Commit | Line | Data |
---|---|---|
252b5132 | 1 | /* This is the Assembler Pre-Processor |
f7e42eb4 | 2 | Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, |
b1ac4c66 | 3 | 1999, 2000, 2002, 2003 |
252b5132 RH |
4 | Free Software Foundation, Inc. |
5 | ||
6 | This file is part of GAS, the GNU Assembler. | |
7 | ||
8 | GAS is free software; you can redistribute it and/or modify | |
9 | it under the terms of the GNU General Public License as published by | |
10 | the Free Software Foundation; either version 2, or (at your option) | |
11 | any later version. | |
12 | ||
13 | GAS 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 GAS; see the file COPYING. If not, write to the Free | |
20 | Software Foundation, 59 Temple Place - Suite 330, Boston, MA | |
21 | 02111-1307, USA. */ | |
22 | ||
204cd129 | 23 | /* Modified by Allen Wirfs-Brock, Instantiations Inc 2/90. */ |
252b5132 RH |
24 | /* App, the assembler pre-processor. This pre-processor strips out excess |
25 | spaces, turns single-quoted characters into a decimal constant, and turns | |
26 | # <number> <filename> <garbage> into a .line <number>\n.file <filename> | |
27 | pair. This needs better error-handling. */ | |
28 | ||
29 | #include <stdio.h> | |
8d9cd6b1 | 30 | #include "as.h" /* For BAD_CASE() only. */ |
252b5132 RH |
31 | |
32 | #if (__STDC__ != 1) | |
33 | #ifndef const | |
34 | #define const /* empty */ | |
35 | #endif | |
36 | #endif | |
37 | ||
abd63a32 | 38 | #ifdef TC_M68K |
252b5132 RH |
39 | /* Whether we are scrubbing in m68k MRI mode. This is different from |
40 | flag_m68k_mri, because the two flags will be affected by the .mri | |
41 | pseudo-op at different times. */ | |
42 | static int scrub_m68k_mri; | |
43 | ||
44 | /* The pseudo-op which switches in and out of MRI mode. See the | |
45 | comment in do_scrub_chars. */ | |
46 | static const char mri_pseudo[] = ".mri 0"; | |
72297628 AM |
47 | #else |
48 | #define scrub_m68k_mri 0 | |
49 | #endif | |
252b5132 RH |
50 | |
51 | #if defined TC_ARM && defined OBJ_ELF | |
3ee4defc | 52 | /* The pseudo-op for which we need to special-case `@' characters. |
252b5132 RH |
53 | See the comment in do_scrub_chars. */ |
54 | static const char symver_pseudo[] = ".symver"; | |
55 | static const char * symver_state; | |
56 | #endif | |
57 | ||
58 | static char lex[256]; | |
59 | static const char symbol_chars[] = | |
60 | "$._ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; | |
61 | ||
62 | #define LEX_IS_SYMBOL_COMPONENT 1 | |
63 | #define LEX_IS_WHITESPACE 2 | |
64 | #define LEX_IS_LINE_SEPARATOR 3 | |
65 | #define LEX_IS_COMMENT_START 4 | |
66 | #define LEX_IS_LINE_COMMENT_START 5 | |
67 | #define LEX_IS_TWOCHAR_COMMENT_1ST 6 | |
68 | #define LEX_IS_STRINGQUOTE 8 | |
69 | #define LEX_IS_COLON 9 | |
70 | #define LEX_IS_NEWLINE 10 | |
71 | #define LEX_IS_ONECHAR_QUOTE 11 | |
72 | #ifdef TC_V850 | |
73 | #define LEX_IS_DOUBLEDASH_1ST 12 | |
74 | #endif | |
75 | #ifdef TC_M32R | |
f28e8eb3 TW |
76 | #define DOUBLEBAR_PARALLEL |
77 | #endif | |
78 | #ifdef DOUBLEBAR_PARALLEL | |
252b5132 RH |
79 | #define LEX_IS_DOUBLEBAR_1ST 13 |
80 | #endif | |
62f65a7b | 81 | #define LEX_IS_PARALLEL_SEPARATOR 14 |
252b5132 RH |
82 | #define IS_SYMBOL_COMPONENT(c) (lex[c] == LEX_IS_SYMBOL_COMPONENT) |
83 | #define IS_WHITESPACE(c) (lex[c] == LEX_IS_WHITESPACE) | |
84 | #define IS_LINE_SEPARATOR(c) (lex[c] == LEX_IS_LINE_SEPARATOR) | |
62f65a7b | 85 | #define IS_PARALLEL_SEPARATOR(c) (lex[c] == LEX_IS_PARALLEL_SEPARATOR) |
252b5132 RH |
86 | #define IS_COMMENT(c) (lex[c] == LEX_IS_COMMENT_START) |
87 | #define IS_LINE_COMMENT(c) (lex[c] == LEX_IS_LINE_COMMENT_START) | |
88 | #define IS_NEWLINE(c) (lex[c] == LEX_IS_NEWLINE) | |
89 | ||
73ee5e4c | 90 | static int process_escape (int); |
252b5132 RH |
91 | |
92 | /* FIXME-soon: The entire lexer/parser thingy should be | |
93 | built statically at compile time rather than dynamically | |
3ee4defc | 94 | each and every time the assembler is run. xoxorich. */ |
252b5132 | 95 | |
3ee4defc | 96 | void |
73ee5e4c | 97 | do_scrub_begin (int m68k_mri ATTRIBUTE_UNUSED) |
252b5132 RH |
98 | { |
99 | const char *p; | |
100 | int c; | |
101 | ||
252b5132 RH |
102 | lex[' '] = LEX_IS_WHITESPACE; |
103 | lex['\t'] = LEX_IS_WHITESPACE; | |
104 | lex['\r'] = LEX_IS_WHITESPACE; | |
105 | lex['\n'] = LEX_IS_NEWLINE; | |
252b5132 RH |
106 | lex[':'] = LEX_IS_COLON; |
107 | ||
abd63a32 AM |
108 | #ifdef TC_M68K |
109 | scrub_m68k_mri = m68k_mri; | |
110 | ||
252b5132 | 111 | if (! m68k_mri) |
abd63a32 | 112 | #endif |
252b5132 RH |
113 | { |
114 | lex['"'] = LEX_IS_STRINGQUOTE; | |
115 | ||
5b93d8bb | 116 | #if ! defined (TC_HPPA) && ! defined (TC_I370) |
204cd129 | 117 | /* I370 uses single-quotes to delimit integer, float constants. */ |
252b5132 RH |
118 | lex['\''] = LEX_IS_ONECHAR_QUOTE; |
119 | #endif | |
120 | ||
121 | #ifdef SINGLE_QUOTE_STRINGS | |
122 | lex['\''] = LEX_IS_STRINGQUOTE; | |
123 | #endif | |
124 | } | |
125 | ||
126 | /* Note: if any other character can be LEX_IS_STRINGQUOTE, the loop | |
127 | in state 5 of do_scrub_chars must be changed. */ | |
128 | ||
129 | /* Note that these override the previous defaults, e.g. if ';' is a | |
130 | comment char, then it isn't a line separator. */ | |
131 | for (p = symbol_chars; *p; ++p) | |
204cd129 | 132 | lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; |
252b5132 RH |
133 | |
134 | for (c = 128; c < 256; ++c) | |
135 | lex[c] = LEX_IS_SYMBOL_COMPONENT; | |
136 | ||
137 | #ifdef tc_symbol_chars | |
138 | /* This macro permits the processor to specify all characters which | |
