Annotation of src/usr.bin/mandoc/read.c, Revision 1.2
1.2 ! schwarze 1: /* $Id: read.c,v 1.1 2011/04/24 16:22:02 schwarze Exp $ */
1.1 schwarze 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: */
18: #include <sys/stat.h>
19: #include <sys/mman.h>
20:
21: #include <assert.h>
22: #include <ctype.h>
23: #include <fcntl.h>
24: #include <stdarg.h>
25: #include <stdio.h>
26: #include <stdlib.h>
27: #include <string.h>
28: #include <unistd.h>
29:
30: #include "mandoc.h"
31: #include "libmandoc.h"
32: #include "mdoc.h"
33: #include "man.h"
34:
35: #define REPARSE_LIMIT 1000
36:
37: struct buf {
38: char *buf; /* binary input buffer */
39: size_t sz; /* size of binary buffer */
40: };
41:
42: struct mparse {
43: enum mandoclevel file_status; /* status of current parse */
44: enum mandoclevel wlevel; /* ignore messages below this */
45: int line; /* line number in the file */
46: enum mparset inttype; /* which parser to use */
47: struct man *pman; /* persistent man parser */
48: struct mdoc *pmdoc; /* persistent mdoc parser */
49: struct man *man; /* man parser */
50: struct mdoc *mdoc; /* mdoc parser */
51: struct roff *roff; /* roff parser (!NULL) */
52: struct regset regs; /* roff registers */
53: int reparse_count; /* finite interp. stack */
54: mandocmsg mmsg; /* warning/error message handler */
55: void *arg; /* argument to mmsg */
56: const char *file;
57: };
58:
59: static void resize_buf(struct buf *, size_t);
60: static void mparse_buf_r(struct mparse *, struct buf, int);
61: static void mparse_readfd_r(struct mparse *, int, const char *, int);
62: static void pset(const char *, int, struct mparse *);
63: static void pdesc(struct mparse *, const char *, int);
64: static int read_whole_file(const char *, int, struct buf *, int *);
65: static void mparse_end(struct mparse *);
66:
67: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
68: MANDOCERR_OK,
69: MANDOCERR_WARNING,
70: MANDOCERR_WARNING,
71: MANDOCERR_ERROR,
72: MANDOCERR_FATAL,
73: MANDOCERR_MAX,
74: MANDOCERR_MAX
75: };
76:
77: static const char * const mandocerrs[MANDOCERR_MAX] = {
78: "ok",
79:
80: "generic warning",
81:
82: /* related to the prologue */
83: "no title in document",
84: "document title should be all caps",
85: "unknown manual section",
86: "date missing, using today's date",
87: "cannot parse date, using it verbatim",
88: "prologue macros out of order",
89: "duplicate prologue macro",
90: "macro not allowed in prologue",
91: "macro not allowed in body",
92:
93: /* related to document structure */
94: ".so is fragile, better use ln(1)",
95: "NAME section must come first",
96: "bad NAME section contents",
97: "manual name not yet set",
98: "sections out of conventional order",
99: "duplicate section name",
100: "section not in conventional manual section",
101:
102: /* related to macros and nesting */
103: "skipping obsolete macro",
104: "skipping paragraph macro",
105: "skipping no-space macro",
106: "blocks badly nested",
107: "child violates parent syntax",
108: "nested displays are not portable",
109: "already in literal mode",
110: "line scope broken",
111:
112: /* related to missing macro arguments */
113: "skipping empty macro",
114: "argument count wrong",
115: "missing display type",
116: "list type must come first",
117: "tag lists require a width argument",
118: "missing font type",
119: "skipping end of block that is not open",
120:
121: /* related to bad macro arguments */
122: "skipping argument",
123: "duplicate argument",
124: "duplicate display type",
125: "duplicate list type",
126: "unknown AT&T UNIX version",
127: "bad Boolean value",
128: "unknown font",
129: "unknown standard specifier",
130: "bad width argument",
131:
132: /* related to plain text */
133: "blank line in non-literal context",
134: "tab in non-literal context",
135: "end of line whitespace",
136: "bad comment style",
1.2 ! schwarze 137: "bad escape sequence",
1.1 schwarze 138: "unterminated quoted string",
139:
140: "generic error",
141:
142: /* related to tables */
143: "bad table syntax",
144: "bad table option",
145: "bad table layout",
