Annotation of mandoc/read.c, Revision 1.8
1.8 ! kristaps 1: /* $Id: read.c,v 1.7 2011/03/28 21:49:42 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>
1.3 kristaps 24: #include <stdarg.h>
1.1 kristaps 25: #include <stdio.h>
26: #include <stdlib.h>
27: #include <string.h>
28: #include <unistd.h>
29:
30: #include "mandoc.h"
1.3 kristaps 31: #include "libmandoc.h"
1.1 kristaps 32: #include "mdoc.h"
33: #include "man.h"
34:
35: #ifndef MAP_FILE
36: #define MAP_FILE 0
37: #endif
38:
39: #define REPARSE_LIMIT 1000
40:
41: struct buf {
42: char *buf; /* binary input buffer */
43: size_t sz; /* size of binary buffer */
44: };
45:
46: struct mparse {
47: enum mandoclevel file_status; /* status of current parse */
1.3 kristaps 48: enum mandoclevel wlevel; /* ignore messages below this */
1.1 kristaps 49: int line; /* line number in the file */
50: enum mparset inttype; /* which parser to use */
51: struct man *pman; /* persistent man parser */
52: struct mdoc *pmdoc; /* persistent mdoc parser */
53: struct man *man; /* man parser */
54: struct mdoc *mdoc; /* mdoc parser */
55: struct roff *roff; /* roff parser (!NULL) */
56: struct regset regs; /* roff registers */
57: int reparse_count; /* finite interp. stack */
58: mandocmsg mmsg; /* warning/error message handler */
59: void *arg; /* argument to mmsg */
1.3 kristaps 60: const char *file;
1.1 kristaps 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:
1.3 kristaps 71: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
72: MANDOCERR_OK,
73: MANDOCERR_WARNING,
74: MANDOCERR_WARNING,
75: MANDOCERR_ERROR,
76: MANDOCERR_FATAL,
77: MANDOCERR_MAX,
78: MANDOCERR_MAX
79: };
80:
1.7 kristaps 81: static const char * const mandocerrs[MANDOCERR_MAX] = {
82: "ok",
83:
84: "generic warning",
85:
86: /* related to the prologue */
87: "no title in document",
88: "document title should be all caps",
89: "unknown manual section",
90: "date missing, using today's date",
91: "cannot parse date, using it verbatim",
92: "prologue macros out of order",
93: "duplicate prologue macro",
94: "macro not allowed in prologue",
95: "macro not allowed in body",
96:
97: /* related to document structure */
98: ".so is fragile, better use ln(1)",
99: "NAME section must come first",
100: "bad NAME section contents",
101: "manual name not yet set",
102: "sections out of conventional order",
103: "duplicate section name",
104: "section not in conventional manual section",
105:
106: /* related to macros and nesting */
107: "skipping obsolete macro",
108: "skipping paragraph macro",
109: "skipping no-space macro",
110: "blocks badly nested",
111: "child violates parent syntax",
112: "nested displays are not portable",
113: "already in literal mode",
114: "line scope broken",
115:
116: /* related to missing macro arguments */
117: "skipping empty macro",
118: "argument count wrong",
119: "missing display type",
120: "list type must come first",
121: "tag lists require a width argument",
122: "missing font type",
123: "skipping end of block that is not open",
124:
125: /* related to bad macro arguments */
126: "skipping argument",
127: "duplicate argument",
128: "duplicate display type",
129: "duplicate list type",
130: "unknown AT&T UNIX version",
131: "bad Boolean value",
132: "unknown font",
133: "unknown standard specifier",
134: "bad width argument",
135:
136: /* related to plain text */
137: "blank line in non-literal context",
138: "tab in non-literal context",
139: "end of line whitespace",
140: "bad comment style",
141: "unknown escape sequence",
142: "unterminated quoted string",
143:
144: "generic error",
145:
146: /* related to tables */
147: "bad table syntax",
148: "bad table option",
149: "bad table layout",
150: "no table layout cells specified",
151: "no table data cells specified",
152: "ignore data in cell",
153: "data block still open",
154: "ignoring extra data cells",
155:
156: "input stack limit exceeded, infinite loop?",
157: "skipping bad character",
158: "escaped character not allowed in a name",
159: "skipping text before the first section header",
160: "skipping unknown macro",
161: "NOT IMPLEMENTED, please use groff: skipping request",
162: "argument count wrong",
163: "skipping end of block that is not open",
