Annotation of mandoc/read.c, Revision 1.2
1.2 ! kristaps 1: /* $Id: read.c,v 1.1 2011/03/20 11:41:24 kristaps Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4: * Copyright (c) 2010, 2011 Ingo Schwarze <schwarze@openbsd.org>
5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
1.2 ! kristaps 18: #include <sys/stat.h>
1.1 kristaps 19: #include <sys/mman.h>
20:
21: #include <assert.h>
22: #include <ctype.h>
23: #include <fcntl.h>
24: #include <stdio.h>
25: #include <stdlib.h>
26: #include <string.h>
27: #include <unistd.h>
28:
29: #include "mandoc.h"
30: #include "mdoc.h"
31: #include "man.h"
32: #include "roff.h"
33:
34: #ifndef MAP_FILE
35: #define MAP_FILE 0
36: #endif
37:
38: #define REPARSE_LIMIT 1000
39:
40: struct buf {
41: char *buf; /* binary input buffer */
42: size_t sz; /* size of binary buffer */
43: };
44:
45: struct mparse {
46: enum mandoclevel file_status; /* status of current parse */
47: int line; /* line number in the file */
48: enum mparset inttype; /* which parser to use */
49: struct man *pman; /* persistent man parser */
50: struct mdoc *pmdoc; /* persistent mdoc parser */
51: struct man *man; /* man parser */
52: struct mdoc *mdoc; /* mdoc parser */
53: struct roff *roff; /* roff parser (!NULL) */
54: struct regset regs; /* roff registers */
55: int reparse_count; /* finite interp. stack */
56: mandocmsg mmsg; /* warning/error message handler */
57: void *arg; /* argument to mmsg */
58: mevt_open evt_open; /* file-open event */
59: mevt_close evt_close; /* file-close event */
60: const char *svfile;
61: };
62:
63: static void resize_buf(struct buf *, size_t);
64: static void mparse_buf_r(struct mparse *, struct buf, int);
65: static void mparse_readfd_r(struct mparse *, int, const char *, int);
66: static void pset(const char *, int, struct mparse *);
67: static void pdesc(struct mparse *, const char *, int);
68: static int read_whole_file(const char *, int, struct buf *, int *);
69: static void mparse_end(struct mparse *);
70:
71: static void
72: resize_buf(struct buf *buf, size_t initial)
73: {
74:
75: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
76: buf->buf = mandoc_realloc(buf->buf, buf->sz);
77: }
78:
79: static void
80: pset(const char *buf, int pos, struct mparse *curp)
81: {
82: int i;
83:
84: /*
85: * Try to intuit which kind of manual parser should be used. If
86: * passed in by command-line (-man, -mdoc), then use that
87: * explicitly. If passed as -mandoc, then try to guess from the
88: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
89: * default to -man, which is more lenient.
90: *
91: * Separate out pmdoc/pman from mdoc/man: the first persists
92: * through all parsers, while the latter is used per-parse.
93: */
94:
95: if ('.' == buf[0] || '\'' == buf[0]) {
96: for (i = 1; buf[i]; i++)
97: if (' ' != buf[i] && '\t' != buf[i])
98: break;
99: if ('\0' == buf[i])
100: return;
101: }
102:
103: switch (curp->inttype) {
104: case (MPARSE_MDOC):
105: if (NULL == curp->pmdoc)
106: curp->pmdoc = mdoc_alloc
107: (&curp->regs, curp->arg, curp->mmsg);
108: assert(curp->pmdoc);
109: curp->mdoc = curp->pmdoc;
110: return;
111: case (MPARSE_MAN):
112: if (NULL == curp->pman)
113: curp->pman = man_alloc
114: (&curp->regs, curp->arg, curp->mmsg);
115: assert(curp->pman);
116: curp->man = curp->pman;
117: return;
118: default:
119: break;
120: }
121:
122: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
123: if (NULL == curp->pmdoc)
124: curp->pmdoc = mdoc_alloc
125: (&curp->regs, curp->arg, curp->mmsg);
126: assert(curp->pmdoc);
127: curp->mdoc = curp->pmdoc;
128: return;
129: }
130:
131: if (NULL == curp->pman)
132: curp->pman = man_alloc
133: (&curp->regs, curp->arg, curp->mmsg);
134: assert(curp->pman);
135: curp->man = curp->pman;
136: }
137:
138: /*
139: * Main parse routine for an opened file. This is called for each
140: * opened file and simply loops around the full input file, possibly
141: * nesting (i.e., with `so').