139 | may appears in an operand. This will prevent the scrubber from | |
140 | discarding meaningful whitespace in certain cases. The i386 | |
141 | backend uses this to support prefixes, which can confuse the | |
142 | scrubber as to whether it is parsing operands or opcodes. */ | |
143 | for (p = tc_symbol_chars; *p; ++p) | |
144 | lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; | |
145 | #endif | |
146 | ||
147 | /* The m68k backend wants to be able to change comment_chars. */ | |
148 | #ifndef tc_comment_chars | |
149 | #define tc_comment_chars comment_chars | |
150 | #endif | |
151 | for (p = tc_comment_chars; *p; p++) | |
204cd129 | 152 | lex[(unsigned char) *p] = LEX_IS_COMMENT_START; |
252b5132 RH |
153 | |
154 | for (p = line_comment_chars; *p; p++) | |
204cd129 | 155 | lex[(unsigned char) *p] = LEX_IS_LINE_COMMENT_START; |
252b5132 RH |
156 | |
157 | for (p = line_separator_chars; *p; p++) | |
204cd129 | 158 | lex[(unsigned char) *p] = LEX_IS_LINE_SEPARATOR; |
252b5132 | 159 | |
62f65a7b DB |
160 | #ifdef tc_parallel_separator_chars |
161 | /* This macro permits the processor to specify all characters which | |
162 | separate parallel insns on the same line. */ | |
163 | for (p = tc_parallel_separator_chars; *p; p++) | |
204cd129 | 164 | lex[(unsigned char) *p] = LEX_IS_PARALLEL_SEPARATOR; |
62f65a7b DB |
165 | #endif |
166 | ||
252b5132 RH |
167 | /* Only allow slash-star comments if slash is not in use. |
168 | FIXME: This isn't right. We should always permit them. */ | |
169 | if (lex['/'] == 0) | |
204cd129 | 170 | lex['/'] = LEX_IS_TWOCHAR_COMMENT_1ST; |
252b5132 | 171 | |
abd63a32 | 172 | #ifdef TC_M68K |
252b5132 RH |
173 | if (m68k_mri) |
174 | { | |
175 | lex['\''] = LEX_IS_STRINGQUOTE; | |
176 | lex[';'] = LEX_IS_COMMENT_START; | |
177 | lex['*'] = LEX_IS_LINE_COMMENT_START; | |
178 | /* The MRI documentation says '!' is LEX_IS_COMMENT_START, but | |
b1ac4c66 | 179 | then it can't be used in an expression. */ |
252b5132 RH |
180 | lex['!'] = LEX_IS_LINE_COMMENT_START; |
181 | } | |
abd63a32 | 182 | #endif |
252b5132 RH |
183 | |
184 | #ifdef TC_V850 | |
185 | lex['-'] = LEX_IS_DOUBLEDASH_1ST; | |
186 | #endif | |
f28e8eb3 | 187 | #ifdef DOUBLEBAR_PARALLEL |
252b5132 RH |
188 | lex['|'] = LEX_IS_DOUBLEBAR_1ST; |
189 | #endif | |
190 | #ifdef TC_D30V | |
204cd129 | 191 | /* Must do this is we want VLIW instruction with "->" or "<-". */ |
252b5132 RH |
192 | lex['-'] = LEX_IS_SYMBOL_COMPONENT; |
193 | #endif | |
204cd129 | 194 | } |
252b5132 | 195 | |
204cd129 | 196 | /* Saved state of the scrubber. */ |
252b5132 RH |
197 | static int state; |
198 | static int old_state; | |
199 | static char *out_string; | |
200 | static char out_buf[20]; | |
201 | static int add_newlines; | |
202 | static char *saved_input; | |
203 | static int saved_input_len; | |
2b47531b | 204 | static char input_buffer[32 * 1024]; |
252b5132 RH |
205 | static const char *mri_state; |
206 | static char mri_last_ch; | |
207 | ||
208 | /* Data structure for saving the state of app across #include's. Note that | |
209 | app is called asynchronously to the parsing of the .include's, so our | |
210 | state at the time .include is interpreted is completely unrelated. | |
211 | That's why we have to save it all. */ | |
212 | ||
204cd129 NC |
213 | struct app_save |
214 | { | |
30a2b4ef KH |
215 | int state; |
216 | int old_state; | |
217 | char * out_string; | |
218 | char out_buf[sizeof (out_buf)]; | |
219 | int add_newlines; | |
220 | char * saved_input; | |
221 | int saved_input_len; | |
abd63a32 | 222 | #ifdef TC_M68K |
30a2b4ef | 223 | int scrub_m68k_mri; |
abd63a32 | 224 | #endif |
30a2b4ef KH |
225 | const char * mri_state; |
226 | char mri_last_ch; | |
252b5132 | 227 | #if defined TC_ARM && defined OBJ_ELF |
30a2b4ef | 228 | const char * symver_state; |
252b5132 | 229 | #endif |
30a2b4ef | 230 | }; |
252b5132 RH |
231 | |
232 | char * | |
73ee5e4c | 233 | app_push (void) |
252b5132 RH |
234 | { |
235 | register struct app_save *saved; | |
236 | ||
237 | saved = (struct app_save *) xmalloc (sizeof (*saved)); | |
238 | saved->state = state; | |
239 | saved->old_state = old_state; | |
240 | saved->out_string = out_string; | |
241 | memcpy (saved->out_buf, out_buf, sizeof (out_buf)); | |
242 | saved->add_newlines = add_newlines; | |
2b47531b ILT |
243 | if (saved_input == NULL) |
244 | saved->saved_input = NULL; | |
245 | else | |
246 | { | |
247 | saved->saved_input = xmalloc (saved_input_len); | |
248 | memcpy (saved->saved_input, saved_input, saved_input_len); | |
249 | saved->saved_input_len = saved_input_len; | |
250 | } | |
abd63a32 | 251 | #ifdef TC_M68K |
252b5132 | 252 | saved->scrub_m68k_mri = scrub_m68k_mri; |
abd63a32 | 253 | #endif |
252b5132 RH |
254 | saved->mri_state = mri_state; |
255 | saved->mri_last_ch = mri_last_ch; | |
256 | #if defined TC_ARM && defined OBJ_ELF | |
257 | saved->symver_state = symver_state; | |
258 | #endif | |
259 | ||
3ee4defc | 260 | /* do_scrub_begin() is not useful, just wastes time. */ |
252b5132 RH |
261 | |
262 | state = 0; | |
263 | saved_input = NULL; | |
264 | ||
265 | return (char *) saved; | |
266 | } | |
267 | ||
3ee4defc | 268 | void |
73ee5e4c | 269 | app_pop (char *arg) |
252b5132 RH |
270 | { |
271 | register struct app_save *saved = (struct app_save *) arg; | |
272 | ||
3ee4defc | 273 | /* There is no do_scrub_end (). */ |
252b5132 RH |
274 | state = saved->state; |
275 | old_state = saved->old_state; | |
276 | out_string = saved->out_string; | |
277 | memcpy (out_buf, saved->out_buf, sizeof (out_buf)); | |
278 | add_newlines = saved->add_newlines; | |
