Commit | Line | Data |
---|---|---|
3bd55f52 RH |
1 | /* |
2 | * Copyright (c) 1982 Regents of the University of California | |
3 | */ | |
4 | #ifndef lint | |
ad67d5e5 | 5 | static char sccsid[] = "@(#)asscan2.c 4.7 %G%"; |
3bd55f52 RH |
6 | #endif not lint |
7 | ||
8 | #include "asscanl.h" | |
9 | static inttoktype oval = NL; | |
10 | ||
ad67d5e5 | 11 | #define ASINBUFSIZ 4096 |
3bd55f52 | 12 | #define NINBUFFERS 2 |
ad67d5e5 | 13 | #define INBUFLG NINBUFFERS*ASINBUFSIZ + 2 |
3bd55f52 | 14 | /* |
ad67d5e5 | 15 | * We have NINBUFFERS input buffers; the first one is reserved |
3bd55f52 | 16 | * for catching the tail of a line split across a buffer |
ad67d5e5 RH |
17 | * boundary; the other ones are used for snarfing a buffer |
18 | * worth of assembly language source. | |
3bd55f52 RH |
19 | */ |
20 | static char inbuffer[INBUFLG]; | |
21 | static char *InBufPtr = 0; | |
22 | ||
23 | /* | |
24 | * fill the inbuffer from the standard input. | |
25 | * Assert: there are always n COMPLETE! lines in the buffer area. | |
26 | * Assert: there is always a \n terminating the last line | |
27 | * in the buffer area. | |
28 | * Assert: after the \n, there is an EOFCHAR (hard end of file) | |
29 | * or a NEEDCHAR (end of buffer) | |
30 | * Assert: fgets always null pads the string it reads. | |
31 | * Assert: no ungetc's are done at the end of a line or at the | |
32 | * beginning of a line. | |
33 | * | |
34 | * We read a complete buffer of characters in one single read. | |
35 | * We then back scan within this buffer to find the end of the | |
36 | * last complete line, and force the assertions, and save a pointer | |
37 | * to the incomplete line. | |
38 | * The next call to fillinbuffer will move the unread characters | |
39 | * to the end of the first buffer, and then read another two buffers, | |
40 | * completing the cycle. | |
41 | */ | |
42 | ||
43 | static char p_swapped = '\0'; | |
ad67d5e5 RH |
44 | static char *p_start = &inbuffer[NINBUFFERS * ASINBUFSIZ]; |
45 | static char *p_stop = &inbuffer[NINBUFFERS * ASINBUFSIZ]; | |
46 | ||
47 | #define MIDDLE &inbuffer[ASINBUFSIZ] | |
8a7e35d9 | 48 | |
3bd55f52 RH |
49 | char *fillinbuffer() |
50 | { | |
3bd55f52 RH |
51 | register char *from; |
52 | char *inbufptr; | |
53 | int nread; | |
8a7e35d9 RH |
54 | static int hadeof; |
55 | int goal; | |
56 | int got; | |
3bd55f52 RH |
57 | |
58 | *p_start = p_swapped; | |
ad67d5e5 RH |
59 | inbufptr = MIDDLE - (p_stop - p_start); |
60 | movestr(inbufptr, p_start, p_stop - p_start); | |
3bd55f52 | 61 | /* |
ad67d5e5 | 62 | * Now, go read up to NINBUFFERS - 1 full buffers |
3bd55f52 | 63 | */ |
8a7e35d9 RH |
64 | if (hadeof){ |
65 | hadeof = 0; | |
66 | return (0); | |
67 | } | |
ad67d5e5 | 68 | goal = (NINBUFFERS - 1)*ASINBUFSIZ; |
8a7e35d9 RH |
69 | nread = 0; |
70 | do { | |
ad67d5e5 | 71 | got = read(stdin->_file, MIDDLE + nread, goal); |
8a7e35d9 RH |
72 | if (got == 0) |
73 | hadeof = 1; | |
74 | if (got <= 0) | |
75 | break; | |
76 | nread += got; | |
77 | goal -= got; | |
78 | } while (goal); | |
79 | ||
3bd55f52 RH |
80 | if (nread == 0) |
81 | return(0); | |
ad67d5e5 RH |
82 | from = MIDDLE + nread; |
83 | p_stop = from; | |
3bd55f52 | 84 | *from = '\0'; |
8a7e35d9 RH |
85 | while (*--from != '\n'){ |
86 | /* | |
87 | * back over the partial line | |
88 | */ | |
ad67d5e5 | 89 | if (from == MIDDLE) { |
8a7e35d9 RH |
