3 * Copyright © 2010 Intel Corporation
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
30 #include "glcpp-parse.h"
32 /* Flex annoyingly generates some functions without making them
33 * static. Let's declare them here. */
34 int glcpp_get_column (yyscan_t yyscanner);
35 void glcpp_set_column (int column_no , yyscan_t yyscanner);
38 #define YY_NO_UNISTD_H
43 #define YY_USER_ACTION \
45 if (parser->has_new_line_number) \
46 yylineno = parser->new_line_number; \
47 if (parser->has_new_source_number) \
48 yylloc->source = parser->new_source_number; \
49 yylloc->first_column = yycolumn + 1; \
50 yylloc->first_line = yylloc->last_line = yylineno; \
52 yylloc->last_column = yycolumn + 1; \
53 parser->has_new_line_number = 0; \
54 parser->has_new_source_number = 0; \
57 #define YY_USER_INIT \
64 /* It's ugly to have macros that have return statements inside of
65 * them, but flex-based lexer generation is all built around the
68 * To mitigate the ugliness, we defer as much of the logic as possible
69 * to an actual function, not a macro (see
70 * glcpplex_update_state_per_token) and we make the word RETURN
71 * prominent in all of the macros which may return.
73 * The most-commonly-used macro is RETURN_TOKEN which will perform all
74 * necessary state updates based on the provided token,, then
75 * conditionally return the token. It will not return a token if the
76 * parser is currently skipping tokens, (such as within #if
79 * The RETURN_TOKEN_NEVER_SKIP macro is a lower-level variant that
80 * makes the token returning unconditional. This is needed for things
81 * like #if and the tokens of its condition, (since these must be
82 * evaluated by the parser even when otherwise skipping).
84 * Finally, RETURN_STRING_TOKEN is a simple convenience wrapper on top
85 * of RETURN_TOKEN that performs a string copy of yytext before the
88 #define RETURN_TOKEN_NEVER_SKIP(token) \
90 if (glcpp_lex_update_state_per_token (parser, token)) \
94 #define RETURN_TOKEN(token) \
96 if (! parser->skipping) { \
97 RETURN_TOKEN_NEVER_SKIP(token); \
101 #define RETURN_STRING_TOKEN(token) \
103 if (! parser->skipping) { \
104 yylval->str = ralloc_strdup (yyextra, yytext); \
105 RETURN_TOKEN_NEVER_SKIP (token); \
110 /* Update all state necessary for each token being returned.
112 * Here we'll be tracking newlines and spaces so that the lexer can
113 * alter its behavior as necessary, (for example, '#' has special
114 * significance if it is the first non-whitespace, non-comment token
115 * in a line, but does not otherwise).
117 * NOTE: If this function returns FALSE, then no token should be
118 * returned at all. This is used to suprress duplicate SPACE tokens.
121 glcpp_lex_update_state_per_token (glcpp_parser_t *parser, int token)
123 /* After the first non-space token in a line, we won't
124 * allow any '#' to introduce a directive. */
125 if (token == NEWLINE) {
126 parser->first_non_space_token_this_line = 1;
127 } else if (token != SPACE) {
128 parser->first_non_space_token_this_line = 0;
131 /* Track newlines just to know whether a newline needs
132 * to be inserted if end-of-file comes early. */
133 if (token == NEWLINE) {
134 parser->last_token_was_newline = 1;
136 parser->last_token_was_newline = 0;
139 /* Track spaces to avoid emitting multiple SPACE
140 * tokens in a row. */
141 if (token == SPACE) {
142 if (! parser->last_token_was_space) {
143 parser->last_token_was_space = 1;
146 parser->last_token_was_space = 1;
150 parser->last_token_was_space = 0;
158 %option bison-bridge bison-locations reentrant noyywrap
159 %option extra-type="glcpp_parser_t *"
160 %option prefix="glcpp_"
