
| 456 | /* Main loop, processing 16 bytes at a time. */ |
| 457 | #ifdef __GCC_ASM_FLAG_OUTPUTS__1 |
| 458 | while (1) |
| 459 | { |
| 460 | char f; |
| 461 | |
| 462 | /* By using inline assembly instead of the builtin, |
| 463 | we can use the result, as well as the flags set. */ |
| 464 | __asm ("%vpcmpestri\t$0, %2, %3" |
| 465 | : "=c"(index), "=@ccc"(f) |
| 466 | : "m"(*s), "x"(search), "a"(4), "d"(16)); |
| 467 | if (f) |
| 468 | break; |
| 469 | |
| 470 | s += 16; |
| 471 | } |
| 472 | #else |
| 473 | s -= 16; |
| 474 | /* By doing the whole loop in inline assembly, |
| 475 | we can make proper use of the flags set. */ |
| 476 | __asm ( ".balign 16\n" |
| 477 | "0: add $16, %1\n" |
| 478 | " %vpcmpestri\t$0, (%1), %2\n" |
| 479 | " jnc 0b" |
| 480 | : "=&c"(index), "+r"(s) |
| 481 | : "x"(search), "a"(4), "d"(16)); |
| 482 | #endif |
| 483 | |
| 484 | found: |
| 485 | return s + index; |
| 486 | } |
| 487 | |
| 488 | #else |
| 489 | /* Work around out-dated assemblers without sse4 support. */ |
| 490 | #define search_line_sse42 search_line_sse2 |
| 491 | #endif |
| 492 | |
| 493 | /* Check the CPU capabilities. */ |
| 494 | |
| 495 | #include "../gcc/config/i386/cpuid.h" |
| 496 | |
| 497 | typedef const uchar * (*search_line_fast_type) (const uchar *, const uchar *); |
| 498 | static search_line_fast_type search_line_fastsearch_line_acc_char; |
| 499 | |
| 500 | #define HAVE_init_vectorized_lexer 1 |
| 501 | static inline void |
| 502 | init_vectorized_lexer (void) |
| 503 | { |
| 504 | unsigned dummy, ecx = 0, edx = 0; |
| 505 | search_line_fast_type impl = search_line_acc_char; |
| 506 | int minimum = 0; |
| 507 | |
| 508 | #if defined(__SSE4_2__) |
| 509 | minimum = 3; |
| 510 | #elif defined(__SSE2__1) |
| 511 | minimum = 2; |
| 512 | #elif defined(__SSE__1) |
| 513 | minimum = 1; |
| 514 | #endif |
| 515 | |
| 516 | if (minimum == 3) |
| 517 | impl = search_line_sse42; |
| 518 | else if (__get_cpuid (1, &dummy, &dummy, &ecx, &edx) || minimum == 2) |
| 519 | { |
| 520 | if (minimum == 3 || (ecx & bit_SSE4_2)) |
| 521 | impl = search_line_sse42; |
| 522 | else if (minimum == 2 || (edx & bit_SSE2)) |
| 523 | impl = search_line_sse2; |
| 524 | else if (minimum == 1 || (edx & bit_SSE)) |
| 525 | impl = search_line_mmx; |
| 526 | } |
| 527 | else if (__get_cpuid (0x80000001, &dummy, &dummy, &dummy, &edx)) |
| 528 | { |
| 529 | if (minimum == 1 |
| 530 | || (edx & (bit_MMXEXT | bit_CMOV)) == (bit_MMXEXT | bit_CMOV)) |
| 531 | impl = search_line_mmx; |
| 532 | } |
| 533 | |
| 534 | search_line_fastsearch_line_acc_char = impl; |
| 535 | } |
| 536 | |
| 537 | #elif (GCC_VERSION(4 * 1000 + 2) >= 4005) && defined(_ARCH_PWR8) && defined(__ALTIVEC__) |
| 538 | |
| 539 | /* A vection of the fast scanner using AltiVec vectorized byte compares |
| 540 | and VSX unaligned loads (when VSX is available). This is otherwise |
| 541 | the same as the AltiVec version. */ |
| 542 | |
| 543 | ATTRIBUTE_NO_SANITIZE_UNDEFINED |
| 544 | static const uchar * |
| 545 | search_line_fastsearch_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 546 | { |
| 547 | typedef __attribute__((altivec(vector))) unsigned char vc; |
| 548 | |
| 549 | const vc repl_nl = { |
| 550 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', |
| 551 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' |
| 552 | }; |
| 553 | const vc repl_cr = { |
| 554 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', |
| 555 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' |
| 556 | }; |
| 557 | const vc repl_bs = { |
| 558 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', |
| 559 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' |
| 560 | }; |
| 561 | const vc repl_qm = { |
| 562 | '?', '?', '?', '?', '?', '?', '?', '?', |
| 563 | '?', '?', '?', '?', '?', '?', '?', '?', |
| 564 | }; |
| 565 | const vc zero = { 0 }; |
| 566 | |
| 567 | vc data, t; |
| 568 | |
| 569 | /* Main loop processing 16 bytes at a time. */ |
| 570 | do |
| 571 | { |
| 572 | vc m_nl, m_cr, m_bs, m_qm; |
| 573 | |
| 574 | data = __builtin_vec_vsx_ld (0, s); |
| 575 | s += 16; |
| 576 | |
| 577 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); |
| 578 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); |
| 579 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); |
| 580 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); |
| 581 | t = (m_nl | m_cr) | (m_bs | m_qm); |
| 582 | |
| 583 | /* T now contains 0xff in bytes for which we matched one of the relevant |
| 584 | characters. We want to exit the loop if any byte in T is non-zero. |
| 585 | Below is the expansion of vec_any_ne(t, zero). */ |
| 586 | } |
| 587 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); |
| 588 | |
| 589 | /* Restore s to to point to the 16 bytes we just processed. */ |
| 590 | s -= 16; |
| 591 | |
| 592 | { |
| 593 | #define N (sizeof(vc) / sizeof(long)) |
| 594 | |
| 595 | union { |
| 596 | vc v; |
| 597 | /* Statically assert that N is 2 or 4. */ |
| 598 | unsigned long l[(N == 2 || N == 4) ? N : -1]; |
| 599 | } u; |
| 600 | unsigned long l, i = 0; |
| 601 | |
| 602 | u.v = t; |
| 603 | |
| 604 | /* Find the first word of T that is non-zero. */ |
| 605 | switch (N) |
| 606 | { |
| 607 | case 4: |
| 608 | l = u.l[i++]; |
| 609 | if (l != 0) |
| 610 | break; |
| 611 | s += sizeof(unsigned long); |
| 612 | l = u.l[i++]; |
| 613 | if (l != 0) |
| 614 | break; |
| 615 | s += sizeof(unsigned long); |
| 616 | /* FALLTHRU */ |
| 617 | case 2: |
| 618 | l = u.l[i++]; |
| 619 | if (l != 0) |
| 620 | break; |
| 621 | s += sizeof(unsigned long); |
| 622 | l = u.l[i]; |
| 623 | } |
| 624 | |
| 625 | /* L now contains 0xff in bytes for which we matched one of the |
| 626 | relevant characters. We can find the byte index by finding |
| 627 | its bit index and dividing by 8. */ |
| 628 | #ifdef __BIG_ENDIAN__ |
| 629 | l = __builtin_clzl(l) >> 3; |
| 630 | #else |
| 631 | l = __builtin_ctzl(l) >> 3; |
| 632 | #endif |
| 633 | return s + l; |
| 634 | |
| 635 | #undef N |
| 636 | } |
| 637 | } |
| 638 | |
| 639 | #elif (GCC_VERSION(4 * 1000 + 2) >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__) |
| 640 | |
| 641 | /* A vection of the fast scanner using AltiVec vectorized byte compares. |
| 642 | This cannot be used for little endian because vec_lvsl/lvsr are |
| 643 | deprecated for little endian and the code won't work properly. */ |
| 644 | /* ??? Unfortunately, attribute(target("altivec")) is not yet supported, |
| 645 | so we can't compile this function without -maltivec on the command line |
| 646 | (or implied by some other switch). */ |
| 647 | |
| 648 | static const uchar * |
| 649 | search_line_fastsearch_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 650 | { |
| 651 | typedef __attribute__((altivec(vector))) unsigned char vc; |
| 652 | |
| 653 | const vc repl_nl = { |
| 654 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n', |
| 655 | '\n', '\n', '\n', '\n', '\n', '\n', '\n', '\n' |
| 656 | }; |
| 657 | const vc repl_cr = { |
| 658 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r', |
| 659 | '\r', '\r', '\r', '\r', '\r', '\r', '\r', '\r' |
| 660 | }; |
| 661 | const vc repl_bs = { |
| 662 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\', |
| 663 | '\\', '\\', '\\', '\\', '\\', '\\', '\\', '\\' |
| 664 | }; |
| 665 | const vc repl_qm = { |
| 666 | '?', '?', '?', '?', '?', '?', '?', '?', |
| 667 | '?', '?', '?', '?', '?', '?', '?', '?', |
| 668 | }; |
| 669 | const vc ones = { |
| 670 | -1, -1, -1, -1, -1, -1, -1, -1, |
| 671 | -1, -1, -1, -1, -1, -1, -1, -1, |
| 672 | }; |
| 673 | const vc zero = { 0 }; |
| 674 | |
| 675 | vc data, mask, t; |
| 676 | |
| 677 | /* Altivec loads automatically mask addresses with -16. This lets us |
| 678 | issue the first load as early as possible. */ |
| 679 | data = __builtin_vec_ld(0, (const vc *)s); |
| 680 | |
| 681 | /* Discard bytes before the beginning of the buffer. Do this by |
| 682 | beginning with all ones and shifting in zeros according to the |
| 683 | mis-alignment. The LVSR instruction pulls the exact shift we |
| 684 | want from the address. */ |
| 685 | mask = __builtin_vec_lvsr(0, s); |
| 686 | mask = __builtin_vec_perm(zero, ones, mask); |
| 687 | data &= mask; |
| 688 | |
| 689 | /* While altivec loads mask addresses, we still need to align S so |
| 690 | that the offset we compute at the end is correct. */ |
| 691 | s = (const uchar *)((uintptr_t)s & -16); |
| 692 | |
| 693 | /* Main loop processing 16 bytes at a time. */ |
| 694 | goto start; |
| 695 | do |
| 696 | { |
| 697 | vc m_nl, m_cr, m_bs, m_qm; |
| 698 | |
| 699 | s += 16; |
| 700 | data = __builtin_vec_ld(0, (const vc *)s); |
| 701 | |
| 702 | start: |
| 703 | m_nl = (vc) __builtin_vec_cmpeq(data, repl_nl); |
| 704 | m_cr = (vc) __builtin_vec_cmpeq(data, repl_cr); |
| 705 | m_bs = (vc) __builtin_vec_cmpeq(data, repl_bs); |
| 706 | m_qm = (vc) __builtin_vec_cmpeq(data, repl_qm); |
| 707 | t = (m_nl | m_cr) | (m_bs | m_qm); |
| 708 | |
| 709 | /* T now contains 0xff in bytes for which we matched one of the relevant |
| 710 | characters. We want to exit the loop if any byte in T is non-zero. |
| 711 | Below is the expansion of vec_any_ne(t, zero). */ |
| 712 | } |
| 713 | while (!__builtin_vec_vcmpeq_p(/*__CR6_LT_REV*/3, t, zero)); |
| 714 | |
| 715 | { |
| 716 | #define N (sizeof(vc) / sizeof(long)) |
| 717 | |
| 718 | union { |
| 719 | vc v; |
| 720 | /* Statically assert that N is 2 or 4. */ |
| 721 | unsigned long l[(N == 2 || N == 4) ? N : -1]; |
| 722 | } u; |
| 723 | unsigned long l, i = 0; |
| 724 | |
| 725 | u.v = t; |
| 726 | |
| 727 | /* Find the first word of T that is non-zero. */ |
| 728 | switch (N) |
| 729 | { |
| 730 | case 4: |
| 731 | l = u.l[i++]; |
| 732 | if (l != 0) |
| 733 | break; |
| 734 | s += sizeof(unsigned long); |
| 735 | l = u.l[i++]; |
| 736 | if (l != 0) |
| 737 | break; |
| 738 | s += sizeof(unsigned long); |
| 739 | /* FALLTHROUGH */ |
| 740 | case 2: |
| 741 | l = u.l[i++]; |
| 742 | if (l != 0) |
| 743 | break; |
| 744 | s += sizeof(unsigned long); |
| 745 | l = u.l[i]; |
| 746 | } |
| 747 | |
| 748 | /* L now contains 0xff in bytes for which we matched one of the |
| 749 | relevant characters. We can find the byte index by finding |
| 750 | its bit index and dividing by 8. */ |
| 751 | l = __builtin_clzl(l) >> 3; |
| 752 | return s + l; |
| 753 | |
| 754 | #undef N |
| 755 | } |
| 756 | } |
| 757 | |
| 758 | #elif defined (__ARM_NEON) && defined (__ARM_64BIT_STATE) |
| 759 | #include "arm_neon.h" |
| 760 | |
| 761 | /* This doesn't have to be the exact page size, but no system may use |
| 762 | a size smaller than this. ARMv8 requires a minimum page size of |
| 763 | 4k. The impact of being conservative here is a small number of |
| 764 | cases will take the slightly slower entry path into the main |
| 765 | loop. */ |
| 766 | |
| 767 | #define AARCH64_MIN_PAGE_SIZE 4096 |
| 768 | |
| 769 | static const uchar * |
| 770 | search_line_fastsearch_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 771 | { |
| 772 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); |
| 773 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); |
| 774 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); |
| 775 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); |
| 776 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); |
| 777 | |
| 778 | #ifdef __ARM_BIG_ENDIAN |
| 779 | const int16x8_t shift = {8, 8, 8, 8, 0, 0, 0, 0}; |
| 780 | #else |
| 781 | const int16x8_t shift = {0, 0, 0, 0, 8, 8, 8, 8}; |
| 782 | #endif |
| 783 | |
| 784 | unsigned int found; |
| 785 | const uint8_t *p; |
| 786 | uint8x16_t data; |
| 787 | uint8x16_t t; |
| 788 | uint16x8_t m; |
| 789 | uint8x16_t u, v, w; |
| 790 | |
| 791 | /* Align the source pointer. */ |
| 792 | p = (const uint8_t *)((uintptr_t)s & -16); |
| 793 | |
| 794 | /* Assuming random string start positions, with a 4k page size we'll take |
| 795 | the slow path about 0.37% of the time. */ |
| 796 | if (__builtin_expect ((AARCH64_MIN_PAGE_SIZE |
| 797 | - (((uintptr_t) s) & (AARCH64_MIN_PAGE_SIZE - 1))) |
| 798 | < 16, 0)) |
| 799 | { |
| 800 | /* Slow path: the string starts near a possible page boundary. */ |
| 801 | uint32_t misalign, mask; |
| 802 | |
| 803 | misalign = (uintptr_t)s & 15; |
| 804 | mask = (-1u << misalign) & 0xffff; |
| 805 | data = vld1q_u8 (p); |
| 806 | t = vceqq_u8 (data, repl_nl); |
| 807 | u = vceqq_u8 (data, repl_cr); |
| 808 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); |
| 809 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); |
| 810 | t = vorrq_u8 (v, w); |
| 811 | t = vandq_u8 (t, xmask); |
| 812 | m = vpaddlq_u8 (t); |
| 813 | m = vshlq_u16 (m, shift); |
| 814 | found = vaddvq_u16 (m); |
| 815 | found &= mask; |
| 816 | if (found) |
| 817 | return (const uchar*)p + __builtin_ctz (found); |
| 818 | } |
| 819 | else |
| 820 | { |
| 821 | data = vld1q_u8 ((const uint8_t *) s); |
| 822 | t = vceqq_u8 (data, repl_nl); |
| 823 | u = vceqq_u8 (data, repl_cr); |
| 824 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); |
| 825 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); |
| 826 | t = vorrq_u8 (v, w); |
| 827 | if (__builtin_expect (vpaddd_u64 ((uint64x2_t)t) != 0, 0)) |
| 828 | goto done; |
| 829 | } |
| 830 | |
| 831 | do |
| 832 | { |
| 833 | p += 16; |
| 834 | data = vld1q_u8 (p); |
| 835 | t = vceqq_u8 (data, repl_nl); |
| 836 | u = vceqq_u8 (data, repl_cr); |
| 837 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); |
| 838 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); |
| 839 | t = vorrq_u8 (v, w); |
| 840 | } while (!vpaddd_u64 ((uint64x2_t)t)); |
| 841 | |
| 842 | done: |
| 843 | /* Now that we've found the terminating substring, work out precisely where |
| 844 | we need to stop. */ |
| 845 | t = vandq_u8 (t, xmask); |
| 846 | m = vpaddlq_u8 (t); |
| 847 | m = vshlq_u16 (m, shift); |
| 848 | found = vaddvq_u16 (m); |
| 849 | return (((((uintptr_t) p) < (uintptr_t) s) ? s : (const uchar *)p) |
| 850 | + __builtin_ctz (found)); |
| 851 | } |
| 852 | |
| 853 | #elif defined (__ARM_NEON) |
| 854 | #include "arm_neon.h" |
| 855 | |
| 856 | static const uchar * |
| 857 | search_line_fastsearch_line_acc_char (const uchar *s, const uchar *end ATTRIBUTE_UNUSED__attribute__ ((__unused__))) |
| 858 | { |
| 859 | const uint8x16_t repl_nl = vdupq_n_u8 ('\n'); |
| 860 | const uint8x16_t repl_cr = vdupq_n_u8 ('\r'); |
| 861 | const uint8x16_t repl_bs = vdupq_n_u8 ('\\'); |
| 862 | const uint8x16_t repl_qm = vdupq_n_u8 ('?'); |
| 863 | const uint8x16_t xmask = (uint8x16_t) vdupq_n_u64 (0x8040201008040201ULL); |
| 864 | |
| 865 | unsigned int misalign, found, mask; |
| 866 | const uint8_t *p; |
| 867 | uint8x16_t data; |
| 868 | |
| 869 | /* Align the source pointer. */ |
| 870 | misalign = (uintptr_t)s & 15; |
| 871 | p = (const uint8_t *)((uintptr_t)s & -16); |
| 872 | data = vld1q_u8 (p); |
| 873 | |
| 874 | /* Create a mask for the bytes that are valid within the first |
| 875 | 16-byte block. The Idea here is that the AND with the mask |
| 876 | within the loop is "free", since we need some AND or TEST |
| 877 | insn in order to set the flags for the branch anyway. */ |
| 878 | mask = (-1u << misalign) & 0xffff; |
| 879 | |
| 880 | /* Main loop, processing 16 bytes at a time. */ |
| 881 | goto start; |
| 882 | |
| 883 | do |
| 884 | { |
| 885 | uint8x8_t l; |
| 886 | uint16x4_t m; |
| 887 | uint32x2_t n; |
| 888 | uint8x16_t t, u, v, w; |
| 889 | |
| 890 | p += 16; |
| 891 | data = vld1q_u8 (p); |
| 892 | mask = 0xffff; |
| 893 | |
| 894 | start: |
| 895 | t = vceqq_u8 (data, repl_nl); |
| 896 | u = vceqq_u8 (data, repl_cr); |
| 897 | v = vorrq_u8 (t, vceqq_u8 (data, repl_bs)); |
| 898 | w = vorrq_u8 (u, vceqq_u8 (data, repl_qm)); |
| 899 | t = vandq_u8 (vorrq_u8 (v, w), xmask); |
| 900 | l = vpadd_u8 (vget_low_u8 (t), vget_high_u8 (t)); |
| 901 | m = vpaddl_u8 (l); |
| 902 | n = vpaddl_u16 (m); |
| 903 | |
| 904 | found = vget_lane_u32 ((uint32x2_t) vorr_u64 ((uint64x1_t) n, |
| 905 | vshr_n_u64 ((uint64x1_t) n, 24)), 0); |
| 906 | found &= mask; |
| 907 | } |
| 908 | while (!found); |
| 909 | |
| 910 | /* FOUND contains 1 in bits for which we matched a relevant |
| 911 | character. Conversion to the byte index is trivial. */ |
| 912 | found = __builtin_ctz (found); |
| 913 | return (const uchar *)p + found; |
| 914 | } |
| 915 | |
| 916 | #else |
| 917 | |
| 918 | /* We only have one accelerated alternative. Use a direct call so that |
| 919 | we encourage inlining. */ |
| 920 | |
| 921 | #define search_line_fastsearch_line_acc_char search_line_acc_char |
| 922 | |
| 923 | #endif |
| 924 | |
| 925 | /* Initialize the lexer if needed. */ |
| 926 | |
| 927 | void |
| 928 | _cpp_init_lexer (void) |
| 929 | { |
| 930 | #ifdef HAVE_init_vectorized_lexer |
| 931 | init_vectorized_lexer (); |
| 932 | #endif |
| 933 | } |
| 934 | |
| 935 | /* Returns with a logical line that contains no escaped newlines or |
| 936 | trigraphs. This is a time-critical inner loop. */ |
| 937 | void |
| 938 | _cpp_clean_line (cpp_reader *pfile) |
| 939 | { |
| 940 | cpp_buffer *buffer; |
| 941 | const uchar *s; |
| 942 | uchar c, *d, *p; |
| 943 | |
| 944 | buffer = pfile->buffer; |
| 945 | buffer->cur_note = buffer->notes_used = 0; |
| 946 | buffer->cur = buffer->line_base = buffer->next_line; |
| 947 | buffer->need_line = false; |
| 948 | s = buffer->next_line; |
| 949 | |
| 950 | if (!buffer->from_stage3) |
| 951 | { |
| 952 | const uchar *pbackslash = NULL__null; |
| 953 | |
| 954 | /* Fast path. This is the common case of an un-escaped line with |
| 955 | no trigraphs. The primary win here is by not writing any |
| 956 | data back to memory until we have to. */ |
| 957 | while (1) |
| 958 | { |
| 959 | /* Perform an optimized search for \n, \r, \\, ?. */ |
| 960 | s = search_line_fastsearch_line_acc_char (s, buffer->rlimit); |
| 961 | |
| 962 | c = *s; |
| 963 | if (c == '\\') |
| 964 | { |
| 965 | /* Record the location of the backslash and continue. */ |
| 966 | pbackslash = s++; |
| 967 | } |
| 968 | else if (__builtin_expect (c == '?', 0)) |
| 969 | { |
| 970 | if (__builtin_expect (s[1] == '?', false) |
| 971 | && _cpp_trigraph_map[s[2]]) |
| 972 | { |
| 973 | /* Have a trigraph. We may or may not have to convert |
| 974 | it. Add a line note regardless, for -Wtrigraphs. */ |
| 975 | add_line_note (buffer, s, s[2]); |
| 976 | if (CPP_OPTION (pfile, trigraphs)((pfile)->opts.trigraphs)) |
| 977 | { |
| 978 | /* We do, and that means we have to switch to the |
| 979 | slow path. */ |
| 980 | d = (uchar *) s; |
| 981 | *d = _cpp_trigraph_map[s[2]]; |
| 982 | s += 2; |
| 983 | goto slow_path; |
| 984 | } |
| 985 | } |
| 986 | /* Not a trigraph. Continue on fast-path. */ |
| 987 | s++; |
| 988 | } |
| 989 | else |
| 990 | break; |
| 991 | } |
| 992 | |
| 993 | /* This must be \r or \n. We're either done, or we'll be forced |
| 994 | to write back to the buffer and continue on the slow path. */ |
| 995 | d = (uchar *) s; |
| 996 | |
| 997 | if (__builtin_expect (s == buffer->rlimit, false)) |
| 998 | goto done; |
| 999 | |
| 1000 | /* DOS line ending? */ |
| 1001 | if (__builtin_expect (c == '\r', false) && s[1] == '\n') |
| 1002 | { |
| 1003 | s++; |
| 1004 | if (s == buffer->rlimit) |
| 1005 | goto done; |
| 1006 | } |
| 1007 | |
| 1008 | if (__builtin_expect (pbackslash == NULL__null, true)) |
| 1009 | goto done; |
| 1010 | |
| 1011 | /* Check for escaped newline. */ |
| 1012 | p = d; |
| 1013 | while (is_nvspace (p[-1])(_sch_istable[(p[-1]) & 0xff] & (unsigned short)(_sch_isnvsp ))) |
| 1014 | p--; |
| 1015 | if (p - 1 != pbackslash) |
| 1016 | goto done; |
| 1017 | |
| 1018 | /* Have an escaped newline; process it and proceed to |
| 1019 | the slow path. */ |
| 1020 | add_line_note (buffer, p - 1, p != d ? ' ' : '\\'); |
| 1021 | d = p - 2; |
| 1022 | buffer->next_line = p - 1; |
| 1023 | |
| 1024 | slow_path: |
| 1025 | while (1) |
| 1026 | { |
| 1027 | c = *++s; |
| 1028 | *++d = c; |
| 1029 | |
| 1030 | if (c == '\n' || c == '\r') |
| 1031 | { |
| 1032 | /* Handle DOS line endings. */ |
| 1033 | if (c == '\r' && s != buffer->rlimit && s[1] == '\n') |
| 1034 | s++; |
| 1035 | if (s == buffer->rlimit) |
| 1036 | break; |
| 1037 | |
| 1038 | /* Escaped? */ |
| 1039 | p = d; |
| 1040 | while (p != buffer->next_line && is_nvspace (p[-1])(_sch_istable[(p[-1]) & 0xff] & (unsigned short)(_sch_isnvsp ))) |
| 1041 | p--; |
| 1042 | if (p == buffer->next_line || p[-1] != '\\') |
| 1043 | break; |
| 1044 | |
| 1045 | add_line_note (buffer, p - 1, p != d ? ' ': '\\'); |
| 1046 | d = p - 2; |
| 1047 | buffer->next_line = p - 1; |
| 1048 | } |
| 1049 | else if (c == '?' && s[1] == '?' && _cpp_trigraph_map[s[2]]) |
| 1050 | { |
| 1051 | /* Add a note regardless, for the benefit of -Wtrigraphs. */ |
| 1052 | add_line_note (buffer, d, s[2]); |
| 1053 | if (CPP_OPTION (pfile, trigraphs)((pfile)->opts.trigraphs)) |
| 1054 | { |
| 1055 | *d = _cpp_trigraph_map[s[2]]; |
| 1056 | s += 2; |
| 1057 | } |
| 1058 | } |
| 1059 | } |
| 1060 | } |
| 1061 | else |
| 1062 | { |
| 1063 | while (*s != '\n' && *s != '\r') |
| 1064 | s++; |
| 1065 | d = (uchar *) s; |
| 1066 | |
| 1067 | /* Handle DOS line endings. */ |
| 1068 | if (*s == '\r' && s + 1 != buffer->rlimit && s[1] == '\n') |
| 1069 | s++; |
| 1070 | } |
| 1071 | |
| 1072 | done: |
| 1073 | *d = '\n'; |
| 1074 | /* A sentinel note that should never be processed. */ |
| 1075 | add_line_note (buffer, d + 1, '\n'); |
| 1076 | buffer->next_line = s + 1; |
| 1077 | } |
| 1078 | |
| 1079 | template <bool lexing_raw_string> |
| 1080 | static bool get_fresh_line_impl (cpp_reader *pfile); |
| 1081 | |
| 1082 | /* Return true if the trigraph indicated by NOTE should be warned |
| 1083 | about in a comment. */ |
| 1084 | static bool |
| 1085 | warn_in_comment (cpp_reader *pfile, _cpp_line_note *note) |
| 1086 | { |
| 1087 | const uchar *p; |
| 1088 | |
| 1089 | /* Within comments we don't warn about trigraphs, unless the |
| 1090 | trigraph forms an escaped newline, as that may change |
| 1091 | behavior. */ |
| 1092 | if (note->type != '/') |
| 1093 | return false; |
| 1094 | |
| 1095 | /* If -trigraphs, then this was an escaped newline iff the next note |
| 1096 | is coincident. */ |
| 1097 | if (CPP_OPTION (pfile, trigraphs)((pfile)->opts.trigraphs)) |
| 1098 | return note[1].pos == note->pos; |
| 1099 | |
| 1100 | /* Otherwise, see if this forms an escaped newline. */ |
| 1101 | p = note->pos + 3; |
| 1102 | while (is_nvspace (*p)(_sch_istable[(*p) & 0xff] & (unsigned short)(_sch_isnvsp ))) |
| 1103 | p++; |
| 1104 | |
| 1105 | /* There might have been escaped newlines between the trigraph and the |
| 1106 | newline we found. Hence the position test. */ |
| 1107 | return (*p == '\n' && p < note[1].pos); |
| 1108 | } |
| 1109 | |
| 1110 | /* Process the notes created by add_line_note as far as the current |
| 1111 | location. */ |
| 1112 | void |
| 1113 | _cpp_process_line_notes (cpp_reader *pfile, int in_comment) |
| 1114 | { |
| 1115 | cpp_buffer *buffer = pfile->buffer; |
| 1116 | |
| 1117 | for (;;) |
| 1118 | { |
| 1119 | _cpp_line_note *note = &buffer->notes[buffer->cur_note]; |
| 1120 | unsigned int col; |
| 1121 | |
| 1122 | if (note->pos > buffer->cur) |
| 1123 | break; |
| 1124 | |
| 1125 | buffer->cur_note++; |
| 1126 | col = CPP_BUF_COLUMN (buffer, note->pos + 1)((note->pos + 1) - (buffer)->line_base); |
| 1127 | |
| 1128 | if (note->type == '\\' || note->type == ' ') |
| 1129 | { |
| 1130 | if (note->type == ' ' && !in_comment) |
| 1131 | cpp_error_with_line (pfile, CPP_DL_WARNING, pfile->line_table->highest_line, col, |
| 1132 | "backslash and newline separated by space"); |
| 1133 | |
| 1134 | if (buffer->next_line > buffer->rlimit) |
| 1135 | { |
| 1136 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, col, |
| 1137 | "backslash-newline at end of file"); |
| 1138 | /* Prevent "no newline at end of file" warning. */ |
| 1139 | buffer->next_line = buffer->rlimit; |
| 1140 | } |
| 1141 | |
| 1142 | buffer->line_base = note->pos; |
| 1143 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 1144 | } |
| 1145 | else if (_cpp_trigraph_map[note->type]) |
| 1146 | { |
| 1147 | if (CPP_OPTION (pfile, warn_trigraphs)((pfile)->opts.warn_trigraphs) |
| 1148 | && (!in_comment || warn_in_comment (pfile, note))) |
| 1149 | { |
| 1150 | if (CPP_OPTION (pfile, trigraphs)((pfile)->opts.trigraphs)) |
| 1151 | cpp_warning_with_line (pfile, CPP_W_TRIGRAPHS, |
| 1152 | pfile->line_table->highest_line, col, |
| 1153 | "trigraph ??%c converted to %c", |
| 1154 | note->type, |
| 1155 | (int) _cpp_trigraph_map[note->type]); |
| 1156 | else |
| 1157 | { |