90 | from = p_stop; |
91 | *p_stop++ = '\n'; | |
92 | break; | |
93 | } else { | |
94 | continue; | |
95 | } | |
96 | } | |
97 | ||
3bd55f52 RH |
98 | from++; /* first char of partial line */ |
99 | p_start = from; | |
100 | p_swapped = *p_start; | |
101 | *p_start = NEEDCHAR; /* force assertion */ | |
102 | return(inbufptr); | |
103 | } | |
104 | ||
105 | scan_dot_s(bufferbox) | |
106 | struct tokbufdesc *bufferbox; | |
107 | { | |
108 | reg int ryylval; /* local copy of lexical value */ | |
109 | extern int yylval; /* global copy of lexical value */ | |
110 | reg int val; /* the value returned */ | |
111 | int i; /* simple counter */ | |
112 | reg char *rcp; | |
113 | char *cp; /* can have address taken */ | |
114 | reg int ch; /* treated as a character */ | |
115 | int ch1; /* shadow value */ | |
116 | reg char *inbufptr; | |
117 | struct symtab *op; | |
118 | ||
119 | reg ptrall bufptr; /* where to stuff tokens */ | |
120 | ptrall lgbackpatch; /* where to stuff a string length */ | |
121 | ptrall bufub; /* where not to stuff tokens */ | |
ad67d5e5 | 122 | reg int maxstrlg; /* how long a string can be */ |
3bd55f52 RH |
123 | long intval; /* value of int */ |
124 | int linescrossed; /* when doing strings and comments */ | |
125 | struct Opcode opstruct; | |
126 | ||
127 | (bytetoktype *)bufptr = (bytetoktype *) & (bufferbox->toks[0]); | |
128 | (bytetoktype *)bufub = &(bufferbox->toks[AVAILTOKS]); | |
129 | ||
130 | inbufptr = InBufPtr; | |
131 | if (inbufptr == 0){ | |
132 | inbufptr = fillinbuffer(); | |
133 | if (inbufptr == 0){ /*end of file*/ | |
134 | endoffile: | |
135 | inbufptr = 0; | |
136 | ptoken(bufptr, PARSEEOF); | |
137 | goto done; | |
138 | } | |
139 | } | |
140 | ||
141 | if (newfflag){ | |
ad67d5e5 RH |
142 | newfflag = 0; |
143 | ryylval = (int)savestr(newfname, strlen(newfname) + 1); | |
144 | ||
3bd55f52 RH |
145 | ptoken(bufptr, IFILE); |
146 | ptoken(bufptr, STRING); | |
ad67d5e5 | 147 | pptr(bufptr, ryylval); |
3bd55f52 RH |
148 | |
149 | ptoken(bufptr, ILINENO); | |
150 | ptoken(bufptr, INT); | |
151 | pint(bufptr, 1); | |
3bd55f52 RH |
152 | } |
153 | ||
154 | while (bufptr < bufub){ | |
155 | loop: | |
156 | switch(ryylval = (type+2)[ch = getchar()]) { | |
157 | case SCANEOF: | |
158 | inbufptr = 0; | |
159 | goto endoffile; | |
160 | ||
161 | case NEEDSBUF: | |
162 | inbufptr = fillinbuffer(); | |
163 | if (inbufptr == 0) | |
164 | goto endoffile; | |
165 | goto loop; | |
166 | ||
167 | case DIV: /*process C style comments*/ | |
168 | if ( (ch = getchar()) == '*') { /*comment prelude*/ | |
169 | int incomment; | |
170 | linescrossed = 0; | |
171 | incomment = 1; | |
172 | ch = getchar(); /*skip over the * */ | |
173 | while(incomment){ | |
174 | switch(ch){ | |
175 | case '*': | |
176 | ch = getchar(); | |
177 | incomment = (ch != '/'); | |
178 | break; | |
179 | case '\n': | |
180 | scanlineno++; | |
181 | linescrossed++; | |
182 | ch = getchar(); | |
183 | break; | |
184 | case EOFCHAR: | |
185 | goto endoffile; | |
186 | case NEEDCHAR: | |
187 | inbufptr = fillinbuffer(); | |
188 | if (inbufptr == 0) | |
189 | goto endoffile; | |
190 | lineno++; | |
191 | ch = getchar(); | |
192 | break; | |
193 | default: | |
194 | ch = getchar(); | |
195 | break; | |
196 | } | |
197 | } | |
198 | val = ILINESKIP; | |
199 | ryylval = linescrossed; | |
200 | goto ret; | |