162 %option never-interactive
163 %option warn nodefault
165 /* Note: When adding any start conditions to this list, you must also
166 * update the "Internal compiler error" catch-all rule near the end of
169 %x COMMENT DEFINE DONE HASH NEWLINE_CATCHUP UNREACHABLE
172 NONSPACE [^[:space:]]
175 NEWLINE (\r\n|\n\r|\r|\n)
176 IDENTIFIER [_a-zA-Z][_a-zA-Z0-9]*
177 PP_NUMBER [.]?[0-9]([._a-zA-Z0-9]|[eEpP][-+])*
178 PUNCTUATION [][(){}.&*~!/%<>^|;,=+-]
180 /* The OTHER class is simply a catch-all for things that the CPP
181 parser just doesn't care about. Since flex regular expressions that
182 match longer strings take priority over those matching shorter
183 strings, we have to be careful to avoid OTHER matching and hiding
184 something that CPP does care about. So we simply exclude all
185 characters that appear in any other expressions. */
187 OTHER [^][_#[:space:]#a-zA-Z0-9(){}.&*~!/%<>^|;,=+-]
190 DECIMAL_INTEGER [1-9][0-9]*[uU]?
191 OCTAL_INTEGER 0[0-7]*[uU]?
192 HEXADECIMAL_INTEGER 0[xX][0-9a-fA-F]+[uU]?
196 glcpp_parser_t *parser = yyextra;
198 /* When we lex a multi-line comment, we replace it (as
199 * specified) with a single space. But if the comment spanned
200 * multiple lines, then subsequent parsing stages will not
201 * count correct line numbers. To avoid this problem we keep
202 * track of all newlines that were commented out by a
203 * multi-line comment, and we emit a NEWLINE token for each at
204 * the next legal opportunity, (which is when the lexer would
205 * be emitting a NEWLINE token anyway).
207 if (YY_START == NEWLINE_CATCHUP) {
208 if (parser->commented_newlines)
209 parser->commented_newlines--;
210 if (parser->commented_newlines == 0)
212 RETURN_TOKEN_NEVER_SKIP (NEWLINE);
215 /* Set up the parser->skipping bit here before doing any lexing.
217 * This bit controls whether tokens are skipped, (as implemented by
218 * RETURN_TOKEN), such as between "#if 0" and "#endif".
220 * The parser maintains a skip_stack indicating whether we should be
221 * skipping, (and nested levels of #if/#ifdef/#ifndef/#endif) will
222 * push and pop items from the stack.
224 * Here are the rules for determining whether we are skipping:
226 * 1. If the skip stack is NULL, we are outside of all #if blocks
227 * and we are not skipping.
229 * 2. If the skip stack is non-NULL, the type of the top node in
230 * the stack determines whether to skip. A type of
231 * SKIP_NO_SKIP is used for blocks wheere we are emitting
232 * tokens, (such as between #if 1 and #endif, or after the
233 * #else of an #if 0, etc.).
235 * 3. The lexing_directive bit overrides the skip stack. This bit
236 * is set when we are actively lexing the expression for a
237 * pre-processor condition, (such as #if, #elif, or #else). In
238 * this case, even if otherwise skipping, we need to emit the
239 * tokens for this condition so that the parser can evaluate
240 * the expression. (For, #else, there's no expression, but we
241 * emit tokens so the parser can generate a nice error message
242 * if there are any tokens here).