| 1158 | cpp_warning_with_line |
| 1159 | (pfile, CPP_W_TRIGRAPHS, |
| 1160 | pfile->line_table->highest_line, col, |
| 1161 | "trigraph ??%c ignored, use -trigraphs to enable", |
| 1162 | note->type); |
| 1163 | } |
| 1164 | } |
| 1165 | } |
| 1166 | else if (note->type == 0) |
| 1167 | /* Already processed in lex_raw_string. */; |
| 1168 | else |
| 1169 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1169, __FUNCTION__); |
| 1170 | } |
| 1171 | } |
| 1172 | |
| 1173 | namespace bidi { |
| 1174 | enum class kind { |
| 1175 | NONE, LRE, RLE, LRO, RLO, LRI, RLI, FSI, PDF, PDI, LTR, RTL |
| 1176 | }; |
| 1177 | |
| 1178 | /* All the UTF-8 encodings of bidi characters start with E2. */ |
| 1179 | constexpr uchar utf8_start = 0xe2; |
| 1180 | |
| 1181 | struct context |
| 1182 | { |
| 1183 | context () {} |
| 1184 | context (location_t loc, kind k, bool pdf, bool ucn) |
| 1185 | : m_loc (loc), m_kind (k), m_pdf (pdf), m_ucn (ucn) |
| 1186 | { |
| 1187 | } |
| 1188 | |
| 1189 | kind get_pop_kind () const |
| 1190 | { |
| 1191 | return m_pdf ? kind::PDF : kind::PDI; |
| 1192 | } |
| 1193 | bool ucn_p () const |
| 1194 | { |
| 1195 | return m_ucn; |
| 1196 | } |
| 1197 | |
| 1198 | location_t m_loc; |
| 1199 | kind m_kind; |
| 1200 | unsigned m_pdf : 1; |
| 1201 | unsigned m_ucn : 1; |
| 1202 | }; |
| 1203 | |
| 1204 | /* A vector holding currently open bidi contexts. We use a char for |
| 1205 | each context, its LSB is 1 if it represents a PDF context, 0 if it |
| 1206 | represents a PDI context. The next bit is 1 if this context was open |
| 1207 | by a bidi character written as a UCN, and 0 when it was UTF-8. */ |
| 1208 | semi_embedded_vec <context, 16> vec; |
| 1209 | |
| 1210 | /* Close the whole comment/identifier/string literal/character constant |
| 1211 | context. */ |
| 1212 | void on_close () |
| 1213 | { |
| 1214 | vec.truncate (0); |
| 1215 | } |
| 1216 | |
| 1217 | /* Pop the last element in the vector. */ |
| 1218 | void pop () |
| 1219 | { |
| 1220 | unsigned int len = vec.count (); |
| 1221 | gcc_checking_assert (len > 0)((void)(!(len > 0) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1221, __FUNCTION__), 0 : 0)); |
| 1222 | vec.truncate (len - 1); |
| 1223 | } |
| 1224 | |
| 1225 | /* Return the pop kind of the context of the Ith element. */ |
| 1226 | kind pop_kind_at (unsigned int i) |
| 1227 | { |
| 1228 | return vec[i].get_pop_kind (); |
| 1229 | } |
| 1230 | |
| 1231 | /* Return the pop kind of the context that is currently opened. */ |
| 1232 | kind current_ctx () |
| 1233 | { |
| 1234 | unsigned int len = vec.count (); |
| 1235 | if (len == 0) |
| 1236 | return kind::NONE; |
| 1237 | return vec[len - 1].get_pop_kind (); |
| 1238 | } |
| 1239 | |
| 1240 | /* Return true if the current context comes from a UCN origin, that is, |
| 1241 | the bidi char which started this bidi context was written as a UCN. */ |
| 1242 | bool current_ctx_ucn_p () |
| 1243 | { |
| 1244 | unsigned int len = vec.count (); |
| 1245 | gcc_checking_assert (len > 0)((void)(!(len > 0) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1245, __FUNCTION__), 0 : 0)); |
| 1246 | return vec[len - 1].m_ucn; |
| 1247 | } |
| 1248 | |
| 1249 | location_t current_ctx_loc () |
| 1250 | { |
| 1251 | unsigned int len = vec.count (); |
| 1252 | gcc_checking_assert (len > 0)((void)(!(len > 0) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1252, __FUNCTION__), 0 : 0)); |
| 1253 | return vec[len - 1].m_loc; |
| 1254 | } |
| 1255 | |
| 1256 | /* We've read a bidi char, update the current vector as necessary. |
| 1257 | LOC is only valid when K is not kind::NONE. */ |
| 1258 | void on_char (kind k, bool ucn_p, location_t loc) |
| 1259 | { |
| 1260 | switch (k) |
| 1261 | { |
| 1262 | case kind::LRE: |
| 1263 | case kind::RLE: |
| 1264 | case kind::LRO: |
| 1265 | case kind::RLO: |
| 1266 | vec.push (context (loc, k, true, ucn_p)); |
| 1267 | break; |
| 1268 | case kind::LRI: |
| 1269 | case kind::RLI: |
| 1270 | case kind::FSI: |
| 1271 | vec.push (context (loc, k, false, ucn_p)); |
| 1272 | break; |
| 1273 | /* PDF terminates the scope of the last LRE, RLE, LRO, or RLO |
| 1274 | whose scope has not yet been terminated. */ |
| 1275 | case kind::PDF: |
| 1276 | if (current_ctx () == kind::PDF) |
| 1277 | pop (); |
| 1278 | break; |
| 1279 | /* PDI terminates the scope of the last LRI, RLI, or FSI whose |
| 1280 | scope has not yet been terminated, as well as the scopes of |
| 1281 | any subsequent LREs, RLEs, LROs, or RLOs whose scopes have not |
| 1282 | yet been terminated. */ |
| 1283 | case kind::PDI: |
| 1284 | for (int i = vec.count () - 1; i >= 0; --i) |
| 1285 | if (pop_kind_at (i) == kind::PDI) |
| 1286 | { |
| 1287 | vec.truncate (i); |
| 1288 | break; |
| 1289 | } |
| 1290 | break; |
| 1291 | case kind::LTR: |
| 1292 | case kind::RTL: |
| 1293 | /* These aren't popped by a PDF/PDI. */ |
| 1294 | break; |
| 1295 | ATTR_LIKELY[[likely]] case kind::NONE: |
| 1296 | break; |
| 1297 | default: |
| 1298 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1298, __FUNCTION__); |
| 1299 | } |
| 1300 | } |
| 1301 | |
| 1302 | /* Return a descriptive string for K. */ |
| 1303 | const char *to_str (kind k) |
| 1304 | { |
| 1305 | switch (k) |
| 1306 | { |
| 1307 | case kind::LRE: |
| 1308 | return "U+202A (LEFT-TO-RIGHT EMBEDDING)"; |
| 1309 | case kind::RLE: |
| 1310 | return "U+202B (RIGHT-TO-LEFT EMBEDDING)"; |
| 1311 | case kind::LRO: |
| 1312 | return "U+202D (LEFT-TO-RIGHT OVERRIDE)"; |
| 1313 | case kind::RLO: |
| 1314 | return "U+202E (RIGHT-TO-LEFT OVERRIDE)"; |
| 1315 | case kind::LRI: |
| 1316 | return "U+2066 (LEFT-TO-RIGHT ISOLATE)"; |
| 1317 | case kind::RLI: |
| 1318 | return "U+2067 (RIGHT-TO-LEFT ISOLATE)"; |
| 1319 | case kind::FSI: |
| 1320 | return "U+2068 (FIRST STRONG ISOLATE)"; |
| 1321 | case kind::PDF: |
| 1322 | return "U+202C (POP DIRECTIONAL FORMATTING)"; |
| 1323 | case kind::PDI: |
| 1324 | return "U+2069 (POP DIRECTIONAL ISOLATE)"; |
| 1325 | case kind::LTR: |
| 1326 | return "U+200E (LEFT-TO-RIGHT MARK)"; |
| 1327 | case kind::RTL: |
| 1328 | return "U+200F (RIGHT-TO-LEFT MARK)"; |
| 1329 | default: |
| 1330 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1330, __FUNCTION__); |
| 1331 | } |
| 1332 | } |
| 1333 | } |
| 1334 | |
| 1335 | /* Get location_t for the range of bytes [START, START + NUM_BYTES) |
| 1336 | within the current line in FILE, with the caret at START. */ |
| 1337 | |
| 1338 | static location_t |
| 1339 | get_location_for_byte_range_in_cur_line (cpp_reader *pfile, |
| 1340 | const unsigned char *const start, |
| 1341 | size_t num_bytes) |
| 1342 | { |
| 1343 | gcc_checking_assert (num_bytes > 0)((void)(!(num_bytes > 0) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1343, __FUNCTION__), 0 : 0)); |
| 1344 | |
| 1345 | /* CPP_BUF_COLUMN and linemap_position_for_column both refer |
| 1346 | to offsets in bytes, but CPP_BUF_COLUMN is 0-based, |
| 1347 | whereas linemap_position_for_column is 1-based. */ |
| 1348 | |
| 1349 | /* Get 0-based offsets within the line. */ |
| 1350 | size_t start_offset = CPP_BUF_COLUMN (pfile->buffer, start)((start) - (pfile->buffer)->line_base); |
| 1351 | size_t end_offset = start_offset + num_bytes - 1; |
| 1352 | |
| 1353 | /* Now convert to location_t, where "columns" are 1-based byte offsets. */ |
| 1354 | location_t start_loc = linemap_position_for_column (pfile->line_table, |
| 1355 | start_offset + 1); |
| 1356 | location_t end_loc = linemap_position_for_column (pfile->line_table, |
| 1357 | end_offset + 1); |
| 1358 | |
| 1359 | if (start_loc == end_loc) |
| 1360 | return start_loc; |
| 1361 | |
| 1362 | source_range src_range; |
| 1363 | src_range.m_start = start_loc; |
| 1364 | src_range.m_finish = end_loc; |
| 1365 | location_t combined_loc = COMBINE_LOCATION_DATA (pfile->line_table, |
| 1366 | start_loc, |
| 1367 | src_range, |
| 1368 | NULL__null, |
| 1369 | 0); |
| 1370 | return combined_loc; |
| 1371 | } |
| 1372 | |
| 1373 | /* Parse a sequence of 3 bytes starting with P and return its bidi code. */ |
| 1374 | |
| 1375 | static bidi::kind |
| 1376 | get_bidi_utf8_1 (const unsigned char *const p) |
| 1377 | { |
| 1378 | gcc_checking_assert (p[0] == bidi::utf8_start)((void)(!(p[0] == bidi::utf8_start) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 1378, __FUNCTION__), 0 : 0)); |
| 1379 | |
| 1380 | if (p[1] == 0x80) |
| 1381 | switch (p[2]) |
| 1382 | { |
| 1383 | case 0xaa: |
| 1384 | return bidi::kind::LRE; |
| 1385 | case 0xab: |
| 1386 | return bidi::kind::RLE; |
| 1387 | case 0xac: |
| 1388 | return bidi::kind::PDF; |
| 1389 | case 0xad: |
| 1390 | return bidi::kind::LRO; |
| 1391 | case 0xae: |
| 1392 | return bidi::kind::RLO; |
| 1393 | case 0x8e: |
| 1394 | return bidi::kind::LTR; |
| 1395 | case 0x8f: |
| 1396 | return bidi::kind::RTL; |
| 1397 | default: |
| 1398 | break; |
| 1399 | } |
| 1400 | else if (p[1] == 0x81) |
| 1401 | switch (p[2]) |
| 1402 | { |
| 1403 | case 0xa6: |
| 1404 | return bidi::kind::LRI; |
| 1405 | case 0xa7: |
| 1406 | return bidi::kind::RLI; |
| 1407 | case 0xa8: |
| 1408 | return bidi::kind::FSI; |
| 1409 | case 0xa9: |
| 1410 | return bidi::kind::PDI; |
| 1411 | default: |
| 1412 | break; |
| 1413 | } |
| 1414 | |
| 1415 | return bidi::kind::NONE; |
| 1416 | } |
| 1417 | |
| 1418 | /* Parse a sequence of 3 bytes starting with P and return its bidi code. |
| 1419 | If the kind is not NONE, write the location to *OUT.*/ |
| 1420 | |
| 1421 | static bidi::kind |
| 1422 | get_bidi_utf8 (cpp_reader *pfile, const unsigned char *const p, location_t *out) |
| 1423 | { |
| 1424 | bidi::kind result = get_bidi_utf8_1 (p); |
| 1425 | if (result != bidi::kind::NONE) |
| 1426 | { |
| 1427 | /* We have a sequence of 3 bytes starting at P. */ |
| 1428 | *out = get_location_for_byte_range_in_cur_line (pfile, p, 3); |
| 1429 | } |
| 1430 | return result; |
| 1431 | } |
| 1432 | |
| 1433 | /* Parse a UCN where P points just past \u or \U and return its bidi code. */ |
| 1434 | |
| 1435 | static bidi::kind |
| 1436 | get_bidi_ucn_1 (const unsigned char *p, bool is_U, const unsigned char **end) |
| 1437 | { |
| 1438 | /* 6.4.3 Universal Character Names |
| 1439 | \u hex-quad |
| 1440 | \U hex-quad hex-quad |
| 1441 | \u { simple-hexadecimal-digit-sequence } |
| 1442 | where \unnnn means \U0000nnnn. */ |
| 1443 | |
| 1444 | *end = p + 4; |
| 1445 | if (is_U) |
| 1446 | { |
| 1447 | if (p[0] != '0' || p[1] != '0' || p[2] != '0' || p[3] != '0') |
| 1448 | return bidi::kind::NONE; |
| 1449 | /* Skip 4B so we can treat \u and \U the same below. */ |
| 1450 | p += 4; |
| 1451 | *end += 4; |
| 1452 | } |
| 1453 | else if (p[0] == '{') |
| 1454 | { |
| 1455 | p++; |
| 1456 | while (*p == '0') |
| 1457 | p++; |
| 1458 | if (p[0] != '2' |
| 1459 | || p[1] != '0' |
| 1460 | || !ISXDIGIT (p[2])(_sch_istable[(p[2]) & 0xff] & (unsigned short)(_sch_isxdigit )) |
| 1461 | || !ISXDIGIT (p[3])(_sch_istable[(p[3]) & 0xff] & (unsigned short)(_sch_isxdigit )) |
| 1462 | || p[4] != '}') |
| 1463 | return bidi::kind::NONE; |
| 1464 | *end = p + 5; |
| 1465 | } |
| 1466 | |
| 1467 | /* All code points we are looking for start with 20xx. */ |
| 1468 | if (p[0] != '2' || p[1] != '0') |
| 1469 | return bidi::kind::NONE; |
| 1470 | else if (p[2] == '2') |
| 1471 | switch (p[3]) |
| 1472 | { |
| 1473 | case 'a': |
| 1474 | case 'A': |
| 1475 | return bidi::kind::LRE; |
| 1476 | case 'b': |
| 1477 | case 'B': |
| 1478 | return bidi::kind::RLE; |
| 1479 | case 'c': |
| 1480 | case 'C': |
| 1481 | return bidi::kind::PDF; |
| 1482 | case 'd': |
| 1483 | case 'D': |
| 1484 | return bidi::kind::LRO; |
| 1485 | case 'e': |
| 1486 | case 'E': |
| 1487 | return bidi::kind::RLO; |
| 1488 | default: |
| 1489 | break; |
| 1490 | } |
| 1491 | else if (p[2] == '6') |
| 1492 | switch (p[3]) |
| 1493 | { |
| 1494 | case '6': |
| 1495 | return bidi::kind::LRI; |
| 1496 | case '7': |
| 1497 | return bidi::kind::RLI; |
| 1498 | case '8': |
| 1499 | return bidi::kind::FSI; |
| 1500 | case '9': |
| 1501 | return bidi::kind::PDI; |
| 1502 | default: |
| 1503 | break; |
| 1504 | } |
| 1505 | else if (p[2] == '0') |
| 1506 | switch (p[3]) |
| 1507 | { |
| 1508 | case 'e': |
| 1509 | case 'E': |
| 1510 | return bidi::kind::LTR; |
| 1511 | case 'f': |
| 1512 | case 'F': |
| 1513 | return bidi::kind::RTL; |
| 1514 | default: |
| 1515 | break; |
| 1516 | } |
| 1517 | |
| 1518 | return bidi::kind::NONE; |
| 1519 | } |
| 1520 | |
| 1521 | /* Parse a UCN where P points just past \u or \U and return its bidi code. |
| 1522 | If the kind is not NONE, write the location to *OUT. */ |
| 1523 | |
| 1524 | static bidi::kind |
| 1525 | get_bidi_ucn (cpp_reader *pfile, const unsigned char *p, bool is_U, |
| 1526 | location_t *out) |
| 1527 | { |
| 1528 | const unsigned char *end; |
| 1529 | bidi::kind result = get_bidi_ucn_1 (p, is_U, &end); |
| 1530 | if (result != bidi::kind::NONE) |
| 1531 | { |
| 1532 | const unsigned char *start = p - 2; |
| 1533 | size_t num_bytes = end - start; |
| 1534 | *out = get_location_for_byte_range_in_cur_line (pfile, start, num_bytes); |
| 1535 | } |
| 1536 | return result; |
| 1537 | } |
| 1538 | |
| 1539 | /* Parse a named universal character escape where P points just past \N and |
| 1540 | return its bidi code. If the kind is not NONE, write the location to |
| 1541 | *OUT. */ |
| 1542 | |
| 1543 | static bidi::kind |
| 1544 | get_bidi_named (cpp_reader *pfile, const unsigned char *p, location_t *out) |
| 1545 | { |
| 1546 | bidi::kind result = bidi::kind::NONE; |
| 1547 | if (*p != '{') |
| 1548 | return bidi::kind::NONE; |
| 1549 | if (strncmp ((const char *) (p + 1), "LEFT-TO-RIGHT ", 14) == 0) |
| 1550 | { |
| 1551 | if (strncmp ((const char *) (p + 15), "MARK}", 5) == 0) |
| 1552 | result = bidi::kind::LTR; |
| 1553 | else if (strncmp ((const char *) (p + 15), "EMBEDDING}", 10) == 0) |
| 1554 | result = bidi::kind::LRE; |
| 1555 | else if (strncmp ((const char *) (p + 15), "OVERRIDE}", 9) == 0) |
| 1556 | result = bidi::kind::LRO; |
| 1557 | else if (strncmp ((const char *) (p + 15), "ISOLATE}", 8) == 0) |
| 1558 | result = bidi::kind::LRI; |
| 1559 | } |
| 1560 | else if (strncmp ((const char *) (p + 1), "RIGHT-TO-LEFT ", 14) == 0) |
| 1561 | { |
| 1562 | if (strncmp ((const char *) (p + 15), "MARK}", 5) == 0) |
| 1563 | result = bidi::kind::RTL; |
| 1564 | else if (strncmp ((const char *) (p + 15), "EMBEDDING}", 10) == 0) |
| 1565 | result = bidi::kind::RLE; |
| 1566 | else if (strncmp ((const char *) (p + 15), "OVERRIDE}", 9) == 0) |
| 1567 | result = bidi::kind::RLO; |
| 1568 | else if (strncmp ((const char *) (p + 15), "ISOLATE}", 8) == 0) |
| 1569 | result = bidi::kind::RLI; |
| 1570 | } |
| 1571 | else if (strncmp ((const char *) (p + 1), "POP DIRECTIONAL ", 16) == 0) |
| 1572 | { |
| 1573 | if (strncmp ((const char *) (p + 16), "FORMATTING}", 11) == 0) |
| 1574 | result = bidi::kind::PDF; |
| 1575 | else if (strncmp ((const char *) (p + 16), "ISOLATE}", 8) == 0) |
| 1576 | result = bidi::kind::PDI; |
| 1577 | } |
| 1578 | else if (strncmp ((const char *) (p + 1), "FIRST STRONG ISOLATE}", 21) == 0) |
| 1579 | result = bidi::kind::FSI; |
| 1580 | if (result != bidi::kind::NONE) |
| 1581 | *out = get_location_for_byte_range_in_cur_line (pfile, p - 2, |
| 1582 | (strchr ((const char *) |
| 1583 | (p + 1), '}') |
| 1584 | - (const char *) p) |
| 1585 | + 3); |
| 1586 | return result; |
| 1587 | } |
| 1588 | |
| 1589 | /* Subclass of rich_location for reporting on unpaired UTF-8 |
| 1590 | bidirectional control character(s). |
| 1591 | Escape the source lines on output, and show all unclosed |
| 1592 | bidi context, labelling everything. */ |
| 1593 | |
| 1594 | class unpaired_bidi_rich_location : public rich_location |
| 1595 | { |
| 1596 | public: |
| 1597 | class custom_range_label : public range_label |
| 1598 | { |
| 1599 | public: |
| 1600 | label_text get_text (unsigned range_idx) const final override |
| 1601 | { |
| 1602 | /* range 0 is the primary location; each subsequent range i + 1 |
| 1603 | is for bidi::vec[i]. */ |
| 1604 | if (range_idx > 0) |
| 1605 | { |
| 1606 | const bidi::context &ctxt (bidi::vec[range_idx - 1]); |
| 1607 | return label_text::borrow (bidi::to_str (ctxt.m_kind)); |
| 1608 | } |
| 1609 | else |
| 1610 | return label_text::borrow (_("end of bidirectional context")("end of bidirectional context")); |
| 1611 | } |
| 1612 | }; |
| 1613 | |
| 1614 | unpaired_bidi_rich_location (cpp_reader *pfile, location_t loc) |
| 1615 | : rich_location (pfile->line_table, loc, &m_custom_label) |
| 1616 | { |
| 1617 | set_escape_on_output (true); |
| 1618 | for (unsigned i = 0; i < bidi::vec.count (); i++) |
| 1619 | add_range (bidi::vec[i].m_loc, |
| 1620 | SHOW_RANGE_WITHOUT_CARET, |
| 1621 | &m_custom_label); |
| 1622 | } |
| 1623 | |
| 1624 | private: |
| 1625 | custom_range_label m_custom_label; |
| 1626 | }; |
| 1627 | |
| 1628 | /* We're closing a bidi context, that is, we've encountered a newline, |
| 1629 | are closing a C-style comment, or are at the end of a string literal, |
| 1630 | character constant, or identifier. Warn if this context was not |
| 1631 | properly terminated by a PDI or PDF. P points to the last character |
| 1632 | in this context. */ |
| 1633 | |
| 1634 | static void |
| 1635 | maybe_warn_bidi_on_close (cpp_reader *pfile, const uchar *p) |
| 1636 | { |
| 1637 | const auto warn_bidi = CPP_OPTION (pfile, cpp_warn_bidirectional)((pfile)->opts.cpp_warn_bidirectional); |
| 1638 | if (bidi::vec.count () > 0 |
| 1639 | && (warn_bidi & bidirectional_unpaired |
| 1640 | && (!bidi::current_ctx_ucn_p () |
| 1641 | || (warn_bidi & bidirectional_ucn)))) |
| 1642 | { |
| 1643 | const location_t loc |
| 1644 | = linemap_position_for_column (pfile->line_table, |
| 1645 | CPP_BUF_COLUMN (pfile->buffer, p)((p) - (pfile->buffer)->line_base)); |
| 1646 | unpaired_bidi_rich_location rich_loc (pfile, loc); |
| 1647 | /* cpp_callbacks doesn't yet have a way to handle singular vs plural |
| 1648 | forms of a diagnostic, so fake it for now. */ |
| 1649 | if (bidi::vec.count () > 1) |
| 1650 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
| 1651 | "unpaired UTF-8 bidirectional control characters " |
| 1652 | "detected"); |
| 1653 | else |
| 1654 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
| 1655 | "unpaired UTF-8 bidirectional control character " |
| 1656 | "detected"); |
| 1657 | } |
| 1658 | /* We're done with this context. */ |
| 1659 | bidi::on_close (); |
| 1660 | } |
| 1661 | |
| 1662 | /* We're at the beginning or in the middle of an identifier/comment/string |
| 1663 | literal/character constant. Warn if we've encountered a bidi character. |
| 1664 | KIND says which bidi control character it was; UCN_P is true iff this bidi |
| 1665 | control character was written as a UCN. LOC is the location of the |
| 1666 | character, but is only valid if KIND != bidi::kind::NONE. */ |
| 1667 | |
| 1668 | static void |
| 1669 | maybe_warn_bidi_on_char (cpp_reader *pfile, bidi::kind kind, |
| 1670 | bool ucn_p, location_t loc) |
| 1671 | { |
| 1672 | if (__builtin_expect (kind == bidi::kind::NONE, 1)) |
| 1673 | return; |
| 1674 | |
| 1675 | const auto warn_bidi = CPP_OPTION (pfile, cpp_warn_bidirectional)((pfile)->opts.cpp_warn_bidirectional); |
| 1676 | |
| 1677 | if (warn_bidi & (bidirectional_unpaired|bidirectional_any)) |
| 1678 | { |
| 1679 | rich_location rich_loc (pfile->line_table, loc); |
| 1680 | rich_loc.set_escape_on_output (true); |
| 1681 | |
| 1682 | /* It seems excessive to warn about a PDI/PDF that is closing |
| 1683 | an opened context because we've already warned about the |
| 1684 | opening character. Except warn when we have a UCN x UTF-8 |
| 1685 | mismatch, if UCN checking is enabled. */ |
| 1686 | if (kind == bidi::current_ctx ()) |
| 1687 | { |
| 1688 | if (warn_bidi == (bidirectional_unpaired|bidirectional_ucn) |
| 1689 | && bidi::current_ctx_ucn_p () != ucn_p) |
| 1690 | { |
| 1691 | rich_loc.add_range (bidi::current_ctx_loc ()); |
| 1692 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
| 1693 | "UTF-8 vs UCN mismatch when closing " |
| 1694 | "a context by \"%s\"", bidi::to_str (kind)); |
| 1695 | } |
| 1696 | } |
| 1697 | else if (warn_bidi & bidirectional_any |
| 1698 | && (!ucn_p || (warn_bidi & bidirectional_ucn))) |
| 1699 | { |
| 1700 | if (kind == bidi::kind::PDF || kind == bidi::kind::PDI) |
| 1701 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
| 1702 | "\"%s\" is closing an unopened context", |
| 1703 | bidi::to_str (kind)); |
| 1704 | else |
| 1705 | cpp_warning_at (pfile, CPP_W_BIDIRECTIONAL, &rich_loc, |
| 1706 | "found problematic Unicode character \"%s\"", |
| 1707 | bidi::to_str (kind)); |
| 1708 | } |
| 1709 | } |
| 1710 | /* We're done with this context. */ |
| 1711 | bidi::on_char (kind, ucn_p, loc); |
| 1712 | } |
| 1713 | |
| 1714 | static const cppchar_t utf8_continuation = 0x80; |
| 1715 | static const cppchar_t utf8_signifier = 0xC0; |
| 1716 | |
| 1717 | /* Emit -Winvalid-utf8 warning on invalid UTF-8 character starting |
| 1718 | at PFILE->buffer->cur. Return a pointer after the diagnosed |
| 1719 | invalid character. */ |
| 1720 | |
| 1721 | static const uchar * |
| 1722 | _cpp_warn_invalid_utf8 (cpp_reader *pfile) |
| 1723 | { |
| 1724 | cpp_buffer *buffer = pfile->buffer; |
| 1725 | const uchar *cur = buffer->cur; |
| 1726 | bool pedantic = (CPP_PEDANTIC (pfile)((pfile)->opts.cpp_pedantic) |
| 1727 | && CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8) == 2); |
| 1728 | |
| 1729 | if (cur[0] < utf8_signifier |
| 1730 | || cur[1] < utf8_continuation || cur[1] >= utf8_signifier) |
| 1731 | { |
| 1732 | if (pedantic) |
| 1733 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, |
| 1734 | pfile->line_table->highest_line, |
| 1735 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1736 | "invalid UTF-8 character <%x>", |
| 1737 | cur[0]); |
| 1738 | else |
| 1739 | cpp_warning_with_line (pfile, CPP_W_INVALID_UTF8, |
| 1740 | pfile->line_table->highest_line, |
| 1741 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1742 | "invalid UTF-8 character <%x>", |
| 1743 | cur[0]); |
| 1744 | return cur + 1; |
| 1745 | } |
| 1746 | else if (cur[2] < utf8_continuation || cur[2] >= utf8_signifier) |
| 1747 | { |
| 1748 | if (pedantic) |
| 1749 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, |
| 1750 | pfile->line_table->highest_line, |
| 1751 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1752 | "invalid UTF-8 character <%x><%x>", |
| 1753 | cur[0], cur[1]); |
| 1754 | else |
| 1755 | cpp_warning_with_line (pfile, CPP_W_INVALID_UTF8, |
| 1756 | pfile->line_table->highest_line, |
| 1757 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1758 | "invalid UTF-8 character <%x><%x>", |
| 1759 | cur[0], cur[1]); |
| 1760 | return cur + 2; |
| 1761 | } |
| 1762 | else if (cur[3] < utf8_continuation || cur[3] >= utf8_signifier) |
| 1763 | { |
| 1764 | if (pedantic) |
| 1765 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, |
| 1766 | pfile->line_table->highest_line, |
| 1767 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1768 | "invalid UTF-8 character <%x><%x><%x>", |
| 1769 | cur[0], cur[1], cur[2]); |
| 1770 | else |
| 1771 | cpp_warning_with_line (pfile, CPP_W_INVALID_UTF8, |
| 1772 | pfile->line_table->highest_line, |
| 1773 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1774 | "invalid UTF-8 character <%x><%x><%x>", |
| 1775 | cur[0], cur[1], cur[2]); |
| 1776 | return cur + 3; |
| 1777 | } |
| 1778 | else |
| 1779 | { |
| 1780 | if (pedantic) |
| 1781 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, |
| 1782 | pfile->line_table->highest_line, |
| 1783 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1784 | "invalid UTF-8 character <%x><%x><%x><%x>", |
| 1785 | cur[0], cur[1], cur[2], cur[3]); |
| 1786 | else |
| 1787 | cpp_warning_with_line (pfile, CPP_W_INVALID_UTF8, |
| 1788 | pfile->line_table->highest_line, |
| 1789 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1790 | "invalid UTF-8 character <%x><%x><%x><%x>", |
| 1791 | cur[0], cur[1], cur[2], cur[3]); |
| 1792 | return cur + 4; |
| 1793 | } |
| 1794 | } |
| 1795 | |
| 1796 | /* Helper function of *skip_*_comment and lex*_string. For C, |
| 1797 | character at CUR[-1] with MSB set handle -Wbidi-chars* and |
| 1798 | -Winvalid-utf8 diagnostics and return pointer to first character |
| 1799 | that should be processed next. */ |
| 1800 | |
| 1801 | static inline const uchar * |
| 1802 | _cpp_handle_multibyte_utf8 (cpp_reader *pfile, uchar c, |
| 1803 | const uchar *cur, bool warn_bidi_p, |
| 1804 | bool warn_invalid_utf8_p) |
| 1805 | { |
| 1806 | /* If this is a beginning of a UTF-8 encoding, it might be |