146: "no table layout cells specified",
147: "no table data cells specified",
148: "ignore data in cell",
149: "data block still open",
150: "ignoring extra data cells",
151:
152: "input stack limit exceeded, infinite loop?",
153: "skipping bad character",
154: "escaped character not allowed in a name",
155: "skipping text before the first section header",
156: "skipping unknown macro",
157: "NOT IMPLEMENTED, please use groff: skipping request",
158: "argument count wrong",
159: "skipping end of block that is not open",
160: "missing end of block",
161: "scope open on exit",
162: "uname(3) system call failed",
163: "macro requires line argument(s)",
164: "macro requires body argument(s)",
165: "macro requires argument(s)",
166: "missing list type",
167: "line argument(s) will be lost",
168: "body argument(s) will be lost",
169:
170: "generic fatal error",
171:
172: "not a manual",
173: "column syntax is inconsistent",
174: "NOT IMPLEMENTED: .Bd -file",
175: "line scope broken, syntax violated",
176: "argument count wrong, violates syntax",
177: "child violates parent syntax",
178: "argument count wrong, violates syntax",
179: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
180: "no document body",
181: "no document prologue",
182: "static buffer exhausted",
183: };
184:
185: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
186: "SUCCESS",
187: "RESERVED",
188: "WARNING",
189: "ERROR",
190: "FATAL",
191: "BADARG",
192: "SYSERR"
193: };
194:
195: static void
196: resize_buf(struct buf *buf, size_t initial)
197: {
198:
199: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
200: buf->buf = mandoc_realloc(buf->buf, buf->sz);
201: }
202:
203: static void
204: pset(const char *buf, int pos, struct mparse *curp)
205: {
206: int i;
207:
208: /*
209: * Try to intuit which kind of manual parser should be used. If
210: * passed in by command-line (-man, -mdoc), then use that
211: * explicitly. If passed as -mandoc, then try to guess from the
212: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
213: * default to -man, which is more lenient.
214: *
215: * Separate out pmdoc/pman from mdoc/man: the first persists
216: * through all parsers, while the latter is used per-parse.
217: */
218:
219: if ('.' == buf[0] || '\'' == buf[0]) {
220: for (i = 1; buf[i]; i++)
221: if (' ' != buf[i] && '\t' != buf[i])
222: break;
223: if ('\0' == buf[i])
224: return;
225: }
226:
227: switch (curp->inttype) {
228: case (MPARSE_MDOC):
229: if (NULL == curp->pmdoc)
230: curp->pmdoc = mdoc_alloc(&curp->regs, curp);
231: assert(curp->pmdoc);
232: curp->mdoc = curp->pmdoc;
233: return;
234: case (MPARSE_MAN):
235: if (NULL == curp->pman)
236: curp->pman = man_alloc(&curp->regs, curp);
237: assert(curp->pman);
238: curp->man = curp->pman;
239: return;
240: default:
241: break;
242: }
243:
244: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
245: if (NULL == curp->pmdoc)
246: curp->pmdoc = mdoc_alloc(&curp->regs, curp);
247: assert(curp->pmdoc);
248: curp->mdoc = curp->pmdoc;
249: return;
250: }
251:
252: if (NULL == curp->pman)
253: curp->pman = man_alloc(&curp->regs, curp);
254: assert(curp->pman);
255: curp->man = curp->pman;
256: }
257:
258: /*
259: * Main parse routine for an opened file. This is called for each
260: * opened file and simply loops around the full input file, possibly
261: * nesting (i.e., with `so').
262: */
263: static void
264: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
265: {
266: const struct tbl_span *span;
267: struct buf ln;
268: enum rofferr rr;
269: int i, of, rc;
270: int pos; /* byte number in the ln buffer */
271: int lnn; /* line number in the real file */
272: unsigned char c;
273:
274: memset(&ln, 0, sizeof(struct buf));
275:
276: lnn = curp->line;
277: pos = 0;
278:
279: for (i = 0; i < (int)blk.sz; ) {
280: if (0 == pos && '\0' == blk.buf[i])
281: break;
282:
283: if (start) {
284: curp->line = lnn;
285: curp->reparse_count = 0;
286: }
287:
288: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
289:
290: /*
291: * When finding an unescaped newline character,
292: * leave the character loop to process the line.