244 if (parser->skip_stack &&
245 parser->skip_stack->type != SKIP_NO_SKIP &&
246 ! parser->lexing_directive)
248 parser->skipping = 1;
250 parser->skipping = 0;
253 /* Single-line comments */
254 <INITIAL,DEFINE,HASH>"//"[^\r\n]* {
257 /* Multi-line comments */
258 <INITIAL,DEFINE,HASH>"/*" { yy_push_state(COMMENT, yyscanner); }
260 <COMMENT>[^*\r\n]*{NEWLINE} { yylineno++; yycolumn = 0; parser->commented_newlines++; }
261 <COMMENT>"*"+[^*/\r\n]*
262 <COMMENT>"*"+[^*/\r\n]*{NEWLINE} { yylineno++; yycolumn = 0; parser->commented_newlines++; }
264 yy_pop_state(yyscanner);
265 /* In the <HASH> start condition, we don't want any SPACE token. */
266 if (yyextra->space_tokens && YY_START != HASH)
267 RETURN_TOKEN (SPACE);
272 /* If the '#' is the first non-whitespace, non-comment token on this
273 * line, then it introduces a directive, switch to the <HASH> start
276 * Otherwise, this is just punctuation, so return the HASH_TOKEN
278 if (parser->first_non_space_token_this_line) {
282 RETURN_TOKEN_NEVER_SKIP (HASH_TOKEN);
285 <HASH>version{HSPACE}+ {
287 yyextra->space_tokens = 0;
288 RETURN_STRING_TOKEN (VERSION_TOKEN);
291 /* Swallow empty #pragma directives, (to avoid confusing the
292 * downstream compiler).
294 * Note: We use a simple regular expression for the lookahead
295 * here. Specifically, we cannot use the complete {NEWLINE} expression
296 * since it uses alternation and we've found that there's a flex bug
297 * where using alternation in the lookahead portion of a pattern
298 * triggers a buffer overrun. */
299 <HASH>pragma{HSPACE}*/[\r\n] {
303 /* glcpp doesn't handle #extension, #version, or #pragma directives.
304 * Simply pass them through to the main compiler's lexer/parser. */
305 <HASH>(extension|pragma)[^\r\n]* {
307 RETURN_STRING_TOKEN (PRAGMA);
310 <HASH>line{HSPACE}+ {
317 RETURN_TOKEN_NEVER_SKIP (NEWLINE);
320 /* For the pre-processor directives, we return these tokens
321 * even when we are otherwise skipping. */
324 yyextra->lexing_directive = 1;
325 yyextra->space_tokens = 0;
326 RETURN_TOKEN_NEVER_SKIP (IFDEF);
331 yyextra->lexing_directive = 1;
332 yyextra->space_tokens = 0;
333 RETURN_TOKEN_NEVER_SKIP (IFNDEF);
336 <HASH>if/[^_a-zA-Z0-9] {
338 yyextra->lexing_directive = 1;
339 yyextra->space_tokens = 0;
340 RETURN_TOKEN_NEVER_SKIP (IF);
343 <HASH>elif/[^_a-zA-Z0-9] {
345 yyextra->lexing_directive = 1;
346 yyextra->space_tokens = 0;
347 RETURN_TOKEN_NEVER_SKIP (ELIF);
352 yyextra->space_tokens = 0;
353 RETURN_TOKEN_NEVER_SKIP (ELSE);
358 yyextra->space_tokens = 0;
359 RETURN_TOKEN_NEVER_SKIP (ENDIF);
362 <HASH>error[^\r\n]* {
364 RETURN_STRING_TOKEN (ERROR_TOKEN);
367 /* After we see a "#define" we enter the <DEFINE> start state
368 * for the lexer. Within <DEFINE> we are looking for the first
369 * identifier and specifically checking whether the identifier
370 * is followed by a '(' or not, (to lex either a
371 * FUNC_IDENTIFIER or an OBJ_IDENITIFIER token).
373 * While in the <DEFINE> state we also need to explicitly
374 * handle a few other things that may appear before the
377 * * Comments, (handled above with the main support for
380 * * Whitespace (simply ignored)
382 * * Anything else, (not an identifier, not a comment,
383 * and not whitespace). This will generate an error.