164: "missing end of block",
165: "scope open on exit",
166: "uname(3) system call failed",
167: "macro requires line argument(s)",
168: "macro requires body argument(s)",
169: "macro requires argument(s)",
170: "missing list type",
171: "line argument(s) will be lost",
172: "body argument(s) will be lost",
173:
174: "generic fatal error",
175:
176: "not a manual",
177: "column syntax is inconsistent",
178: "NOT IMPLEMENTED: .Bd -file",
179: "line scope broken, syntax violated",
180: "argument count wrong, violates syntax",
181: "child violates parent syntax",
182: "argument count wrong, violates syntax",
183: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
184: "no document body",
185: "no document prologue",
186: "static buffer exhausted",
187: };
188:
189: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
190: "SUCCESS",
191: "RESERVED",
192: "WARNING",
193: "ERROR",
194: "FATAL",
195: "BADARG",
196: "SYSERR"
197: };
198:
1.1 kristaps 199: static void
200: resize_buf(struct buf *buf, size_t initial)
201: {
202:
203: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
204: buf->buf = mandoc_realloc(buf->buf, buf->sz);
205: }
206:
207: static void
208: pset(const char *buf, int pos, struct mparse *curp)
209: {
210: int i;
211:
212: /*
213: * Try to intuit which kind of manual parser should be used. If
214: * passed in by command-line (-man, -mdoc), then use that
215: * explicitly. If passed as -mandoc, then try to guess from the
216: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
217: * default to -man, which is more lenient.
218: *
219: * Separate out pmdoc/pman from mdoc/man: the first persists
220: * through all parsers, while the latter is used per-parse.
221: */
222:
223: if ('.' == buf[0] || '\'' == buf[0]) {
224: for (i = 1; buf[i]; i++)
225: if (' ' != buf[i] && '\t' != buf[i])
226: break;
227: if ('\0' == buf[i])
228: return;
229: }
230:
231: switch (curp->inttype) {
232: case (MPARSE_MDOC):
233: if (NULL == curp->pmdoc)
1.3 kristaps 234: curp->pmdoc = mdoc_alloc(&curp->regs, curp);
1.1 kristaps 235: assert(curp->pmdoc);
236: curp->mdoc = curp->pmdoc;
237: return;
238: case (MPARSE_MAN):
239: if (NULL == curp->pman)
1.3 kristaps 240: curp->pman = man_alloc(&curp->regs, curp);
1.1 kristaps 241: assert(curp->pman);
242: curp->man = curp->pman;
243: return;
244: default:
245: break;
246: }
247:
248: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
249: if (NULL == curp->pmdoc)
1.3 kristaps 250: curp->pmdoc = mdoc_alloc(&curp->regs, curp);
1.1 kristaps 251: assert(curp->pmdoc);
252: curp->mdoc = curp->pmdoc;
253: return;
254: }
255:
256: if (NULL == curp->pman)
1.3 kristaps 257: curp->pman = man_alloc(&curp->regs, curp);
1.1 kristaps 258: assert(curp->pman);
259: curp->man = curp->pman;
260: }
261:
262: /*
263: * Main parse routine for an opened file. This is called for each
264: * opened file and simply loops around the full input file, possibly
265: * nesting (i.e., with `so').
266: */
267: static void
268: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
269: {
270: const struct tbl_span *span;
271: struct buf ln;
272: enum rofferr rr;
273: int i, of, rc;
274: int pos; /* byte number in the ln buffer */
275: int lnn; /* line number in the real file */
276: unsigned char c;
277:
278: memset(&ln, 0, sizeof(struct buf));
279:
280: lnn = curp->line;
281: pos = 0;
282:
283: for (i = 0; i < (int)blk.sz; ) {
284: if (0 == pos && '\0' == blk.buf[i])
285: break;
286:
287: if (start) {
288: curp->line = lnn;
289: curp->reparse_count = 0;
290: }
291:
292: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
293:
294: /*
295: * When finding an unescaped newline character,
296: * leave the character loop to process the line.
297: * Skip a preceding carriage return, if any.
298: */
299:
300: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
301: '\n' == blk.buf[i + 1])
302: ++i;
303: if ('\n' == blk.buf[i]) {
304: ++i;
305: ++lnn;
306: break;
307: }
308:
309: /*
310: * Warn about bogus characters. If you're using
311: * non-ASCII encoding, you're screwing your
312: * readers. Since I'd rather this not happen,
313: * I'll be helpful and drop these characters so
314: * we don't display gibberish. Note to manual
315: * writers: use special characters.