2b47531b ILT |
279 | if (saved->saved_input == NULL) |
280 | saved_input = NULL; | |
281 | else | |
282 | { | |
ab9da554 | 283 | assert (saved->saved_input_len <= (int) (sizeof input_buffer)); |
2b47531b ILT |
284 | memcpy (input_buffer, saved->saved_input, saved->saved_input_len); |
285 | saved_input = input_buffer; | |
286 | saved_input_len = saved->saved_input_len; | |
287 | free (saved->saved_input); | |
288 | } | |
abd63a32 | 289 | #ifdef TC_M68K |
252b5132 | 290 | scrub_m68k_mri = saved->scrub_m68k_mri; |
abd63a32 | 291 | #endif |
252b5132 RH |
292 | mri_state = saved->mri_state; |
293 | mri_last_ch = saved->mri_last_ch; | |
294 | #if defined TC_ARM && defined OBJ_ELF | |
295 | symver_state = saved->symver_state; | |
296 | #endif | |
297 | ||
298 | free (arg); | |
204cd129 | 299 | } |
252b5132 RH |
300 | |
301 | /* @@ This assumes that \n &c are the same on host and target. This is not | |
302 | necessarily true. */ | |
204cd129 | 303 | |
3ee4defc | 304 | static int |
73ee5e4c | 305 | process_escape (int ch) |
252b5132 RH |
306 | { |
307 | switch (ch) | |
308 | { | |
309 | case 'b': | |
310 | return '\b'; | |
311 | case 'f': | |
312 | return '\f'; | |
313 | case 'n': | |
314 | return '\n'; | |
315 | case 'r': | |
316 | return '\r'; | |
317 | case 't': | |
318 | return '\t'; | |
319 | case '\'': | |
320 | return '\''; | |
321 | case '"': | |
322 | return '\"'; | |
323 | default: | |
324 | return ch; | |
325 | } | |
326 | } | |
327 | ||
328 | /* This function is called to process input characters. The GET | |
329 | parameter is used to retrieve more input characters. GET should | |
330 | set its parameter to point to a buffer, and return the length of | |
331 | the buffer; it should return 0 at end of file. The scrubbed output | |
332 | characters are put into the buffer starting at TOSTART; the TOSTART | |
333 | buffer is TOLEN bytes in length. The function returns the number | |
334 | of scrubbed characters put into TOSTART. This will be TOLEN unless | |
335 | end of file was seen. This function is arranged as a state | |
336 | machine, and saves its state so that it may return at any point. | |
337 | This is the way the old code used to work. */ | |
338 | ||
339 | int | |
73ee5e4c | 340 | do_scrub_chars (int (*get) (char *, int), char *tostart, int tolen) |
252b5132 RH |
341 | { |
342 | char *to = tostart; | |
343 | char *toend = tostart + tolen; | |
344 | char *from; | |
345 | char *fromend; | |
346 | int fromlen; | |
347 | register int ch, ch2 = 0; | |
348 | ||
349 | /*State 0: beginning of normal line | |
350 | 1: After first whitespace on line (flush more white) | |
351 | 2: After first non-white (opcode) on line (keep 1white) | |
352 | 3: after second white on line (into operands) (flush white) | |
353 | 4: after putting out a .line, put out digits | |
354 | 5: parsing a string, then go to old-state | |
355 | 6: putting out \ escape in a "d string. | |
356 | 7: After putting out a .appfile, put out string. | |
357 | 8: After putting out a .appfile string, flush until newline. | |
358 | 9: After seeing symbol char in state 3 (keep 1white after symchar) | |
359 | 10: After seeing whitespace in state 9 (keep white before symchar) | |
360 | 11: After seeing a symbol character in state 0 (eg a label definition) | |
361 | -1: output string in out_string and go to the state in old_state | |
362 | -2: flush text until a '*' '/' is seen, then go to state old_state | |
363 | #ifdef TC_V850 | |
b1ac4c66 AM |
364 | 12: After seeing a dash, looking for a second dash as a start |
365 | of comment. | |
252b5132 | 366 | #endif |
f28e8eb3 | 367 | #ifdef DOUBLEBAR_PARALLEL |
b1ac4c66 AM |
368 | 13: After seeing a vertical bar, looking for a second |
369 | vertical bar as a parallel expression separator. | |
52628315 L |
370 | #endif |
371 | #ifdef TC_IA64 | |
372 | 14: After seeing a `(' at state 0, looking for a `)' as | |
373 | predicate. | |
374 | 15: After seeing a `(' at state 1, looking for a `)' as | |
375 | predicate. | |
252b5132 RH |
376 | #endif |
377 | */ | |
378 | ||
379 | /* I added states 9 and 10 because the MIPS ECOFF assembler uses | |
380 | constructs like ``.loc 1 20''. This was turning into ``.loc | |
381 | 120''. States 9 and 10 ensure that a space is never dropped in | |
3b37fd66 | 382 | between characters which could appear in an identifier. Ian |
252b5132 RH |
383 | Taylor, ian@cygnus.com. |
384 | ||
385 | I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works | |
386 | correctly on the PA (and any other target where colons are optional). | |
387 | Jeff Law, law@cs.utah.edu. | |
388 | ||
389 | I added state 13 so that something like "cmp r1, r2 || trap #1" does not | |
390 | get squashed into "cmp r1,r2||trap#1", with the all important space | |
391 | between the 'trap' and the '#1' being eliminated. nickc@cygnus.com */ | |
392 | ||
393 | /* This macro gets the next input character. */ | |
394 | ||
2b47531b ILT |
395 | #define GET() \ |
396 | (from < fromend \ | |
397 | ? * (unsigned char *) (from++) \ | |
398 | : (saved_input = NULL, \ | |
399 | fromlen = (*get) (input_buffer, sizeof input_buffer), \ | |
400 | from = input_buffer, \ | |
401 | fromend = from + fromlen, \ | |
402 | (fromlen == 0 \ | |
403 | ? EOF \ | |
252b5132 RH |
404 | : * (unsigned char *) (from++)))) |
405 | ||
406 | /* This macro pushes a character back on the input stream. */ | |
407 | ||
408 | #define UNGET(uch) (*--from = (uch)) | |
409 | ||
410 | /* This macro puts a character into the output buffer. If this | |
411 | character fills the output buffer, this macro jumps to the label | |