142: */
143: static void
144: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
145: {
146: const struct tbl_span *span;
147: struct buf ln;
148: enum rofferr rr;
149: int i, of, rc;
150: int pos; /* byte number in the ln buffer */
151: int lnn; /* line number in the real file */
152: unsigned char c;
153:
154: memset(&ln, 0, sizeof(struct buf));
155:
156: lnn = curp->line;
157: pos = 0;
158:
159: for (i = 0; i < (int)blk.sz; ) {
160: if (0 == pos && '\0' == blk.buf[i])
161: break;
162:
163: if (start) {
164: curp->line = lnn;
165: curp->reparse_count = 0;
166: }
167:
168: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
169:
170: /*
171: * When finding an unescaped newline character,
172: * leave the character loop to process the line.
173: * Skip a preceding carriage return, if any.
174: */
175:
176: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
177: '\n' == blk.buf[i + 1])
178: ++i;
179: if ('\n' == blk.buf[i]) {
180: ++i;
181: ++lnn;
182: break;
183: }
184:
185: /*
186: * Warn about bogus characters. If you're using
187: * non-ASCII encoding, you're screwing your
188: * readers. Since I'd rather this not happen,
189: * I'll be helpful and drop these characters so
190: * we don't display gibberish. Note to manual
191: * writers: use special characters.
192: */
193:
194: c = (unsigned char) blk.buf[i];
195:
196: if ( ! (isascii(c) &&
197: (isgraph(c) || isblank(c)))) {
198: curp->mmsg(MANDOCERR_BADCHAR, curp->arg,
199: curp->line, pos, "ignoring byte");
200: i++;
201: continue;
202: }
203:
204: /* Trailing backslash = a plain char. */
205:
206: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
207: if (pos >= (int)ln.sz)
208: resize_buf(&ln, 256);
209: ln.buf[pos++] = blk.buf[i++];
210: continue;
211: }
212:
213: /*
214: * Found escape and at least one other character.
215: * When it's a newline character, skip it.
216: * When there is a carriage return in between,
217: * skip that one as well.
218: */
219:
220: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
221: '\n' == blk.buf[i + 2])
222: ++i;
223: if ('\n' == blk.buf[i + 1]) {
224: i += 2;
225: ++lnn;
226: continue;
227: }
228:
229: if ('"' == blk.buf[i + 1]) {
230: i += 2;
231: /* Comment, skip to end of line */
232: for (; i < (int)blk.sz; ++i) {
233: if ('\n' == blk.buf[i]) {
234: ++i;
235: ++lnn;
236: break;
237: }
238: }
239:
240: /* Backout trailing whitespaces */
241: for (; pos > 0; --pos) {
242: if (ln.buf[pos - 1] != ' ')
243: break;
244: if (pos > 2 && ln.buf[pos - 2] == '\\')
245: break;
246: }
247: break;
248: }
249:
250: /* Some other escape sequence, copy & cont. */
251:
252: if (pos + 1 >= (int)ln.sz)
253: resize_buf(&ln, 256);
254:
255: ln.buf[pos++] = blk.buf[i++];
256: ln.buf[pos++] = blk.buf[i++];
257: }
258:
259: if (pos >= (int)ln.sz)
260: resize_buf(&ln, 256);
261:
262: ln.buf[pos] = '\0';
263:
264: /*
265: * A significant amount of complexity is contained by
266: * the roff preprocessor. It's line-oriented but can be
267: * expressed on one line, so we need at times to
268: * readjust our starting point and re-run it. The roff
269: * preprocessor can also readjust the buffers with new
270: * data, so we pass them in wholesale.
271: */
272:
273: of = 0;
274:
275: rerun:
276: rr = roff_parseln
277: (curp->roff, curp->line,
278: &ln.buf, &ln.sz, of, &of);
279:
280: switch (rr) {
281: case (ROFF_REPARSE):
282: if (REPARSE_LIMIT >= ++curp->reparse_count)
283: mparse_buf_r(curp, ln, 0);
284: else
285: curp->mmsg(MANDOCERR_ROFFLOOP, curp->arg,
286: curp->line, pos, NULL);
287: pos = 0;
288: continue;
289: case (ROFF_APPEND):
290: pos = (int)strlen(ln.buf);
291: continue;
292: case (ROFF_RERUN):
293: goto rerun;
294: case (ROFF_IGN):
295: pos = 0;
296: continue;
297: case (ROFF_ERR):
298: assert(MANDOCLEVEL_FATAL <= curp->file_status);
299: break;
300: case (ROFF_SO):
301: mparse_readfd_r(curp, -1, ln.buf + of, 1);
302: if (MANDOCLEVEL_FATAL <= curp->file_status)
303: break;
304: pos = 0;
305: continue;
306: default:
307: break;
308: }
309:
310: /*
311: * If we encounter errors in the recursive parse, make
312: * sure we don't continue parsing.