| 1807 | a bidirectional control character. */ |
| 1808 | if (c == bidi::utf8_start && warn_bidi_p) |
| 1809 | { |
| 1810 | location_t loc; |
| 1811 | bidi::kind kind = get_bidi_utf8 (pfile, cur - 1, &loc); |
| 1812 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); |
| 1813 | } |
| 1814 | if (!warn_invalid_utf8_p) |
| 1815 | return cur; |
| 1816 | if (c >= utf8_signifier) |
| 1817 | { |
| 1818 | cppchar_t s; |
| 1819 | const uchar *pstr = cur - 1; |
| 1820 | if (_cpp_valid_utf8 (pfile, &pstr, pfile->buffer->rlimit, 0, NULL__null, &s) |
| 1821 | && s <= UCS_LIMIT0x10FFFF) |
| 1822 | return pstr; |
| 1823 | } |
| 1824 | pfile->buffer->cur = cur - 1; |
| 1825 | return _cpp_warn_invalid_utf8 (pfile); |
| 1826 | } |
| 1827 | |
| 1828 | /* Skip a C-style block comment. We find the end of the comment by |
| 1829 | seeing if an asterisk is before every '/' we encounter. Returns |
| 1830 | nonzero if comment terminated by EOF, zero otherwise. |
| 1831 | |
| 1832 | Buffer->cur points to the initial asterisk of the comment. */ |
| 1833 | bool |
| 1834 | _cpp_skip_block_comment (cpp_reader *pfile) |
| 1835 | { |
| 1836 | cpp_buffer *buffer = pfile->buffer; |
| 1837 | const uchar *cur = buffer->cur; |
| 1838 | uchar c; |
| 1839 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 1840 | const bool warn_invalid_utf8_p = CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8); |
| 1841 | const bool warn_bidi_or_invalid_utf8_p = warn_bidi_p | warn_invalid_utf8_p; |
| 1842 | |
| 1843 | cur++; |
| 1844 | if (*cur == '/') |
| 1845 | cur++; |
| 1846 | |
| 1847 | for (;;) |
| 1848 | { |
| 1849 | /* People like decorating comments with '*', so check for '/' |
| 1850 | instead for efficiency. */ |
| 1851 | c = *cur++; |
| 1852 | |
| 1853 | if (c == '/') |
| 1854 | { |
| 1855 | if (cur[-2] == '*') |
| 1856 | { |
| 1857 | if (warn_bidi_p) |
| 1858 | maybe_warn_bidi_on_close (pfile, cur); |
| 1859 | break; |
| 1860 | } |
| 1861 | |
| 1862 | /* Warn about potential nested comments, but not if the '/' |
| 1863 | comes immediately before the true comment delimiter. |
| 1864 | Don't bother to get it right across escaped newlines. */ |
| 1865 | if (CPP_OPTION (pfile, warn_comments)((pfile)->opts.warn_comments) |
| 1866 | && cur[0] == '*' && cur[1] != '/') |
| 1867 | { |
| 1868 | buffer->cur = cur; |
| 1869 | cpp_warning_with_line (pfile, CPP_W_COMMENTS, |
| 1870 | pfile->line_table->highest_line, |
| 1871 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1872 | "\"/*\" within comment"); |
| 1873 | } |
| 1874 | } |
| 1875 | else if (c == '\n') |
| 1876 | { |
| 1877 | unsigned int cols; |
| 1878 | buffer->cur = cur - 1; |
| 1879 | if (warn_bidi_p) |
| 1880 | maybe_warn_bidi_on_close (pfile, cur); |
| 1881 | _cpp_process_line_notes (pfile, true); |
| 1882 | if (buffer->next_line >= buffer->rlimit) |
| 1883 | return true; |
| 1884 | _cpp_clean_line (pfile); |
| 1885 | |
| 1886 | cols = buffer->next_line - buffer->line_base; |
| 1887 | CPP_INCREMENT_LINE (pfile, cols)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, cols); } while (0); |
| 1888 | |
| 1889 | cur = buffer->cur; |
| 1890 | } |
| 1891 | else if (__builtin_expect (c >= utf8_continuation, 0) |
| 1892 | && warn_bidi_or_invalid_utf8_p) |
| 1893 | cur = _cpp_handle_multibyte_utf8 (pfile, c, cur, warn_bidi_p, |
| 1894 | warn_invalid_utf8_p); |
| 1895 | } |
| 1896 | |
| 1897 | buffer->cur = cur; |
| 1898 | _cpp_process_line_notes (pfile, true); |
| 1899 | return false; |
| 1900 | } |
| 1901 | |
| 1902 | /* Skip a C++ line comment, leaving buffer->cur pointing to the |
| 1903 | terminating newline. Handles escaped newlines. Returns nonzero |
| 1904 | if a multiline comment. */ |
| 1905 | static int |
| 1906 | skip_line_comment (cpp_reader *pfile) |
| 1907 | { |
| 1908 | cpp_buffer *buffer = pfile->buffer; |
| 1909 | location_t orig_line = pfile->line_table->highest_line; |
| 1910 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 1911 | const bool warn_invalid_utf8_p = CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8); |
| 1912 | const bool warn_bidi_or_invalid_utf8_p = warn_bidi_p | warn_invalid_utf8_p; |
| 1913 | |
| 1914 | if (!warn_bidi_or_invalid_utf8_p) |
| 1915 | while (*buffer->cur != '\n') |
| 1916 | buffer->cur++; |
| 1917 | else if (!warn_invalid_utf8_p) |
| 1918 | { |
| 1919 | while (*buffer->cur != '\n' |
| 1920 | && *buffer->cur != bidi::utf8_start) |
| 1921 | buffer->cur++; |
| 1922 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0)) |
| 1923 | { |
| 1924 | while (*buffer->cur != '\n') |
| 1925 | { |
| 1926 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0)) |
| 1927 | { |
| 1928 | location_t loc; |
| 1929 | bidi::kind kind = get_bidi_utf8 (pfile, buffer->cur, &loc); |
| 1930 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); |
| 1931 | } |
| 1932 | buffer->cur++; |
| 1933 | } |
| 1934 | maybe_warn_bidi_on_close (pfile, buffer->cur); |
| 1935 | } |
| 1936 | } |
| 1937 | else |
| 1938 | { |
| 1939 | while (*buffer->cur != '\n') |
| 1940 | { |
| 1941 | if (*buffer->cur < utf8_continuation) |
| 1942 | { |
| 1943 | buffer->cur++; |
| 1944 | continue; |
| 1945 | } |
| 1946 | buffer->cur |
| 1947 | = _cpp_handle_multibyte_utf8 (pfile, *buffer->cur, buffer->cur + 1, |
| 1948 | warn_bidi_p, warn_invalid_utf8_p); |
| 1949 | } |
| 1950 | if (warn_bidi_p) |
| 1951 | maybe_warn_bidi_on_close (pfile, buffer->cur); |
| 1952 | } |
| 1953 | |
| 1954 | _cpp_process_line_notes (pfile, true); |
| 1955 | return orig_line != pfile->line_table->highest_line; |
| 1956 | } |
| 1957 | |
| 1958 | /* Skips whitespace, saving the next non-whitespace character. */ |
| 1959 | static void |
| 1960 | skip_whitespace (cpp_reader *pfile, cppchar_t c) |
| 1961 | { |
| 1962 | cpp_buffer *buffer = pfile->buffer; |
| 1963 | bool saw_NUL = false; |
| 1964 | |
| 1965 | do |
| 1966 | { |
| 1967 | /* Horizontal space always OK. */ |
| 1968 | if (c == ' ' || c == '\t') |
| 1969 | ; |
| 1970 | /* Just \f \v or \0 left. */ |
| 1971 | else if (c == '\0') |
| 1972 | saw_NUL = true; |
| 1973 | else if (pfile->state.in_directive && CPP_PEDANTIC (pfile)((pfile)->opts.cpp_pedantic)) |
| 1974 | cpp_error_with_line (pfile, CPP_DL_PEDWARN, pfile->line_table->highest_line, |
| 1975 | CPP_BUF_COL (buffer)(((buffer)->cur) - (buffer)->line_base), |
| 1976 | "%s in preprocessing directive", |
| 1977 | c == '\f' ? "form feed" : "vertical tab"); |
| 1978 | |
| 1979 | c = *buffer->cur++; |
| 1980 | } |
| 1981 | /* We only want non-vertical space, i.e. ' ' \t \f \v \0. */ |
| 1982 | while (is_nvspace (c)(_sch_istable[(c) & 0xff] & (unsigned short)(_sch_isnvsp ))); |
| 1983 | |
| 1984 | if (saw_NUL) |
| 1985 | { |
| 1986 | encoding_rich_location rich_loc (pfile); |
| 1987 | cpp_error_at (pfile, CPP_DL_WARNING, &rich_loc, |
| 1988 | "null character(s) ignored"); |
| 1989 | } |
| 1990 | |
| 1991 | buffer->cur--; |
| 1992 | } |
| 1993 | |
| 1994 | /* See if the characters of a number token are valid in a name (no |
| 1995 | '.', '+' or '-'). */ |
| 1996 | static int |
| 1997 | name_p (cpp_reader *pfile, const cpp_string *string) |
| 1998 | { |
| 1999 | unsigned int i; |
| 2000 | |
| 2001 | for (i = 0; i < string->len; i++) |
| 2002 | if (!is_idchar (string->text[i])((_sch_istable[(string->text[i]) & 0xff] & (unsigned short)(_sch_isidnum)) || ((string->text[i]) == '$' && ((pfile)->opts.dollars_in_ident)))) |
| 2003 | return 0; |
| 2004 | |
| 2005 | return 1; |
| 2006 | } |
| 2007 | |
| 2008 | /* After parsing an identifier or other sequence, produce a warning about |
| 2009 | sequences not in NFC/NFKC. */ |
| 2010 | static void |
| 2011 | warn_about_normalization (cpp_reader *pfile, |
| 2012 | const cpp_token *token, |
| 2013 | const struct normalize_state *s, |
| 2014 | bool identifier) |
| 2015 | { |
| 2016 | if (CPP_OPTION (pfile, warn_normalize)((pfile)->opts.warn_normalize) < NORMALIZE_STATE_RESULT (s)((s)->level) |
| 2017 | && !pfile->state.skipping) |
| 2018 | { |
| 2019 | location_t loc = token->src_loc; |
| 2020 | |
| 2021 | /* If possible, create a location range for the token. */ |
| 2022 | if (loc >= RESERVED_LOCATION_COUNT |
| 2023 | && token->type != CPP_EOF |
| 2024 | /* There must be no line notes to process. */ |
| 2025 | && (!(pfile->buffer->cur |
| 2026 | >= pfile->buffer->notes[pfile->buffer->cur_note].pos |
| 2027 | && !pfile->overlaid_buffer))) |
| 2028 | { |
| 2029 | source_range tok_range; |
| 2030 | tok_range.m_start = loc; |
| 2031 | tok_range.m_finish |
| 2032 | = linemap_position_for_column (pfile->line_table, |
| 2033 | CPP_BUF_COLUMN (pfile->buffer,((pfile->buffer->cur) - (pfile->buffer)->line_base ) |
| 2034 | pfile->buffer->cur)((pfile->buffer->cur) - (pfile->buffer)->line_base )); |
| 2035 | loc = COMBINE_LOCATION_DATA (pfile->line_table, |
| 2036 | loc, tok_range, NULL__null, 0); |
| 2037 | } |
| 2038 | |
| 2039 | encoding_rich_location rich_loc (pfile, loc); |
| 2040 | |
| 2041 | /* Make sure that the token is printed using UCNs, even |
| 2042 | if we'd otherwise happily print UTF-8. */ |
| 2043 | unsigned char *buf = XNEWVEC (unsigned char, cpp_token_len (token))((unsigned char *) xmalloc (sizeof (unsigned char) * (cpp_token_len (token)))); |
| 2044 | size_t sz; |
| 2045 | |
| 2046 | sz = cpp_spell_token (pfile, token, buf, false) - buf; |
| 2047 | if (NORMALIZE_STATE_RESULT (s)((s)->level) == normalized_C) |
| 2048 | cpp_warning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
| 2049 | "`%.*s' is not in NFKC", (int) sz, buf); |
| 2050 | else if (identifier && CPP_OPTION (pfile, xid_identifiers)((pfile)->opts.xid_identifiers)) |
| 2051 | cpp_pedwarning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
| 2052 | "`%.*s' is not in NFC", (int) sz, buf); |
| 2053 | else |
| 2054 | cpp_warning_at (pfile, CPP_W_NORMALIZE, &rich_loc, |
| 2055 | "`%.*s' is not in NFC", (int) sz, buf); |
| 2056 | free (buf); |
| 2057 | } |
| 2058 | } |
| 2059 | |
| 2060 | /* Returns TRUE if the sequence starting at buffer->cur is valid in |
| 2061 | an identifier. FIRST is TRUE if this starts an identifier. */ |
| 2062 | |
| 2063 | static bool |
| 2064 | forms_identifier_p (cpp_reader *pfile, int first, |
| 2065 | struct normalize_state *state) |
| 2066 | { |
| 2067 | cpp_buffer *buffer = pfile->buffer; |
| 2068 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 2069 | |
| 2070 | if (*buffer->cur == '$') |
| 2071 | { |
| 2072 | if (!CPP_OPTION (pfile, dollars_in_ident)((pfile)->opts.dollars_in_ident)) |
| 2073 | return false; |
| 2074 | |
| 2075 | buffer->cur++; |
| 2076 | if (CPP_OPTION (pfile, warn_dollars)((pfile)->opts.warn_dollars) && !pfile->state.skipping) |
| 2077 | { |
| 2078 | CPP_OPTION (pfile, warn_dollars)((pfile)->opts.warn_dollars) = 0; |
| 2079 | cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number"); |
| 2080 | } |
| 2081 | |
| 2082 | return true; |
| 2083 | } |
| 2084 | |
| 2085 | /* Is this a syntactically valid UCN or a valid UTF-8 char? */ |
| 2086 | if (CPP_OPTION (pfile, extended_identifiers)((pfile)->opts.extended_identifiers)) |
| 2087 | { |
| 2088 | cppchar_t s; |
| 2089 | if (*buffer->cur >= utf8_signifier) |
| 2090 | { |
| 2091 | if (__builtin_expect (*buffer->cur == bidi::utf8_start, 0) |
| 2092 | && warn_bidi_p) |
| 2093 | { |
| 2094 | location_t loc; |
| 2095 | bidi::kind kind = get_bidi_utf8 (pfile, buffer->cur, &loc); |
| 2096 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/false, loc); |
| 2097 | } |
| 2098 | if (_cpp_valid_utf8 (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
| 2099 | state, &s)) |
| 2100 | return true; |
| 2101 | } |
| 2102 | else if (*buffer->cur == '\\' |
| 2103 | && (buffer->cur[1] == 'u' |
| 2104 | || buffer->cur[1] == 'U' |
| 2105 | || buffer->cur[1] == 'N')) |
| 2106 | { |
| 2107 | buffer->cur += 2; |
| 2108 | if (warn_bidi_p) |
| 2109 | { |
| 2110 | location_t loc; |
| 2111 | bidi::kind kind; |
| 2112 | if (buffer->cur[-1] == 'N') |
| 2113 | kind = get_bidi_named (pfile, buffer->cur, &loc); |
| 2114 | else |
| 2115 | kind = get_bidi_ucn (pfile, buffer->cur, |
| 2116 | buffer->cur[-1] == 'U', &loc); |
| 2117 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/true, loc); |
| 2118 | } |
| 2119 | if (_cpp_valid_ucn (pfile, &buffer->cur, buffer->rlimit, 1 + !first, |
| 2120 | state, &s, NULL__null, NULL__null)) |
| 2121 | return true; |
| 2122 | buffer->cur -= 2; |
| 2123 | } |
| 2124 | } |
| 2125 | |
| 2126 | return false; |
| 2127 | } |
| 2128 | |
| 2129 | /* Helper function to issue error about improper __VA_OPT__ use. */ |
| 2130 | static void |
| 2131 | maybe_va_opt_error (cpp_reader *pfile) |
| 2132 | { |
| 2133 | if (CPP_PEDANTIC (pfile)((pfile)->opts.cpp_pedantic) && !CPP_OPTION (pfile, va_opt)((pfile)->opts.va_opt)) |
| 2134 | { |
| 2135 | /* __VA_OPT__ should not be accepted at all, but allow it in |
| 2136 | system headers. */ |
| 2137 | if (!_cpp_in_system_header (pfile)) |
| 2138 | { |
| 2139 | if (CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus)) |
| 2140 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2141 | "__VA_OPT__ is not available until C++20"); |
| 2142 | else |
| 2143 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2144 | "__VA_OPT__ is not available until C2X"); |
| 2145 | } |
| 2146 | } |
| 2147 | else if (!pfile->state.va_args_ok) |
| 2148 | { |
| 2149 | /* __VA_OPT__ should only appear in the replacement list of a |
| 2150 | variadic macro. */ |
| 2151 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2152 | "__VA_OPT__ can only appear in the expansion" |
| 2153 | " of a C++20 variadic macro"); |
| 2154 | } |
| 2155 | } |
| 2156 | |
| 2157 | /* Helper function to get the cpp_hashnode of the identifier BASE. */ |
| 2158 | static cpp_hashnode * |
| 2159 | lex_identifier_intern (cpp_reader *pfile, const uchar *base) |
| 2160 | { |
| 2161 | cpp_hashnode *result; |
| 2162 | const uchar *cur; |
| 2163 | unsigned int len; |
| 2164 | unsigned int hash = HT_HASHSTEP (0, *base)((0) * 67 + ((*base) - 113));; |
| 2165 | |
| 2166 | cur = base + 1; |
| 2167 | while (ISIDNUM (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 2168 | { |
| 2169 | hash = HT_HASHSTEP (hash, *cur)((hash) * 67 + ((*cur) - 113));; |
| 2170 | cur++; |
| 2171 | } |
| 2172 | len = cur - base; |
| 2173 | hash = HT_HASHFINISH (hash, len)((hash) + (len)); |
| 2174 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table,((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC))) |
| 2175 | base, len, hash, HT_ALLOC))((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC))); |
| 2176 | |
| 2177 | /* Rarely, identifiers require diagnostics when lexed. */ |
| 2178 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC(1 << 2)) |
| 2179 | && !pfile->state.skipping, 0)) |
| 2180 | { |
| 2181 | /* It is allowed to poison the same identifier twice. */ |
| 2182 | if ((result->flags & NODE_POISONED(1 << 1)) && !pfile->state.poisoned_ok) |
| 2183 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
| 2184 | NODE_NAME (result)(((&(result)->ident))->str)); |
| 2185 | |
| 2186 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the |
| 2187 | replacement list of a variadic macro. */ |
| 2188 | if (result == pfile->spec_nodes.n__VA_ARGS__ |
| 2189 | && !pfile->state.va_args_ok) |
| 2190 | { |
| 2191 | if (CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus)) |
| 2192 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2193 | "__VA_ARGS__ can only appear in the expansion" |
| 2194 | " of a C++11 variadic macro"); |
| 2195 | else |
| 2196 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2197 | "__VA_ARGS__ can only appear in the expansion" |
| 2198 | " of a C99 variadic macro"); |
| 2199 | } |
| 2200 | |
| 2201 | if (result == pfile->spec_nodes.n__VA_OPT__) |
| 2202 | maybe_va_opt_error (pfile); |
| 2203 | |
| 2204 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
| 2205 | if (result->flags & NODE_WARN_OPERATOR(1 << 7)) |
| 2206 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
| 2207 | "identifier \"%s\" is a special operator name in C++", |
| 2208 | NODE_NAME (result)(((&(result)->ident))->str)); |
| 2209 | } |
| 2210 | |
| 2211 | return result; |
| 2212 | } |
| 2213 | |
| 2214 | /* Get the cpp_hashnode of an identifier specified by NAME in |
| 2215 | the current cpp_reader object. If none is found, NULL is returned. */ |
| 2216 | cpp_hashnode * |
| 2217 | _cpp_lex_identifier (cpp_reader *pfile, const char *name) |
| 2218 | { |
| 2219 | cpp_hashnode *result; |
| 2220 | result = lex_identifier_intern (pfile, (uchar *) name); |
| 2221 | return result; |
| 2222 | } |
| 2223 | |
| 2224 | /* Lex an identifier starting at BUFFER->CUR - 1. */ |
| 2225 | static cpp_hashnode * |
| 2226 | lex_identifier (cpp_reader *pfile, const uchar *base, bool starts_ucn, |
| 2227 | struct normalize_state *nst, cpp_hashnode **spelling) |
| 2228 | { |
| 2229 | cpp_hashnode *result; |
| 2230 | const uchar *cur; |
| 2231 | unsigned int len; |
| 2232 | unsigned int hash = HT_HASHSTEP (0, *base)((0) * 67 + ((*base) - 113));; |
| 2233 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 2234 | |
| 2235 | cur = pfile->buffer->cur; |
| 2236 | if (! starts_ucn) |
| 2237 | { |
| 2238 | while (ISIDNUM (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 2239 | { |
| 2240 | hash = HT_HASHSTEP (hash, *cur)((hash) * 67 + ((*cur) - 113));; |
| 2241 | cur++; |
| 2242 | } |
| 2243 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *(cur - 1))((nst)->previous = (*(cur - 1)), (nst)->prev_class = 0); |
| 2244 | } |
| 2245 | pfile->buffer->cur = cur; |
| 2246 | if (starts_ucn || forms_identifier_p (pfile, false, nst)) |
| 2247 | { |
| 2248 | /* Slower version for identifiers containing UCNs |
| 2249 | or extended chars (including $). */ |
| 2250 | do { |
| 2251 | while (ISIDNUM (*pfile->buffer->cur)(_sch_istable[(*pfile->buffer->cur) & 0xff] & ( unsigned short)(_sch_isidnum))) |
| 2252 | { |
| 2253 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *pfile->buffer->cur)((nst)->previous = (*pfile->buffer->cur), (nst)-> prev_class = 0); |
| 2254 | pfile->buffer->cur++; |
| 2255 | } |
| 2256 | } while (forms_identifier_p (pfile, false, nst)); |
| 2257 | if (warn_bidi_p) |
| 2258 | maybe_warn_bidi_on_close (pfile, pfile->buffer->cur); |
| 2259 | result = _cpp_interpret_identifier (pfile, base, |
| 2260 | pfile->buffer->cur - base); |
| 2261 | *spelling = cpp_lookup (pfile, base, pfile->buffer->cur - base); |
| 2262 | } |
| 2263 | else |
| 2264 | { |
| 2265 | len = cur - base; |
| 2266 | hash = HT_HASHFINISH (hash, len)((hash) + (len)); |
| 2267 | |
| 2268 | result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table,((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC))) |
| 2269 | base, len, hash, HT_ALLOC))((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, len, hash, HT_ALLOC))); |
| 2270 | *spelling = result; |
| 2271 | } |
| 2272 | |
| 2273 | /* Rarely, identifiers require diagnostics when lexed. */ |
| 2274 | if (__builtin_expect ((result->flags & NODE_DIAGNOSTIC(1 << 2)) |
| 2275 | && !pfile->state.skipping, 0)) |
| 2276 | { |
| 2277 | /* It is allowed to poison the same identifier twice. */ |
| 2278 | if ((result->flags & NODE_POISONED(1 << 1)) && !pfile->state.poisoned_ok) |
| 2279 | cpp_error (pfile, CPP_DL_ERROR, "attempt to use poisoned \"%s\"", |
| 2280 | NODE_NAME (result)(((&(result)->ident))->str)); |
| 2281 | |
| 2282 | /* Constraint 6.10.3.5: __VA_ARGS__ should only appear in the |
| 2283 | replacement list of a variadic macro. */ |
| 2284 | if (result == pfile->spec_nodes.n__VA_ARGS__ |
| 2285 | && !pfile->state.va_args_ok) |
| 2286 | { |
| 2287 | if (CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus)) |
| 2288 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2289 | "__VA_ARGS__ can only appear in the expansion" |
| 2290 | " of a C++11 variadic macro"); |
| 2291 | else |
| 2292 | cpp_error (pfile, CPP_DL_PEDWARN, |
| 2293 | "__VA_ARGS__ can only appear in the expansion" |
| 2294 | " of a C99 variadic macro"); |
| 2295 | } |
| 2296 | |
| 2297 | /* __VA_OPT__ should only appear in the replacement list of a |
| 2298 | variadic macro. */ |
| 2299 | if (result == pfile->spec_nodes.n__VA_OPT__) |
| 2300 | maybe_va_opt_error (pfile); |
| 2301 | |
| 2302 | /* For -Wc++-compat, warn about use of C++ named operators. */ |
| 2303 | if (result->flags & NODE_WARN_OPERATOR(1 << 7)) |
| 2304 | cpp_warning (pfile, CPP_W_CXX_OPERATOR_NAMES, |
| 2305 | "identifier \"%s\" is a special operator name in C++", |
| 2306 | NODE_NAME (result)(((&(result)->ident))->str)); |
| 2307 | } |
| 2308 | |
| 2309 | return result; |
| 2310 | } |
| 2311 | |
| 2312 | /* Lex a number to NUMBER starting at BUFFER->CUR - 1. */ |
| 2313 | static void |
| 2314 | lex_number (cpp_reader *pfile, cpp_string *number, |
| 2315 | struct normalize_state *nst) |
| 2316 | { |
| 2317 | const uchar *cur; |
| 2318 | const uchar *base; |
| 2319 | uchar *dest; |
| 2320 | |
| 2321 | base = pfile->buffer->cur - 1; |
| 2322 | do |
| 2323 | { |
| 2324 | const uchar *adj_digit_sep = NULL__null; |
| 2325 | cur = pfile->buffer->cur; |
| 2326 | |
| 2327 | /* N.B. ISIDNUM does not include $. */ |
| 2328 | while (ISIDNUM (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidnum )) |
| 2329 | || (*cur == '.' && !DIGIT_SEP (cur[-1])((cur[-1]) == '\'' && ((pfile)->opts.digit_separators ))) |
| 2330 | || DIGIT_SEP (*cur)((*cur) == '\'' && ((pfile)->opts.digit_separators )) |
| 2331 | || (VALID_SIGN (*cur, cur[-1])(((*cur) == '+' || (*cur) == '-') && ((cur[-1]) == 'e' || (cur[-1]) == 'E' || (((cur[-1]) == 'p' || (cur[-1]) == 'P' ) && ((pfile)->opts.extended_numbers)))) && !DIGIT_SEP (cur[-2])((cur[-2]) == '\'' && ((pfile)->opts.digit_separators )))) |
| 2332 | { |
| 2333 | NORMALIZE_STATE_UPDATE_IDNUM (nst, *cur)((nst)->previous = (*cur), (nst)->prev_class = 0); |
| 2334 | /* Adjacent digit separators do not form part of the pp-number syntax. |
| 2335 | However, they can safely be diagnosed here as an error, since '' is |
| 2336 | not a valid preprocessing token. */ |
| 2337 | if (DIGIT_SEP (*cur)((*cur) == '\'' && ((pfile)->opts.digit_separators )) && DIGIT_SEP (cur[-1])((cur[-1]) == '\'' && ((pfile)->opts.digit_separators )) && !adj_digit_sep) |
| 2338 | adj_digit_sep = cur; |
| 2339 | cur++; |
| 2340 | } |
| 2341 | /* A number can't end with a digit separator. */ |
| 2342 | while (cur > pfile->buffer->cur && DIGIT_SEP (cur[-1])((cur[-1]) == '\'' && ((pfile)->opts.digit_separators ))) |
| 2343 | --cur; |
| 2344 | if (adj_digit_sep && adj_digit_sep < cur) |
| 2345 | cpp_error (pfile, CPP_DL_ERROR, "adjacent digit separators"); |
| 2346 | |
| 2347 | pfile->buffer->cur = cur; |
| 2348 | } |
| 2349 | while (forms_identifier_p (pfile, false, nst)); |
| 2350 | |
| 2351 | number->len = cur - base; |
| 2352 | dest = _cpp_unaligned_alloc (pfile, number->len + 1); |
| 2353 | memcpy (dest, base, number->len); |
| 2354 | dest[number->len] = '\0'; |
| 2355 | number->text = dest; |
| 2356 | } |
| 2357 | |
| 2358 | /* Create a token of type TYPE with a literal spelling. */ |
| 2359 | static void |
| 2360 | create_literal (cpp_reader *pfile, cpp_token *token, const uchar *base, |
| 2361 | unsigned int len, enum cpp_ttype type) |
| 2362 | { |
| 2363 | token->type = type; |
| 2364 | token->val.str.len = len; |
| 2365 | token->val.str.text = cpp_alloc_token_string (pfile, base, len); |
| 2366 | } |
| 2367 | |
| 2368 | const uchar * |
| 2369 | cpp_alloc_token_string (cpp_reader *pfile, |
| 2370 | const unsigned char *ptr, unsigned len) |
| 2371 | { |
| 2372 | uchar *dest = _cpp_unaligned_alloc (pfile, len + 1); |
| 2373 | |
| 2374 | dest[len] = 0; |
| 2375 | memcpy (dest, ptr, len); |
| 2376 | return dest; |
| 2377 | } |
| 2378 | |
| 2379 | /* A pair of raw buffer pointers. The currently open one is [1], the |
| 2380 | first one is [0]. Used for string literal lexing. */ |
| 2381 | struct lit_accum { |
| 2382 | _cpp_buff *first; |
| 2383 | _cpp_buff *last; |
| 2384 | const uchar *rpos; |
| 2385 | size_t accum; |
| 2386 | |
| 2387 | lit_accum () |
| 2388 | : first (NULL__null), last (NULL__null), rpos (0), accum (0) |
| 2389 | { |
| 2390 | } |
| 2391 | |
| 2392 | void append (cpp_reader *, const uchar *, size_t); |
| 2393 | |
| 2394 | void read_begin (cpp_reader *); |
| 2395 | bool reading_p () const |
| 2396 | { |
| 2397 | return rpos != NULL__null; |
| 2398 | } |
| 2399 | char read_char () |
| 2400 | { |
| 2401 | char c = *rpos++; |
| 2402 | if (rpos == BUFF_FRONT (last)((last)->cur)) |
| 2403 | rpos = NULL__null; |
| 2404 | return c; |
| 2405 | } |
| 2406 | }; |
| 2407 | |
| 2408 | /* Subroutine of lex_raw_string: Append LEN chars from BASE to the buffer |
| 2409 | sequence from *FIRST_BUFF_P to LAST_BUFF_P. */ |
| 2410 | |