293: * Skip a preceding carriage return, if any.
294: */
295:
296: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
297: '\n' == blk.buf[i + 1])
298: ++i;
299: if ('\n' == blk.buf[i]) {
300: ++i;
301: ++lnn;
302: break;
303: }
304:
305: /*
306: * Warn about bogus characters. If you're using
307: * non-ASCII encoding, you're screwing your
308: * readers. Since I'd rather this not happen,
309: * I'll be helpful and drop these characters so
310: * we don't display gibberish. Note to manual
311: * writers: use special characters.
312: */
313:
314: c = (unsigned char) blk.buf[i];
315:
316: if ( ! (isascii(c) &&
317: (isgraph(c) || isblank(c)))) {
318: mandoc_msg(MANDOCERR_BADCHAR, curp,
319: curp->line, pos, "ignoring byte");
320: i++;
321: continue;
322: }
323:
324: /* Trailing backslash = a plain char. */
325:
326: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
327: if (pos >= (int)ln.sz)
328: resize_buf(&ln, 256);
329: ln.buf[pos++] = blk.buf[i++];
330: continue;
331: }
332:
333: /*
334: * Found escape and at least one other character.
335: * When it's a newline character, skip it.
336: * When there is a carriage return in between,
337: * skip that one as well.
338: */
339:
340: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
341: '\n' == blk.buf[i + 2])
342: ++i;
343: if ('\n' == blk.buf[i + 1]) {
344: i += 2;
345: ++lnn;
346: continue;
347: }
348:
349: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
350: i += 2;
351: /* Comment, skip to end of line */
352: for (; i < (int)blk.sz; ++i) {
353: if ('\n' == blk.buf[i]) {
354: ++i;
355: ++lnn;
356: break;
357: }
358: }
359:
360: /* Backout trailing whitespaces */
361: for (; pos > 0; --pos) {
362: if (ln.buf[pos - 1] != ' ')
363: break;
364: if (pos > 2 && ln.buf[pos - 2] == '\\')
365: break;
366: }
367: break;
368: }
369:
370: /* Some other escape sequence, copy & cont. */
371:
372: if (pos + 1 >= (int)ln.sz)
373: resize_buf(&ln, 256);
374:
375: ln.buf[pos++] = blk.buf[i++];
376: ln.buf[pos++] = blk.buf[i++];
377: }
378:
379: if (pos >= (int)ln.sz)
380: resize_buf(&ln, 256);
381:
382: ln.buf[pos] = '\0';
383:
384: /*
385: * A significant amount of complexity is contained by
386: * the roff preprocessor. It's line-oriented but can be
387: * expressed on one line, so we need at times to
388: * readjust our starting point and re-run it. The roff
389: * preprocessor can also readjust the buffers with new
390: * data, so we pass them in wholesale.
391: */
392:
393: of = 0;
394:
395: rerun:
396: rr = roff_parseln
397: (curp->roff, curp->line,
398: &ln.buf, &ln.sz, of, &of);
399:
400: switch (rr) {
401: case (ROFF_REPARSE):
402: if (REPARSE_LIMIT >= ++curp->reparse_count)
403: mparse_buf_r(curp, ln, 0);
404: else
405: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
406: curp->line, pos, NULL);
407: pos = 0;
408: continue;
409: case (ROFF_APPEND):
410: pos = (int)strlen(ln.buf);
411: continue;
412: case (ROFF_RERUN):
413: goto rerun;
414: case (ROFF_IGN):
415: pos = 0;
416: continue;
417: case (ROFF_ERR):
418: assert(MANDOCLEVEL_FATAL <= curp->file_status);
419: break;
420: case (ROFF_SO):
421: mparse_readfd_r(curp, -1, ln.buf + of, 1);
422: if (MANDOCLEVEL_FATAL <= curp->file_status)
423: break;
424: pos = 0;
425: continue;
426: default:
427: break;
428: }
429:
430: /*
431: * If we encounter errors in the recursive parse, make
432: * sure we don't continue parsing.
433: */
434:
435: if (MANDOCLEVEL_FATAL <= curp->file_status)
436: break;
437:
438: /*
439: * If input parsers have not been allocated, do so now.