412 | TOFULL. We use this rather ugly approach because we need to | |
413 | handle two different termination conditions: EOF on the input | |
414 | stream, and a full output buffer. It would be simpler if we | |
415 | always read in the entire input stream before processing it, but | |
416 | I don't want to make such a significant change to the assembler's | |
417 | memory usage. */ | |
418 | ||
411863a4 KH |
419 | #define PUT(pch) \ |
420 | do \ | |
421 | { \ | |
422 | *to++ = (pch); \ | |
423 | if (to >= toend) \ | |
424 | goto tofull; \ | |
425 | } \ | |
252b5132 RH |
426 | while (0) |
427 | ||
428 | if (saved_input != NULL) | |
429 | { | |
430 | from = saved_input; | |
431 | fromend = from + saved_input_len; | |
432 | } | |
433 | else | |
434 | { | |
2b47531b | 435 | fromlen = (*get) (input_buffer, sizeof input_buffer); |
252b5132 RH |
436 | if (fromlen == 0) |
437 | return 0; | |
2b47531b | 438 | from = input_buffer; |
252b5132 RH |
439 | fromend = from + fromlen; |
440 | } | |
441 | ||
442 | while (1) | |
443 | { | |
444 | /* The cases in this switch end with continue, in order to | |
b1ac4c66 AM |
445 | branch back to the top of this while loop and generate the |
446 | next output character in the appropriate state. */ | |
252b5132 RH |
447 | switch (state) |
448 | { | |
449 | case -1: | |
450 | ch = *out_string++; | |
451 | if (*out_string == '\0') | |
452 | { | |
453 | state = old_state; | |
454 | old_state = 3; | |
455 | } | |
456 | PUT (ch); | |
457 | continue; | |
458 | ||
459 | case -2: | |
460 | for (;;) | |
461 | { | |
462 | do | |
463 | { | |
464 | ch = GET (); | |
465 | ||
466 | if (ch == EOF) | |
467 | { | |
468 | as_warn (_("end of file in comment")); | |
469 | goto fromeof; | |
470 | } | |
471 | ||
472 | if (ch == '\n') | |
473 | PUT ('\n'); | |
474 | } | |
475 | while (ch != '*'); | |
476 | ||
477 | while ((ch = GET ()) == '*') | |
478 | ; | |
479 | ||
480 | if (ch == EOF) | |
481 | { | |
482 | as_warn (_("end of file in comment")); | |
483 | goto fromeof; | |
484 | } | |
485 | ||
486 | if (ch == '/') | |
487 | break; | |
488 | ||
489 | UNGET (ch); | |
490 | } | |
491 | ||
492 | state = old_state; | |
493 | UNGET (' '); | |
494 | continue; | |
495 | ||
496 | case 4: | |
497 | ch = GET (); | |
498 | if (ch == EOF) | |
499 | goto fromeof; | |
500 | else if (ch >= '0' && ch <= '9') | |
501 | PUT (ch); | |
502 | else | |
503 | { | |
504 | while (ch != EOF && IS_WHITESPACE (ch)) | |
505 | ch = GET (); | |
506 | if (ch == '"') | |
507 | { | |
508 | UNGET (ch); | |
509 | if (scrub_m68k_mri) | |
510 | out_string = "\n\tappfile "; | |
511 | else | |
512 | out_string = "\n\t.appfile "; | |
513 | old_state = 7; | |
514 | state = -1; | |
515 | PUT (*out_string++); | |
516 | } | |
517 | else | |
518 | { | |
519 | while (ch != EOF && ch != '\n') | |
520 | ch = GET (); | |
521 | state = 0; | |
522 | PUT (ch); | |
523 | } | |
524 | } | |
525 | continue; | |
526 | ||
527 | case 5: | |
528 | /* We are going to copy everything up to a quote character, | |
b1ac4c66 AM |
529 | with special handling for a backslash. We try to |
530 | optimize the copying in the simple case without using the | |
531 | GET and PUT macros. */ | |
252b5132 RH |
532 | { |
533 | char *s; | |
534 | int len; | |
535 | ||
536 | for (s = from; s < fromend; s++) | |
537 | { | |
538 | ch = *s; | |
539 | /* This condition must be changed if the type of any | |
b1ac4c66 | 540 | other character can be LEX_IS_STRINGQUOTE. */ |
252b5132 RH |
541 | if (ch == '\\' |
542 | || ch == '"' | |
543 | || ch == '\'' | |
544 | || ch == '\n') | |
545 | break; | |
546 | } | |
547 | len = s - from; | |
548 | if (len > toend - to) | |
549 | len = toend - to; | |
550 | if (len > 0) | |
551 | { | |
552 | memcpy (to, from, len); | |
553 | to += len; | |
554 | from += len; | |
555 | } | |
556 | } | |
557 | ||
558 | ch = GET (); | |
559 | if (ch == EOF) | |
560 | { | |
0e389e77 | 561 | as_warn (_("end of file in string; inserted '\"'")); |
252b5132 RH |
562 | state = old_state; |
563 | UNGET ('\n'); | |
564 | PUT ('"'); | |
565 | } | |
566 | else if (lex[ch] == LEX_IS_STRINGQUOTE) | |
567 | { | |
568 | state = old_state; | |
569 | PUT (ch); | |
570 | } | |
571 | #ifndef NO_STRING_ESCAPES | |
572 | else if (ch == '\\') | |
573 | { | |
574 | state = 6; | |
575 | PUT (ch); | |
576 | } | |
577 | #endif | |
578 | else if (scrub_m68k_mri && ch == '\n') | |
579 | { | |
580 | /* Just quietly terminate the string. This permits lines like | |
204cd129 | 581 | bne label loop if we haven't reach end yet. */ |
252b5132 RH |
582 | state = old_state; |
583 | UNGET (ch); | |
584 | PUT ('\''); | |
585 | } | |
586 | else | |
587 | { | |
588 | PUT (ch); | |
589 | } | |
590 | continue; | |
591 | ||
592 | case 6: | |
593 | state = 5; | |
594 | ch = GET (); | |
595 | switch (ch) | |
596 | { | |
597 | /* Handle strings broken across lines, by turning '\n' into | |
598 | '\\' and 'n'. */ | |
599 | case '\n': | |
600 | UNGET ('n'); | |
601 | add_newlines++; | |
602 | PUT ('\\'); | |
603 | continue; | |
604 | ||
4252e537 AM |
605 | case EOF: |
606 | as_warn (_("end of file in string; '\"' inserted")); | |
607 | PUT ('"'); | |
608 | continue; | |
609 | ||
252b5132 RH |
610 | case '"': |
611 | case '\\': | |
612 | case 'b': | |
613 | case 'f': | |
614 | case 'n': | |
615 | case 'r': | |
616 | case 't': | |
617 | case 'v': | |
618 | case 'x': | |
619 | case 'X': | |
620 | case '0': | |
621 | case '1': | |
622 | case '2': | |
623 | case '3': | |
624 | case '4': | |
625 | case '5': | |
626 | case '6': | |
627 | case '7': | |
628 | break; | |
4252e537 | 629 | |