313: */
314:
315: if (MANDOCLEVEL_FATAL <= curp->file_status)
316: break;
317:
318: /*
319: * If input parsers have not been allocated, do so now.
320: * We keep these instanced betwen parsers, but set them
321: * locally per parse routine since we can use different
322: * parsers with each one.
323: */
324:
325: if ( ! (curp->man || curp->mdoc))
326: pset(ln.buf + of, pos - of, curp);
327:
328: /*
329: * Lastly, push down into the parsers themselves. One
330: * of these will have already been set in the pset()
331: * routine.
332: * If libroff returns ROFF_TBL, then add it to the
333: * currently open parse. Since we only get here if
334: * there does exist data (see tbl_data.c), we're
335: * guaranteed that something's been allocated.
336: * Do the same for ROFF_EQN.
337: */
338:
339: rc = -1;
340:
341: if (ROFF_TBL == rr)
342: while (NULL != (span = roff_span(curp->roff))) {
343: rc = curp->man ?
344: man_addspan(curp->man, span) :
345: mdoc_addspan(curp->mdoc, span);
346: if (0 == rc)
347: break;
348: }
349: else if (ROFF_EQN == rr)
350: rc = curp->mdoc ?
351: mdoc_addeqn(curp->mdoc,
352: roff_eqn(curp->roff)) :
353: man_addeqn(curp->man,
354: roff_eqn(curp->roff));
355: else if (curp->man || curp->mdoc)
356: rc = curp->man ?
357: man_parseln(curp->man,
358: curp->line, ln.buf, of) :
359: mdoc_parseln(curp->mdoc,
360: curp->line, ln.buf, of);
361:
362: if (0 == rc) {
363: assert(MANDOCLEVEL_FATAL <= curp->file_status);
364: break;
365: }
366:
367: /* Temporary buffers typically are not full. */
368:
369: if (0 == start && '\0' == blk.buf[i])
370: break;
371:
372: /* Start the next input line. */
373:
374: pos = 0;
375: }
376:
377: free(ln.buf);
378: }
379:
380: static void
381: pdesc(struct mparse *curp, const char *file, int fd)
382: {
383: struct buf blk;
384: int with_mmap;
385:
386: /*
387: * Run for each opened file; may be called more than once for
388: * each full parse sequence if the opened file is nested (i.e.,
389: * from `so'). Simply sucks in the whole file and moves into
390: * the parse phase for the file.
391: */
392:
393: if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
394: curp->file_status = MANDOCLEVEL_SYSERR;
395: return;
396: }
397:
398: /* Line number is per-file. */
399:
400: curp->line = 1;
401:
402: mparse_buf_r(curp, blk, 1);
403:
404: if (with_mmap)
405: munmap(blk.buf, blk.sz);
406: else
407: free(blk.buf);
408: }
409:
410: static int
411: read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
412: {
413: struct stat st;
414: size_t off;
415: ssize_t ssz;
416:
417: if (-1 == fstat(fd, &st)) {
418: perror(file);
419: return(0);
420: }
421:
422: /*
423: * If we're a regular file, try just reading in the whole entry
424: * via mmap(). This is faster than reading it into blocks, and
425: * since each file is only a few bytes to begin with, I'm not
426: * concerned that this is going to tank any machines.
427: */
428:
429: if (S_ISREG(st.st_mode)) {
430: if (st.st_size >= (1U << 31)) {
431: fprintf(stderr, "%s: input too large\n", file);
432: return(0);
433: }
434: *with_mmap = 1;
435: fb->sz = (size_t)st.st_size;
436: fb->buf = mmap(NULL, fb->sz, PROT_READ,
437: MAP_FILE|MAP_SHARED, fd, 0);
438: if (fb->buf != MAP_FAILED)
439: return(1);
440: }
441:
442: /*
443: * If this isn't a regular file (like, say, stdin), then we must
444: * go the old way and just read things in bit by bit.