1.2 ! schwarze 440: * We keep these instanced between parsers, but set them
1.1 schwarze 441: * locally per parse routine since we can use different
442: * parsers with each one.
443: */
444:
445: if ( ! (curp->man || curp->mdoc))
446: pset(ln.buf + of, pos - of, curp);
447:
448: /*
449: * Lastly, push down into the parsers themselves. One
450: * of these will have already been set in the pset()
451: * routine.
452: * If libroff returns ROFF_TBL, then add it to the
453: * currently open parse. Since we only get here if
454: * there does exist data (see tbl_data.c), we're
455: * guaranteed that something's been allocated.
456: * Do the same for ROFF_EQN.
457: */
458:
459: rc = -1;
460:
461: if (ROFF_TBL == rr)
462: while (NULL != (span = roff_span(curp->roff))) {
463: rc = curp->man ?
464: man_addspan(curp->man, span) :
465: mdoc_addspan(curp->mdoc, span);
466: if (0 == rc)
467: break;
468: }
469: else if (ROFF_EQN == rr)
470: rc = curp->mdoc ?
471: mdoc_addeqn(curp->mdoc,
472: roff_eqn(curp->roff)) :
473: man_addeqn(curp->man,
474: roff_eqn(curp->roff));
475: else if (curp->man || curp->mdoc)
476: rc = curp->man ?
477: man_parseln(curp->man,
478: curp->line, ln.buf, of) :
479: mdoc_parseln(curp->mdoc,
480: curp->line, ln.buf, of);
481:
482: if (0 == rc) {
483: assert(MANDOCLEVEL_FATAL <= curp->file_status);
484: break;
485: }
486:
487: /* Temporary buffers typically are not full. */
488:
489: if (0 == start && '\0' == blk.buf[i])
490: break;
491:
492: /* Start the next input line. */
493:
494: pos = 0;
495: }
496:
497: free(ln.buf);
498: }
499:
500: static void
501: pdesc(struct mparse *curp, const char *file, int fd)
502: {
503: struct buf blk;
504: int with_mmap;
505:
506: /*
507: * Run for each opened file; may be called more than once for
508: * each full parse sequence if the opened file is nested (i.e.,
509: * from `so'). Simply sucks in the whole file and moves into
510: * the parse phase for the file.
511: */
512:
513: if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
514: curp->file_status = MANDOCLEVEL_SYSERR;
515: return;
516: }
517:
518: /* Line number is per-file. */
519:
520: curp->line = 1;
521:
522: mparse_buf_r(curp, blk, 1);
523:
524: if (with_mmap)
525: munmap(blk.buf, blk.sz);
526: else
527: free(blk.buf);
528: }
529:
530: static int
531: read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
532: {
533: struct stat st;
534: size_t off;
535: ssize_t ssz;
536:
537: if (-1 == fstat(fd, &st)) {
538: perror(file);
539: return(0);
540: }
541:
542: /*
543: * If we're a regular file, try just reading in the whole entry
544: * via mmap(). This is faster than reading it into blocks, and
545: * since each file is only a few bytes to begin with, I'm not
546: * concerned that this is going to tank any machines.
547: */
548:
549: if (S_ISREG(st.st_mode)) {
550: if (st.st_size >= (1U << 31)) {
551: fprintf(stderr, "%s: input too large\n", file);
552: return(0);
553: }
554: *with_mmap = 1;
555: fb->sz = (size_t)st.st_size;
556: fb->buf = mmap(NULL, fb->sz, PROT_READ,
557: MAP_FILE|MAP_SHARED, fd, 0);
558: if (fb->buf != MAP_FAILED)
559: return(1);
560: }
561:
562: /*
563: * If this isn't a regular file (like, say, stdin), then we must
564: * go the old way and just read things in bit by bit.