252b5132 | 630 | default: |
4252e537 | 631 | #ifdef ONLY_STANDARD_ESCAPES |
0e389e77 | 632 | as_warn (_("unknown escape '\\%c' in string; ignored"), ch); |
4252e537 | 633 | #endif |
252b5132 | 634 | break; |
252b5132 RH |
635 | } |
636 | PUT (ch); | |
637 | continue; | |
638 | ||
639 | case 7: | |
640 | ch = GET (); | |
641 | state = 5; | |
642 | old_state = 8; | |
643 | if (ch == EOF) | |
644 | goto fromeof; | |
645 | PUT (ch); | |
646 | continue; | |
647 | ||
648 | case 8: | |
649 | do | |
650 | ch = GET (); | |
651 | while (ch != '\n' && ch != EOF); | |
652 | if (ch == EOF) | |
653 | goto fromeof; | |
654 | state = 0; | |
655 | PUT (ch); | |
656 | continue; | |
b1ac4c66 AM |
657 | |
658 | #ifdef DOUBLEBAR_PARALLEL | |
659 | case 13: | |
660 | ch = GET (); | |
661 | if (ch != '|') | |
662 | abort (); | |
663 | ||
664 | /* Reset back to state 1 and pretend that we are parsing a | |
665 | line from just after the first white space. */ | |
666 | state = 1; | |
667 | PUT ('|'); | |
668 | continue; | |
669 | #endif | |
252b5132 RH |
670 | } |
671 | ||
204cd129 | 672 | /* OK, we are somewhere in states 0 through 4 or 9 through 11. */ |
252b5132 RH |
673 | |
674 | /* flushchar: */ | |
675 | ch = GET (); | |
676 | ||
52628315 L |
677 | #ifdef TC_IA64 |
678 | if (ch == '(' && (state == 0 || state == 1)) | |
679 | { | |
680 | state += 14; | |
681 | PUT (ch); | |
682 | continue; | |
683 | } | |
684 | else if (state == 14 || state == 15) | |
685 | { | |
686 | if (ch == ')') | |
687 | state -= 14; | |
688 | else | |
689 | { | |
690 | PUT (ch); | |
691 | continue; | |
692 | } | |
693 | } | |
694 | #endif | |
695 | ||
252b5132 RH |
696 | recycle: |
697 | ||
698 | #if defined TC_ARM && defined OBJ_ELF | |
699 | /* We need to watch out for .symver directives. See the comment later | |
700 | in this function. */ | |
701 | if (symver_state == NULL) | |
702 | { | |
703 | if ((state == 0 || state == 1) && ch == symver_pseudo[0]) | |
704 | symver_state = symver_pseudo + 1; | |
705 | } | |
706 | else | |
707 | { | |
708 | /* We advance to the next state if we find the right | |
709 | character. */ | |
710 | if (ch != '\0' && (*symver_state == ch)) | |
711 | ++symver_state; | |
712 | else if (*symver_state != '\0') | |
713 | /* We did not get the expected character, or we didn't | |
714 | get a valid terminating character after seeing the | |
715 | entire pseudo-op, so we must go back to the beginning. */ | |
716 | symver_state = NULL; | |
717 | else | |
718 | { | |
719 | /* We've read the entire pseudo-op. If this is the end | |
720 | of the line, go back to the beginning. */ | |
721 | if (IS_NEWLINE (ch)) | |
722 | symver_state = NULL; | |
723 | } | |
724 | } | |
725 | #endif /* TC_ARM && OBJ_ELF */ | |
726 | ||
727 | #ifdef TC_M68K | |
728 | /* We want to have pseudo-ops which control whether we are in | |
b1ac4c66 AM |
729 | MRI mode or not. Unfortunately, since m68k MRI mode affects |
730 | the scrubber, that means that we need a special purpose | |
731 | recognizer here. */ | |
252b5132 RH |
732 | if (mri_state == NULL) |
733 | { | |
734 | if ((state == 0 || state == 1) | |
735 | && ch == mri_pseudo[0]) | |
736 | mri_state = mri_pseudo + 1; | |
737 | } | |
738 | else | |
739 | { | |
740 | /* We advance to the next state if we find the right | |
741 | character, or if we need a space character and we get any | |
742 | whitespace character, or if we need a '0' and we get a | |
743 | '1' (this is so that we only need one state to handle | |
744 | ``.mri 0'' and ``.mri 1''). */ | |
745 | if (ch != '\0' | |
746 | && (*mri_state == ch | |
747 | || (*mri_state == ' ' | |
748 | && lex[ch] == LEX_IS_WHITESPACE) | |
749 | || (*mri_state == '0' | |
750 | && ch == '1'))) | |
751 | { | |
752 | mri_last_ch = ch; | |
753 | ++mri_state; | |
754 | } | |
755 | else if (*mri_state != '\0' | |
756 | || (lex[ch] != LEX_IS_WHITESPACE | |
757 | && lex[ch] != LEX_IS_NEWLINE)) | |
758 | { | |
759 | /* We did not get the expected character, or we didn't | |
760 | get a valid terminating character after seeing the | |
761 | entire pseudo-op, so we must go back to the | |
762 | beginning. */ | |
763 | mri_state = NULL; | |
764 | } | |
765 | else | |
766 | { | |
767 | /* We've read the entire pseudo-op. mips_last_ch is | |
b1ac4c66 AM |
768 | either '0' or '1' indicating whether to enter or |
769 | leave MRI mode. */ | |
252b5132 RH |
770 | do_scrub_begin (mri_last_ch == '1'); |
771 | mri_state = NULL; | |
772 | ||
773 | /* We continue handling the character as usual. The | |
b1ac4c66 AM |
774 | main gas reader must also handle the .mri pseudo-op |
775 | to control expression parsing and the like. */ | |
252b5132 RH |
776 | } |
777 | } | |
778 | #endif | |
779 | ||
780 | if (ch == EOF) | |
781 | { | |
782 | if (state != 0) | |
783 | { | |
784 | as_warn (_("end of file not at end of a line; newline inserted")); | |
785 | state = 0; | |
786 | PUT ('\n'); | |
787 | } | |
788 | goto fromeof; | |
789 | } | |
790 | ||
791 | switch (lex[ch]) | |
792 | { | |
793 | case LEX_IS_WHITESPACE: | |
794 | do | |
795 | { | |
796 | ch = GET (); | |
797 | } | |
798 | while (ch != EOF && IS_WHITESPACE (ch)); | |
799 | if (ch == EOF) | |
800 | goto fromeof; | |
801 | ||
802 | if (state == 0) | |
803 | { | |
804 | /* Preserve a single whitespace character at the | |
805 | beginning of a line. */ | |
806 | state = 1; | |
807 | UNGET (ch); | |
808 | PUT (' '); | |
809 | break; | |
810 | } | |
811 | ||
f28e8eb3 | 812 | #ifdef KEEP_WHITE_AROUND_COLON |
30a2b4ef KH |
813 | if (lex[ch] == LEX_IS_COLON) |
814 | { | |