445: */
446:
447: *with_mmap = 0;
448: off = 0;
449: fb->sz = 0;
450: fb->buf = NULL;
451: for (;;) {
452: if (off == fb->sz) {
453: if (fb->sz == (1U << 31)) {
454: fprintf(stderr, "%s: input too large\n", file);
455: break;
456: }
457: resize_buf(fb, 65536);
458: }
459: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
460: if (ssz == 0) {
461: fb->sz = off;
462: return(1);
463: }
464: if (ssz == -1) {
465: perror(file);
466: break;
467: }
468: off += (size_t)ssz;
469: }
470:
471: free(fb->buf);
472: fb->buf = NULL;
473: return(0);
474: }
475:
476: static void
477: mparse_end(struct mparse *curp)
478: {
479:
480: if (MANDOCLEVEL_FATAL <= curp->file_status)
481: return;
482:
483: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
484: assert(MANDOCLEVEL_FATAL <= curp->file_status);
485: return;
486: }
487:
488: if (curp->man && ! man_endparse(curp->man)) {
489: assert(MANDOCLEVEL_FATAL <= curp->file_status);
490: return;
491: }
492:
493: #if 0
494: /* NOTE a parser may not have been assigned, yet. */
495:
496: if ( ! (curp->man || curp->mdoc)) {
497: /* FIXME: make into an mandoc.h error. */
498: fprintf(stderr, "%s: Not a manual\n", curp->file);
499: curp->file_status = MANDOCLEVEL_FATAL;
500: goto cleanup;
501: }
502: #endif
503:
504: roff_endparse(curp->roff);
505: }
506:
507: static void
508: mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
509: {
510: const char *svfile;
511:
512: if ( ! (*curp->evt_open)(curp->arg, file)) {
513: curp->file_status = MANDOCLEVEL_SYSERR;
514: return;
515: }
516:
517: if (-1 == fd)
518: if (-1 == (fd = open(file, O_RDONLY, 0))) {
519: perror(file);
520: curp->file_status = MANDOCLEVEL_SYSERR;
521: return;
522: }
523:
524: svfile = curp->svfile;
525: curp->svfile = file;
526:
527: pdesc(curp, file, fd);
528:
529: if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
530: mparse_end(curp);
531:
532: if (STDIN_FILENO != fd && -1 == close(fd))
533: perror(file);
534:
535: (*curp->evt_close)(curp->arg, svfile);
536: curp->svfile = svfile;
537: }
538:
539: enum mandoclevel
540: mparse_readfd(struct mparse *curp, int fd, const char *file)
541: {
542:
543: mparse_readfd_r(curp, fd, file, 0);
544: return(curp->file_status);
545: }
546:
547: void
548: mparse_setstatus(struct mparse *curp, enum mandoclevel lvl)
549: {
550:
551: if (curp->file_status < lvl)
552: curp->file_status = lvl;
553: }
554:
555: struct mparse *
556: mparse_alloc(enum mparset inttype, mevt_open eopen,
557: mevt_close eclose, mandocmsg mmsg, void *arg)
558: {
559: struct mparse *curp;
560:
561: curp = mandoc_calloc(1, sizeof(struct mparse));
562:
563: curp->mmsg = mmsg;
564: curp->arg = arg;
565: curp->inttype = inttype;
566: curp->evt_open = eopen;
567: curp->evt_close = eclose;
568:
569: curp->roff = roff_alloc(&curp->regs, arg, mmsg);
570: return(curp);
571: }
572:
573: void
574: mparse_reset(struct mparse *curp)
575: {
576:
577: memset(&curp->regs, 0, sizeof(struct regset));
578:
579: roff_reset(curp->roff);
580:
581: if (curp->mdoc)
582: mdoc_reset(curp->mdoc);
583: if (curp->man)
584: man_reset(curp->man);
585:
586: curp->file_status = MANDOCLEVEL_OK;
587: curp->mdoc = NULL;
588: curp->man = NULL;
589: }
590:
591: void
592: mparse_free(struct mparse *curp)
593: {
594:
595: if (curp->pmdoc)
596: mdoc_free(curp->pmdoc);
597: if (curp->pman)
598: man_free(curp->pman);
599: if (curp->roff)
600: roff_free(curp->roff);
601:
602: free(curp);
603: }
604:
605: void
606: mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
607: {
608:
609: *mdoc = curp->mdoc;
610: *man = curp->man;
611: }
CVSweb