| 2411 | void |
| 2412 | lit_accum::append (cpp_reader *pfile, const uchar *base, size_t len) |
| 2413 | { |
| 2414 | if (!last) |
| 2415 | /* Starting. */ |
| 2416 | first = last = _cpp_get_buff (pfile, len); |
| 2417 | else if (len > BUFF_ROOM (last)(size_t) ((last)->limit - (last)->cur)) |
| 2418 | { |
| 2419 | /* There is insufficient room in the buffer. Copy what we can, |
| 2420 | and then either extend or create a new one. */ |
| 2421 | size_t room = BUFF_ROOM (last)(size_t) ((last)->limit - (last)->cur); |
| 2422 | memcpy (BUFF_FRONT (last)((last)->cur), base, room); |
| 2423 | BUFF_FRONT (last)((last)->cur) += room; |
| 2424 | base += room; |
| 2425 | len -= room; |
| 2426 | accum += room; |
| 2427 | |
| 2428 | gcc_checking_assert (!rpos)((void)(!(!rpos) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2428, __FUNCTION__), 0 : 0)); |
| 2429 | |
| 2430 | last = _cpp_append_extend_buff (pfile, last, len); |
| 2431 | } |
| 2432 | |
| 2433 | memcpy (BUFF_FRONT (last)((last)->cur), base, len); |
| 2434 | BUFF_FRONT (last)((last)->cur) += len; |
| 2435 | accum += len; |
| 2436 | } |
| 2437 | |
| 2438 | void |
| 2439 | lit_accum::read_begin (cpp_reader *pfile) |
| 2440 | { |
| 2441 | /* We never accumulate more than 4 chars to read. */ |
| 2442 | if (BUFF_ROOM (last)(size_t) ((last)->limit - (last)->cur) < 4) |
| 2443 | |
| 2444 | last = _cpp_append_extend_buff (pfile, last, 4); |
| 2445 | rpos = BUFF_FRONT (last)((last)->cur); |
| 2446 | } |
| 2447 | |
| 2448 | /* Returns true if a macro has been defined. |
| 2449 | This might not work if compile with -save-temps, |
| 2450 | or preprocess separately from compilation. */ |
| 2451 | |
| 2452 | static bool |
| 2453 | is_macro(cpp_reader *pfile, const uchar *base) |
| 2454 | { |
| 2455 | const uchar *cur = base; |
| 2456 | if (! ISIDST (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidst ))) |
| 2457 | return false; |
| 2458 | unsigned int hash = HT_HASHSTEP (0, *cur)((0) * 67 + ((*cur) - 113));; |
| 2459 | ++cur; |
| 2460 | while (ISIDNUM (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 2461 | { |
| 2462 | hash = HT_HASHSTEP (hash, *cur)((hash) * 67 + ((*cur) - 113));; |
| 2463 | ++cur; |
| 2464 | } |
| 2465 | hash = HT_HASHFINISH (hash, cur - base)((hash) + (cur - base)); |
| 2466 | |
| 2467 | cpp_hashnode *result = CPP_HASHNODE (ht_lookup_with_hash (pfile->hash_table,((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, cur - base, hash, HT_NO_INSERT))) |
| 2468 | base, cur - base, hash, HT_NO_INSERT))((cpp_hashnode *) (ht_lookup_with_hash (pfile->hash_table, base, cur - base, hash, HT_NO_INSERT))); |
| 2469 | |
| 2470 | return result && cpp_macro_p (result); |
| 2471 | } |
| 2472 | |
| 2473 | /* Returns true if a literal suffix does not have the expected form |
| 2474 | and is defined as a macro. */ |
| 2475 | |
| 2476 | static bool |
| 2477 | is_macro_not_literal_suffix(cpp_reader *pfile, const uchar *base) |
| 2478 | { |
| 2479 | /* User-defined literals outside of namespace std must start with a single |
| 2480 | underscore, so assume anything of that form really is a UDL suffix. |
| 2481 | We don't need to worry about UDLs defined inside namespace std because |
| 2482 | their names are reserved, so cannot be used as macro names in valid |
| 2483 | programs. */ |
| 2484 | if (base[0] == '_' && base[1] != '_') |
| 2485 | return false; |
| 2486 | return is_macro (pfile, base); |
| 2487 | } |
| 2488 | |
| 2489 | /* Lexes a raw string. The stored string contains the spelling, |
| 2490 | including double quotes, delimiter string, '(' and ')', any leading |
| 2491 | 'L', 'u', 'U' or 'u8' and 'R' modifier. The created token contains |
| 2492 | the type of the literal, or CPP_OTHER if it was not properly |
| 2493 | terminated. |
| 2494 | |
| 2495 | BASE is the start of the token. Updates pfile->buffer->cur to just |
| 2496 | after the lexed string. |
| 2497 | |
| 2498 | The spelling is NUL-terminated, but it is not guaranteed that this |
| 2499 | is the first NUL since embedded NULs are preserved. */ |
| 2500 | |
| 2501 | static void |
| 2502 | lex_raw_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
| 2503 | { |
| 2504 | const uchar *pos = base; |
| 2505 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 2506 | const bool warn_invalid_utf8_p = CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8); |
| 2507 | const bool warn_bidi_or_invalid_utf8_p = warn_bidi_p | warn_invalid_utf8_p; |
| 2508 | |
| 2509 | /* 'tis a pity this information isn't passed down from the lexer's |
| 2510 | initial categorization of the token. */ |
| 2511 | enum cpp_ttype type = CPP_STRING; |
| 2512 | |
| 2513 | if (*pos == 'L') |
| 2514 | { |
| 2515 | type = CPP_WSTRING; |
| 2516 | pos++; |
| 2517 | } |
| 2518 | else if (*pos == 'U') |
| 2519 | { |
| 2520 | type = CPP_STRING32; |
| 2521 | pos++; |
| 2522 | } |
| 2523 | else if (*pos == 'u') |
| 2524 | { |
| 2525 | if (pos[1] == '8') |
| 2526 | { |
| 2527 | type = CPP_UTF8STRING; |
| 2528 | pos++; |
| 2529 | } |
| 2530 | else |
| 2531 | type = CPP_STRING16; |
| 2532 | pos++; |
| 2533 | } |
| 2534 | |
| 2535 | gcc_checking_assert (pos[0] == 'R' && pos[1] == '"')((void)(!(pos[0] == 'R' && pos[1] == '"') ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2535, __FUNCTION__), 0 : 0)); |
| 2536 | pos += 2; |
| 2537 | |
| 2538 | _cpp_line_note *note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
| 2539 | |
| 2540 | /* Skip notes before the ". */ |
| 2541 | while (note->pos < pos) |
| 2542 | ++note; |
| 2543 | |
| 2544 | lit_accum accum; |
| 2545 | |
| 2546 | uchar prefix[17]; |
| 2547 | unsigned prefix_len = 0; |
| 2548 | enum Phase |
| 2549 | { |
| 2550 | PHASE_PREFIX = -2, |
| 2551 | PHASE_NONE = -1, |
| 2552 | PHASE_SUFFIX = 0 |
| 2553 | } phase = PHASE_PREFIX; |
| 2554 | |
| 2555 | for (;;) |
| 2556 | { |
| 2557 | gcc_checking_assert (note->pos >= pos)((void)(!(note->pos >= pos) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2557, __FUNCTION__), 0 : 0)); |
| 2558 | |
| 2559 | /* Undo any escaped newlines and trigraphs. */ |
| 2560 | if (!accum.reading_p () && note->pos == pos) |
| 2561 | switch (note->type) |
| 2562 | { |
| 2563 | case '\\': |
| 2564 | case ' ': |
| 2565 | /* Restore backslash followed by newline. */ |
| 2566 | accum.append (pfile, base, pos - base); |
| 2567 | base = pos; |
| 2568 | accum.read_begin (pfile); |
| 2569 | accum.append (pfile, UC(const uchar *)"\\", 1); |
| 2570 | |
| 2571 | after_backslash: |
| 2572 | if (note->type == ' ') |
| 2573 | /* GNU backslash whitespace newline extension. FIXME |
| 2574 | could be any sequence of non-vertical space. When we |
| 2575 | can properly restore any such sequence, we should |
| 2576 | mark this note as handled so _cpp_process_line_notes |
| 2577 | doesn't warn. */ |
| 2578 | accum.append (pfile, UC(const uchar *)" ", 1); |
| 2579 | |
| 2580 | accum.append (pfile, UC(const uchar *)"\n", 1); |
| 2581 | note++; |
| 2582 | break; |
| 2583 | |
| 2584 | case '\n': |
| 2585 | /* This can happen for ??/<NEWLINE> when trigraphs are not |
| 2586 | being interpretted. */ |
| 2587 | gcc_checking_assert (!CPP_OPTION (pfile, trigraphs))((void)(!(!((pfile)->opts.trigraphs)) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2587, __FUNCTION__), 0 : 0)); |
| 2588 | note->type = 0; |
| 2589 | note++; |
| 2590 | break; |
| 2591 | |
| 2592 | default: |
| 2593 | gcc_checking_assert (_cpp_trigraph_map[note->type])((void)(!(_cpp_trigraph_map[note->type]) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2593, __FUNCTION__), 0 : 0)); |
| 2594 | |
| 2595 | /* Don't warn about this trigraph in |
| 2596 | _cpp_process_line_notes, since trigraphs show up as |
| 2597 | trigraphs in raw strings. */ |
| 2598 | uchar type = note->type; |
| 2599 | note->type = 0; |
| 2600 | |
| 2601 | if (CPP_OPTION (pfile, trigraphs)((pfile)->opts.trigraphs)) |
| 2602 | { |
| 2603 | accum.append (pfile, base, pos - base); |
| 2604 | base = pos; |
| 2605 | accum.read_begin (pfile); |
| 2606 | accum.append (pfile, UC(const uchar *)"??", 2); |
| 2607 | accum.append (pfile, &type, 1); |
| 2608 | |
| 2609 | /* ??/ followed by newline gets two line notes, one for |
| 2610 | the trigraph and one for the backslash/newline. */ |
| 2611 | if (type == '/' && note[1].pos == pos) |
| 2612 | { |
| 2613 | note++; |
| 2614 | gcc_assert (note->type == '\\' || note->type == ' ')((void)(!(note->type == '\\' || note->type == ' ') ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2614, __FUNCTION__), 0 : 0)); |
| 2615 | goto after_backslash; |
| 2616 | } |
| 2617 | /* Skip the replacement character. */ |
| 2618 | base = ++pos; |
| 2619 | } |
| 2620 | |
| 2621 | note++; |
| 2622 | break; |
| 2623 | } |
| 2624 | |
| 2625 | /* Now get a char to process. Either from an expanded note, or |
| 2626 | from the line buffer. */ |
| 2627 | bool read_note = accum.reading_p (); |
| 2628 | char c = read_note ? accum.read_char () : *pos++; |
| 2629 | |
| 2630 | if (phase == PHASE_PREFIX) |
| 2631 | { |
| 2632 | if (c == '(') |
| 2633 | { |
| 2634 | /* Done. */ |
| 2635 | phase = PHASE_NONE; |
| 2636 | prefix[prefix_len++] = '"'; |
| 2637 | } |
| 2638 | else if (prefix_len < 16 |
| 2639 | /* Prefix chars are any of the basic character set, |
| 2640 | [lex.charset] except for ' |
| 2641 | ()\\\t\v\f\n'. Optimized for a contiguous |
| 2642 | alphabet. */ |
| 2643 | /* Unlike a switch, this collapses down to one or |
| 2644 | two shift and bitmask operations on an ASCII |
| 2645 | system, with an outlier or two. */ |
| 2646 | && (('Z' - 'A' == 25 |
| 2647 | ? ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) |
| 2648 | : ISIDST (c)(_sch_istable[(c) & 0xff] & (unsigned short)(_sch_isidst ))) |
| 2649 | || (c >= '0' && c <= '9') |
| 2650 | || c == '_' || c == '{' || c == '}' |
| 2651 | || c == '[' || c == ']' || c == '#' |
| 2652 | || c == '<' || c == '>' || c == '%' |
| 2653 | || c == ':' || c == ';' || c == '.' || c == '?' |
| 2654 | || c == '*' || c == '+' || c == '-' || c == '/' |
| 2655 | || c == '^' || c == '&' || c == '|' || c == '~' |
| 2656 | || c == '!' || c == '=' || c == ',' |
| 2657 | || c == '"' || c == '\'')) |
| 2658 | prefix[prefix_len++] = c; |
| 2659 | else |
| 2660 | { |
| 2661 | /* Something is wrong. */ |
| 2662 | int col = CPP_BUF_COLUMN (pfile->buffer, pos)((pos) - (pfile->buffer)->line_base) + read_note; |
| 2663 | if (prefix_len == 16) |
| 2664 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, |
| 2665 | col, "raw string delimiter longer " |
| 2666 | "than 16 characters"); |
| 2667 | else if (c == '\n') |
| 2668 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, |
| 2669 | col, "invalid new-line in raw " |
| 2670 | "string delimiter"); |
| 2671 | else |
| 2672 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, |
| 2673 | col, "invalid character '%c' in " |
| 2674 | "raw string delimiter", c); |
| 2675 | type = CPP_OTHER; |
| 2676 | phase = PHASE_NONE; |
| 2677 | /* Continue until we get a close quote, that's probably |
| 2678 | the best failure mode. */ |
| 2679 | prefix_len = 0; |
| 2680 | } |
| 2681 | if (c != '\n') |
| 2682 | continue; |
| 2683 | } |
| 2684 | |
| 2685 | if (phase != PHASE_NONE) |
| 2686 | { |
| 2687 | if (prefix[phase] != c) |
| 2688 | phase = PHASE_NONE; |
| 2689 | else if (unsigned (phase + 1) == prefix_len) |
| 2690 | break; |
| 2691 | else |
| 2692 | { |
| 2693 | phase = Phase (phase + 1); |
| 2694 | continue; |
| 2695 | } |
| 2696 | } |
| 2697 | |
| 2698 | if (!prefix_len && c == '"') |
| 2699 | /* Failure mode lexing. */ |
| 2700 | goto out; |
| 2701 | else if (prefix_len && c == ')') |
| 2702 | phase = PHASE_SUFFIX; |
| 2703 | else if (!read_note && c == '\n') |
| 2704 | { |
| 2705 | pos--; |
| 2706 | pfile->buffer->cur = pos; |
| 2707 | if ((pfile->state.in_directive || pfile->state.parsing_args) |
| 2708 | && pfile->buffer->next_line >= pfile->buffer->rlimit) |
| 2709 | { |
| 2710 | cpp_error_with_line (pfile, CPP_DL_ERROR, token->src_loc, 0, |
| 2711 | "unterminated raw string"); |
| 2712 | type = CPP_OTHER; |
| 2713 | goto out; |
| 2714 | } |
| 2715 | |
| 2716 | accum.append (pfile, base, pos - base + 1); |
| 2717 | _cpp_process_line_notes (pfile, false); |
| 2718 | |
| 2719 | if (pfile->buffer->next_line < pfile->buffer->rlimit) |
| 2720 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 2721 | pfile->buffer->need_line = true; |
| 2722 | |
| 2723 | if (!get_fresh_line_impl<true> (pfile)) |
| 2724 | { |
| 2725 | /* We ran out of file and failed to get a line. */ |
| 2726 | location_t src_loc = token->src_loc; |
| 2727 | token->type = CPP_EOF; |
| 2728 | /* Tell the compiler the line number of the EOF token. */ |
| 2729 | token->src_loc = pfile->line_table->highest_line; |
| 2730 | token->flags = BOL(1 << 12); |
| 2731 | if (accum.first) |
| 2732 | _cpp_release_buff (pfile, accum.first); |
| 2733 | cpp_error_with_line (pfile, CPP_DL_ERROR, src_loc, 0, |
| 2734 | "unterminated raw string"); |
| 2735 | |
| 2736 | /* Now pop the buffer that get_fresh_line_impl() did not. Popping |
| 2737 | is not safe if processing a directive, however this cannot |
| 2738 | happen as we already checked above that a line would be |
| 2739 | available, and get_fresh_line_impl() can't fail in this |
| 2740 | case. */ |
| 2741 | gcc_assert (!pfile->state.in_directive)((void)(!(!pfile->state.in_directive) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 2741, __FUNCTION__), 0 : 0)); |
| 2742 | _cpp_pop_buffer (pfile); |
| 2743 | |
| 2744 | return; |
| 2745 | } |
| 2746 | |
| 2747 | pos = base = pfile->buffer->cur; |
| 2748 | note = &pfile->buffer->notes[pfile->buffer->cur_note]; |
| 2749 | } |
| 2750 | else if (__builtin_expect ((unsigned char) c >= utf8_continuation, 0) |
| 2751 | && warn_bidi_or_invalid_utf8_p) |
| 2752 | pos = _cpp_handle_multibyte_utf8 (pfile, c, pos, warn_bidi_p, |
| 2753 | warn_invalid_utf8_p); |
| 2754 | } |
| 2755 | |
| 2756 | if (warn_bidi_p) |
| 2757 | maybe_warn_bidi_on_close (pfile, pos); |
| 2758 | |
| 2759 | if (CPP_OPTION (pfile, user_literals)((pfile)->opts.user_literals)) |
| 2760 | { |
| 2761 | /* If a string format macro, say from inttypes.h, is placed touching |
| 2762 | a string literal it could be parsed as a C++11 user-defined string |
| 2763 | literal thus breaking the program. */ |
| 2764 | if (is_macro_not_literal_suffix (pfile, pos)) |
| 2765 | { |
| 2766 | /* Raise a warning, but do not consume subsequent tokens. */ |
| 2767 | if (CPP_OPTION (pfile, warn_literal_suffix)((pfile)->opts.warn_literal_suffix) && !pfile->state.skipping) |
| 2768 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
| 2769 | token->src_loc, 0, |
| 2770 | "invalid suffix on literal; C++11 requires " |
| 2771 | "a space between literal and string macro"); |
| 2772 | } |
| 2773 | /* Grab user defined literal suffix. */ |
| 2774 | else if (ISIDST (*pos)(_sch_istable[(*pos) & 0xff] & (unsigned short)(_sch_isidst ))) |
| 2775 | { |
| 2776 | type = cpp_userdef_string_add_type (type); |
| 2777 | ++pos; |
| 2778 | |
| 2779 | while (ISIDNUM (*pos)(_sch_istable[(*pos) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 2780 | ++pos; |
| 2781 | } |
| 2782 | } |
| 2783 | |
| 2784 | out: |
| 2785 | pfile->buffer->cur = pos; |
| 2786 | if (!accum.accum) |
| 2787 | create_literal (pfile, token, base, pos - base, type); |
| 2788 | else |
| 2789 | { |
| 2790 | size_t extra_len = pos - base; |
| 2791 | uchar *dest = _cpp_unaligned_alloc (pfile, accum.accum + extra_len + 1); |
| 2792 | |
| 2793 | token->type = type; |
| 2794 | token->val.str.len = accum.accum + extra_len; |
| 2795 | token->val.str.text = dest; |
| 2796 | for (_cpp_buff *buf = accum.first; buf; buf = buf->next) |
| 2797 | { |
| 2798 | size_t len = BUFF_FRONT (buf)((buf)->cur) - buf->base; |
| 2799 | memcpy (dest, buf->base, len); |
| 2800 | dest += len; |
| 2801 | } |
| 2802 | _cpp_release_buff (pfile, accum.first); |
| 2803 | memcpy (dest, base, extra_len); |
| 2804 | dest[extra_len] = '\0'; |
| 2805 | } |
| 2806 | } |
| 2807 | |
| 2808 | /* Lexes a string, character constant, or angle-bracketed header file |
| 2809 | name. The stored string contains the spelling, including opening |
| 2810 | quote and any leading 'L', 'u', 'U' or 'u8' and optional |
| 2811 | 'R' modifier. It returns the type of the literal, or CPP_OTHER |
| 2812 | if it was not properly terminated, or CPP_LESS for an unterminated |
| 2813 | header name which must be relexed as normal tokens. |
| 2814 | |
| 2815 | The spelling is NUL-terminated, but it is not guaranteed that this |
| 2816 | is the first NUL since embedded NULs are preserved. */ |
| 2817 | static void |
| 2818 | lex_string (cpp_reader *pfile, cpp_token *token, const uchar *base) |
| 2819 | { |
| 2820 | bool saw_NUL = false; |
| 2821 | const uchar *cur; |
| 2822 | cppchar_t terminator; |
| 2823 | enum cpp_ttype type; |
| 2824 | |
| 2825 | cur = base; |
| 2826 | terminator = *cur++; |
| 2827 | if (terminator == 'L' || terminator == 'U') |
| 2828 | terminator = *cur++; |
| 2829 | else if (terminator == 'u') |
| 2830 | { |
| 2831 | terminator = *cur++; |
| 2832 | if (terminator == '8') |
| 2833 | terminator = *cur++; |
| 2834 | } |
| 2835 | if (terminator == 'R') |
| 2836 | { |
| 2837 | lex_raw_string (pfile, token, base); |
| 2838 | return; |
| 2839 | } |
| 2840 | if (terminator == '"') |
| 2841 | type = (*base == 'L' ? CPP_WSTRING : |
| 2842 | *base == 'U' ? CPP_STRING32 : |
| 2843 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8STRING : CPP_STRING16) |
| 2844 | : CPP_STRING); |
| 2845 | else if (terminator == '\'') |
| 2846 | type = (*base == 'L' ? CPP_WCHAR : |
| 2847 | *base == 'U' ? CPP_CHAR32 : |
| 2848 | *base == 'u' ? (base[1] == '8' ? CPP_UTF8CHAR : CPP_CHAR16) |
| 2849 | : CPP_CHAR); |
| 2850 | else |
| 2851 | terminator = '>', type = CPP_HEADER_NAME; |
| 2852 | |
| 2853 | const bool warn_bidi_p = pfile->warn_bidi_p (); |
| 2854 | const bool warn_invalid_utf8_p = CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8); |
| 2855 | const bool warn_bidi_or_invalid_utf8_p = warn_bidi_p | warn_invalid_utf8_p; |
| 2856 | for (;;) |
| 2857 | { |
| 2858 | cppchar_t c = *cur++; |
| 2859 | |
| 2860 | /* In #include-style directives, terminators are not escapable. */ |
| 2861 | if (c == '\\' && !pfile->state.angled_headers && *cur != '\n') |
| 2862 | { |
| 2863 | if ((cur[0] == 'u' || cur[0] == 'U' || cur[0] == 'N') && warn_bidi_p) |
| 2864 | { |
| 2865 | location_t loc; |
| 2866 | bidi::kind kind; |
| 2867 | if (cur[0] == 'N') |
| 2868 | kind = get_bidi_named (pfile, cur + 1, &loc); |
| 2869 | else |
| 2870 | kind = get_bidi_ucn (pfile, cur + 1, cur[0] == 'U', &loc); |
| 2871 | maybe_warn_bidi_on_char (pfile, kind, /*ucn_p=*/true, loc); |
| 2872 | } |
| 2873 | cur++; |
| 2874 | } |
| 2875 | else if (c == terminator) |
| 2876 | { |
| 2877 | if (warn_bidi_p) |
| 2878 | maybe_warn_bidi_on_close (pfile, cur - 1); |
| 2879 | break; |
| 2880 | } |
| 2881 | else if (c == '\n') |
| 2882 | { |
| 2883 | cur--; |
| 2884 | /* Unmatched quotes always yield undefined behavior, but |
| 2885 | greedy lexing means that what appears to be an unterminated |
| 2886 | header name may actually be a legitimate sequence of tokens. */ |
| 2887 | if (terminator == '>') |
| 2888 | { |
| 2889 | token->type = CPP_LESS; |
| 2890 | return; |
| 2891 | } |
| 2892 | type = CPP_OTHER; |
| 2893 | break; |
| 2894 | } |
| 2895 | else if (c == '\0') |
| 2896 | saw_NUL = true; |
| 2897 | else if (__builtin_expect (c >= utf8_continuation, 0) |
| 2898 | && warn_bidi_or_invalid_utf8_p) |
| 2899 | cur = _cpp_handle_multibyte_utf8 (pfile, c, cur, warn_bidi_p, |
| 2900 | warn_invalid_utf8_p); |
| 2901 | } |
| 2902 | |
| 2903 | if (saw_NUL && !pfile->state.skipping) |
| 2904 | cpp_error (pfile, CPP_DL_WARNING, |
| 2905 | "null character(s) preserved in literal"); |
| 2906 | |
| 2907 | if (type == CPP_OTHER && CPP_OPTION (pfile, lang)((pfile)->opts.lang) != CLK_ASM) |
| 2908 | cpp_error (pfile, CPP_DL_PEDWARN, "missing terminating %c character", |
| 2909 | (int) terminator); |
| 2910 | |
| 2911 | if (CPP_OPTION (pfile, user_literals)((pfile)->opts.user_literals)) |
| 2912 | { |
| 2913 | /* If a string format macro, say from inttypes.h, is placed touching |
| 2914 | a string literal it could be parsed as a C++11 user-defined string |
| 2915 | literal thus breaking the program. */ |
| 2916 | if (is_macro_not_literal_suffix (pfile, cur)) |
| 2917 | { |
| 2918 | /* Raise a warning, but do not consume subsequent tokens. */ |
| 2919 | if (CPP_OPTION (pfile, warn_literal_suffix)((pfile)->opts.warn_literal_suffix) && !pfile->state.skipping) |
| 2920 | cpp_warning_with_line (pfile, CPP_W_LITERAL_SUFFIX, |
| 2921 | token->src_loc, 0, |
| 2922 | "invalid suffix on literal; C++11 requires " |
| 2923 | "a space between literal and string macro"); |
| 2924 | } |
| 2925 | /* Grab user defined literal suffix. */ |
| 2926 | else if (ISIDST (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidst ))) |
| 2927 | { |
| 2928 | type = cpp_userdef_char_add_type (type); |
| 2929 | type = cpp_userdef_string_add_type (type); |
| 2930 | ++cur; |
| 2931 | |
| 2932 | while (ISIDNUM (*cur)(_sch_istable[(*cur) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 2933 | ++cur; |
| 2934 | } |
| 2935 | } |
| 2936 | else if (CPP_OPTION (pfile, cpp_warn_cxx11_compat)((pfile)->opts.cpp_warn_cxx11_compat) |
| 2937 | && is_macro (pfile, cur) |
| 2938 | && !pfile->state.skipping) |
| 2939 | cpp_warning_with_line (pfile, CPP_W_CXX11_COMPAT, |
| 2940 | token->src_loc, 0, "C++11 requires a space " |
| 2941 | "between string literal and macro"); |
| 2942 | |
| 2943 | pfile->buffer->cur = cur; |
| 2944 | create_literal (pfile, token, base, cur - base, type); |
| 2945 | } |
| 2946 | |
| 2947 | /* Return the comment table. The client may not make any assumption |
| 2948 | about the ordering of the table. */ |
| 2949 | cpp_comment_table * |
| 2950 | cpp_get_comments (cpp_reader *pfile) |
| 2951 | { |
| 2952 | return &pfile->comments; |
| 2953 | } |
| 2954 | |
| 2955 | /* Append a comment to the end of the comment table. */ |
| 2956 | static void |
| 2957 | store_comment (cpp_reader *pfile, cpp_token *token) |
| 2958 | { |
| 2959 | int len; |
| 2960 | |
| 2961 | if (pfile->comments.allocated == 0) |
| 2962 | { |
| 2963 | pfile->comments.allocated = 256; |
| 2964 | pfile->comments.entries = (cpp_comment *) xmalloc |
| 2965 | (pfile->comments.allocated * sizeof (cpp_comment)); |
| 2966 | } |
| 2967 | |
| 2968 | if (pfile->comments.count == pfile->comments.allocated) |
| 2969 | { |
| 2970 | pfile->comments.allocated *= 2; |
| 2971 | pfile->comments.entries = (cpp_comment *) xrealloc |
| 2972 | (pfile->comments.entries, |
| 2973 | pfile->comments.allocated * sizeof (cpp_comment)); |
| 2974 | } |
| 2975 | |
| 2976 | len = token->val.str.len; |
| 2977 | |
| 2978 | /* Copy comment. Note, token may not be NULL terminated. */ |
| 2979 | pfile->comments.entries[pfile->comments.count].comment = |
| 2980 | (char *) xmalloc (sizeof (char) * (len + 1)); |
| 2981 | memcpy (pfile->comments.entries[pfile->comments.count].comment, |
| 2982 | token->val.str.text, len); |
| 2983 | pfile->comments.entries[pfile->comments.count].comment[len] = '\0'; |
| 2984 | |
| 2985 | /* Set source location. */ |
| 2986 | pfile->comments.entries[pfile->comments.count].sloc = token->src_loc; |
| 2987 | |
| 2988 | /* Increment the count of entries in the comment table. */ |
| 2989 | pfile->comments.count++; |
| 2990 | } |
| 2991 | |
| 2992 | /* The stored comment includes the comment start and any terminator. */ |
| 2993 | static void |
| 2994 | save_comment (cpp_reader *pfile, cpp_token *token, const unsigned char *from, |
| 2995 | cppchar_t type) |
| 2996 | { |
| 2997 | unsigned char *buffer; |
| 2998 | unsigned int len, clen, i; |
| 2999 | |
| 3000 | len = pfile->buffer->cur - from + 1; /* + 1 for the initial '/'. */ |
| 3001 | |
| 3002 | /* C++ comments probably (not definitely) have moved past a new |
| 3003 | line, which we don't want to save in the comment. */ |
| 3004 | if (is_vspace (pfile->buffer->cur[-1])(_sch_istable[(pfile->buffer->cur[-1]) & 0xff] & (unsigned short)(_sch_isvsp))) |
| 3005 | len--; |
| 3006 | |
| 3007 | /* If we are currently in a directive or in argument parsing, then |
| 3008 | we need to store all C++ comments as C comments internally, and |
| 3009 | so we need to allocate a little extra space in that case. |
| 3010 | |
| 3011 | Note that the only time we encounter a directive here is |
| 3012 | when we are saving comments in a "#define". */ |
| 3013 | clen = ((pfile->state.in_directive || pfile->state.parsing_args) |
| 3014 | && type == '/') ? len + 2 : len; |
| 3015 | |
| 3016 | buffer = _cpp_unaligned_alloc (pfile, clen); |