815 | /* Only keep this white if there's no white *after* the | |
b1ac4c66 | 816 | colon. */ |
30a2b4ef KH |
817 | ch2 = GET (); |
818 | UNGET (ch2); | |
819 | if (!IS_WHITESPACE (ch2)) | |
820 | { | |
821 | state = 9; | |
822 | UNGET (ch); | |
823 | PUT (' '); | |
824 | break; | |
825 | } | |
826 | } | |
f28e8eb3 | 827 | #endif |
252b5132 RH |
828 | if (IS_COMMENT (ch) |
829 | || ch == '/' | |
62f65a7b DB |
830 | || IS_LINE_SEPARATOR (ch) |
831 | || IS_PARALLEL_SEPARATOR (ch)) | |
252b5132 RH |
832 | { |
833 | if (scrub_m68k_mri) | |
834 | { | |
835 | /* In MRI mode, we keep these spaces. */ | |
836 | UNGET (ch); | |
837 | PUT (' '); | |
838 | break; | |
839 | } | |
840 | goto recycle; | |
841 | } | |
842 | ||
843 | /* If we're in state 2 or 11, we've seen a non-white | |
844 | character followed by whitespace. If the next character | |
845 | is ':', this is whitespace after a label name which we | |
846 | normally must ignore. In MRI mode, though, spaces are | |
847 | not permitted between the label and the colon. */ | |
848 | if ((state == 2 || state == 11) | |
849 | && lex[ch] == LEX_IS_COLON | |
850 | && ! scrub_m68k_mri) | |
851 | { | |
852 | state = 1; | |
853 | PUT (ch); | |
854 | break; | |
855 | } | |
856 | ||
857 | switch (state) | |
858 | { | |
859 | case 0: | |
860 | state++; | |
861 | goto recycle; /* Punted leading sp */ | |
862 | case 1: | |
863 | /* We can arrive here if we leave a leading whitespace | |
864 | character at the beginning of a line. */ | |
865 | goto recycle; | |
866 | case 2: | |
867 | state = 3; | |
868 | if (to + 1 < toend) | |
869 | { | |
870 | /* Optimize common case by skipping UNGET/GET. */ | |
871 | PUT (' '); /* Sp after opco */ | |
872 | goto recycle; | |
873 | } | |
874 | UNGET (ch); | |
875 | PUT (' '); | |
876 | break; | |
877 | case 3: | |
878 | if (scrub_m68k_mri) | |
879 | { | |
880 | /* In MRI mode, we keep these spaces. */ | |
881 | UNGET (ch); | |
882 | PUT (' '); | |
883 | break; | |
884 | } | |
885 | goto recycle; /* Sp in operands */ | |
886 | case 9: | |
887 | case 10: | |
888 | if (scrub_m68k_mri) | |
889 | { | |
890 | /* In MRI mode, we keep these spaces. */ | |
891 | state = 3; | |
892 | UNGET (ch); | |
893 | PUT (' '); | |
894 | break; | |
895 | } | |
896 | state = 10; /* Sp after symbol char */ | |
897 | goto recycle; | |
898 | case 11: | |
abd63a32 | 899 | if (LABELS_WITHOUT_COLONS || flag_m68k_mri) |
252b5132 RH |
900 | state = 1; |
901 | else | |
902 | { | |
903 | /* We know that ch is not ':', since we tested that | |
b1ac4c66 AM |
904 | case above. Therefore this is not a label, so it |
905 | must be the opcode, and we've just seen the | |
906 | whitespace after it. */ | |
252b5132 RH |
907 | state = 3; |
908 | } | |
909 | UNGET (ch); | |
910 | PUT (' '); /* Sp after label definition. */ | |
911 | break; | |
912 | default: | |
913 | BAD_CASE (state); | |
914 | } | |
915 | break; | |
916 | ||
917 | case LEX_IS_TWOCHAR_COMMENT_1ST: | |
918 | ch2 = GET (); | |
919 | if (ch2 == '*') | |
920 | { | |
921 | for (;;) | |
922 | { | |
923 | do | |
924 | { | |
925 | ch2 = GET (); | |
926 | if (ch2 != EOF && IS_NEWLINE (ch2)) | |
927 | add_newlines++; | |
928 | } | |
929 | while (ch2 != EOF && ch2 != '*'); | |
930 | ||
931 | while (ch2 == '*') | |
932 | ch2 = GET (); | |
933 | ||
934 | if (ch2 == EOF || ch2 == '/') | |
935 | break; | |
936 | ||
937 | /* This UNGET will ensure that we count newlines | |
b1ac4c66 | 938 | correctly. */ |
252b5132 RH |
939 | UNGET (ch2); |
940 | } | |
941 | ||
942 | if (ch2 == EOF) | |
943 | as_warn (_("end of file in multiline comment")); | |
944 | ||
945 | ch = ' '; | |
946 | goto recycle; | |
947 | } | |
800eeca4 JW |
948 | #ifdef DOUBLESLASH_LINE_COMMENTS |
949 | else if (ch2 == '/') | |
950 | { | |
951 | do | |
952 | { | |
953 | ch = GET (); | |
954 | } | |
955 | while (ch != EOF && !IS_NEWLINE (ch)); | |
956 | if (ch == EOF) | |
957 | as_warn ("end of file in comment; newline inserted"); | |
958 | state = 0; | |
959 | PUT ('\n'); | |
960 | break; | |
961 | } | |
962 | #endif | |
252b5132 RH |
963 | else |
964 | { | |
965 | if (ch2 != EOF) | |
966 | UNGET (ch2); | |
967 | if (state == 9 || state == 10) | |
968 | state = 3; | |
969 | PUT (ch); | |
970 | } | |
971 | break; | |
972 | ||
973 | case LEX_IS_STRINGQUOTE: | |
974 | if (state == 10) | |
975 | { | |
204cd129 | 976 | /* Preserve the whitespace in foo "bar". */ |
252b5132 RH |
977 | UNGET (ch); |
978 | state = 3; | |
979 | PUT (' '); | |
980 | ||
981 | /* PUT didn't jump out. We could just break, but we | |
b1ac4c66 | 982 | know what will happen, so optimize a bit. */ |
252b5132 RH |
983 | ch = GET (); |
984 | old_state = 3; | |
985 | } | |
986 | else if (state == 9) | |
987 | old_state = 3; | |
988 | else | |
989 | old_state = state; | |
990 | state = 5; | |
991 | PUT (ch); | |
992 | break; | |
993 | ||
994 | #ifndef IEEE_STYLE | |
995 | case LEX_IS_ONECHAR_QUOTE: | |
996 | if (state == 10) | |
997 | { | |
204cd129 | 998 | /* Preserve the whitespace in foo 'b'. */ |
252b5132 RH |
999 | UNGET (ch); |
1000 | state = 3; | |
1001 | PUT (' '); | |
1002 | break; | |
1003 | } | |
1004 | ch = GET (); | |
1005 | if (ch == EOF) | |
1006 | { | |
1007 | as_warn (_("end of file after a one-character quote; \\0 inserted")); | |
1008 | ch = 0; | |
1009 | } | |
1010 | if (ch == '\\') | |
1011 | { | |
1012 | ch = GET (); | |
1013 | if (ch == EOF) | |
1014 | { | |
1015 | as_warn (_("end of file in escape character")); | |
1016 | ch = '\\'; | |
1017 | } | |
1018 | else | |
1019 | ch = process_escape (ch); | |
1020 | } | |
1021 | sprintf (out_buf, "%d", (int) (unsigned char) ch); | |
1022 | ||