565: */
566:
567: *with_mmap = 0;
568: off = 0;
569: fb->sz = 0;
570: fb->buf = NULL;
571: for (;;) {
572: if (off == fb->sz) {
573: if (fb->sz == (1U << 31)) {
574: fprintf(stderr, "%s: input too large\n", file);
575: break;
576: }
577: resize_buf(fb, 65536);
578: }
579: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
580: if (ssz == 0) {
581: fb->sz = off;
582: return(1);
583: }
584: if (ssz == -1) {
585: perror(file);
586: break;
587: }
588: off += (size_t)ssz;
589: }
590:
591: free(fb->buf);
592: fb->buf = NULL;
593: return(0);
594: }
595:
596: static void
597: mparse_end(struct mparse *curp)
598: {
599:
600: if (MANDOCLEVEL_FATAL <= curp->file_status)
601: return;
602:
603: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
604: assert(MANDOCLEVEL_FATAL <= curp->file_status);
605: return;
606: }
607:
608: if (curp->man && ! man_endparse(curp->man)) {
609: assert(MANDOCLEVEL_FATAL <= curp->file_status);
610: return;
611: }
612:
613: if ( ! (curp->man || curp->mdoc)) {
614: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
615: curp->file_status = MANDOCLEVEL_FATAL;
616: return;
617: }
618:
619: roff_endparse(curp->roff);
620: }
621:
622: static void
623: mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
624: {
625: const char *svfile;
626:
627: if (-1 == fd)
628: if (-1 == (fd = open(file, O_RDONLY, 0))) {
629: perror(file);
630: curp->file_status = MANDOCLEVEL_SYSERR;
631: return;
632: }
633:
634: svfile = curp->file;
635: curp->file = file;
636:
637: pdesc(curp, file, fd);
638:
639: if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
640: mparse_end(curp);
641:
642: if (STDIN_FILENO != fd && -1 == close(fd))
643: perror(file);
644:
645: curp->file = svfile;
646: }
647:
648: enum mandoclevel
649: mparse_readfd(struct mparse *curp, int fd, const char *file)
650: {
651:
652: mparse_readfd_r(curp, fd, file, 0);
653: return(curp->file_status);
654: }
655:
656: struct mparse *
657: mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
658: {
659: struct mparse *curp;
660:
661: assert(wlevel <= MANDOCLEVEL_FATAL);
662:
663: curp = mandoc_calloc(1, sizeof(struct mparse));
664:
665: curp->wlevel = wlevel;
666: curp->mmsg = mmsg;
667: curp->arg = arg;
668: curp->inttype = inttype;
669:
670: curp->roff = roff_alloc(&curp->regs, curp);
671: return(curp);
672: }
673:
674: void
675: mparse_reset(struct mparse *curp)
676: {
677:
678: memset(&curp->regs, 0, sizeof(struct regset));
679:
680: roff_reset(curp->roff);
681:
682: if (curp->mdoc)
683: mdoc_reset(curp->mdoc);
684: if (curp->man)
685: man_reset(curp->man);
686:
687: curp->file_status = MANDOCLEVEL_OK;
688: curp->mdoc = NULL;
689: curp->man = NULL;
690: }
691:
692: void
693: mparse_free(struct mparse *curp)
694: {
695:
696: if (curp->pmdoc)
697: mdoc_free(curp->pmdoc);
698: if (curp->pman)
699: man_free(curp->pman);
700: if (curp->roff)
701: roff_free(curp->roff);
702:
703: free(curp);
704: }
705:
706: void
707: mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
708: {
709:
710: if (mdoc)
711: *mdoc = curp->mdoc;
712: if (man)
713: *man = curp->man;
714: }
715:
716: void
717: mandoc_vmsg(enum mandocerr t, struct mparse *m,
718: int ln, int pos, const char *fmt, ...)
719: {
720: char buf[256];
721: va_list ap;
722:
723: va_start(ap, fmt);
724: vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
725: va_end(ap);
726:
727: mandoc_msg(t, m, ln, pos, buf);
728: }
729:
730: void
731: mandoc_msg(enum mandocerr er, struct mparse *m,
732: int ln, int col, const char *msg)
733: {
734: enum mandoclevel level;
735:
736: level = MANDOCLEVEL_FATAL;
737: while (er < mandoclimits[level])
738: level--;
739:
740: if (level < m->wlevel)
741: return;
742:
743: if (m->mmsg)
744: (*m->mmsg)(er, level, m->file, ln, col, msg);
745:
746: if (m->file_status < level)
747: m->file_status = level;
748: }
749:
750: const char *
751: mparse_strerror(enum mandocerr er)
752: {
753:
754: return(mandocerrs[er]);
755: }
756:
757: const char *
758: mparse_strlevel(enum mandoclevel lvl)
759: {
760: return(mandoclevels[lvl]);
761: }