| 3017 | |
| 3018 | token->type = CPP_COMMENT; |
| 3019 | token->val.str.len = clen; |
| 3020 | token->val.str.text = buffer; |
| 3021 | |
| 3022 | buffer[0] = '/'; |
| 3023 | memcpy (buffer + 1, from, len - 1); |
| 3024 | |
| 3025 | /* Finish conversion to a C comment, if necessary. */ |
| 3026 | if ((pfile->state.in_directive || pfile->state.parsing_args) && type == '/') |
| 3027 | { |
| 3028 | buffer[1] = '*'; |
| 3029 | buffer[clen - 2] = '*'; |
| 3030 | buffer[clen - 1] = '/'; |
| 3031 | /* As there can be in a C++ comments illegal sequences for C comments |
| 3032 | we need to filter them out. */ |
| 3033 | for (i = 2; i < (clen - 2); i++) |
| 3034 | if (buffer[i] == '/' && (buffer[i - 1] == '*' || buffer[i + 1] == '*')) |
| 3035 | buffer[i] = '|'; |
| 3036 | } |
| 3037 | |
| 3038 | /* Finally store this comment for use by clients of libcpp. */ |
| 3039 | store_comment (pfile, token); |
| 3040 | } |
| 3041 | |
| 3042 | /* Returns true if comment at COMMENT_START is a recognized FALLTHROUGH |
| 3043 | comment. */ |
| 3044 | |
| 3045 | static bool |
| 3046 | fallthrough_comment_p (cpp_reader *pfile, const unsigned char *comment_start) |
| 3047 | { |
| 3048 | const unsigned char *from = comment_start + 1; |
| 3049 | |
| 3050 | switch (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough)((pfile)->opts.cpp_warn_implicit_fallthrough)) |
| 3051 | { |
| 3052 | /* For both -Wimplicit-fallthrough=0 and -Wimplicit-fallthrough=5 we |
| 3053 | don't recognize any comments. The latter only checks attributes, |
| 3054 | the former doesn't warn. */ |
| 3055 | case 0: |
| 3056 | default: |
| 3057 | return false; |
| 3058 | /* -Wimplicit-fallthrough=1 considers any comment, no matter what |
| 3059 | content it has. */ |
| 3060 | case 1: |
| 3061 | return true; |
| 3062 | case 2: |
| 3063 | /* -Wimplicit-fallthrough=2 looks for (case insensitive) |
| 3064 | .*falls?[ \t-]*thr(u|ough).* regex. */ |
| 3065 | for (; (size_t) (pfile->buffer->cur - from) >= sizeof "fallthru" - 1; |
| 3066 | from++) |
| 3067 | { |
| 3068 | /* Is there anything like strpbrk with upper boundary, or |
| 3069 | memchr looking for 2 characters rather than just one? */ |
| 3070 | if (from[0] != 'f' && from[0] != 'F') |
| 3071 | continue; |
| 3072 | if (from[1] != 'a' && from[1] != 'A') |
| 3073 | continue; |
| 3074 | if (from[2] != 'l' && from[2] != 'L') |
| 3075 | continue; |
| 3076 | if (from[3] != 'l' && from[3] != 'L') |
| 3077 | continue; |
| 3078 | from += sizeof "fall" - 1; |
| 3079 | if (from[0] == 's' || from[0] == 'S') |
| 3080 | from++; |
| 3081 | while (*from == ' ' || *from == '\t' || *from == '-') |
| 3082 | from++; |
| 3083 | if (from[0] != 't' && from[0] != 'T') |
| 3084 | continue; |
| 3085 | if (from[1] != 'h' && from[1] != 'H') |
| 3086 | continue; |
| 3087 | if (from[2] != 'r' && from[2] != 'R') |
| 3088 | continue; |
| 3089 | if (from[3] == 'u' || from[3] == 'U') |
| 3090 | return true; |
| 3091 | if (from[3] != 'o' && from[3] != 'O') |
| 3092 | continue; |
| 3093 | if (from[4] != 'u' && from[4] != 'U') |
| 3094 | continue; |
| 3095 | if (from[5] != 'g' && from[5] != 'G') |
| 3096 | continue; |
| 3097 | if (from[6] != 'h' && from[6] != 'H') |
| 3098 | continue; |
| 3099 | return true; |
| 3100 | } |
| 3101 | return false; |
| 3102 | case 3: |
| 3103 | case 4: |
| 3104 | break; |
| 3105 | } |
| 3106 | |
| 3107 | /* Whole comment contents: |
| 3108 | -fallthrough |
| 3109 | @fallthrough@ |
| 3110 | */ |
| 3111 | if (*from == '-' || *from == '@') |
| 3112 | { |
| 3113 | size_t len = sizeof "fallthrough" - 1; |
| 3114 | if ((size_t) (pfile->buffer->cur - from - 1) < len) |
| 3115 | return false; |
| 3116 | if (memcmp (from + 1, "fallthrough", len)) |
| 3117 | return false; |
| 3118 | if (*from == '@') |
| 3119 | { |
| 3120 | if (from[len + 1] != '@') |
| 3121 | return false; |
| 3122 | len++; |
| 3123 | } |
| 3124 | from += 1 + len; |
| 3125 | } |
| 3126 | /* Whole comment contents (regex): |
| 3127 | lint -fallthrough[ \t]* |
| 3128 | */ |
| 3129 | else if (*from == 'l') |
| 3130 | { |
| 3131 | size_t len = sizeof "int -fallthrough" - 1; |
| 3132 | if ((size_t) (pfile->buffer->cur - from - 1) < len) |
| 3133 | return false; |
| 3134 | if (memcmp (from + 1, "int -fallthrough", len)) |
| 3135 | return false; |
| 3136 | from += 1 + len; |
| 3137 | while (*from == ' ' || *from == '\t') |
| 3138 | from++; |
| 3139 | } |
| 3140 | /* Whole comment contents (regex): |
| 3141 | [ \t]*FALLTHR(U|OUGH)[ \t]* |
| 3142 | */ |
| 3143 | else if (CPP_OPTION (pfile, cpp_warn_implicit_fallthrough)((pfile)->opts.cpp_warn_implicit_fallthrough) == 4) |
| 3144 | { |
| 3145 | while (*from == ' ' || *from == '\t') |
| 3146 | from++; |
| 3147 | if ((size_t) (pfile->buffer->cur - from) < sizeof "FALLTHRU" - 1) |
| 3148 | return false; |
| 3149 | if (memcmp (from, "FALLTHR", sizeof "FALLTHR" - 1)) |
| 3150 | return false; |
| 3151 | from += sizeof "FALLTHR" - 1; |
| 3152 | if (*from == 'U') |
| 3153 | from++; |
| 3154 | else if ((size_t) (pfile->buffer->cur - from) < sizeof "OUGH" - 1) |
| 3155 | return false; |
| 3156 | else if (memcmp (from, "OUGH", sizeof "OUGH" - 1)) |
| 3157 | return false; |
| 3158 | else |
| 3159 | from += sizeof "OUGH" - 1; |
| 3160 | while (*from == ' ' || *from == '\t') |
| 3161 | from++; |
| 3162 | } |
| 3163 | /* Whole comment contents (regex): |
| 3164 | [ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)? |
| 3165 | [ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)? |
| 3166 | [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)? |
| 3167 | */ |
| 3168 | else |
| 3169 | { |
| 3170 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
| 3171 | from++; |
| 3172 | unsigned char f = *from; |
| 3173 | bool all_upper = false; |
| 3174 | if (f == 'E' || f == 'e') |
| 3175 | { |
| 3176 | if ((size_t) (pfile->buffer->cur - from) |
| 3177 | < sizeof "else fallthru" - 1) |
| 3178 | return false; |
| 3179 | if (f == 'E' && memcmp (from + 1, "LSE", sizeof "LSE" - 1) == 0) |
| 3180 | all_upper = true; |
| 3181 | else if (memcmp (from + 1, "lse", sizeof "lse" - 1)) |
| 3182 | return false; |
| 3183 | from += sizeof "else" - 1; |
| 3184 | if (*from == ',') |
| 3185 | from++; |
| 3186 | if (*from != ' ') |
| 3187 | return false; |
| 3188 | from++; |
| 3189 | if (all_upper && *from == 'f') |
| 3190 | return false; |
| 3191 | if (f == 'e' && *from == 'F') |
| 3192 | return false; |
| 3193 | f = *from; |
| 3194 | } |
| 3195 | else if (f == 'I' || f == 'i') |
| 3196 | { |
| 3197 | if ((size_t) (pfile->buffer->cur - from) |
| 3198 | < sizeof "intentional fallthru" - 1) |
| 3199 | return false; |
| 3200 | if (f == 'I' && memcmp (from + 1, "NTENTIONAL", |
| 3201 | sizeof "NTENTIONAL" - 1) == 0) |
| 3202 | all_upper = true; |
| 3203 | else if (memcmp (from + 1, "ntentional", |
| 3204 | sizeof "ntentional" - 1)) |
| 3205 | return false; |
| 3206 | from += sizeof "intentional" - 1; |
| 3207 | if (*from == ' ') |
| 3208 | { |
| 3209 | from++; |
| 3210 | if (all_upper && *from == 'f') |
| 3211 | return false; |
| 3212 | } |
| 3213 | else if (all_upper) |
| 3214 | { |
| 3215 | if (memcmp (from, "LY F", sizeof "LY F" - 1)) |
| 3216 | return false; |
| 3217 | from += sizeof "LY " - 1; |
| 3218 | } |
| 3219 | else |
| 3220 | { |
| 3221 | if (memcmp (from, "ly ", sizeof "ly " - 1)) |
| 3222 | return false; |
| 3223 | from += sizeof "ly " - 1; |
| 3224 | } |
| 3225 | if (f == 'i' && *from == 'F') |
| 3226 | return false; |
| 3227 | f = *from; |
| 3228 | } |
| 3229 | if (f != 'F' && f != 'f') |
| 3230 | return false; |
| 3231 | if ((size_t) (pfile->buffer->cur - from) < sizeof "fallthru" - 1) |
| 3232 | return false; |
| 3233 | if (f == 'F' && memcmp (from + 1, "ALL", sizeof "ALL" - 1) == 0) |
| 3234 | all_upper = true; |
| 3235 | else if (all_upper) |
| 3236 | return false; |
| 3237 | else if (memcmp (from + 1, "all", sizeof "all" - 1)) |
| 3238 | return false; |
| 3239 | from += sizeof "fall" - 1; |
| 3240 | if (*from == (all_upper ? 'S' : 's') && from[1] == ' ') |
| 3241 | from += 2; |
| 3242 | else if (*from == ' ' || *from == '-') |
| 3243 | from++; |
| 3244 | else if (*from != (all_upper ? 'T' : 't')) |
| 3245 | return false; |
| 3246 | if ((f == 'f' || *from != 'T') && (all_upper || *from != 't')) |
| 3247 | return false; |
| 3248 | if ((size_t) (pfile->buffer->cur - from) < sizeof "thru" - 1) |
| 3249 | return false; |
| 3250 | if (memcmp (from + 1, all_upper ? "HRU" : "hru", sizeof "hru" - 1)) |
| 3251 | { |
| 3252 | if ((size_t) (pfile->buffer->cur - from) < sizeof "through" - 1) |
| 3253 | return false; |
| 3254 | if (memcmp (from + 1, all_upper ? "HROUGH" : "hrough", |
| 3255 | sizeof "hrough" - 1)) |
| 3256 | return false; |
| 3257 | from += sizeof "through" - 1; |
| 3258 | } |
| 3259 | else |
| 3260 | from += sizeof "thru" - 1; |
| 3261 | while (*from == ' ' || *from == '\t' || *from == '.' || *from == '!') |
| 3262 | from++; |
| 3263 | if (*from == '-') |
| 3264 | { |
| 3265 | from++; |
| 3266 | if (*comment_start == '*') |
| 3267 | { |
| 3268 | do |
| 3269 | { |
| 3270 | while (*from && *from != '*' |
| 3271 | && *from != '\n' && *from != '\r') |
| 3272 | from++; |
| 3273 | if (*from != '*' || from[1] == '/') |
| 3274 | break; |
| 3275 | from++; |
| 3276 | } |
| 3277 | while (1); |
| 3278 | } |
| 3279 | else |
| 3280 | while (*from && *from != '\n' && *from != '\r') |
| 3281 | from++; |
| 3282 | } |
| 3283 | } |
| 3284 | /* C block comment. */ |
| 3285 | if (*comment_start == '*') |
| 3286 | { |
| 3287 | if (*from != '*' || from[1] != '/') |
| 3288 | return false; |
| 3289 | } |
| 3290 | /* C++ line comment. */ |
| 3291 | else if (*from != '\n') |
| 3292 | return false; |
| 3293 | |
| 3294 | return true; |
| 3295 | } |
| 3296 | |
| 3297 | /* Allocate COUNT tokens for RUN. */ |
| 3298 | void |
| 3299 | _cpp_init_tokenrun (tokenrun *run, unsigned int count) |
| 3300 | { |
| 3301 | run->base = XNEWVEC (cpp_token, count)((cpp_token *) xmalloc (sizeof (cpp_token) * (count))); |
| 3302 | run->limit = run->base + count; |
| 3303 | run->next = NULL__null; |
| 3304 | } |
| 3305 | |
| 3306 | /* Returns the next tokenrun, or creates one if there is none. */ |
| 3307 | static tokenrun * |
| 3308 | next_tokenrun (tokenrun *run) |
| 3309 | { |
| 3310 | if (run->next == NULL__null) |
| 3311 | { |
| 3312 | run->next = XNEW (tokenrun)((tokenrun *) xmalloc (sizeof (tokenrun))); |
| 3313 | run->next->prev = run; |
| 3314 | _cpp_init_tokenrun (run->next, 250); |
| 3315 | } |
| 3316 | |
| 3317 | return run->next; |
| 3318 | } |
| 3319 | |
| 3320 | /* Return the number of not yet processed token in a given |
| 3321 | context. */ |
| 3322 | int |
| 3323 | _cpp_remaining_tokens_num_in_context (cpp_context *context) |
| 3324 | { |
| 3325 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
| 3326 | return (LAST (context)((context)->u.iso.last).token - FIRST (context)((context)->u.iso.first).token); |
| 3327 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
| 3328 | || context->tokens_kind == TOKENS_KIND_EXTENDED) |
| 3329 | return (LAST (context)((context)->u.iso.last).ptoken - FIRST (context)((context)->u.iso.first).ptoken); |
| 3330 | else |
| 3331 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3331, __FUNCTION__); |
| 3332 | } |
| 3333 | |
| 3334 | /* Returns the token present at index INDEX in a given context. If |
| 3335 | INDEX is zero, the next token to be processed is returned. */ |
| 3336 | static const cpp_token* |
| 3337 | _cpp_token_from_context_at (cpp_context *context, int index) |
| 3338 | { |
| 3339 | if (context->tokens_kind == TOKENS_KIND_DIRECT) |
| 3340 | return &(FIRST (context)((context)->u.iso.first).token[index]); |
| 3341 | else if (context->tokens_kind == TOKENS_KIND_INDIRECT |
| 3342 | || context->tokens_kind == TOKENS_KIND_EXTENDED) |
| 3343 | return FIRST (context)((context)->u.iso.first).ptoken[index]; |
| 3344 | else |
| 3345 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3345, __FUNCTION__); |
| 3346 | } |
| 3347 | |
| 3348 | /* Look ahead in the input stream. */ |
| 3349 | const cpp_token * |
| 3350 | cpp_peek_token (cpp_reader *pfile, int index) |
| 3351 | { |
| 3352 | cpp_context *context = pfile->context; |
| 3353 | const cpp_token *peektok; |
| 3354 | int count; |
| 3355 | |
| 3356 | /* First, scan through any pending cpp_context objects. */ |
| 3357 | while (context->prev) |
| 3358 | { |
| 3359 | ptrdiff_t sz = _cpp_remaining_tokens_num_in_context (context); |
| 3360 | |
| 3361 | if (index < (int) sz) |
| 3362 | return _cpp_token_from_context_at (context, index); |
| 3363 | index -= (int) sz; |
| 3364 | context = context->prev; |
| 3365 | } |
| 3366 | |
| 3367 | /* We will have to read some new tokens after all (and do so |
| 3368 | without invalidating preceding tokens). */ |
| 3369 | count = index; |
| 3370 | pfile->keep_tokens++; |
| 3371 | |
| 3372 | /* For peeked tokens temporarily disable line_change reporting, |
| 3373 | until the tokens are parsed for real. */ |
| 3374 | void (*line_change) (cpp_reader *, const cpp_token *, int) |
| 3375 | = pfile->cb.line_change; |
| 3376 | pfile->cb.line_change = NULL__null; |
| 3377 | |
| 3378 | do |
| 3379 | { |
| 3380 | peektok = _cpp_lex_token (pfile); |
| 3381 | if (peektok->type == CPP_EOF) |
| 3382 | { |
| 3383 | index--; |
| 3384 | break; |
| 3385 | } |
| 3386 | else if (peektok->type == CPP_PRAGMA) |
| 3387 | { |
| 3388 | /* Don't peek past a pragma. */ |
| 3389 | if (peektok == &pfile->directive_result) |
| 3390 | /* Save the pragma in the buffer. */ |
| 3391 | *pfile->cur_token++ = *peektok; |
| 3392 | index--; |
| 3393 | break; |
| 3394 | } |
| 3395 | } |
| 3396 | while (index--); |
| 3397 | |
| 3398 | _cpp_backup_tokens_direct (pfile, count - index); |
| 3399 | pfile->keep_tokens--; |
| 3400 | pfile->cb.line_change = line_change; |
| 3401 | |
| 3402 | return peektok; |
| 3403 | } |
| 3404 | |
| 3405 | /* Allocate a single token that is invalidated at the same time as the |
| 3406 | rest of the tokens on the line. Has its line and col set to the |
| 3407 | same as the last lexed token, so that diagnostics appear in the |
| 3408 | right place. */ |
| 3409 | cpp_token * |
| 3410 | _cpp_temp_token (cpp_reader *pfile) |
| 3411 | { |
| 3412 | cpp_token *old, *result; |
| 3413 | ptrdiff_t sz = pfile->cur_run->limit - pfile->cur_token; |
| 3414 | ptrdiff_t la = (ptrdiff_t) pfile->lookaheads; |
| 3415 | |
| 3416 | old = pfile->cur_token - 1; |
| 3417 | /* Any pre-existing lookaheads must not be clobbered. */ |
| 3418 | if (la) |
| 3419 | { |
| 3420 | if (sz <= la) |
| 3421 | { |
| 3422 | tokenrun *next = next_tokenrun (pfile->cur_run); |
| 3423 | |
| 3424 | if (sz < la) |
| 3425 | memmove (next->base + 1, next->base, |
| 3426 | (la - sz) * sizeof (cpp_token)); |
| 3427 | |
| 3428 | next->base[0] = pfile->cur_run->limit[-1]; |
| 3429 | } |
| 3430 | |
| 3431 | if (sz > 1) |
| 3432 | memmove (pfile->cur_token + 1, pfile->cur_token, |
| 3433 | MIN (la, sz - 1)((la) < (sz - 1) ? (la) : (sz - 1)) * sizeof (cpp_token)); |
| 3434 | } |
| 3435 | |
| 3436 | if (!sz && pfile->cur_token == pfile->cur_run->limit) |
| 3437 | { |
| 3438 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
| 3439 | pfile->cur_token = pfile->cur_run->base; |
| 3440 | } |
| 3441 | |
| 3442 | result = pfile->cur_token++; |
| 3443 | result->src_loc = old->src_loc; |
| 3444 | return result; |
| 3445 | } |
| 3446 | |
| 3447 | /* We're at the beginning of a logical line (so not in |
| 3448 | directives-mode) and RESULT is a CPP_NAME with NODE_MODULE set. See |
| 3449 | if we should enter deferred_pragma mode to tokenize the rest of the |
| 3450 | line as a module control-line. */ |
| 3451 | |
| 3452 | static void |
| 3453 | cpp_maybe_module_directive (cpp_reader *pfile, cpp_token *result) |
| 3454 | { |
| 3455 | unsigned backup = 0; /* Tokens we peeked. */ |
| 3456 | cpp_hashnode *node = result->val.node.node; |
| 3457 | cpp_token *peek = result; |
| 3458 | cpp_token *keyword = peek; |
| 3459 | cpp_hashnode *(&n_modules)[spec_nodes::M_HWM][2] = pfile->spec_nodes.n_modules; |
| 3460 | int header_count = 0; |
| 3461 | |
| 3462 | /* Make sure the incoming state is as we expect it. This way we |
| 3463 | can restore it using constants. */ |
| 3464 | gcc_checking_assert (!pfile->state.in_deferred_pragma((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)) |
| 3465 | && !pfile->state.skipping((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)) |
| 3466 | && !pfile->state.parsing_args((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)) |
| 3467 | && !pfile->state.angled_headers((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)) |
| 3468 | && (pfile->state.save_comments((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)) |
| 3469 | == !CPP_OPTION (pfile, discard_comments)))((void)(!(!pfile->state.in_deferred_pragma && !pfile ->state.skipping && !pfile->state.parsing_args && !pfile->state.angled_headers && (pfile->state. save_comments == !((pfile)->opts.discard_comments))) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3469, __FUNCTION__), 0 : 0)); |
| 3470 | |
| 3471 | /* Enter directives mode sufficiently for peeking. We don't have |
| 3472 | to actually set in_directive. */ |
| 3473 | pfile->state.in_deferred_pragma = true; |
| 3474 | |
| 3475 | /* These two fields are needed to process tokenization in deferred |
| 3476 | pragma mode. They are not used outside deferred pragma mode or |
| 3477 | directives mode. */ |
| 3478 | pfile->state.pragma_allow_expansion = true; |
| 3479 | pfile->directive_line = result->src_loc; |
| 3480 | |
| 3481 | /* Saving comments is incompatible with directives mode. */ |
| 3482 | pfile->state.save_comments = 0; |
| 3483 | |
| 3484 | if (node == n_modules[spec_nodes::M_EXPORT][0]) |
| 3485 | { |
| 3486 | peek = _cpp_lex_direct (pfile); |
| 3487 | keyword = peek; |
| 3488 | backup++; |
| 3489 | if (keyword->type != CPP_NAME) |
| 3490 | goto not_module; |
| 3491 | node = keyword->val.node.node; |
| 3492 | if (!(node->flags & NODE_MODULE(1 << 8))) |
| 3493 | goto not_module; |
| 3494 | } |
| 3495 | |
| 3496 | if (node == n_modules[spec_nodes::M__IMPORT][0]) |
| 3497 | /* __import */ |
| 3498 | header_count = backup + 2 + 16; |
| 3499 | else if (node == n_modules[spec_nodes::M_IMPORT][0]) |
| 3500 | /* import */ |
| 3501 | header_count = backup + 2 + (CPP_OPTION (pfile, preprocessed)((pfile)->opts.preprocessed) ? 16 : 0); |
| 3502 | else if (node == n_modules[spec_nodes::M_MODULE][0]) |
| 3503 | ; /* module */ |
| 3504 | else |
| 3505 | goto not_module; |
| 3506 | |
| 3507 | /* We've seen [export] {module|import|__import}. Check the next token. */ |
| 3508 | if (header_count) |
| 3509 | /* After '{,__}import' a header name may appear. */ |
| 3510 | pfile->state.angled_headers = true; |
| 3511 | peek = _cpp_lex_direct (pfile); |
| 3512 | backup++; |
| 3513 | |
| 3514 | /* ... import followed by identifier, ':', '<' or |
| 3515 | header-name preprocessing tokens, or module |
| 3516 | followed by cpp-identifier, ':' or ';' preprocessing |
| 3517 | tokens. C++ keywords are not yet relevant. */ |
| 3518 | if (peek->type == CPP_NAME |
| 3519 | || peek->type == CPP_COLON |
| 3520 | || (header_count |
| 3521 | ? (peek->type == CPP_LESS |
| 3522 | || (peek->type == CPP_STRING && peek->val.str.text[0] != 'R') |
| 3523 | || peek->type == CPP_HEADER_NAME) |
| 3524 | : peek->type == CPP_SEMICOLON)) |
| 3525 | { |
| 3526 | pfile->state.pragma_allow_expansion = !CPP_OPTION (pfile, preprocessed)((pfile)->opts.preprocessed); |
| 3527 | if (!pfile->state.pragma_allow_expansion) |
| 3528 | pfile->state.prevent_expansion++; |
| 3529 | |
| 3530 | if (!header_count && linemap_included_from |
| 3531 | (LINEMAPS_LAST_ORDINARY_MAP (pfile->line_table))) |
| 3532 | cpp_error_with_line (pfile, CPP_DL_ERROR, keyword->src_loc, 0, |
| 3533 | "module control-line cannot be in included file"); |
| 3534 | |
| 3535 | /* The first one or two tokens cannot be macro names. */ |
| 3536 | for (int ix = backup; ix--;) |
| 3537 | { |
| 3538 | cpp_token *tok = ix ? keyword : result; |
| 3539 | cpp_hashnode *node = tok->val.node.node; |
| 3540 | |
| 3541 | /* Don't attempt to expand the token. */ |
| 3542 | tok->flags |= NO_EXPAND(1 << 10); |
| 3543 | if (_cpp_defined_macro_p (node) |
| 3544 | && _cpp_maybe_notify_macro_use (pfile, node, tok->src_loc) |
| 3545 | && !cpp_fun_like_macro_p (node)) |
| 3546 | cpp_error_with_line (pfile, CPP_DL_ERROR, tok->src_loc, 0, |
| 3547 | "module control-line \"%s\" cannot be" |
| 3548 | " an object-like macro", |
| 3549 | NODE_NAME (node)(((&(node)->ident))->str)); |
| 3550 | } |
| 3551 | |
| 3552 | /* Map to underbar variants. */ |
| 3553 | keyword->val.node.node = n_modules[header_count |
| 3554 | ? spec_nodes::M_IMPORT |
| 3555 | : spec_nodes::M_MODULE][1]; |
| 3556 | if (backup != 1) |
| 3557 | result->val.node.node = n_modules[spec_nodes::M_EXPORT][1]; |
| 3558 | |
| 3559 | /* Maybe tell the tokenizer we expect a header-name down the |
| 3560 | road. */ |
| 3561 | pfile->state.directive_file_token = header_count; |
| 3562 | } |
| 3563 | else |
| 3564 | { |
| 3565 | not_module: |
| 3566 | /* Drop out of directive mode. */ |
| 3567 | /* We aaserted save_comments had this value upon entry. */ |
| 3568 | pfile->state.save_comments |
| 3569 | = !CPP_OPTION (pfile, discard_comments)((pfile)->opts.discard_comments); |
| 3570 | pfile->state.in_deferred_pragma = false; |
| 3571 | /* Do not let this remain on. */ |
| 3572 | pfile->state.angled_headers = false; |
| 3573 | } |
| 3574 | |
| 3575 | /* In either case we want to backup the peeked tokens. */ |
| 3576 | if (backup) |
| 3577 | { |
| 3578 | /* If we saw EOL, we should drop it, because this isn't a module |
| 3579 | control-line after all. */ |
| 3580 | bool eol = peek->type == CPP_PRAGMA_EOL; |
| 3581 | if (!eol || backup > 1) |
| 3582 | { |
| 3583 | /* Put put the peeked tokens back */ |
| 3584 | _cpp_backup_tokens_direct (pfile, backup); |
| 3585 | /* But if the last one was an EOL, forget it. */ |
| 3586 | if (eol) |
| 3587 | pfile->lookaheads--; |
| 3588 | } |
| 3589 | } |
| 3590 | } |
| 3591 | |
| 3592 | /* Lex a token into RESULT (external interface). Takes care of issues |
| 3593 | like directive handling, token lookahead, multiple include |
| 3594 | optimization and skipping. */ |
| 3595 | const cpp_token * |
| 3596 | _cpp_lex_token (cpp_reader *pfile) |
| 3597 | { |
| 3598 | cpp_token *result; |
| 3599 | |
| 3600 | for (;;) |
| 3601 | { |
| 3602 | if (pfile->cur_token == pfile->cur_run->limit) |
| 3603 | { |
| 3604 | pfile->cur_run = next_tokenrun (pfile->cur_run); |
| 3605 | pfile->cur_token = pfile->cur_run->base; |
| 3606 | } |
| 3607 | /* We assume that the current token is somewhere in the current |
| 3608 | run. */ |
| 3609 | if (pfile->cur_token < pfile->cur_run->base |
| 3610 | || pfile->cur_token >= pfile->cur_run->limit) |
| 3611 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3611, __FUNCTION__); |
| 3612 | |
| 3613 | if (pfile->lookaheads) |
| 3614 | { |
| 3615 | pfile->lookaheads--; |
| 3616 | result = pfile->cur_token++; |
| 3617 | } |
| 3618 | else |
| 3619 | result = _cpp_lex_direct (pfile); |
| 3620 | |
| 3621 | if (result->flags & BOL(1 << 12)) |
| 3622 | { |
| 3623 | /* Is this a directive. If _cpp_handle_directive returns |
| 3624 | false, it is an assembler #. */ |
| 3625 | if (result->type == CPP_HASH |
| 3626 | /* 6.10.3 p 11: Directives in a list of macro arguments |
| 3627 | gives undefined behavior. This implementation |
| 3628 | handles the directive as normal. */ |
| 3629 | && pfile->state.parsing_args != 1) |
| 3630 | { |
| 3631 | if (_cpp_handle_directive (pfile, result->flags & PREV_WHITE(1 << 0))) |
| 3632 | { |
| 3633 | if (pfile->directive_result.type == CPP_PADDING) |
| 3634 | continue; |
| 3635 | result = &pfile->directive_result; |
| 3636 | } |
| 3637 | } |
| 3638 | else if (pfile->state.in_deferred_pragma) |
| 3639 | result = &pfile->directive_result; |
| 3640 | else if (result->type == CPP_NAME |
| 3641 | && (result->val.node.node->flags & NODE_MODULE(1 << 8)) |
| 3642 | && !pfile->state.skipping |
| 3643 | /* Unlike regular directives, we do not deal with |
| 3644 | tokenizing module directives as macro arguments. |
| 3645 | That's not permitted. */ |
| 3646 | && !pfile->state.parsing_args) |
| 3647 | { |
| 3648 | /* P1857. Before macro expansion, At start of logical |
| 3649 | line ... */ |
| 3650 | /* We don't have to consider lookaheads at this point. */ |
| 3651 | gcc_checking_assert (!pfile->lookaheads)((void)(!(!pfile->lookaheads) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 3651, __FUNCTION__), 0 : 0)); |
| 3652 | |
| 3653 | cpp_maybe_module_directive (pfile, result); |
| 3654 | } |