1023 | /* None of these 'x constants for us. We want 'x'. */ | |
1024 | if ((ch = GET ()) != '\'') | |
1025 | { | |
1026 | #ifdef REQUIRE_CHAR_CLOSE_QUOTE | |
0e389e77 | 1027 | as_warn (_("missing close quote; (assumed)")); |
252b5132 RH |
1028 | #else |
1029 | if (ch != EOF) | |
1030 | UNGET (ch); | |
1031 | #endif | |
1032 | } | |
1033 | if (strlen (out_buf) == 1) | |
1034 | { | |
1035 | PUT (out_buf[0]); | |
1036 | break; | |
1037 | } | |
1038 | if (state == 9) | |
1039 | old_state = 3; | |
1040 | else | |
1041 | old_state = state; | |
1042 | state = -1; | |
1043 | out_string = out_buf; | |
1044 | PUT (*out_string++); | |
1045 | break; | |
1046 | #endif | |
1047 | ||
1048 | case LEX_IS_COLON: | |
f28e8eb3 | 1049 | #ifdef KEEP_WHITE_AROUND_COLON |
30a2b4ef | 1050 | state = 9; |
f28e8eb3 | 1051 | #else |
252b5132 RH |
1052 | if (state == 9 || state == 10) |
1053 | state = 3; | |
1054 | else if (state != 3) | |
1055 | state = 1; | |
f28e8eb3 | 1056 | #endif |
252b5132 RH |
1057 | PUT (ch); |
1058 | break; | |
1059 | ||
1060 | case LEX_IS_NEWLINE: | |
1061 | /* Roll out a bunch of newlines from inside comments, etc. */ | |
1062 | if (add_newlines) | |
1063 | { | |
1064 | --add_newlines; | |
1065 | UNGET (ch); | |
1066 | } | |
3ee4defc | 1067 | /* Fall through. */ |
252b5132 RH |
1068 | |
1069 | case LEX_IS_LINE_SEPARATOR: | |
1070 | state = 0; | |
1071 | PUT (ch); | |
1072 | break; | |
1073 | ||
62f65a7b DB |
1074 | case LEX_IS_PARALLEL_SEPARATOR: |
1075 | state = 1; | |
1076 | PUT (ch); | |
1077 | break; | |
1078 | ||
252b5132 RH |
1079 | #ifdef TC_V850 |
1080 | case LEX_IS_DOUBLEDASH_1ST: | |
30a2b4ef | 1081 | ch2 = GET (); |
252b5132 RH |
1082 | if (ch2 != '-') |
1083 | { | |
1084 | UNGET (ch2); | |
1085 | goto de_fault; | |
1086 | } | |
3ee4defc | 1087 | /* Read and skip to end of line. */ |
252b5132 RH |
1088 | do |
1089 | { | |
1090 | ch = GET (); | |
1091 | } | |
1092 | while (ch != EOF && ch != '\n'); | |
204cd129 | 1093 | |
252b5132 | 1094 | if (ch == EOF) |
204cd129 NC |
1095 | as_warn (_("end of file in comment; newline inserted")); |
1096 | ||
252b5132 RH |
1097 | state = 0; |
1098 | PUT ('\n'); | |
1099 | break; | |
3ee4defc | 1100 | #endif |
f28e8eb3 | 1101 | #ifdef DOUBLEBAR_PARALLEL |
252b5132 | 1102 | case LEX_IS_DOUBLEBAR_1ST: |
30a2b4ef | 1103 | ch2 = GET (); |
b1ac4c66 | 1104 | UNGET (ch2); |
252b5132 | 1105 | if (ch2 != '|') |
204cd129 NC |
1106 | goto de_fault; |
1107 | ||
b1ac4c66 AM |
1108 | /* Handle '||' in two states as invoking PUT twice might |
1109 | result in the first one jumping out of this loop. We'd | |
1110 | then lose track of the state and one '|' char. */ | |
1111 | state = 13; | |
252b5132 RH |
1112 | PUT ('|'); |
1113 | break; | |
3ee4defc | 1114 | #endif |
252b5132 RH |
1115 | case LEX_IS_LINE_COMMENT_START: |
1116 | /* FIXME-someday: The two character comment stuff was badly | |
1117 | thought out. On i386, we want '/' as line comment start | |
1118 | AND we want C style comments. hence this hack. The | |
1119 | whole lexical process should be reworked. xoxorich. */ | |
1120 | if (ch == '/') | |
1121 | { | |
1122 | ch2 = GET (); | |
1123 | if (ch2 == '*') | |
1124 | { | |
1125 | old_state = 3; | |
1126 | state = -2; | |
1127 | break; | |
1128 | } | |
1129 | else | |
1130 | { | |
1131 | UNGET (ch2); | |
1132 | } | |
204cd129 | 1133 | } |
252b5132 RH |
1134 | |
1135 | if (state == 0 || state == 1) /* Only comment at start of line. */ | |
1136 | { | |
1137 | int startch; | |
1138 | ||
1139 | startch = ch; | |
1140 | ||
1141 | do | |
1142 | { | |
1143 | ch = GET (); | |
1144 | } | |
1145 | while (ch != EOF && IS_WHITESPACE (ch)); | |
204cd129 | 1146 | |
252b5132 RH |
1147 | if (ch == EOF) |
1148 | { | |
1149 | as_warn (_("end of file in comment; newline inserted")); | |
1150 | PUT ('\n'); | |
1151 | break; | |
1152 | } | |
204cd129 | 1153 | |
252b5132 RH |
1154 | if (ch < '0' || ch > '9' || state != 0 || startch != '#') |
1155 | { | |
1156 | /* Not a cpp line. */ | |
1157 | while (ch != EOF && !IS_NEWLINE (ch)) | |
1158 | ch = GET (); | |
1159 | if (ch == EOF) | |
0e389e77 | 1160 | as_warn (_("end of file in comment; newline inserted")); |
252b5132 RH |
1161 | state = 0; |
1162 | PUT ('\n'); | |
1163 | break; | |
1164 | } | |
3ee4defc | 1165 | /* Looks like `# 123 "filename"' from cpp. */ |
252b5132 RH |
1166 | UNGET (ch); |
1167 | old_state = 4; | |
1168 | state = -1; | |
1169 | if (scrub_m68k_mri) | |
1170 | out_string = "\tappline "; | |
1171 | else | |
1172 | out_string = "\t.appline "; | |
1173 | PUT (*out_string++); | |
1174 | break; | |
1175 | } | |
1176 | ||
1177 | #ifdef TC_D10V | |
1178 | /* All insns end in a char for which LEX_IS_SYMBOL_COMPONENT is true. | |
1179 | Trap is the only short insn that has a first operand that is | |
1180 | neither register nor label. | |
1181 | We must prevent exef0f ||trap #1 to degenerate to exef0f ||trap#1 . | |
30a2b4ef KH |
1182 | We can't make '#' LEX_IS_SYMBOL_COMPONENT because it is |
1183 | already LEX_IS_LINE_COMMENT_START. However, it is the | |
1184 | only character in line_comment_chars for d10v, hence we | |
1185 | can recognize it as such. */ | |
252b5132 RH |
1186 | /* An alternative approach would be to reset the state to 1 when |
1187 | we see '||', '<'- or '->', but that seems to be overkill. */ | |
30a2b4ef KH |
1188 | if (state == 10) |
1189 | PUT (' '); | |
252b5132 RH |
1190 | #endif |
1191 | /* We have a line comment character which is not at the | |
1192 | start of a line. If this is also a normal comment | |
1193 | character, fall through. Otherwise treat it as a default | |