385 <HASH>define{HSPACE}* {
386 if (! parser->skipping) {
388 yyextra->space_tokens = 0;
389 RETURN_TOKEN (DEFINE_TOKEN);
395 yyextra->space_tokens = 0;
396 RETURN_TOKEN (UNDEF);
400 /* Nothing to do here. Importantly, don't leave the <HASH>
401 * start condition, since it's legal to have space between the
402 * '#' and the directive.. */
405 /* This will catch any non-directive garbage after a HASH */
408 RETURN_TOKEN (GARBAGE);
411 /* An identifier immediately followed by '(' */
412 <DEFINE>{IDENTIFIER}/"(" {
414 RETURN_STRING_TOKEN (FUNC_IDENTIFIER);
417 /* An identifier not immediately followed by '(' */
418 <DEFINE>{IDENTIFIER} {
420 RETURN_STRING_TOKEN (OBJ_IDENTIFIER);
425 /* Just ignore it. Nothing to do here. */
428 /* '/' not followed by '*', so not a comment. This is an error. */
429 <DEFINE>[/][^*]{NONSPACE}* {
431 glcpp_error(yylloc, yyextra, "#define followed by a non-identifier: %s", yytext);
432 RETURN_STRING_TOKEN (INTEGER_STRING);
435 /* A character that can't start an identifier, comment, or
436 * space. This is an error. */
437 <DEFINE>[^_a-zA-Z/[:space:]]{NONSPACE}* {
439 glcpp_error(yylloc, yyextra, "#define followed by a non-identifier: %s", yytext);
440 RETURN_STRING_TOKEN (INTEGER_STRING);
444 RETURN_STRING_TOKEN (INTEGER_STRING);
448 RETURN_STRING_TOKEN (INTEGER_STRING);
451 {HEXADECIMAL_INTEGER} {
452 RETURN_STRING_TOKEN (INTEGER_STRING);
456 RETURN_TOKEN (LEFT_SHIFT);
460 RETURN_TOKEN (RIGHT_SHIFT);
464 RETURN_TOKEN (LESS_OR_EQUAL);
468 RETURN_TOKEN (GREATER_OR_EQUAL);
472 RETURN_TOKEN (EQUAL);
476 RETURN_TOKEN (NOT_EQUAL);
488 RETURN_TOKEN (PLUS_PLUS);
492 RETURN_TOKEN (MINUS_MINUS);
496 if (! parser->skipping) {
498 glcpp_error(yylloc, yyextra, "Token pasting (##) is illegal in GLES");
499 RETURN_TOKEN (PASTE);
504 RETURN_TOKEN (DEFINED);
508 RETURN_STRING_TOKEN (IDENTIFIER);
512 RETURN_STRING_TOKEN (OTHER);
516 RETURN_TOKEN (yytext[0]);
520 RETURN_STRING_TOKEN (OTHER);
524 if (yyextra->space_tokens) {
525 RETURN_TOKEN (SPACE);
529 /* We preserve all newlines, even between #if 0..#endif, so no
532 if (parser->commented_newlines) {
533 BEGIN NEWLINE_CATCHUP;
537 yyextra->space_tokens = 1;
538 yyextra->lexing_directive = 0;
541 RETURN_TOKEN_NEVER_SKIP (NEWLINE);
544 <INITIAL,COMMENT,DEFINE,HASH><<EOF>> {
545 if (YY_START == COMMENT)
546 glcpp_error(yylloc, yyextra, "Unterminated comment");
547 BEGIN DONE; /* Don't keep matching this rule forever. */
548 yyextra->lexing_directive = 0;
549 if (! parser->last_token_was_newline)
550 RETURN_TOKEN (NEWLINE);
553 /* This is a catch-all to avoid the annoying default flex action which
554 * matches any character and prints it. If any input ever matches this
555 * rule, then we have made a mistake above and need to fix one or more
556 * of the preceding patterns to match that input. */
559 glcpp_error(yylloc, yyextra, "Internal compiler error: Unexpected character: %s", yytext);
561 /* We don't actually use the UNREACHABLE start condition. We
562 only have this block here so that we can pretend to call some
563 generated functions, (to avoid "defined but not used"
565 if (YY_START == UNREACHABLE) {
567 yy_top_state(yyextra);
574 glcpp_lex_set_source_string(glcpp_parser_t *parser, const char *shader)
576 yy_scan_string(shader, parser->scanner);