| 3655 | |
| 3656 | if (pfile->cb.line_change && !pfile->state.skipping) |
| 3657 | pfile->cb.line_change (pfile, result, pfile->state.parsing_args); |
| 3658 | } |
| 3659 | |
| 3660 | /* We don't skip tokens in directives. */ |
| 3661 | if (pfile->state.in_directive || pfile->state.in_deferred_pragma) |
| 3662 | break; |
| 3663 | |
| 3664 | /* Outside a directive, invalidate controlling macros. At file |
| 3665 | EOF, _cpp_lex_direct takes care of popping the buffer, so we never |
| 3666 | get here and MI optimization works. */ |
| 3667 | pfile->mi_valid = false; |
| 3668 | |
| 3669 | if (!pfile->state.skipping || result->type == CPP_EOF) |
| 3670 | break; |
| 3671 | } |
| 3672 | |
| 3673 | return result; |
| 3674 | } |
| 3675 | |
| 3676 | /* Returns true if a fresh line has been loaded. */ |
| 3677 | template <bool lexing_raw_string> |
| 3678 | static bool |
| 3679 | get_fresh_line_impl (cpp_reader *pfile) |
| 3680 | { |
| 3681 | /* We can't get a new line until we leave the current directive, unless we |
| 3682 | are lexing a raw string, in which case it will be OK as long as we don't |
| 3683 | pop the current buffer. */ |
| 3684 | if (!lexing_raw_string && pfile->state.in_directive) |
| 3685 | return false; |
| 3686 | |
| 3687 | for (;;) |
| 3688 | { |
| 3689 | cpp_buffer *buffer = pfile->buffer; |
| 3690 | |
| 3691 | if (!buffer->need_line) |
| 3692 | return true; |
| 3693 | |
| 3694 | if (buffer->next_line < buffer->rlimit) |
| 3695 | { |
| 3696 | _cpp_clean_line (pfile); |
| 3697 | return true; |
| 3698 | } |
| 3699 | |
| 3700 | /* We can't change buffers until we leave the current directive. */ |
| 3701 | if (lexing_raw_string && pfile->state.in_directive) |
| 3702 | return false; |
| 3703 | |
| 3704 | /* First, get out of parsing arguments state. */ |
| 3705 | if (pfile->state.parsing_args) |
| 3706 | return false; |
| 3707 | |
| 3708 | /* End of buffer. Non-empty files should end in a newline. */ |
| 3709 | if (buffer->buf != buffer->rlimit |
| 3710 | && buffer->next_line > buffer->rlimit |
| 3711 | && !buffer->from_stage3) |
| 3712 | { |
| 3713 | /* Clip to buffer size. */ |
| 3714 | buffer->next_line = buffer->rlimit; |
| 3715 | } |
| 3716 | |
| 3717 | if (buffer->prev && !buffer->return_at_eof) |
| 3718 | _cpp_pop_buffer (pfile); |
| 3719 | else |
| 3720 | { |
| 3721 | /* End of translation. Do not pop the buffer yet. Increment |
| 3722 | line number so that the EOF token is on a line of its own |
| 3723 | (_cpp_lex_direct doesn't increment in that case, because |
| 3724 | it's hard for it to distinguish this special case). */ |
| 3725 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 3726 | return false; |
| 3727 | } |
| 3728 | } |
| 3729 | } |
| 3730 | |
| 3731 | bool |
| 3732 | _cpp_get_fresh_line (cpp_reader *pfile) |
| 3733 | { |
| 3734 | return get_fresh_line_impl<false> (pfile); |
| 3735 | } |
| 3736 | |
| 3737 | |
| 3738 | #define IF_NEXT_IS(CHAR, THEN_TYPE, ELSE_TYPE)do { result->type = ELSE_TYPE; if (*buffer->cur == CHAR ) buffer->cur++, result->type = THEN_TYPE; } while (0) \ |
| 3739 | do \ |
| 3740 | { \ |
| 3741 | result->type = ELSE_TYPE; \ |
| 3742 | if (*buffer->cur == CHAR) \ |
| 3743 | buffer->cur++, result->type = THEN_TYPE; \ |
| 3744 | } \ |
| 3745 | while (0) |
| 3746 | |
| 3747 | /* Lex a token into pfile->cur_token, which is also incremented, to |
| 3748 | get diagnostics pointing to the correct location. |
| 3749 | |
| 3750 | Does not handle issues such as token lookahead, multiple-include |
| 3751 | optimization, directives, skipping etc. This function is only |
| 3752 | suitable for use by _cpp_lex_token, and in special cases like |
| 3753 | lex_expansion_token which doesn't care for any of these issues. |
| 3754 | |
| 3755 | When meeting a newline, returns CPP_EOF if parsing a directive, |
| 3756 | otherwise returns to the start of the token buffer if permissible. |
| 3757 | Returns the location of the lexed token. */ |
| 3758 | cpp_token * |
| 3759 | _cpp_lex_direct (cpp_reader *pfile) |
| 3760 | { |
| 3761 | cppchar_t c; |
| 3762 | cpp_buffer *buffer; |
| 3763 | const unsigned char *comment_start; |
| 3764 | bool fallthrough_comment = false; |
| 3765 | cpp_token *result = pfile->cur_token++; |
| 3766 | |
| 3767 | fresh_line: |
| 3768 | result->flags = 0; |
| 3769 | buffer = pfile->buffer; |
| 3770 | if (buffer->need_line) |
| 3771 | { |
| 3772 | if (pfile->state.in_deferred_pragma) |
| 3773 | { |
| 3774 | /* This can happen in cases like: |
| 3775 | #define loop(x) whatever |
| 3776 | #pragma omp loop |
| 3777 | where when trying to expand loop we need to peek |
| 3778 | next token after loop, but aren't still in_deferred_pragma |
| 3779 | mode but are in in_directive mode, so buffer->need_line |
| 3780 | is set, a CPP_EOF is peeked. */ |
| 3781 | result->type = CPP_PRAGMA_EOL; |
| 3782 | pfile->state.in_deferred_pragma = false; |
| 3783 | if (!pfile->state.pragma_allow_expansion) |
| 3784 | pfile->state.prevent_expansion--; |
| 3785 | return result; |
| 3786 | } |
| 3787 | if (!_cpp_get_fresh_line (pfile)) |
| 3788 | { |
| 3789 | result->type = CPP_EOF; |
| 3790 | /* Not a real EOF in a directive or arg parsing -- we refuse |
| 3791 | to advance to the next file now, and will once we're out |
| 3792 | of those modes. */ |
| 3793 | if (!pfile->state.in_directive && !pfile->state.parsing_args) |
| 3794 | { |
| 3795 | /* Tell the compiler the line number of the EOF token. */ |
| 3796 | result->src_loc = pfile->line_table->highest_line; |
| 3797 | result->flags = BOL(1 << 12); |
| 3798 | /* Now pop the buffer that _cpp_get_fresh_line did not. */ |
| 3799 | _cpp_pop_buffer (pfile); |
| 3800 | } |
| 3801 | return result; |
| 3802 | } |
| 3803 | if (buffer != pfile->buffer) |
| 3804 | fallthrough_comment = false; |
| 3805 | if (!pfile->keep_tokens) |
| 3806 | { |
| 3807 | pfile->cur_run = &pfile->base_run; |
| 3808 | result = pfile->base_run.base; |
| 3809 | pfile->cur_token = result + 1; |
| 3810 | } |
| 3811 | result->flags = BOL(1 << 12); |
| 3812 | if (pfile->state.parsing_args == 2) |
| 3813 | result->flags |= PREV_WHITE(1 << 0); |
| 3814 | } |
| 3815 | buffer = pfile->buffer; |
| 3816 | update_tokens_line: |
| 3817 | result->src_loc = pfile->line_table->highest_line; |
| 3818 | |
| 3819 | skipped_white: |
| 3820 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
| 3821 | && !pfile->overlaid_buffer) |
| 3822 | { |
| 3823 | _cpp_process_line_notes (pfile, false); |
| 3824 | result->src_loc = pfile->line_table->highest_line; |
| 3825 | } |
| 3826 | c = *buffer->cur++; |
| 3827 | |
| 3828 | if (pfile->forced_token_location) |
| 3829 | result->src_loc = pfile->forced_token_location; |
| 3830 | else |
| 3831 | result->src_loc = linemap_position_for_column (pfile->line_table, |
| 3832 | CPP_BUF_COLUMN (buffer, buffer->cur)((buffer->cur) - (buffer)->line_base)); |
| 3833 | |
| 3834 | switch (c) |
| 3835 | { |
| 3836 | case ' ': case '\t': case '\f': case '\v': case '\0': |
| 3837 | result->flags |= PREV_WHITE(1 << 0); |
| 3838 | skip_whitespace (pfile, c); |
| 3839 | goto skipped_white; |
| 3840 | |
| 3841 | case '\n': |
| 3842 | /* Increment the line, unless this is the last line ... */ |
| 3843 | if (buffer->cur < buffer->rlimit |
| 3844 | /* ... or this is a #include, (where _cpp_stack_file needs to |
| 3845 | unwind by one line) ... */ |
| 3846 | || (pfile->state.in_directive > 1 |
| 3847 | /* ... except traditional-cpp increments this elsewhere. */ |
| 3848 | && !CPP_OPTION (pfile, traditional)((pfile)->opts.traditional))) |
| 3849 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 3850 | buffer->need_line = true; |
| 3851 | if (pfile->state.in_deferred_pragma) |
| 3852 | { |
| 3853 | /* Produce the PRAGMA_EOL on this line. File reading |
| 3854 | ensures there is always a \n at end of the buffer, thus |
| 3855 | in a deferred pragma we always see CPP_PRAGMA_EOL before |
| 3856 | any CPP_EOF. */ |
| 3857 | result->type = CPP_PRAGMA_EOL; |
| 3858 | result->flags &= ~PREV_WHITE(1 << 0); |
| 3859 | pfile->state.in_deferred_pragma = false; |
| 3860 | if (!pfile->state.pragma_allow_expansion) |
| 3861 | pfile->state.prevent_expansion--; |
| 3862 | return result; |
| 3863 | } |
| 3864 | goto fresh_line; |
| 3865 | |
| 3866 | case '0': case '1': case '2': case '3': case '4': |
| 3867 | case '5': case '6': case '7': case '8': case '9': |
| 3868 | { |
| 3869 | struct normalize_state nst = INITIAL_NORMALIZE_STATE{ 0, 0, normalized_KC }; |
| 3870 | result->type = CPP_NUMBER; |
| 3871 | lex_number (pfile, &result->val.str, &nst); |
| 3872 | warn_about_normalization (pfile, result, &nst, false); |
| 3873 | break; |
| 3874 | } |
| 3875 | |
| 3876 | case 'L': |
| 3877 | case 'u': |
| 3878 | case 'U': |
| 3879 | case 'R': |
| 3880 | /* 'L', 'u', 'U', 'u8' or 'R' may introduce wide characters, |
| 3881 | wide strings or raw strings. */ |
| 3882 | if (c == 'L' || CPP_OPTION (pfile, rliterals)((pfile)->opts.rliterals) |
| 3883 | || (c != 'R' && CPP_OPTION (pfile, uliterals)((pfile)->opts.uliterals))) |
| 3884 | { |
| 3885 | if ((*buffer->cur == '\'' && c != 'R') |
| 3886 | || *buffer->cur == '"' |
| 3887 | || (*buffer->cur == 'R' |
| 3888 | && c != 'R' |
| 3889 | && buffer->cur[1] == '"' |
| 3890 | && CPP_OPTION (pfile, rliterals)((pfile)->opts.rliterals)) |
| 3891 | || (*buffer->cur == '8' |
| 3892 | && c == 'u' |
| 3893 | && ((buffer->cur[1] == '"' || (buffer->cur[1] == '\'' |
| 3894 | && CPP_OPTION (pfile, utf8_char_literals)((pfile)->opts.utf8_char_literals))) |
| 3895 | || (buffer->cur[1] == 'R' && buffer->cur[2] == '"' |
| 3896 | && CPP_OPTION (pfile, rliterals)((pfile)->opts.rliterals))))) |
| 3897 | { |
| 3898 | lex_string (pfile, result, buffer->cur - 1); |
| 3899 | break; |
| 3900 | } |
| 3901 | } |
| 3902 | /* Fall through. */ |
| 3903 | |
| 3904 | case '_': |
| 3905 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': |
| 3906 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': |
| 3907 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': |
| 3908 | case 's': case 't': case 'v': case 'w': case 'x': |
| 3909 | case 'y': case 'z': |
| 3910 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': |
| 3911 | case 'G': case 'H': case 'I': case 'J': case 'K': |
| 3912 | case 'M': case 'N': case 'O': case 'P': case 'Q': |
| 3913 | case 'S': case 'T': case 'V': case 'W': case 'X': |
| 3914 | case 'Y': case 'Z': |
| 3915 | result->type = CPP_NAME; |
| 3916 | { |
| 3917 | struct normalize_state nst = INITIAL_NORMALIZE_STATE{ 0, 0, normalized_KC }; |
| 3918 | result->val.node.node = lex_identifier (pfile, buffer->cur - 1, false, |
| 3919 | &nst, |
| 3920 | &result->val.node.spelling); |
| 3921 | warn_about_normalization (pfile, result, &nst, true); |
| 3922 | } |
| 3923 | |
| 3924 | /* Convert named operators to their proper types. */ |
| 3925 | if (result->val.node.node->flags & NODE_OPERATOR(1 << 0)) |
| 3926 | { |
| 3927 | result->flags |= NAMED_OP(1 << 4); |
| 3928 | result->type = (enum cpp_ttype) result->val.node.node->directive_index; |
| 3929 | } |
| 3930 | |
| 3931 | /* Signal FALLTHROUGH comment followed by another token. */ |
| 3932 | if (fallthrough_comment) |
| 3933 | result->flags |= PREV_FALLTHROUGH(1 << 5); |
| 3934 | break; |
| 3935 | |
| 3936 | case '\'': |
| 3937 | case '"': |
| 3938 | lex_string (pfile, result, buffer->cur - 1); |
| 3939 | break; |
| 3940 | |
| 3941 | case '/': |
| 3942 | /* A potential block or line comment. */ |
| 3943 | comment_start = buffer->cur; |
| 3944 | c = *buffer->cur; |
| 3945 | |
| 3946 | if (c == '*') |
| 3947 | { |
| 3948 | if (_cpp_skip_block_comment (pfile)) |
| 3949 | cpp_error (pfile, CPP_DL_ERROR, "unterminated comment"); |
| 3950 | } |
| 3951 | else if (c == '/' && ! CPP_OPTION (pfile, traditional)((pfile)->opts.traditional)) |
| 3952 | { |
| 3953 | /* Don't warn for system headers. */ |
| 3954 | if (_cpp_in_system_header (pfile)) |
| 3955 | ; |
| 3956 | /* Warn about comments if pedantically GNUC89, and not |
| 3957 | in system headers. */ |
| 3958 | else if (CPP_OPTION (pfile, lang)((pfile)->opts.lang) == CLK_GNUC89 |
| 3959 | && CPP_PEDANTIC (pfile)((pfile)->opts.cpp_pedantic) |
| 3960 | && ! buffer->warned_cplusplus_comments) |
| 3961 | { |
| 3962 | if (cpp_error (pfile, CPP_DL_PEDWARN, |
| 3963 | "C++ style comments are not allowed in ISO C90")) |
| 3964 | cpp_error (pfile, CPP_DL_NOTE, |
| 3965 | "(this will be reported only once per input file)"); |
| 3966 | buffer->warned_cplusplus_comments = 1; |
| 3967 | } |
| 3968 | /* Or if specifically desired via -Wc90-c99-compat. */ |
| 3969 | else if (CPP_OPTION (pfile, cpp_warn_c90_c99_compat)((pfile)->opts.cpp_warn_c90_c99_compat) > 0 |
| 3970 | && ! CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus) |
| 3971 | && ! buffer->warned_cplusplus_comments) |
| 3972 | { |
| 3973 | if (cpp_error (pfile, CPP_DL_WARNING, |
| 3974 | "C++ style comments are incompatible with C90")) |
| 3975 | cpp_error (pfile, CPP_DL_NOTE, |
| 3976 | "(this will be reported only once per input file)"); |
| 3977 | buffer->warned_cplusplus_comments = 1; |
| 3978 | } |
| 3979 | /* In C89/C94, C++ style comments are forbidden. */ |
| 3980 | else if ((CPP_OPTION (pfile, lang)((pfile)->opts.lang) == CLK_STDC89 |
| 3981 | || CPP_OPTION (pfile, lang)((pfile)->opts.lang) == CLK_STDC94)) |
| 3982 | { |
| 3983 | /* But don't be confused about valid code such as |
| 3984 | - // immediately followed by *, |
| 3985 | - // in a preprocessing directive, |
| 3986 | - // in an #if 0 block. */ |
| 3987 | if (buffer->cur[1] == '*' |
| 3988 | || pfile->state.in_directive |
| 3989 | || pfile->state.skipping) |
| 3990 | { |
| 3991 | result->type = CPP_DIV; |
| 3992 | break; |
| 3993 | } |
| 3994 | else if (! buffer->warned_cplusplus_comments) |
| 3995 | { |
| 3996 | if (cpp_error (pfile, CPP_DL_ERROR, |
| 3997 | "C++ style comments are not allowed in " |
| 3998 | "ISO C90")) |
| 3999 | cpp_error (pfile, CPP_DL_NOTE, |
| 4000 | "(this will be reported only once per input " |
| 4001 | "file)"); |
| 4002 | buffer->warned_cplusplus_comments = 1; |
| 4003 | } |
| 4004 | } |
| 4005 | if (skip_line_comment (pfile) && CPP_OPTION (pfile, warn_comments)((pfile)->opts.warn_comments)) |
| 4006 | cpp_warning (pfile, CPP_W_COMMENTS, "multi-line comment"); |
| 4007 | } |
| 4008 | else if (c == '=') |
| 4009 | { |
| 4010 | buffer->cur++; |
| 4011 | result->type = CPP_DIV_EQ; |
| 4012 | break; |
| 4013 | } |
| 4014 | else |
| 4015 | { |
| 4016 | result->type = CPP_DIV; |
| 4017 | break; |
| 4018 | } |
| 4019 | |
| 4020 | if (fallthrough_comment_p (pfile, comment_start)) |
| 4021 | fallthrough_comment = true; |
| 4022 | |
| 4023 | if (pfile->cb.comment) |
| 4024 | { |
| 4025 | size_t len = pfile->buffer->cur - comment_start; |
| 4026 | pfile->cb.comment (pfile, result->src_loc, comment_start - 1, |
| 4027 | len + 1); |
| 4028 | } |
| 4029 | |
| 4030 | if (!pfile->state.save_comments) |
| 4031 | { |
| 4032 | result->flags |= PREV_WHITE(1 << 0); |
| 4033 | goto update_tokens_line; |
| 4034 | } |
| 4035 | |
| 4036 | if (fallthrough_comment) |
| 4037 | result->flags |= PREV_FALLTHROUGH(1 << 5); |
| 4038 | |
| 4039 | /* Save the comment as a token in its own right. */ |
| 4040 | save_comment (pfile, result, comment_start, c); |
| 4041 | break; |
| 4042 | |
| 4043 | case '<': |
| 4044 | if (pfile->state.angled_headers) |
| 4045 | { |
| 4046 | lex_string (pfile, result, buffer->cur - 1); |
| 4047 | if (result->type != CPP_LESS) |
| 4048 | break; |
| 4049 | } |
| 4050 | |
| 4051 | result->type = CPP_LESS; |
| 4052 | if (*buffer->cur == '=') |
| 4053 | { |
| 4054 | buffer->cur++, result->type = CPP_LESS_EQ; |
| 4055 | if (*buffer->cur == '>' |
| 4056 | && CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus) |
| 4057 | && CPP_OPTION (pfile, lang)((pfile)->opts.lang) >= CLK_GNUCXX20) |
| 4058 | buffer->cur++, result->type = CPP_SPACESHIP; |
| 4059 | } |
| 4060 | else if (*buffer->cur == '<') |
| 4061 | { |
| 4062 | buffer->cur++; |
| 4063 | IF_NEXT_IS ('=', CPP_LSHIFT_EQ, CPP_LSHIFT)do { result->type = CPP_LSHIFT; if (*buffer->cur == '=' ) buffer->cur++, result->type = CPP_LSHIFT_EQ; } while ( 0); |
| 4064 | } |
| 4065 | else if (CPP_OPTION (pfile, digraphs)((pfile)->opts.digraphs)) |
| 4066 | { |
| 4067 | if (*buffer->cur == ':') |
| 4068 | { |
| 4069 | /* C++11 [2.5/3 lex.pptoken], "Otherwise, if the next |
| 4070 | three characters are <:: and the subsequent character |
| 4071 | is neither : nor >, the < is treated as a preprocessor |
| 4072 | token by itself". */ |
| 4073 | if (CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus) |
| 4074 | && CPP_OPTION (pfile, lang)((pfile)->opts.lang) != CLK_CXX98 |
| 4075 | && CPP_OPTION (pfile, lang)((pfile)->opts.lang) != CLK_GNUCXX |
| 4076 | && buffer->cur[1] == ':' |
| 4077 | && buffer->cur[2] != ':' && buffer->cur[2] != '>') |
| 4078 | break; |
| 4079 | |
| 4080 | buffer->cur++; |
| 4081 | result->flags |= DIGRAPH(1 << 1); |
| 4082 | result->type = CPP_OPEN_SQUARE; |
| 4083 | } |
| 4084 | else if (*buffer->cur == '%') |
| 4085 | { |
| 4086 | buffer->cur++; |
| 4087 | result->flags |= DIGRAPH(1 << 1); |
| 4088 | result->type = CPP_OPEN_BRACE; |
| 4089 | } |
| 4090 | } |
| 4091 | break; |
| 4092 | |
| 4093 | case '>': |
| 4094 | result->type = CPP_GREATER; |
| 4095 | if (*buffer->cur == '=') |
| 4096 | buffer->cur++, result->type = CPP_GREATER_EQ; |
| 4097 | else if (*buffer->cur == '>') |
| 4098 | { |
| 4099 | buffer->cur++; |
| 4100 | IF_NEXT_IS ('=', CPP_RSHIFT_EQ, CPP_RSHIFT)do { result->type = CPP_RSHIFT; if (*buffer->cur == '=' ) buffer->cur++, result->type = CPP_RSHIFT_EQ; } while ( 0); |
| 4101 | } |
| 4102 | break; |
| 4103 | |
| 4104 | case '%': |
| 4105 | result->type = CPP_MOD; |
| 4106 | if (*buffer->cur == '=') |
| 4107 | buffer->cur++, result->type = CPP_MOD_EQ; |
| 4108 | else if (CPP_OPTION (pfile, digraphs)((pfile)->opts.digraphs)) |
| 4109 | { |
| 4110 | if (*buffer->cur == ':') |
| 4111 | { |
| 4112 | buffer->cur++; |
| 4113 | result->flags |= DIGRAPH(1 << 1); |
| 4114 | result->type = CPP_HASH; |
| 4115 | if (*buffer->cur == '%' && buffer->cur[1] == ':') |
| 4116 | buffer->cur += 2, result->type = CPP_PASTE, result->val.token_no = 0; |
| 4117 | } |
| 4118 | else if (*buffer->cur == '>') |
| 4119 | { |
| 4120 | buffer->cur++; |
| 4121 | result->flags |= DIGRAPH(1 << 1); |
| 4122 | result->type = CPP_CLOSE_BRACE; |
| 4123 | } |
| 4124 | } |
| 4125 | break; |
| 4126 | |
| 4127 | case '.': |
| 4128 | result->type = CPP_DOT; |
| 4129 | if (ISDIGIT (*buffer->cur)(_sch_istable[(*buffer->cur) & 0xff] & (unsigned short )(_sch_isdigit))) |
| 4130 | { |
| 4131 | struct normalize_state nst = INITIAL_NORMALIZE_STATE{ 0, 0, normalized_KC }; |
| 4132 | result->type = CPP_NUMBER; |
| 4133 | lex_number (pfile, &result->val.str, &nst); |
| 4134 | warn_about_normalization (pfile, result, &nst, false); |
| 4135 | } |
| 4136 | else if (*buffer->cur == '.' && buffer->cur[1] == '.') |
| 4137 | buffer->cur += 2, result->type = CPP_ELLIPSIS; |
| 4138 | else if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus)) |
| 4139 | buffer->cur++, result->type = CPP_DOT_STAR; |
| 4140 | break; |
| 4141 | |
| 4142 | case '+': |
| 4143 | result->type = CPP_PLUS; |
| 4144 | if (*buffer->cur == '+') |
| 4145 | buffer->cur++, result->type = CPP_PLUS_PLUS; |
| 4146 | else if (*buffer->cur == '=') |
| 4147 | buffer->cur++, result->type = CPP_PLUS_EQ; |
| 4148 | break; |
| 4149 | |
| 4150 | case '-': |
| 4151 | result->type = CPP_MINUS; |
| 4152 | if (*buffer->cur == '>') |
| 4153 | { |
| 4154 | buffer->cur++; |
| 4155 | result->type = CPP_DEREF; |
| 4156 | if (*buffer->cur == '*' && CPP_OPTION (pfile, cplusplus)((pfile)->opts.cplusplus)) |
| 4157 | buffer->cur++, result->type = CPP_DEREF_STAR; |
| 4158 | } |
| 4159 | else if (*buffer->cur == '-') |
| 4160 | buffer->cur++, result->type = CPP_MINUS_MINUS; |
| 4161 | else if (*buffer->cur == '=') |
| 4162 | buffer->cur++, result->type = CPP_MINUS_EQ; |
| 4163 | break; |
| 4164 | |
| 4165 | case '&': |
| 4166 | result->type = CPP_AND; |
| 4167 | if (*buffer->cur == '&') |
| 4168 | buffer->cur++, result->type = CPP_AND_AND; |
| 4169 | else if (*buffer->cur == '=') |
| 4170 | buffer->cur++, result->type = CPP_AND_EQ; |
| 4171 | break; |
| 4172 | |
| 4173 | case '|': |
| 4174 | result->type = CPP_OR; |
| 4175 | if (*buffer->cur == '|') |
| 4176 | buffer->cur++, result->type = CPP_OR_OR; |
| 4177 | else if (*buffer->cur == '=') |
| 4178 | buffer->cur++, result->type = CPP_OR_EQ; |
| 4179 | break; |
| 4180 | |
| 4181 | case ':': |
| 4182 | result->type = CPP_COLON; |
| 4183 | if (*buffer->cur == ':' && CPP_OPTION (pfile, scope)((pfile)->opts.scope)) |
| 4184 | buffer->cur++, result->type = CPP_SCOPE; |
| 4185 | else if (*buffer->cur == '>' && CPP_OPTION (pfile, digraphs)((pfile)->opts.digraphs)) |
| 4186 | { |
| 4187 | buffer->cur++; |
| 4188 | result->flags |= DIGRAPH(1 << 1); |
| 4189 | result->type = CPP_CLOSE_SQUARE; |
| 4190 | } |
| 4191 | break; |
| 4192 | |
| 4193 | case '*': IF_NEXT_IS ('=', CPP_MULT_EQ, CPP_MULT)do { result->type = CPP_MULT; if (*buffer->cur == '=') buffer ->cur++, result->type = CPP_MULT_EQ; } while (0); break; |
| 4194 | case '=': IF_NEXT_IS ('=', CPP_EQ_EQ, CPP_EQ)do { result->type = CPP_EQ; if (*buffer->cur == '=') buffer ->cur++, result->type = CPP_EQ_EQ; } while (0); break; |
| 4195 | case '!': IF_NEXT_IS ('=', CPP_NOT_EQ, CPP_NOT)do { result->type = CPP_NOT; if (*buffer->cur == '=') buffer ->cur++, result->type = CPP_NOT_EQ; } while (0); break; |
| 4196 | case '^': IF_NEXT_IS ('=', CPP_XOR_EQ, CPP_XOR)do { result->type = CPP_XOR; if (*buffer->cur == '=') buffer ->cur++, result->type = CPP_XOR_EQ; } while (0); break; |
| 4197 | case '#': IF_NEXT_IS ('#', CPP_PASTE, CPP_HASH)do { result->type = CPP_HASH; if (*buffer->cur == '#') buffer ->cur++, result->type = CPP_PASTE; } while (0); result->val.token_no = 0; break; |
| 4198 | |
| 4199 | case '?': result->type = CPP_QUERY; break; |
| 4200 | case '~': result->type = CPP_COMPL; break; |
| 4201 | case ',': result->type = CPP_COMMA; break; |
| 4202 | case '(': result->type = CPP_OPEN_PAREN; break; |
| 4203 | case ')': result->type = CPP_CLOSE_PAREN; break; |
| 4204 | case '[': result->type = CPP_OPEN_SQUARE; break; |
| 4205 | case ']': result->type = CPP_CLOSE_SQUARE; break; |
| 4206 | case '{': result->type = CPP_OPEN_BRACE; break; |
| 4207 | case '}': result->type = CPP_CLOSE_BRACE; break; |
| 4208 | case ';': result->type = CPP_SEMICOLON; break; |
| 4209 | |
| 4210 | /* @ is a punctuator in Objective-C. */ |
| 4211 | case '@': result->type = CPP_ATSIGN; break; |
| 4212 | |
| 4213 | default: |
| 4214 | { |
| 4215 | const uchar *base = --buffer->cur; |
| 4216 | static int no_warn_cnt; |
| 4217 | |
| 4218 | /* Check for an extended identifier ($ or UCN or UTF-8). */ |
| 4219 | struct normalize_state nst = INITIAL_NORMALIZE_STATE{ 0, 0, normalized_KC }; |
| 4220 | if (forms_identifier_p (pfile, true, &nst)) |
| 4221 | { |
| 4222 | result->type = CPP_NAME; |
| 4223 | result->val.node.node = lex_identifier (pfile, base, true, &nst, |
| 4224 | &result->val.node.spelling); |
| 4225 | warn_about_normalization (pfile, result, &nst, true); |
| 4226 | break; |
| 4227 | } |
| 4228 | |
| 4229 | /* Otherwise this will form a CPP_OTHER token. Parse valid UTF-8 as a |
| 4230 | single token. */ |
| 4231 | buffer->cur++; |
| 4232 | if (c >= utf8_signifier) |
| 4233 | { |
| 4234 | const uchar *pstr = base; |
| 4235 | cppchar_t s; |
| 4236 | if (_cpp_valid_utf8 (pfile, &pstr, buffer->rlimit, 0, NULL__null, &s)) |
| 4237 | { |
| 4238 | if (s > UCS_LIMIT0x10FFFF && CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8)) |
| 4239 | { |
| 4240 | buffer->cur = base; |
| 4241 | _cpp_warn_invalid_utf8 (pfile); |
| 4242 | } |
| 4243 | buffer->cur = pstr; |
| 4244 | } |
| 4245 | else if (CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8)) |
| 4246 | { |
| 4247 | buffer->cur = base; |
| 4248 | const uchar *end = _cpp_warn_invalid_utf8 (pfile); |
| 4249 | buffer->cur = base + 1; |
| 4250 | no_warn_cnt = end - buffer->cur; |
| 4251 | } |
| 4252 | } |
| 4253 | else if (c >= utf8_continuation |
| 4254 | && CPP_OPTION (pfile, cpp_warn_invalid_utf8)((pfile)->opts.cpp_warn_invalid_utf8)) |
| 4255 | { |
| 4256 | if (no_warn_cnt) |
| 4257 | --no_warn_cnt; |
| 4258 | else |
| 4259 | { |
| 4260 | buffer->cur = base; |
| 4261 | _cpp_warn_invalid_utf8 (pfile); |
| 4262 | buffer->cur = base + 1; |
| 4263 | } |
| 4264 | } |
| 4265 | create_literal (pfile, result, base, buffer->cur - base, CPP_OTHER); |
| 4266 | break; |
| 4267 | } |
| 4268 | |
| 4269 | } |
| 4270 | |
| 4271 | /* Potentially convert the location of the token to a range. */ |
| 4272 | if (result->src_loc >= RESERVED_LOCATION_COUNT |