1194 | character. */ | |
1195 | if (strchr (tc_comment_chars, ch) == NULL | |
1196 | && (! scrub_m68k_mri | |
1197 | || (ch != '!' && ch != '*'))) | |
1198 | goto de_fault; | |
1199 | if (scrub_m68k_mri | |
1200 | && (ch == '!' || ch == '*' || ch == '#') | |
1201 | && state != 1 | |
1202 | && state != 10) | |
1203 | goto de_fault; | |
1204 | /* Fall through. */ | |
1205 | case LEX_IS_COMMENT_START: | |
1206 | #if defined TC_ARM && defined OBJ_ELF | |
1207 | /* On the ARM, `@' is the comment character. | |
1208 | Unfortunately this is also a special character in ELF .symver | |
30a2b4ef KH |
1209 | directives (and .type, though we deal with those another way). |
1210 | So we check if this line is such a directive, and treat | |
1211 | the character as default if so. This is a hack. */ | |
252b5132 RH |
1212 | if ((symver_state != NULL) && (*symver_state == 0)) |
1213 | goto de_fault; | |
4c400d5e AM |
1214 | #endif |
1215 | #ifdef WARN_COMMENTS | |
1216 | if (!found_comment) | |
1217 | as_where (&found_comment_file, &found_comment); | |
252b5132 RH |
1218 | #endif |
1219 | do | |
1220 | { | |
1221 | ch = GET (); | |
1222 | } | |
1223 | while (ch != EOF && !IS_NEWLINE (ch)); | |
1224 | if (ch == EOF) | |
1225 | as_warn (_("end of file in comment; newline inserted")); | |
1226 | state = 0; | |
1227 | PUT ('\n'); | |
1228 | break; | |
1229 | ||
1230 | case LEX_IS_SYMBOL_COMPONENT: | |
1231 | if (state == 10) | |
1232 | { | |
1233 | /* This is a symbol character following another symbol | |
1234 | character, with whitespace in between. We skipped | |
1235 | the whitespace earlier, so output it now. */ | |
1236 | UNGET (ch); | |
1237 | state = 3; | |
1238 | PUT (' '); | |
1239 | break; | |
1240 | } | |
1241 | ||
1242 | if (state == 3) | |
1243 | state = 9; | |
1244 | ||
1245 | /* This is a common case. Quickly copy CH and all the | |
b1ac4c66 | 1246 | following symbol component or normal characters. */ |
252b5132 RH |
1247 | if (to + 1 < toend |
1248 | && mri_state == NULL | |
1249 | #if defined TC_ARM && defined OBJ_ELF | |
1250 | && symver_state == NULL | |
1251 | #endif | |
1252 | ) | |
1253 | { | |
1254 | char *s; | |
1255 | int len; | |
1256 | ||
1257 | for (s = from; s < fromend; s++) | |
1258 | { | |
1259 | int type; | |
1260 | ||
30a2b4ef | 1261 | ch2 = *(unsigned char *) s; |
252b5132 RH |
1262 | type = lex[ch2]; |
1263 | if (type != 0 | |
1264 | && type != LEX_IS_SYMBOL_COMPONENT) | |
1265 | break; | |
1266 | } | |
204cd129 | 1267 | |
252b5132 | 1268 | if (s > from) |
204cd129 NC |
1269 | /* Handle the last character normally, for |
1270 | simplicity. */ | |
1271 | --s; | |
1272 | ||
252b5132 | 1273 | len = s - from; |
204cd129 | 1274 | |
252b5132 RH |
1275 | if (len > (toend - to) - 1) |
1276 | len = (toend - to) - 1; | |
204cd129 | 1277 | |
252b5132 RH |
1278 | if (len > 0) |
1279 | { | |
1280 | PUT (ch); | |
1281 | if (len > 8) | |
1282 | { | |
1283 | memcpy (to, from, len); | |
1284 | to += len; | |
1285 | from += len; | |
1286 | } | |
1287 | else | |
1288 | { | |
1289 | switch (len) | |
1290 | { | |
1291 | case 8: *to++ = *from++; | |
1292 | case 7: *to++ = *from++; | |
1293 | case 6: *to++ = *from++; | |
1294 | case 5: *to++ = *from++; | |
1295 | case 4: *to++ = *from++; | |
1296 | case 3: *to++ = *from++; | |
1297 | case 2: *to++ = *from++; | |
1298 | case 1: *to++ = *from++; | |
1299 | } | |
3ee4defc | 1300 | } |
252b5132 RH |
1301 | ch = GET (); |
1302 | } | |
1303 | } | |
1304 | ||
1305 | /* Fall through. */ | |
1306 | default: | |
1307 | de_fault: | |
1308 | /* Some relatively `normal' character. */ | |
1309 | if (state == 0) | |
1310 | { | |
2cdb18a7 | 1311 | if (IS_SYMBOL_COMPONENT (ch)) |
204cd129 | 1312 | state = 11; /* Now seeing label definition. */ |
252b5132 RH |
1313 | } |
1314 | else if (state == 1) | |
1315 | { | |
204cd129 NC |
1316 | if (IS_SYMBOL_COMPONENT (ch)) |
1317 | state = 2; /* Ditto. */ | |
252b5132 RH |
1318 | } |
1319 | else if (state == 9) | |
1320 | { | |
2cdb18a7 | 1321 | if (!IS_SYMBOL_COMPONENT (ch)) |
252b5132 RH |
1322 | state = 3; |
1323 | } | |
1324 | else if (state == 10) | |
1325 | { | |
c5c834aa AH |
1326 | if (ch == '\\') |
1327 | { | |
1328 | /* Special handling for backslash: a backslash may | |
1329 | be the beginning of a formal parameter (of a | |
1330 | macro) following another symbol character, with | |
1331 | whitespace in between. If that is the case, we | |
1332 | output a space before the parameter. Strictly | |
1333 | speaking, correct handling depends upon what the | |
1334 | macro parameter expands into; if the parameter | |
1335 | expands into something which does not start with | |
1336 | an operand character, then we don't want to keep | |
1337 | the space. We don't have enough information to | |
1338 | make the right choice, so here we are making the | |
1339 | choice which is more likely to be correct. */ | |
1340 | PUT (' '); | |
1341 | } | |
1342 | ||
252b5132 RH |
1343 | state = 3; |
1344 | } | |
1345 | PUT (ch); | |
1346 | break; | |
1347 | } | |
1348 | } | |
1349 | ||
1350 | /*NOTREACHED*/ | |
1351 | ||
1352 | fromeof: | |
1353 | /* We have reached the end of the input. */ | |
1354 | return to - tostart; | |
1355 | ||
1356 | tofull: | |
1357 | /* The output buffer is full. Save any input we have not yet | |
1358 | processed. */ | |
1359 | if (fromend > from) | |
1360 | { | |
2b47531b | 1361 | saved_input = from; |
252b5132 RH |
1362 | saved_input_len = fromend - from; |
1363 | } | |
1364 | else | |
2b47531b ILT |
1365 | saved_input = NULL; |
1366 | ||
252b5132 RH |
1367 | return to - tostart; |
1368 | } | |
1369 |