201 | } else { /*just an ordinary DIV*/ | |
202 | ungetc(ch); | |
203 | val = ryylval = DIV; | |
204 | goto ret; | |
205 | } | |
206 | case SH: | |
207 | if (oval == NL){ | |
208 | /* | |
209 | * Attempt to recognize a C preprocessor | |
210 | * style comment '^#[ \t]*[0-9]*[ \t]*".*" | |
211 | */ | |
212 | ch = getchar(); /*bump the #*/ | |
213 | while (INCHARSET(ch, SPACE)) | |
214 | ch = getchar();/*bump white */ | |
215 | if (INCHARSET(ch, DIGIT)){ | |
216 | intval = 0; | |
217 | while(INCHARSET(ch, DIGIT)){ | |
218 | intval = intval*10 + ch - '0'; | |
219 | ch = getchar(); | |
220 | } | |
221 | while (INCHARSET(ch, SPACE)) | |
222 | ch = getchar(); | |
223 | if (ch == '"'){ | |
224 | ptoken(bufptr, ILINENO); | |
225 | ptoken(bufptr, INT); | |
226 | pint(bufptr, intval - 1); | |
227 | ptoken(bufptr, IFILE); | |
228 | /* | |
229 | * The '"' has already been | |
230 | * munched | |
231 | * | |
232 | * eatstr will not eat | |
233 | * the trailing \n, so | |
234 | * it is given to the parser | |
235 | * and counted. | |
236 | */ | |
237 | goto eatstr; | |
238 | } | |
239 | } | |
240 | } | |
241 | /* | |
242 | * Well, its just an ordinary decadent comment | |
243 | */ | |
244 | while ((ch != '\n') && (ch != EOFCHAR)) | |
245 | ch = getchar(); | |
246 | if (ch == EOFCHAR) | |
247 | goto endoffile; | |
248 | val = ryylval = oval = NL; | |
249 | scanlineno++; | |
250 | goto ret; | |
251 | ||
252 | case NL: | |
253 | scanlineno++; | |
254 | val = ryylval; | |
255 | goto ret; | |
256 | ||
257 | case SP: | |
258 | oval = SP; /*invalidate ^# meta comments*/ | |
259 | goto loop; | |
260 | ||
261 | case REGOP: /* % , could be used as modulo, or register*/ | |
262 | ch = getchar(); | |
263 | if (INCHARSET(ch, DIGIT)){ | |
264 | ryylval = ch-'0'; | |
265 | if (ch=='1') { | |
266 | if (INCHARSET( (ch = getchar()), REGDIGIT)) | |
267 | ryylval = 10+ch-'0'; | |
268 | else | |
269 | ungetc(ch); | |
270 | } | |
271 | /* | |
272 | * God only knows what the original author | |
273 | * wanted this undocumented feature to | |
274 | * do. | |
275 | * %5++ is really r7 | |
276 | */ | |
277 | while(INCHARSET( (ch = getchar()), SIGN)) { | |
278 | if (ch=='+') | |
279 | ryylval++; | |
280 | else | |
281 | ryylval--; | |
282 | } | |
283 | ungetc(ch); | |
284 | val = REG; | |
285 | } else { | |
286 | ungetc(ch); | |
287 | val = REGOP; | |
288 | } | |
289 | goto ret; | |
290 | ||
291 | case ALPH: | |
292 | ch1 = ch; | |
293 | if (INCHARSET(ch, SZSPECBEGIN)){ | |
294 | if( (ch = getchar()) == '`' || ch == '^'){ | |
295 | ch1 |= 0100; /*convert to lower*/ | |
296 | switch(ch1){ | |
297 | case 'b': ryylval = 1; break; | |
298 | case 'w': ryylval = 2; break; | |
299 | case 'l': ryylval = 4; break; | |
300 | default: ryylval = d124; break; | |
301 | } | |
302 | val = SIZESPEC; | |
303 | goto ret; | |
304 | } else { | |
305 | ungetc(ch); | |
306 | ch = ch1; /*restore first character*/ | |
307 | } | |
308 | } | |
309 | rcp = yytext; | |
310 | do { | |
311 | if (rcp < &yytext[NCPS]) | |
312 | *rcp++ = ch; | |
313 | } while (INCHARSET ( (ch = getchar()), ALPHA | DIGIT)); | |
314 | *rcp = '\0'; | |
315 | while (INCHARSET(ch, SPACE)) | |
316 | ch = getchar(); | |
317 | ungetc(ch); | |
318 | ||
319 | switch((op = *lookup(1))->s_tag){ | |
320 | case 0: | |
321 | case LABELID: | |
322 | /* | |
323 | * Its a name... (Labels are subsets ofname) | |
324 | */ | |