| 4273 | && result->type != CPP_EOF) |
| 4274 | { |
| 4275 | /* Ensure that any line notes are processed, so that we have the |
| 4276 | correct physical line/column for the end-point of the token even |
| 4277 | when a logical line is split via one or more backslashes. */ |
| 4278 | if (buffer->cur >= buffer->notes[buffer->cur_note].pos |
| 4279 | && !pfile->overlaid_buffer) |
| 4280 | _cpp_process_line_notes (pfile, false); |
| 4281 | |
| 4282 | source_range tok_range; |
| 4283 | tok_range.m_start = result->src_loc; |
| 4284 | tok_range.m_finish |
| 4285 | = linemap_position_for_column (pfile->line_table, |
| 4286 | CPP_BUF_COLUMN (buffer, buffer->cur)((buffer->cur) - (buffer)->line_base)); |
| 4287 | |
| 4288 | result->src_loc = COMBINE_LOCATION_DATA (pfile->line_table, |
| 4289 | result->src_loc, |
| 4290 | tok_range, NULL__null, 0); |
| 4291 | } |
| 4292 | |
| 4293 | return result; |
| 4294 | } |
| 4295 | |
| 4296 | /* An upper bound on the number of bytes needed to spell TOKEN. |
| 4297 | Does not include preceding whitespace. */ |
| 4298 | unsigned int |
| 4299 | cpp_token_len (const cpp_token *token) |
| 4300 | { |
| 4301 | unsigned int len; |
| 4302 | |
| 4303 | switch (TOKEN_SPELL (token)(token_spellings[(token)->type].category)) |
| 4304 | { |
| 4305 | default: len = 6; break; |
| 4306 | case SPELL_LITERAL: len = token->val.str.len; break; |
| 4307 | case SPELL_IDENT: len = NODE_LEN (token->val.node.node)(((&(token->val.node.node)->ident))->len) * 10; break; |
| 4308 | } |
| 4309 | |
| 4310 | return len; |
| 4311 | } |
| 4312 | |
| 4313 | /* Parse UTF-8 out of NAMEP and place a \U escape in BUFFER. |
| 4314 | Return the number of bytes read out of NAME. (There are always |
| 4315 | 10 bytes written to BUFFER.) */ |
| 4316 | |
| 4317 | static size_t |
| 4318 | utf8_to_ucn (unsigned char *buffer, const unsigned char *name) |
| 4319 | { |
| 4320 | int j; |
| 4321 | int ucn_len = 0; |
| 4322 | int ucn_len_c; |
| 4323 | unsigned t; |
| 4324 | unsigned long utf32; |
| 4325 | |
| 4326 | /* Compute the length of the UTF-8 sequence. */ |
| 4327 | for (t = *name; t & 0x80; t <<= 1) |
| 4328 | ucn_len++; |
| 4329 | |
| 4330 | utf32 = *name & (0x7F >> ucn_len); |
| 4331 | for (ucn_len_c = 1; ucn_len_c < ucn_len; ucn_len_c++) |
| 4332 | { |
| 4333 | utf32 = (utf32 << 6) | (*++name & 0x3F); |
| 4334 | |
| 4335 | /* Ill-formed UTF-8. */ |
| 4336 | if ((*name & ~0x3F) != 0x80) |
| 4337 | abort ()fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 4337, __FUNCTION__); |
| 4338 | } |
| 4339 | |
| 4340 | *buffer++ = '\\'; |
| 4341 | *buffer++ = 'U'; |
| 4342 | for (j = 7; j >= 0; j--) |
| 4343 | *buffer++ = "0123456789abcdef"[(utf32 >> (4 * j)) & 0xF]; |
| 4344 | return ucn_len; |
| 4345 | } |
| 4346 | |
| 4347 | /* Given a token TYPE corresponding to a digraph, return a pointer to |
| 4348 | the spelling of the digraph. */ |
| 4349 | static const unsigned char * |
| 4350 | cpp_digraph2name (enum cpp_ttype type) |
| 4351 | { |
| 4352 | return digraph_spellings[(int) type - (int) CPP_FIRST_DIGRAPH]; |
| 4353 | } |
| 4354 | |
| 4355 | /* Write the spelling of an identifier IDENT, using UCNs, to BUFFER. |
| 4356 | The buffer must already contain the enough space to hold the |
| 4357 | token's spelling. Returns a pointer to the character after the |
| 4358 | last character written. */ |
| 4359 | unsigned char * |
| 4360 | _cpp_spell_ident_ucns (unsigned char *buffer, cpp_hashnode *ident) |
| 4361 | { |
| 4362 | size_t i; |
| 4363 | const unsigned char *name = NODE_NAME (ident)(((&(ident)->ident))->str); |
| 4364 | |
| 4365 | for (i = 0; i < NODE_LEN (ident)(((&(ident)->ident))->len); i++) |
| 4366 | if (name[i] & ~0x7F) |
| 4367 | { |
| 4368 | i += utf8_to_ucn (buffer, name + i) - 1; |
| 4369 | buffer += 10; |
| 4370 | } |
| 4371 | else |
| 4372 | *buffer++ = name[i]; |
| 4373 | |
| 4374 | return buffer; |
| 4375 | } |
| 4376 | |
| 4377 | /* Write the spelling of a token TOKEN to BUFFER. The buffer must |
| 4378 | already contain the enough space to hold the token's spelling. |
| 4379 | Returns a pointer to the character after the last character written. |
| 4380 | FORSTRING is true if this is to be the spelling after translation |
| 4381 | phase 1 (with the original spelling of extended identifiers), false |
| 4382 | if extended identifiers should always be written using UCNs (there is |
| 4383 | no option for always writing them in the internal UTF-8 form). |
| 4384 | FIXME: Would be nice if we didn't need the PFILE argument. */ |
| 4385 | unsigned char * |
| 4386 | cpp_spell_token (cpp_reader *pfile, const cpp_token *token, |
| 4387 | unsigned char *buffer, bool forstring) |
| 4388 | { |
| 4389 | switch (TOKEN_SPELL (token)(token_spellings[(token)->type].category)) |
| 4390 | { |
| 4391 | case SPELL_OPERATOR: |
| 4392 | { |
| 4393 | const unsigned char *spelling; |
| 4394 | unsigned char c; |
| 4395 | |
| 4396 | if (token->flags & DIGRAPH(1 << 1)) |
| 4397 | spelling = cpp_digraph2name (token->type); |
| 4398 | else if (token->flags & NAMED_OP(1 << 4)) |
| 4399 | goto spell_ident; |
| 4400 | else |
| 4401 | spelling = TOKEN_NAME (token)(token_spellings[(token)->type].name); |
| 4402 | |
| 4403 | while ((c = *spelling++) != '\0') |
| 4404 | *buffer++ = c; |
| 4405 | } |
| 4406 | break; |
| 4407 | |
| 4408 | spell_ident: |
| 4409 | case SPELL_IDENT: |
| 4410 | if (forstring) |
| 4411 | { |
| 4412 | memcpy (buffer, NODE_NAME (token->val.node.spelling)(((&(token->val.node.spelling)->ident))->str), |
| 4413 | NODE_LEN (token->val.node.spelling)(((&(token->val.node.spelling)->ident))->len)); |
| 4414 | buffer += NODE_LEN (token->val.node.spelling)(((&(token->val.node.spelling)->ident))->len); |
| 4415 | } |
| 4416 | else |
| 4417 | buffer = _cpp_spell_ident_ucns (buffer, token->val.node.node); |
| 4418 | break; |
| 4419 | |
| 4420 | case SPELL_LITERAL: |
| 4421 | memcpy (buffer, token->val.str.text, token->val.str.len); |
| 4422 | buffer += token->val.str.len; |
| 4423 | break; |
| 4424 | |
| 4425 | case SPELL_NONE: |
| 4426 | cpp_error (pfile, CPP_DL_ICE, |
| 4427 | "unspellable token %s", TOKEN_NAME (token)(token_spellings[(token)->type].name)); |
| 4428 | break; |
| 4429 | } |
| 4430 | |
| 4431 | return buffer; |
| 4432 | } |
| 4433 | |
| 4434 | /* Returns TOKEN spelt as a null-terminated string. The string is |
| 4435 | freed when the reader is destroyed. Useful for diagnostics. */ |
| 4436 | unsigned char * |
| 4437 | cpp_token_as_text (cpp_reader *pfile, const cpp_token *token) |
| 4438 | { |
| 4439 | unsigned int len = cpp_token_len (token) + 1; |
| 4440 | unsigned char *start = _cpp_unaligned_alloc (pfile, len), *end; |
| 4441 | |
| 4442 | end = cpp_spell_token (pfile, token, start, false); |
| 4443 | end[0] = '\0'; |
| 4444 | |
| 4445 | return start; |
| 4446 | } |
| 4447 | |
| 4448 | /* Returns a pointer to a string which spells the token defined by |
| 4449 | TYPE and FLAGS. Used by C front ends, which really should move to |
| 4450 | using cpp_token_as_text. */ |
| 4451 | const char * |
| 4452 | cpp_type2name (enum cpp_ttype type, unsigned char flags) |
| 4453 | { |
| 4454 | if (flags & DIGRAPH(1 << 1)) |
| 4455 | return (const char *) cpp_digraph2name (type); |
| 4456 | else if (flags & NAMED_OP(1 << 4)) |
| 4457 | return cpp_named_operator2name (type); |
| 4458 | |
| 4459 | return (const char *) token_spellings[type].name; |
| 4460 | } |
| 4461 | |
| 4462 | /* Writes the spelling of token to FP, without any preceding space. |
| 4463 | Separated from cpp_spell_token for efficiency - to avoid stdio |
| 4464 | double-buffering. */ |
| 4465 | void |
| 4466 | cpp_output_token (const cpp_token *token, FILE *fp) |
| 4467 | { |
| 4468 | switch (TOKEN_SPELL (token)(token_spellings[(token)->type].category)) |
| 4469 | { |
| 4470 | case SPELL_OPERATOR: |
| 4471 | { |
| 4472 | const unsigned char *spelling; |
| 4473 | int c; |
| 4474 | |
| 4475 | if (token->flags & DIGRAPH(1 << 1)) |
| 4476 | spelling = cpp_digraph2name (token->type); |
| 4477 | else if (token->flags & NAMED_OP(1 << 4)) |
| 4478 | goto spell_ident; |
| 4479 | else |
| 4480 | spelling = TOKEN_NAME (token)(token_spellings[(token)->type].name); |
| 4481 | |
| 4482 | c = *spelling; |
| 4483 | do |
| 4484 | putc (c, fp)putc_unlocked (c, fp); |
| 4485 | while ((c = *++spelling) != '\0'); |
| 4486 | } |
| 4487 | break; |
| 4488 | |
| 4489 | spell_ident: |
| 4490 | case SPELL_IDENT: |
| 4491 | { |
| 4492 | size_t i; |
| 4493 | const unsigned char * name = NODE_NAME (token->val.node.node)(((&(token->val.node.node)->ident))->str); |
| 4494 | |
| 4495 | for (i = 0; i < NODE_LEN (token->val.node.node)(((&(token->val.node.node)->ident))->len); i++) |
| 4496 | if (name[i] & ~0x7F) |
| 4497 | { |
| 4498 | unsigned char buffer[10]; |
| 4499 | i += utf8_to_ucn (buffer, name + i) - 1; |
| 4500 | fwrite (buffer, 1, 10, fp)fwrite_unlocked (buffer, 1, 10, fp); |
| 4501 | } |
| 4502 | else |
| 4503 | fputc (NODE_NAME (token->val.node.node)[i], fp)fputc_unlocked ((((&(token->val.node.node)->ident)) ->str)[i], fp); |
| 4504 | } |
| 4505 | break; |
| 4506 | |
| 4507 | case SPELL_LITERAL: |
| 4508 | if (token->type == CPP_HEADER_NAME) |
| 4509 | fputc ('"', fp)fputc_unlocked ('"', fp); |
| 4510 | fwrite (token->val.str.text, 1, token->val.str.len, fp)fwrite_unlocked (token->val.str.text, 1, token->val.str .len, fp); |
| 4511 | if (token->type == CPP_HEADER_NAME) |
| 4512 | fputc ('"', fp)fputc_unlocked ('"', fp); |
| 4513 | break; |
| 4514 | |
| 4515 | case SPELL_NONE: |
| 4516 | /* An error, most probably. */ |
| 4517 | break; |
| 4518 | } |
| 4519 | } |
| 4520 | |
| 4521 | /* Compare two tokens. */ |
| 4522 | int |
| 4523 | _cpp_equiv_tokens (const cpp_token *a, const cpp_token *b) |
| 4524 | { |
| 4525 | if (a->type == b->type && a->flags == b->flags) |
| 4526 | switch (TOKEN_SPELL (a)(token_spellings[(a)->type].category)) |
| 4527 | { |
| 4528 | default: /* Keep compiler happy. */ |
| 4529 | case SPELL_OPERATOR: |
| 4530 | /* token_no is used to track where multiple consecutive ## |
| 4531 | tokens were originally located. */ |
| 4532 | return (a->type != CPP_PASTE || a->val.token_no == b->val.token_no); |
| 4533 | case SPELL_NONE: |
| 4534 | return (a->type != CPP_MACRO_ARG |
| 4535 | || (a->val.macro_arg.arg_no == b->val.macro_arg.arg_no |
| 4536 | && a->val.macro_arg.spelling == b->val.macro_arg.spelling)); |
| 4537 | case SPELL_IDENT: |
| 4538 | return (a->val.node.node == b->val.node.node |
| 4539 | && a->val.node.spelling == b->val.node.spelling); |
| 4540 | case SPELL_LITERAL: |
| 4541 | return (a->val.str.len == b->val.str.len |
| 4542 | && !memcmp (a->val.str.text, b->val.str.text, |
| 4543 | a->val.str.len)); |
| 4544 | } |
| 4545 | |
| 4546 | return 0; |
| 4547 | } |
| 4548 | |
| 4549 | /* Returns nonzero if a space should be inserted to avoid an |
| 4550 | accidental token paste for output. For simplicity, it is |
| 4551 | conservative, and occasionally advises a space where one is not |
| 4552 | needed, e.g. "." and ".2". */ |
| 4553 | int |
| 4554 | cpp_avoid_paste (cpp_reader *pfile, const cpp_token *token1, |
| 4555 | const cpp_token *token2) |
| 4556 | { |
| 4557 | enum cpp_ttype a = token1->type, b = token2->type; |
| 4558 | cppchar_t c; |
| 4559 | |
| 4560 | if (token1->flags & NAMED_OP(1 << 4)) |
| 4561 | a = CPP_NAME; |
| 4562 | if (token2->flags & NAMED_OP(1 << 4)) |
| 4563 | b = CPP_NAME; |
| 4564 | |
| 4565 | c = EOF(-1); |
| 4566 | if (token2->flags & DIGRAPH(1 << 1)) |
| 4567 | c = digraph_spellings[(int) b - (int) CPP_FIRST_DIGRAPH][0]; |
| 4568 | else if (token_spellings[b].category == SPELL_OPERATOR) |
| 4569 | c = token_spellings[b].name[0]; |
| 4570 | |
| 4571 | /* Quickly get everything that can paste with an '='. */ |
| 4572 | if ((int) a <= (int) CPP_LAST_EQ && c == '=') |
| 4573 | return 1; |
| 4574 | |
| 4575 | switch (a) |
| 4576 | { |
| 4577 | case CPP_GREATER: return c == '>'; |
| 4578 | case CPP_LESS: return c == '<' || c == '%' || c == ':'; |
| 4579 | case CPP_PLUS: return c == '+'; |
| 4580 | case CPP_MINUS: return c == '-' || c == '>'; |
| 4581 | case CPP_DIV: return c == '/' || c == '*'; /* Comments. */ |
| 4582 | case CPP_MOD: return c == ':' || c == '>'; |
| 4583 | case CPP_AND: return c == '&'; |
| 4584 | case CPP_OR: return c == '|'; |
| 4585 | case CPP_COLON: return c == ':' || c == '>'; |
| 4586 | case CPP_DEREF: return c == '*'; |
| 4587 | case CPP_DOT: return c == '.' || c == '%' || b == CPP_NUMBER; |
| 4588 | case CPP_HASH: return c == '#' || c == '%'; /* Digraph form. */ |
| 4589 | case CPP_PRAGMA: |
| 4590 | case CPP_NAME: return ((b == CPP_NUMBER |
| 4591 | && name_p (pfile, &token2->val.str)) |
| 4592 | || b == CPP_NAME |
| 4593 | || b == CPP_CHAR || b == CPP_STRING); /* L */ |
| 4594 | case CPP_NUMBER: return (b == CPP_NUMBER || b == CPP_NAME |
| 4595 | || b == CPP_CHAR |
| 4596 | || c == '.' || c == '+' || c == '-'); |
| 4597 | /* UCNs */ |
| 4598 | case CPP_OTHER: return ((token1->val.str.text[0] == '\\' |
| 4599 | && b == CPP_NAME) |
| 4600 | || (CPP_OPTION (pfile, objc)((pfile)->opts.objc) |
| 4601 | && token1->val.str.text[0] == '@' |
| 4602 | && (b == CPP_NAME || b == CPP_STRING))); |
| 4603 | case CPP_LESS_EQ: return c == '>'; |
| 4604 | case CPP_STRING: |
| 4605 | case CPP_WSTRING: |
| 4606 | case CPP_UTF8STRING: |
| 4607 | case CPP_STRING16: |
| 4608 | case CPP_STRING32: return (CPP_OPTION (pfile, user_literals)((pfile)->opts.user_literals) |
| 4609 | && (b == CPP_NAME |
| 4610 | || (TOKEN_SPELL (token2)(token_spellings[(token2)->type].category) == SPELL_LITERAL |
| 4611 | && ISIDST (token2->val.str.text[0])(_sch_istable[(token2->val.str.text[0]) & 0xff] & ( unsigned short)(_sch_isidst))))); |
| 4612 | |
| 4613 | default: break; |
| 4614 | } |
| 4615 | |
| 4616 | return 0; |
| 4617 | } |
| 4618 | |
| 4619 | /* Output all the remaining tokens on the current line, and a newline |
| 4620 | character, to FP. Leading whitespace is removed. If there are |
| 4621 | macros, special token padding is not performed. */ |
| 4622 | void |
| 4623 | cpp_output_line (cpp_reader *pfile, FILE *fp) |
| 4624 | { |
| 4625 | const cpp_token *token; |
| 4626 | |
| 4627 | token = cpp_get_token (pfile); |
| 4628 | while (token->type != CPP_EOF) |
| 4629 | { |
| 4630 | cpp_output_token (token, fp); |
| 4631 | token = cpp_get_token (pfile); |
| 4632 | if (token->flags & PREV_WHITE(1 << 0)) |
| 4633 | putc (' ', fp)putc_unlocked (' ', fp); |
| 4634 | } |
| 4635 | |
| 4636 | putc ('\n', fp)putc_unlocked ('\n', fp); |
| 4637 | } |
| 4638 | |
| 4639 | /* Return a string representation of all the remaining tokens on the |
| 4640 | current line. The result is allocated using xmalloc and must be |
| 4641 | freed by the caller. */ |
| 4642 | unsigned char * |
| 4643 | cpp_output_line_to_string (cpp_reader *pfile, const unsigned char *dir_name) |
| 4644 | { |
| 4645 | const cpp_token *token; |
| 4646 | unsigned int out = dir_name ? ustrlen (dir_name) : 0; |
| 4647 | unsigned int alloced = 120 + out; |
| 4648 | unsigned char *result = (unsigned char *) xmalloc (alloced); |
| 4649 | |
| 4650 | /* If DIR_NAME is empty, there are no initial contents. */ |
| 4651 | if (dir_name) |
| 4652 | { |
| 4653 | sprintf ((char *) result, "#%s ", dir_name); |
| 4654 | out += 2; |
| 4655 | } |
| 4656 | |
| 4657 | token = cpp_get_token (pfile); |
| 4658 | while (token->type != CPP_EOF) |
| 4659 | { |
| 4660 | unsigned char *last; |
| 4661 | /* Include room for a possible space and the terminating nul. */ |
| 4662 | unsigned int len = cpp_token_len (token) + 2; |
| 4663 | |
| 4664 | if (out + len > alloced) |
| 4665 | { |
| 4666 | alloced *= 2; |
| 4667 | if (out + len > alloced) |
| 4668 | alloced = out + len; |
| 4669 | result = (unsigned char *) xrealloc (result, alloced); |
| 4670 | } |
| 4671 | |
| 4672 | last = cpp_spell_token (pfile, token, &result[out], 0); |
| 4673 | out = last - result; |
| 4674 | |
| 4675 | token = cpp_get_token (pfile); |
| 4676 | if (token->flags & PREV_WHITE(1 << 0)) |
| 4677 | result[out++] = ' '; |
| 4678 | } |
| 4679 | |
| 4680 | result[out] = '\0'; |
| 4681 | return result; |
| 4682 | } |
| 4683 | |
| 4684 | /* Memory buffers. Changing these three constants can have a dramatic |
| 4685 | effect on performance. The values here are reasonable defaults, |
| 4686 | but might be tuned. If you adjust them, be sure to test across a |
| 4687 | range of uses of cpplib, including heavy nested function-like macro |
| 4688 | expansion. Also check the change in peak memory usage (NJAMD is a |
| 4689 | good tool for this). */ |
| 4690 | #define MIN_BUFF_SIZE8000 8000 |
| 4691 | #define BUFF_SIZE_UPPER_BOUND(MIN_SIZE)(8000 + (MIN_SIZE) * 3 / 2) (MIN_BUFF_SIZE8000 + (MIN_SIZE) * 3 / 2) |
| 4692 | #define EXTENDED_BUFF_SIZE(BUFF, MIN_EXTRA)(MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) \ |
| 4693 | (MIN_EXTRA + ((BUFF)->limit - (BUFF)->cur) * 2) |
| 4694 | |
| 4695 | #if MIN_BUFF_SIZE8000 > BUFF_SIZE_UPPER_BOUND (0)(8000 + (0) * 3 / 2) |
| 4696 | #error BUFF_SIZE_UPPER_BOUND must be at least as large as MIN_BUFF_SIZE8000! |
| 4697 | #endif |
| 4698 | |
| 4699 | /* Create a new allocation buffer. Place the control block at the end |
| 4700 | of the buffer, so that buffer overflows will cause immediate chaos. */ |
| 4701 | static _cpp_buff * |
| 4702 | new_buff (size_t len) |
| 4703 | { |
| 4704 | _cpp_buff *result; |
| 4705 | unsigned char *base; |
| 4706 | |
| 4707 | if (len < MIN_BUFF_SIZE8000) |
| 4708 | len = MIN_BUFF_SIZE8000; |
| 4709 | len = CPP_ALIGN (len)(((len) + ((__builtin_offsetof(struct dummy, u)) - 1)) & ~ ((__builtin_offsetof(struct dummy, u)) - 1)); |
| 4710 | |
| 4711 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
| 4712 | /* Valgrind warns about uses of interior pointers, so put _cpp_buff |
| 4713 | struct first. */ |
| 4714 | size_t slen = CPP_ALIGN2 (sizeof (_cpp_buff), 2 * DEFAULT_ALIGNMENT)(((sizeof (_cpp_buff)) + ((2 * __builtin_offsetof(struct dummy , u)) - 1)) & ~((2 * __builtin_offsetof(struct dummy, u)) - 1)); |
| 4715 | base = XNEWVEC (unsigned char, len + slen)((unsigned char *) xmalloc (sizeof (unsigned char) * (len + slen ))); |
| 4716 | result = (_cpp_buff *) base; |
| 4717 | base += slen; |
| 4718 | #else |
| 4719 | base = XNEWVEC (unsigned char, len + sizeof (_cpp_buff))((unsigned char *) xmalloc (sizeof (unsigned char) * (len + sizeof (_cpp_buff)))); |
| 4720 | result = (_cpp_buff *) (base + len); |
| 4721 | #endif |
| 4722 | result->base = base; |
| 4723 | result->cur = base; |
| 4724 | result->limit = base + len; |
| 4725 | result->next = NULL__null; |
| 4726 | return result; |
| 4727 | } |
| 4728 | |
| 4729 | /* Place a chain of unwanted allocation buffers on the free list. */ |
| 4730 | void |
| 4731 | _cpp_release_buff (cpp_reader *pfile, _cpp_buff *buff) |
| 4732 | { |
| 4733 | _cpp_buff *end = buff; |
| 4734 | |
| 4735 | while (end->next) |
| 4736 | end = end->next; |
| 4737 | end->next = pfile->free_buffs; |
| 4738 | pfile->free_buffs = buff; |
| 4739 | } |
| 4740 | |
| 4741 | /* Return a free buffer of size at least MIN_SIZE. */ |
| 4742 | _cpp_buff * |
| 4743 | _cpp_get_buff (cpp_reader *pfile, size_t min_size) |
| 4744 | { |
| 4745 | _cpp_buff *result, **p; |
| 4746 | |
| 4747 | for (p = &pfile->free_buffs;; p = &(*p)->next) |
| 4748 | { |
| 4749 | size_t size; |
| 4750 | |
| 4751 | if (*p == NULL__null) |
| 4752 | return new_buff (min_size); |
| 4753 | result = *p; |
| 4754 | size = result->limit - result->base; |
| 4755 | /* Return a buffer that's big enough, but don't waste one that's |
| 4756 | way too big. */ |
| 4757 | if (size >= min_size && size <= BUFF_SIZE_UPPER_BOUND (min_size)(8000 + (min_size) * 3 / 2)) |
| 4758 | break; |
| 4759 | } |
| 4760 | |
| 4761 | *p = result->next; |
| 4762 | result->next = NULL__null; |
| 4763 | result->cur = result->base; |
| 4764 | return result; |
| 4765 | } |
| 4766 | |
| 4767 | /* Creates a new buffer with enough space to hold the uncommitted |
| 4768 | remaining bytes of BUFF, and at least MIN_EXTRA more bytes. Copies |
| 4769 | the excess bytes to the new buffer. Chains the new buffer after |
| 4770 | BUFF, and returns the new buffer. */ |
| 4771 | _cpp_buff * |
| 4772 | _cpp_append_extend_buff (cpp_reader *pfile, _cpp_buff *buff, size_t min_extra) |
| 4773 | { |
| 4774 | size_t size = EXTENDED_BUFF_SIZE (buff, min_extra)(min_extra + ((buff)->limit - (buff)->cur) * 2); |
| 4775 | _cpp_buff *new_buff = _cpp_get_buff (pfile, size); |
| 4776 | |
| 4777 | buff->next = new_buff; |
| 4778 | memcpy (new_buff->base, buff->cur, BUFF_ROOM (buff)(size_t) ((buff)->limit - (buff)->cur)); |
| 4779 | return new_buff; |
| 4780 | } |
| 4781 | |
| 4782 | /* Creates a new buffer with enough space to hold the uncommitted |
| 4783 | remaining bytes of the buffer pointed to by BUFF, and at least |
| 4784 | MIN_EXTRA more bytes. Copies the excess bytes to the new buffer. |
| 4785 | Chains the new buffer before the buffer pointed to by BUFF, and |
| 4786 | updates the pointer to point to the new buffer. */ |
| 4787 | void |
| 4788 | _cpp_extend_buff (cpp_reader *pfile, _cpp_buff **pbuff, size_t min_extra) |
| 4789 | { |
| 4790 | _cpp_buff *new_buff, *old_buff = *pbuff; |
| 4791 | size_t size = EXTENDED_BUFF_SIZE (old_buff, min_extra)(min_extra + ((old_buff)->limit - (old_buff)->cur) * 2); |
| 4792 | |
| 4793 | new_buff = _cpp_get_buff (pfile, size); |
| 4794 | memcpy (new_buff->base, old_buff->cur, BUFF_ROOM (old_buff)(size_t) ((old_buff)->limit - (old_buff)->cur)); |
| 4795 | new_buff->next = old_buff; |
| 4796 | *pbuff = new_buff; |
| 4797 | } |
| 4798 | |
| 4799 | /* Free a chain of buffers starting at BUFF. */ |
| 4800 | void |
| 4801 | _cpp_free_buff (_cpp_buff *buff) |
| 4802 | { |
| 4803 | _cpp_buff *next; |
| 4804 | |
| 4805 | for (; buff; buff = next) |
| 4806 | { |
| 4807 | next = buff->next; |
| 4808 | #ifdef ENABLE_VALGRIND_ANNOTATIONS |
| 4809 | free (buff); |
| 4810 | #else |
| 4811 | free (buff->base); |
| 4812 | #endif |
| 4813 | } |
| 4814 | } |
| 4815 | |
| 4816 | /* Allocate permanent, unaligned storage of length LEN. */ |
| 4817 | unsigned char * |
| 4818 | _cpp_unaligned_alloc (cpp_reader *pfile, size_t len) |
| 4819 | { |
| 4820 | _cpp_buff *buff = pfile->u_buff; |
| 4821 | unsigned char *result = buff->cur; |
| 4822 | |
| 4823 | if (len > (size_t) (buff->limit - result)) |
| 4824 | { |
| 4825 | buff = _cpp_get_buff (pfile, len); |
| 4826 | buff->next = pfile->u_buff; |
| 4827 | pfile->u_buff = buff; |
| 4828 | result = buff->cur; |
| 4829 | } |
| 4830 | |
| 4831 | buff->cur = result + len; |
| 4832 | return result; |
| 4833 | } |
| 4834 | |
| 4835 | /* Allocate permanent, unaligned storage of length LEN from a_buff. |
| 4836 | That buffer is used for growing allocations when saving macro |
| 4837 | replacement lists in a #define, and when parsing an answer to an |
| 4838 | assertion in #assert, #unassert or #if (and therefore possibly |
| 4839 | whilst expanding macros). It therefore must not be used by any |
| 4840 | code that they might call: specifically the lexer and the guts of |
| 4841 | the macro expander. |
| 4842 | |
| 4843 | All existing other uses clearly fit this restriction: storing |
| 4844 | registered pragmas during initialization. */ |
| 4845 | unsigned char * |
| 4846 | _cpp_aligned_alloc (cpp_reader *pfile, size_t len) |
| 4847 | { |
| 4848 | _cpp_buff *buff = pfile->a_buff; |
| 4849 | unsigned char *result = buff->cur; |
| 4850 | |
| 4851 | if (len > (size_t) (buff->limit - result)) |
| 4852 | { |
| 4853 | buff = _cpp_get_buff (pfile, len); |
| 4854 | buff->next = pfile->a_buff; |
| 4855 | pfile->a_buff = buff; |
| 4856 | result = buff->cur; |
| 4857 | } |
| 4858 | |
| 4859 | buff->cur = result + len; |
| 4860 | return result; |
| 4861 | } |
| 4862 | |
| 4863 | /* Commit or allocate storage from a buffer. */ |
| 4864 | |
| 4865 | void * |
| 4866 | _cpp_commit_buff (cpp_reader *pfile, size_t size) |
| 4867 | { |
| 4868 | void *ptr = BUFF_FRONT (pfile->a_buff)((pfile->a_buff)->cur); |
| 4869 | |
| 4870 | if (pfile->hash_table->alloc_subobject) |
| 4871 | { |
| 4872 | void *copy = pfile->hash_table->alloc_subobject (size); |
| 4873 | memcpy (copy, ptr, size); |
| 4874 | ptr = copy; |
| 4875 | } |
| 4876 | else |
| 4877 | BUFF_FRONT (pfile->a_buff)((pfile->a_buff)->cur) += size; |
| 4878 | |
| 4879 | return ptr; |
| 4880 | } |
| 4881 | |
| 4882 | /* Say which field of TOK is in use. */ |
| 4883 | |
| 4884 | enum cpp_token_fld_kind |
| 4885 | cpp_token_val_index (const cpp_token *tok) |
| 4886 | { |
| 4887 | switch (TOKEN_SPELL (tok)(token_spellings[(tok)->type].category)) |
| 4888 | { |
| 4889 | case SPELL_IDENT: |
| 4890 | return CPP_TOKEN_FLD_NODE; |
| 4891 | case SPELL_LITERAL: |
| 4892 | return CPP_TOKEN_FLD_STR; |
| 4893 | case SPELL_OPERATOR: |
| 4894 | /* Operands which were originally spelled as ident keep around |
| 4895 | the node for the exact spelling. */ |
| 4896 | if (tok->flags & NAMED_OP(1 << 4)) |
| 4897 | return CPP_TOKEN_FLD_NODE; |
| 4898 | else if (tok->type == CPP_PASTE) |
| 4899 | return CPP_TOKEN_FLD_TOKEN_NO; |
| 4900 | else |
| 4901 | return CPP_TOKEN_FLD_NONE; |
| 4902 | case SPELL_NONE: |
| 4903 | if (tok->type == CPP_MACRO_ARG) |
| 4904 | return CPP_TOKEN_FLD_ARG_NO; |
| 4905 | else if (tok->type == CPP_PADDING) |
| 4906 | return CPP_TOKEN_FLD_SOURCE; |
| 4907 | else if (tok->type == CPP_PRAGMA) |
| 4908 | return CPP_TOKEN_FLD_PRAGMA; |
| 4909 | /* fall through */ |
| 4910 | default: |
| 4911 | return CPP_TOKEN_FLD_NONE; |
| 4912 | } |
| 4913 | } |
| 4914 | |
| 4915 | /* All tokens lexed in R after calling this function will be forced to |
| 4916 | have their location_t to be P, until |
| 4917 | cpp_stop_forcing_token_locations is called for R. */ |
| 4918 | |
| 4919 | void |
| 4920 | cpp_force_token_locations (cpp_reader *r, location_t loc) |
| 4921 | { |
| 4922 | r->forced_token_location = loc; |
| 4923 | } |
| 4924 | |
| 4925 | /* Go back to assigning locations naturally for lexed tokens. */ |
| 4926 | |
| 4927 | void |
| 4928 | cpp_stop_forcing_token_locations (cpp_reader *r) |
| 4929 | { |
| 4930 | r->forced_token_location = 0; |
| 4931 | } |
| 4932 | |
| 4933 | /* We're looking at \, if it's escaping EOL, look past it. If at |
| 4934 | LIMIT, don't advance. */ |
| 4935 | |
| 4936 | static const unsigned char * |
| 4937 | do_peek_backslash (const unsigned char *peek, const unsigned char *limit) |
| 4938 | { |
| 4939 | const unsigned char *probe = peek; |
| 4940 | |
| 4941 | if (__builtin_expect (peek[1] == '\n', true)) |
| 4942 | { |
| 4943 | eol: |
| 4944 | probe += 2; |
| 4945 | if (__builtin_expect (probe < limit, true)) |
| 4946 | { |
| 4947 | peek = probe; |
| 4948 | if (*peek == '\\') |
| 4949 | /* The user might be perverse. */ |
| 4950 | return do_peek_backslash (peek, limit); |
| 4951 | } |
| 4952 | } |
| 4953 | else if (__builtin_expect (peek[1] == '\r', false)) |
| 4954 | { |
| 4955 | if (probe[2] == '\n') |
| 4956 | probe++; |
| 4957 | goto eol; |
| 4958 | } |
| 4959 | |
| 4960 | return peek; |
| 4961 | } |
| 4962 | |
| 4963 | static const unsigned char * |
| 4964 | do_peek_next (const unsigned char *peek, const unsigned char *limit) |
| 4965 | { |
| 4966 | if (__builtin_expect (*peek == '\\', false)) |
| 4967 | peek = do_peek_backslash (peek, limit); |
| 4968 | return peek; |
| 4969 | } |
| 4970 | |
| 4971 | static const unsigned char * |
| 4972 | do_peek_prev (const unsigned char *peek, const unsigned char *bound) |
| 4973 | { |
| 4974 | if (peek == bound) |
| 4975 | return NULL__null; |
| 4976 | |
| 4977 | unsigned char c = *--peek; |
| 4978 | if (__builtin_expect (c == '\n', false) |
| 4979 | || __builtin_expect (c == 'r', false)) |
| 4980 | { |
| 4981 | if (peek == bound) |
| 4982 | return peek; |
| 4983 | int ix = -1; |
| 4984 | if (c == '\n' && peek[ix] == '\r') |
| 4985 | { |
| 4986 | if (peek + ix == bound) |
| 4987 | return peek; |
| 4988 | ix--; |
| 4989 | } |
| 4990 | |
| 4991 | if (peek[ix] == '\\') |
| 4992 | return do_peek_prev (peek + ix, bound); |
| 4993 | |
| 4994 | return peek; |
| 4995 | } |
| 4996 | else |
| 4997 | return peek; |
| 4998 | } |
| 4999 | |
| 5000 | /* If PEEK[-1] is identifier MATCH, scan past it and trailing white |
| 5001 | space. Otherwise return NULL. */ |
| 5002 | |
| 5003 | static const unsigned char * |
| 5004 | do_peek_ident (const char *match, const unsigned char *peek, |
| 5005 | const unsigned char *limit) |
| 5006 | { |
| 5007 | for (; *++match; peek++) |
| 5008 | if (*peek != *match) |
| 5009 | { |
| 5010 | peek = do_peek_next (peek, limit); |
| 5011 | if (*peek != *match) |
| 5012 | return NULL__null; |
| 5013 | } |
| 5014 | |
| 5015 | /* Must now not be looking at an identifier char. */ |
| 5016 | peek = do_peek_next (peek, limit); |
| 5017 | if (ISIDNUM (*peek)(_sch_istable[(*peek) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 5018 | return NULL__null; |
| 5019 | |
| 5020 | /* Skip control-line whitespace. */ |
| 5021 | ws: |
| 5022 | while (*peek == ' ' || *peek == '\t') |
| 5023 | peek++; |
| 5024 | if (__builtin_expect (*peek == '\\', false)) |
| 5025 | { |
| 5026 | peek = do_peek_backslash (peek, limit); |
| 5027 | if (*peek != '\\') |
| 5028 | goto ws; |
| 5029 | } |
| 5030 | |
| 5031 | return peek; |
| 5032 | } |
| 5033 | |
| 5034 | /* Are we looking at a module control line starting as PEEK - 1? */ |
| 5035 | |
| 5036 | static bool |
| 5037 | do_peek_module (cpp_reader *pfile, unsigned char c, |
| 5038 | const unsigned char *peek, const unsigned char *limit) |
| 5039 | { |
| 5040 | bool import = false; |
| 5041 | |
| 5042 | if (__builtin_expect (c == 'e', false)) |
| 5043 | { |
| 5044 | if (!((peek[0] == 'x' || peek[0] == '\\') |
| 5045 | && (peek = do_peek_ident ("export", peek, limit)))) |
| 5046 | return false; |
| 5047 | |
| 5048 | /* export, peek for import or module. No need to peek __import |
| 5049 | here. */ |
| 5050 | if (peek[0] == 'i') |
| 5051 | { |
| 5052 | if (!((peek[1] == 'm' || peek[1] == '\\') |
| 5053 | && (peek = do_peek_ident ("import", peek + 1, limit)))) |
| 5054 | return false; |
| 5055 | import = true; |
| 5056 | } |
| 5057 | else if (peek[0] == 'm') |
| 5058 | { |
| 5059 | if (!((peek[1] == 'o' || peek[1] == '\\') |
| 5060 | && (peek = do_peek_ident ("module", peek + 1, limit)))) |
| 5061 | return false; |
| 5062 | } |
| 5063 | else |
| 5064 | return false; |
| 5065 | } |
| 5066 | else if (__builtin_expect (c == 'i', false)) |
| 5067 | { |
| 5068 | if (!((peek[0] == 'm' || peek[0] == '\\') |
| 5069 | && (peek = do_peek_ident ("import", peek, limit)))) |
| 5070 | return false; |
| 5071 | import = true; |
| 5072 | } |
| 5073 | else if (__builtin_expect (c == '_', false)) |
| 5074 | { |
| 5075 | /* Needed for translated includes. */ |
| 5076 | if (!((peek[0] == '_' || peek[0] == '\\') |
| 5077 | && (peek = do_peek_ident ("__import", peek, limit)))) |
| 5078 | return false; |
| 5079 | import = true; |
| 5080 | } |
| 5081 | else if (__builtin_expect (c == 'm', false)) |
| 5082 | { |
| 5083 | if (!((peek[0] == 'o' || peek[0] == '\\') |
| 5084 | && (peek = do_peek_ident ("module", peek, limit)))) |
| 5085 | return false; |
| 5086 | } |
| 5087 | else |
| 5088 | return false; |
| 5089 | |
| 5090 | /* Peek the next character to see if it's good enough. We'll be at |
| 5091 | the first non-whitespace char, including skipping an escaped |
| 5092 | newline. */ |
| 5093 | /* ... import followed by identifier, ':', '<' or header-name |
| 5094 | preprocessing tokens, or module followed by identifier, ':' or |
| 5095 | ';' preprocessing tokens. */ |
| 5096 | unsigned char p = *peek++; |
| 5097 | |
| 5098 | /* A character literal is ... single quotes, ... optionally preceded |
| 5099 | by u8, u, U, or L */ |
| 5100 | /* A string-literal is a ... double quotes, optionally prefixed by |
| 5101 | R, u8, u8R, u, uR, U, UR, L, or LR */ |
| 5102 | if (p == 'u') |
| 5103 | { |
| 5104 | peek = do_peek_next (peek, limit); |
| 5105 | if (*peek == '8') |
| 5106 | { |
| 5107 | peek++; |
| 5108 | goto peek_u8; |
| 5109 | } |
| 5110 | goto peek_u; |
| 5111 | } |
| 5112 | else if (p == 'U' || p == 'L') |
| 5113 | { |
| 5114 | peek_u8: |
| 5115 | peek = do_peek_next (peek, limit); |
| 5116 | peek_u: |
| 5117 | if (*peek == '\"' || *peek == '\'') |
| 5118 | return false; |
| 5119 | |
| 5120 | if (*peek == 'R') |
| 5121 | goto peek_R; |
| 5122 | /* Identifier. Ok. */ |
| 5123 | } |
| 5124 | else if (p == 'R') |
| 5125 | { |
| 5126 | peek_R: |
| 5127 | if (CPP_OPTION (pfile, rliterals)((pfile)->opts.rliterals)) |
| 5128 | { |
| 5129 | peek = do_peek_next (peek, limit); |
| 5130 | if (*peek == '\"') |
| 5131 | return false; |
| 5132 | } |
| 5133 | /* Identifier. Ok. */ |
| 5134 | } |
| 5135 | else if ('Z' - 'A' == 25 |
| 5136 | ? ((p >= 'A' && p <= 'Z') || (p >= 'a' && p <= 'z') || p == '_') |
| 5137 | : ISIDST (p)(_sch_istable[(p) & 0xff] & (unsigned short)(_sch_isidst ))) |
| 5138 | { |
| 5139 | /* Identifier. Ok. */ |
| 5140 | } |
| 5141 | else if (p == '<') |
| 5142 | { |
| 5143 | /* Maybe angle header, ok for import. Reject |
| 5144 | '<=', '<<' digraph:'<:'. */ |
| 5145 | if (!import) |
| 5146 | return false; |
| 5147 | peek = do_peek_next (peek, limit); |
| 5148 | if (*peek == '=' || *peek == '<' |
| 5149 | || (*peek == ':' && CPP_OPTION (pfile, digraphs)((pfile)->opts.digraphs))) |
| 5150 | return false; |
| 5151 | } |
| 5152 | else if (p == ';') |
| 5153 | { |
| 5154 | /* SEMICOLON, ok for module. */ |
| 5155 | if (import) |
| 5156 | return false; |
| 5157 | } |
| 5158 | else if (p == '"') |
| 5159 | { |
| 5160 | /* STRING, ok for import. */ |
| 5161 | if (!import) |
| 5162 | return false; |
| 5163 | } |
| 5164 | else if (p == ':') |
| 5165 | { |
| 5166 | /* Maybe COLON, ok. Reject '::', digraph:':>'. */ |
| 5167 | peek = do_peek_next (peek, limit); |
| 5168 | if (*peek == ':' || (*peek == '>' && CPP_OPTION (pfile, digraphs)((pfile)->opts.digraphs))) |
| 5169 | return false; |
| 5170 | } |
| 5171 | else |
| 5172 | /* FIXME: Detect a unicode character, excluding those not |
| 5173 | permitted as the initial character. [lex.name]/1. I presume |
| 5174 | we need to check the \[uU] spellings, and directly using |
| 5175 | Unicode in say UTF8 form? Or perhaps we do the phase-1 |
| 5176 | conversion of UTF8 to universal-character-names? */ |
| 5177 | return false; |
| 5178 | |
| 5179 | return true; |
| 5180 | } |
| 5181 | |
| 5182 | /* Directives-only scanning. Somewhat more relaxed than correct |
| 5183 | parsing -- some ill-formed programs will not be rejected. */ |
| 5184 | |
| 5185 | void |
| 5186 | cpp_directive_only_process (cpp_reader *pfile, |
| 5187 | void *data, |
| 5188 | void (*cb) (cpp_reader *, CPP_DO_task, void *, ...)) |
| 5189 | { |
| 5190 | bool module_p = CPP_OPTION (pfile, module_directives)((pfile)->opts.module_directives); |
| 5191 | |
| 5192 | do |
| 5193 | { |
| 5194 | restart: |
| 5195 | /* Buffer initialization, but no line cleaning. */ |
| 5196 | cpp_buffer *buffer = pfile->buffer; |
| 5197 | buffer->cur_note = buffer->notes_used = 0; |
| 5198 | buffer->cur = buffer->line_base = buffer->next_line; |
| 5199 | buffer->need_line = false; |
| 5200 | /* Files always end in a newline or carriage return. We rely on this for |
| 5201 | character peeking safety. */ |
| 5202 | gcc_assert (buffer->rlimit[0] == '\n' || buffer->rlimit[0] == '\r')((void)(!(buffer->rlimit[0] == '\n' || buffer->rlimit[0 ] == '\r') ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 5202, __FUNCTION__), 0 : 0)); |
| 5203 | |
| 5204 | const unsigned char *base = buffer->cur; |
| 5205 | unsigned line_count = 0; |
| 5206 | const unsigned char *line_start = base; |
| 5207 | |
| 5208 | bool bol = true; |
| 5209 | bool raw = false; |
| 5210 | |
| 5211 | const unsigned char *lwm = base; |
| 5212 | for (const unsigned char *pos = base, *limit = buffer->rlimit; |
| 5213 | pos < limit;) |
| 5214 | { |
| 5215 | unsigned char c = *pos++; |
| 5216 | /* This matches the switch in _cpp_lex_direct. */ |
| 5217 | switch (c) |
| 5218 | { |
| 5219 | case ' ': case '\t': case '\f': case '\v': |
| 5220 | /* Whitespace, do nothing. */ |
| 5221 | break; |
| 5222 | |
| 5223 | case '\r': /* MAC line ending, or Windows \r\n */ |
| 5224 | if (*pos == '\n') |
| 5225 | pos++; |
| 5226 | /* FALLTHROUGH */ |
| 5227 | |
| 5228 | case '\n': |
| 5229 | bol = true; |
| 5230 | |
| 5231 | next_line: |
| 5232 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 5233 | line_count++; |
| 5234 | line_start = pos; |
| 5235 | break; |
| 5236 | |
| 5237 | case '\\': |
| 5238 | /* <backslash><newline> is removed, and doesn't undo any |
| 5239 | preceeding escape or whatnot. */ |
| 5240 | if (*pos == '\n') |
| 5241 | { |
| 5242 | pos++; |
| 5243 | goto next_line; |
| 5244 | } |
| 5245 | else if (*pos == '\r') |
| 5246 | { |
| 5247 | if (pos[1] == '\n') |
| 5248 | pos++; |
| 5249 | pos++; |
| 5250 | goto next_line; |
| 5251 | } |
| 5252 | goto dflt; |
| 5253 | |
| 5254 | case '#': |
| 5255 | if (bol) |
| 5256 | { |
| 5257 | /* Line directive. */ |
| 5258 | if (pos - 1 > base && !pfile->state.skipping) |
| 5259 | cb (pfile, CPP_DO_print, data, |
| 5260 | line_count, base, pos - 1 - base); |
| 5261 | |
| 5262 | /* Prep things for directive handling. */ |
| 5263 | buffer->next_line = pos; |
| 5264 | buffer->need_line = true; |
| 5265 | bool ok = _cpp_get_fresh_line (pfile); |
| 5266 | gcc_checking_assert (ok)((void)(!(ok) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 5266, __FUNCTION__), 0 : 0)); |
| 5267 | |
| 5268 | /* Ensure proper column numbering for generated |
| 5269 | error messages. */ |
| 5270 | buffer->line_base -= pos - line_start; |
| 5271 | |
| 5272 | _cpp_handle_directive (pfile, line_start + 1 != pos); |
| 5273 | |
| 5274 | /* Sanitize the line settings. Duplicate #include's can |
| 5275 | mess things up. */ |
| 5276 | // FIXME: Necessary? |
| 5277 | pfile->line_table->highest_location |
| 5278 | = pfile->line_table->highest_line; |
| 5279 | |
| 5280 | if (!pfile->state.skipping |
| 5281 | && pfile->buffer->next_line < pfile->buffer->rlimit) |
| 5282 | cb (pfile, CPP_DO_location, data, |
| 5283 | pfile->line_table->highest_line); |
| 5284 | |
| 5285 | goto restart; |
| 5286 | } |
| 5287 | goto dflt; |
| 5288 | |
| 5289 | case '/': |
| 5290 | { |
| 5291 | const unsigned char *peek = do_peek_next (pos, limit); |
| 5292 | if (!(*peek == '/' || *peek == '*')) |
| 5293 | goto dflt; |
| 5294 | |
| 5295 | /* Line or block comment */ |
| 5296 | bool is_block = *peek == '*'; |
| 5297 | bool star = false; |
| 5298 | bool esc = false; |
| 5299 | location_t sloc |
| 5300 | = linemap_position_for_column (pfile->line_table, |
| 5301 | pos - line_start); |
| 5302 | |
| 5303 | while (pos < limit) |
| 5304 | { |
| 5305 | char c = *pos++; |
| 5306 | switch (c) |
| 5307 | { |
| 5308 | case '\\': |
| 5309 | esc = true; |
| 5310 | break; |
| 5311 | |
| 5312 | case '\r': |
| 5313 | if (*pos == '\n') |
| 5314 | pos++; |
| 5315 | /* FALLTHROUGH */ |
| 5316 | |
| 5317 | case '\n': |
| 5318 | { |
| 5319 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 5320 | line_count++; |
| 5321 | line_start = pos; |
| 5322 | if (!esc && !is_block) |
| 5323 | { |
| 5324 | bol = true; |
| 5325 | goto done_comment; |
| 5326 | } |
| 5327 | } |
| 5328 | if (!esc) |
| 5329 | star = false; |
| 5330 | esc = false; |
| 5331 | break; |
| 5332 | |
| 5333 | case '*': |
| 5334 | if (pos > peek) |
| 5335 | star = is_block; |
| 5336 | esc = false; |
| 5337 | break; |
| 5338 | |
| 5339 | case '/': |
| 5340 | if (star) |
| 5341 | goto done_comment; |
| 5342 | /* FALLTHROUGH */ |
| 5343 | |
| 5344 | default: |
| 5345 | star = false; |
| 5346 | esc = false; |
| 5347 | break; |
| 5348 | } |
| 5349 | } |
| 5350 | if (pos < limit || is_block) |
| 5351 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, |
| 5352 | "unterminated comment"); |
| 5353 | done_comment: |
| 5354 | lwm = pos; |
| 5355 | break; |
| 5356 | } |
| 5357 | |
| 5358 | case '\'': |
| 5359 | if (!CPP_OPTION (pfile, digit_separators)((pfile)->opts.digit_separators)) |
| 5360 | goto delimited_string; |
| 5361 | |
| 5362 | /* Possibly a number punctuator. */ |
| 5363 | if (!ISIDNUM (*do_peek_next (pos, limit))(_sch_istable[(*do_peek_next (pos, limit)) & 0xff] & ( unsigned short)(_sch_isidnum))) |
| 5364 | goto delimited_string; |
| 5365 | |
| 5366 | goto quote_peek; |
| 5367 | |
| 5368 | case '\"': |
| 5369 | if (!CPP_OPTION (pfile, rliterals)((pfile)->opts.rliterals)) |
| 5370 | goto delimited_string; |
| 5371 | |
| 5372 | quote_peek: |
| 5373 | { |
| 5374 | /* For ' see if it's a number punctuator |
| 5375 | \.?<digit>(<digit>|<identifier-nondigit> |
| 5376 | |'<digit>|'<nondigit>|[eEpP]<sign>|\.)* */ |
| 5377 | /* For " see if it's a raw string |
| 5378 | {U,L,u,u8}R. This includes CPP_NUMBER detection, |
| 5379 | because that could be 0e+R. */ |
| 5380 | const unsigned char *peek = pos - 1; |
| 5381 | bool quote_first = c == '"'; |
| 5382 | bool quote_eight = false; |
| 5383 | bool maybe_number_start = false; |
| 5384 | bool want_number = false; |
| 5385 | |
| 5386 | while ((peek = do_peek_prev (peek, lwm))) |
| 5387 | { |
| 5388 | unsigned char p = *peek; |
| 5389 | if (quote_first) |
| 5390 | { |
| 5391 | if (!raw) |
| 5392 | { |
| 5393 | if (p != 'R') |
| 5394 | break; |
| 5395 | raw = true; |
| 5396 | continue; |
| 5397 | } |
| 5398 | |
| 5399 | quote_first = false; |
| 5400 | if (p == 'L' || p == 'U' || p == 'u') |
| 5401 | ; |
| 5402 | else if (p == '8') |
| 5403 | quote_eight = true; |
| 5404 | else |
| 5405 | goto second_raw; |
| 5406 | } |
| 5407 | else if (quote_eight) |
| 5408 | { |
| 5409 | if (p != 'u') |
| 5410 | { |
| 5411 | raw = false; |
| 5412 | break; |
| 5413 | } |
| 5414 | quote_eight = false; |
| 5415 | } |
| 5416 | else if (c == '"') |
| 5417 | { |
| 5418 | second_raw:; |
| 5419 | if (!want_number && ISIDNUM (p)(_sch_istable[(p) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 5420 | { |
| 5421 | raw = false; |
| 5422 | break; |
| 5423 | } |
| 5424 | } |
| 5425 | |
| 5426 | if (ISDIGIT (p)(_sch_istable[(p) & 0xff] & (unsigned short)(_sch_isdigit ))) |
| 5427 | maybe_number_start = true; |
| 5428 | else if (p == '.') |
| 5429 | want_number = true; |
| 5430 | else if (ISIDNUM (p)(_sch_istable[(p) & 0xff] & (unsigned short)(_sch_isidnum ))) |
| 5431 | maybe_number_start = false; |
| 5432 | else if (p == '+' || p == '-') |
| 5433 | { |
| 5434 | if (const unsigned char *peek_prev |
| 5435 | = do_peek_prev (peek, lwm)) |
| 5436 | { |
| 5437 | p = *peek_prev; |
| 5438 | if (p == 'e' || p == 'E' |
| 5439 | || p == 'p' || p == 'P') |
| 5440 | { |
| 5441 | want_number = true; |
| 5442 | maybe_number_start = false; |
| 5443 | } |
| 5444 | else |
| 5445 | break; |
| 5446 | } |
| 5447 | else |
| 5448 | break; |
| 5449 | } |
| 5450 | else if (p == '\'' || p == '\"') |
| 5451 | { |
| 5452 | /* If this is lwm, this must be the end of a |
| 5453 | previous string. So this is a trailing |
| 5454 | literal type, (a) if those are allowed, |
| 5455 | and (b) maybe_start is false. Otherwise |
| 5456 | this must be a CPP_NUMBER because we've |
| 5457 | met another ', and we'd have checked that |
| 5458 | in its own right. */ |
| 5459 | if (peek == lwm && CPP_OPTION (pfile, uliterals)((pfile)->opts.uliterals)) |
| 5460 | { |
| 5461 | if (!maybe_number_start && !want_number) |
| 5462 | /* Must be a literal type. */ |
| 5463 | raw = false; |
| 5464 | } |
| 5465 | else if (p == '\'' |
| 5466 | && CPP_OPTION (pfile, digit_separators)((pfile)->opts.digit_separators)) |
| 5467 | maybe_number_start = true; |
| 5468 | break; |
| 5469 | } |
| 5470 | else if (c == '\'') |
| 5471 | break; |
| 5472 | else if (!quote_first && !quote_eight) |
| 5473 | break; |
| 5474 | } |
| 5475 | |
| 5476 | if (maybe_number_start) |
| 5477 | { |
| 5478 | if (c == '\'') |
| 5479 | /* A CPP NUMBER. */ |
| 5480 | goto dflt; |
| 5481 | raw = false; |
| 5482 | } |
| 5483 | |
| 5484 | goto delimited_string; |
| 5485 | } |
| 5486 | |
| 5487 | delimited_string: |
| 5488 | { |
| 5489 | /* (Possibly raw) string or char literal. */ |
| 5490 | unsigned char end = c; |
| 5491 | int delim_len = -1; |
| 5492 | const unsigned char *delim = NULL__null; |
| 5493 | location_t sloc = linemap_position_for_column (pfile->line_table, |
| 5494 | pos - line_start); |
| 5495 | int esc = 0; |
| 5496 | |
| 5497 | if (raw) |
| 5498 | { |
| 5499 | /* There can be no line breaks in the delimiter. */ |
| 5500 | delim = pos; |
| 5501 | for (delim_len = 0; (c = *pos++) != '('; delim_len++) |
| 5502 | { |
| 5503 | if (delim_len == 16) |
| 5504 | { |
| 5505 | cpp_error_with_line (pfile, CPP_DL_ERROR, |
| 5506 | sloc, 0, |
| 5507 | "raw string delimiter" |
| 5508 | " longer than %d" |
| 5509 | " characters", |
| 5510 | delim_len); |
| 5511 | raw = false; |
| 5512 | pos = delim; |
| 5513 | break; |
| 5514 | } |
| 5515 | if (strchr (") \\\t\v\f\n", c)) |
| 5516 | { |
| 5517 | cpp_error_with_line (pfile, CPP_DL_ERROR, |
| 5518 | sloc, 0, |
| 5519 | "invalid character '%c'" |
| 5520 | " in raw string" |
| 5521 | " delimiter", c); |
| 5522 | raw = false; |
| 5523 | pos = delim; |
| 5524 | break; |
| 5525 | } |
| 5526 | if (pos >= limit) |
| 5527 | goto bad_string; |
| 5528 | } |
| 5529 | } |
| 5530 | |
| 5531 | while (pos < limit) |
| 5532 | { |
| 5533 | char c = *pos++; |
| 5534 | switch (c) |
| 5535 | { |
| 5536 | case '\\': |
| 5537 | if (!raw) |
| 5538 | esc++; |
| 5539 | break; |
| 5540 | |
| 5541 | case '\r': |
| 5542 | if (*pos == '\n') |
| 5543 | pos++; |
| 5544 | /* FALLTHROUGH */ |
| 5545 | |
| 5546 | case '\n': |
| 5547 | { |
| 5548 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 5549 | line_count++; |
| 5550 | line_start = pos; |
| 5551 | } |
| 5552 | if (esc) |
| 5553 | esc--; |
| 5554 | break; |
| 5555 | |
| 5556 | case ')': |
| 5557 | if (raw |
| 5558 | && pos + delim_len + 1 < limit |
| 5559 | && pos[delim_len] == end |
| 5560 | && !memcmp (delim, pos, delim_len)) |
| 5561 | { |
| 5562 | pos += delim_len + 1; |
| 5563 | raw = false; |
Value stored to 'raw' is never read | |
| 5564 | goto done_string; |
| 5565 | } |
| 5566 | break; |
| 5567 | |
| 5568 | default: |
| 5569 | if (!raw && !(esc & 1) && c == end) |
| 5570 | goto done_string; |
| 5571 | esc = 0; |
| 5572 | break; |
| 5573 | } |
| 5574 | } |
| 5575 | bad_string: |
| 5576 | cpp_error_with_line (pfile, CPP_DL_ERROR, sloc, 0, |
| 5577 | "unterminated literal"); |
| 5578 | |
| 5579 | done_string: |
| 5580 | raw = false; |
| 5581 | lwm = pos - 1; |
| 5582 | } |
| 5583 | goto dflt; |
| 5584 | |
| 5585 | case '_': |
| 5586 | case 'e': |
| 5587 | case 'i': |
| 5588 | case 'm': |
| 5589 | if (bol && module_p && !pfile->state.skipping |
| 5590 | && do_peek_module (pfile, c, pos, limit)) |
| 5591 | { |
| 5592 | /* We've seen the start of a module control line. |
| 5593 | Start up the tokenizer. */ |
| 5594 | pos--; /* Backup over the first character. */ |
| 5595 | |
| 5596 | /* Backup over whitespace to start of line. */ |
| 5597 | while (pos > line_start |
| 5598 | && (pos[-1] == ' ' || pos[-1] == '\t')) |
| 5599 | pos--; |
| 5600 | |
| 5601 | if (pos > base) |
| 5602 | cb (pfile, CPP_DO_print, data, line_count, base, pos - base); |
| 5603 | |
| 5604 | /* Prep things for directive handling. */ |
| 5605 | buffer->next_line = pos; |
| 5606 | buffer->need_line = true; |
| 5607 | |
| 5608 | /* Now get tokens until the PRAGMA_EOL. */ |
| 5609 | do |
| 5610 | { |
| 5611 | location_t spelling; |
| 5612 | const cpp_token *tok |
| 5613 | = cpp_get_token_with_location (pfile, &spelling); |
| 5614 | |
| 5615 | gcc_assert (pfile->state.in_deferred_pragma((void)(!(pfile->state.in_deferred_pragma || tok->type == CPP_PRAGMA_EOL) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 5616, __FUNCTION__), 0 : 0)) |
| 5616 | || tok->type == CPP_PRAGMA_EOL)((void)(!(pfile->state.in_deferred_pragma || tok->type == CPP_PRAGMA_EOL) ? fancy_abort ("/buildworker/marxinbox-gcc-clang-static-analyzer/build/libcpp/lex.cc" , 5616, __FUNCTION__), 0 : 0)); |
| 5617 | cb (pfile, CPP_DO_token, data, tok, spelling); |
| 5618 | } |
| 5619 | while (pfile->state.in_deferred_pragma); |
| 5620 | |
| 5621 | if (pfile->buffer->next_line < pfile->buffer->rlimit) |
| 5622 | cb (pfile, CPP_DO_location, data, |
| 5623 | pfile->line_table->highest_line); |
| 5624 | |
| 5625 | pfile->mi_valid = false; |
| 5626 | goto restart; |
| 5627 | } |
| 5628 | goto dflt; |
| 5629 | |
| 5630 | default: |
| 5631 | dflt: |
| 5632 | bol = false; |
| 5633 | pfile->mi_valid = false; |
| 5634 | break; |
| 5635 | } |
| 5636 | } |
| 5637 | |
| 5638 | if (buffer->rlimit > base && !pfile->state.skipping) |
| 5639 | { |
| 5640 | const unsigned char *limit = buffer->rlimit; |
| 5641 | /* If the file was not newline terminated, add rlimit, which is |
| 5642 | guaranteed to point to a newline, to the end of our range. */ |
| 5643 | if (limit[-1] != '\n') |
| 5644 | { |
| 5645 | limit++; |
| 5646 | CPP_INCREMENT_LINE (pfile, 0)do { const class line_maps *line_table = pfile->line_table ; const struct line_map_ordinary *map = LINEMAPS_LAST_ORDINARY_MAP (line_table); linenum_type line = SOURCE_LINE (map, line_table ->highest_line); linemap_line_start (pfile->line_table, line + 1, 0); } while (0); |
| 5647 | line_count++; |
| 5648 | } |
| 5649 | cb (pfile, CPP_DO_print, data, line_count, base, limit - base); |
| 5650 | } |
| 5651 | |
| 5652 | _cpp_pop_buffer (pfile); |
| 5653 | } |
| 5654 | while (pfile->buffer); |
| 5655 | } |