325 | ryylval = (int)op; | |
326 | val = NAME; | |
327 | break; | |
328 | case INST0: | |
329 | case INSTn: | |
330 | case IJXXX: | |
331 | opstruct.Op_popcode = ( (struct instab *)op)->i_popcode; | |
332 | opstruct.Op_eopcode = ( (struct instab *)op)->i_eopcode; | |
333 | val = op->s_tag; | |
334 | break; | |
335 | default: | |
336 | ryylval = ( (struct instab *)op)->i_popcode; | |
337 | val = op->s_tag; | |
338 | break; | |
339 | } | |
340 | goto ret; | |
341 | ||
342 | case DIG: | |
343 | /* | |
344 | * Implement call by reference on a reg variable | |
345 | */ | |
346 | cp = inbufptr; | |
347 | val = number(ch, &cp); | |
348 | /* | |
349 | * yylval or yybignum has been stuffed as a side | |
350 | * effect to number(); get the global yylval | |
351 | * into our fast local copy in case it was an INT. | |
352 | */ | |
353 | ryylval = yylval; | |
354 | inbufptr = cp; | |
355 | goto ret; | |
356 | ||
357 | case LSH: | |
358 | case RSH: | |
359 | /* | |
360 | * We allow the C style operators | |
361 | * << and >>, as well as < and > | |
362 | */ | |
363 | if ( (ch1 = getchar()) != ch) | |
364 | ungetc(ch1); | |
365 | val = ryylval; | |
366 | goto ret; | |
367 | ||
368 | case MINUS: | |
369 | if ( (ch = getchar()) =='(') | |
370 | ryylval=val=MP; | |
371 | else { | |
372 | ungetc(ch); | |
373 | val=MINUS; | |
374 | } | |
375 | goto ret; | |
376 | ||
377 | case SQ: | |
378 | if ((ryylval = getchar()) == '\n') | |
379 | scanlineno++; /*not entirely correct*/ | |
380 | val = INT; | |
381 | goto ret; | |
382 | ||
383 | case DQ: | |
384 | eatstr: | |
385 | linescrossed = 0; | |
ad67d5e5 RH |
386 | for(rcp = yytext, maxstrlg = NCPS; maxstrlg > 0; --maxstrlg){ |
387 | switch(ch = getchar()){ | |
388 | case '"': | |
389 | goto tailDQ; | |
390 | default: | |
391 | stuff: | |
392 | pchar(rcp, ch); | |
393 | break; | |
394 | case '\n': | |
395 | yywarning("New line in a string constant"); | |
3bd55f52 RH |
396 | scanlineno++; |
397 | linescrossed++; | |
398 | ch = getchar(); | |
ad67d5e5 RH |
399 | switch(ch){ |
400 | case NEEDCHAR: | |
401 | if ( (inbufptr = fillinbuffer()) != 0){ | |
402 | ch = '\n'; | |
403 | goto stuff; | |
404 | } | |
405 | /*FALLTHROUGH*/ | |
406 | case EOFCHAR: | |
407 | pchar(rcp, '\n'); | |
3bd55f52 | 408 | ungetc(EOFCHAR); |
ad67d5e5 RH |
409 | goto tailDQ; |
410 | default: | |
3bd55f52 RH |
411 | ungetc(ch); |
412 | ch = '\n'; | |
413 | goto stuff; | |
414 | } | |
ad67d5e5 RH |
415 | break; |
416 | ||
417 | case '\\': | |
3bd55f52 RH |
418 | ch = getchar(); /*skip the '\\'*/ |
419 | if ( INCHARSET(ch, BSESCAPE)){ | |
420 | switch (ch){ | |
421 | case 'b': ch = '\b'; goto stuff; | |
422 | case 'f': ch = '\f'; goto stuff; | |
423 | case 'n': ch = '\n'; goto stuff; | |
424 | case 'r': ch = '\r'; goto stuff; | |
425 | case 't': ch = '\t'; goto stuff; | |
426 | } | |
427 | } | |
ad67d5e5 RH |
428 | if ( !(INCHARSET(ch, OCTDIGIT)) ) |
429 | goto stuff; | |
3bd55f52 RH |
430 | i = 0; |
431 | intval = 0; | |
432 | while ( (i < 3) && (INCHARSET(ch, OCTDIGIT))){ | |
ad67d5e5 RH |
433 | i++; |
434 | intval <<= 3; | |
435 | intval += ch - '0'; | |
3bd55f52 RH |
436 | ch = getchar(); |
437 | } | |
438 | ungetc(ch); | |
8a7e35d9 | 439 | ch = (char)intval; |
3bd55f52 | 440 | goto stuff; |
ad67d5e5 | 441 | } |
3bd55f52 | 442 | } |
ad67d5e5 | 443 | tailDQ: ; |
3bd55f52 | 444 | /* |
ad67d5e5 | 445 | * account for any lines that were crossed |
3bd55f52 | 446 | */ |
3bd55f52 | 447 | if (linescrossed){ |
ad67d5e5 RH |
448 | ptoken(bufptr, ILINESKIP); |
449 | pint(bufptr, linescrossed); | |
450 | } | |
451 | /* | |
452 | * put the string in yytext into the string pool | |
453 | * | |
454 | * The value in ryylval points to the string; | |
455 | * the previous 2 bytes is the length of the string | |
456 | * | |
457 | * Cheat: append a trailing null to the string | |
458 | * and then adjust the string length to ignore | |
459 | * the trailing null. If any STRING client requires | |
460 | * the trailing null, the client can just change STRLEN | |
461 | */ | |
462 | val = STRING; | |
463 | *rcp++ = 0; | |
464 | ryylval = (int)savestr(yytext, rcp - yytext); | |
465 | STRLEN(((char *)ryylval)) -= 1; | |
466 | goto ret; | |
3bd55f52 RH |
467 | |
468 | case BADCHAR: | |
469 | linescrossed = lineno; | |
470 | lineno = scanlineno; | |
471 | yyerror("Illegal character mapped: %d, char read:(octal) %o", | |
472 | ryylval, ch); | |
473 | lineno = linescrossed; | |
474 | val = BADCHAR; | |
475 | goto ret; | |
476 | ||
477 | default: | |
478 | val = ryylval; | |
479 | goto ret; | |
480 | } /*end of the switch*/ | |
481 | /* | |
482 | * here with one token, so stuff it | |
483 | */ | |
484 | ret: | |
485 | oval = val; | |
486 | ptoken(bufptr, val); | |
487 | switch(val){ | |
488 | case ILINESKIP: | |
489 | pint(bufptr, ryylval); | |
490 | break; | |
491 | case SIZESPEC: | |
492 | pchar(bufptr, ryylval); | |
493 | break; | |
494 | case BFINT: plong(bufptr, ryylval); | |
495 | break; | |
496 | case INT: plong(bufptr, ryylval); | |
497 | break; | |
498 | case BIGNUM: pnumber(bufptr, yybignum); | |
499 | break; | |
ad67d5e5 RH |
500 | case STRING: pptr(bufptr, (int)(char *)ryylval); |
501 | break; | |
3bd55f52 RH |
502 | case NAME: pptr(bufptr, (int)(struct symtab *)ryylval); |
503 | break; | |
504 | case REG: pchar(bufptr, ryylval); | |
505 | break; | |
506 | case INST0: | |
507 | case INSTn: | |
508 | popcode(bufptr, opstruct); | |
509 | break; | |
510 | case IJXXX: | |
511 | popcode(bufptr, opstruct); | |
512 | pptr(bufptr, (int)(struct symtab *)symalloc()); | |
513 | break; | |
514 | case ISTAB: | |
515 | case ISTABSTR: | |
516 | case ISTABNONE: | |
517 | case ISTABDOT: | |
518 | case IALIGN: | |
519 | pptr(bufptr, (int)(struct symtab *)symalloc()); | |
520 | break; | |
521 | /* | |
522 | * default: | |
523 | */ | |
524 | } | |
525 | builtval: ; | |
526 | } /*end of the while to stuff the buffer*/ | |
527 | done: | |
528 | bufferbox->tok_count = (bytetoktype *)bufptr - &(bufferbox->toks[0]); | |
529 | ||
530 | /* | |
531 | * This is a real kludge: | |
532 | * | |
533 | * We put the last token in the buffer to be a MINUS | |
534 | * symbol. This last token will never be picked up | |
535 | * in the normal way, but can be looked at during | |
536 | * a peekahead look that the short circuit expression | |
537 | * evaluator uses to see if an expression is complicated. | |
538 | * | |
539 | * Consider the following situation: | |
540 | * | |
541 | * .word 45 + 47 | |
542 | * buffer 1 | buffer 0 | |
543 | * the peekahead would want to look across the buffer, | |
544 | * but will look in the buffer end zone, see the minus, and | |
545 | * fail. | |
546 | */ | |
547 | ptoken(bufptr, MINUS); | |
548 | InBufPtr = inbufptr; /*copy this back*/ | |
549 | } |