316: */
317:
318: c = (unsigned char) blk.buf[i];
319:
320: if ( ! (isascii(c) &&
321: (isgraph(c) || isblank(c)))) {
1.3 kristaps 322: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.1 kristaps 323: curp->line, pos, "ignoring byte");
324: i++;
325: continue;
326: }
327:
328: /* Trailing backslash = a plain char. */
329:
330: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
331: if (pos >= (int)ln.sz)
332: resize_buf(&ln, 256);
333: ln.buf[pos++] = blk.buf[i++];
334: continue;
335: }
336:
337: /*
338: * Found escape and at least one other character.
339: * When it's a newline character, skip it.
340: * When there is a carriage return in between,
341: * skip that one as well.
342: */
343:
344: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
345: '\n' == blk.buf[i + 2])
346: ++i;
347: if ('\n' == blk.buf[i + 1]) {
348: i += 2;
349: ++lnn;
350: continue;
351: }
352:
353: if ('"' == blk.buf[i + 1]) {
354: i += 2;
355: /* Comment, skip to end of line */
356: for (; i < (int)blk.sz; ++i) {
357: if ('\n' == blk.buf[i]) {
358: ++i;
359: ++lnn;
360: break;
361: }
362: }
363:
364: /* Backout trailing whitespaces */
365: for (; pos > 0; --pos) {
366: if (ln.buf[pos - 1] != ' ')
367: break;
368: if (pos > 2 && ln.buf[pos - 2] == '\\')
369: break;
370: }
371: break;
372: }
373:
374: /* Some other escape sequence, copy & cont. */
375:
376: if (pos + 1 >= (int)ln.sz)
377: resize_buf(&ln, 256);
378:
379: ln.buf[pos++] = blk.buf[i++];
380: ln.buf[pos++] = blk.buf[i++];
381: }
382:
383: if (pos >= (int)ln.sz)
384: resize_buf(&ln, 256);
385:
386: ln.buf[pos] = '\0';
387:
388: /*
389: * A significant amount of complexity is contained by
390: * the roff preprocessor. It's line-oriented but can be
391: * expressed on one line, so we need at times to
392: * readjust our starting point and re-run it. The roff
393: * preprocessor can also readjust the buffers with new
394: * data, so we pass them in wholesale.
395: */
396:
397: of = 0;
398:
399: rerun:
400: rr = roff_parseln
401: (curp->roff, curp->line,
402: &ln.buf, &ln.sz, of, &of);
403:
404: switch (rr) {
405: case (ROFF_REPARSE):
406: if (REPARSE_LIMIT >= ++curp->reparse_count)
407: mparse_buf_r(curp, ln, 0);
408: else
1.3 kristaps 409: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.1 kristaps 410: curp->line, pos, NULL);
411: pos = 0;
412: continue;
413: case (ROFF_APPEND):
414: pos = (int)strlen(ln.buf);
415: continue;
416: case (ROFF_RERUN):
417: goto rerun;
418: case (ROFF_IGN):
419: pos = 0;
420: continue;
421: case (ROFF_ERR):
422: assert(MANDOCLEVEL_FATAL <= curp->file_status);
423: break;
424: case (ROFF_SO):
425: mparse_readfd_r(curp, -1, ln.buf + of, 1);
426: if (MANDOCLEVEL_FATAL <= curp->file_status)
427: break;
428: pos = 0;
429: continue;
430: default:
431: break;
432: }
433:
434: /*
435: * If we encounter errors in the recursive parse, make
436: * sure we don't continue parsing.
437: */
438:
439: if (MANDOCLEVEL_FATAL <= curp->file_status)
440: break;
441:
442: /*
443: * If input parsers have not been allocated, do so now.
444: * We keep these instanced betwen parsers, but set them
445: * locally per parse routine since we can use different
446: * parsers with each one.
447: */
448:
449: if ( ! (curp->man || curp->mdoc))
450: pset(ln.buf + of, pos - of, curp);
451:
452: /*
453: * Lastly, push down into the parsers themselves. One
454: * of these will have already been set in the pset()
455: * routine.
456: * If libroff returns ROFF_TBL, then add it to the
457: * currently open parse. Since we only get here if
458: * there does exist data (see tbl_data.c), we're
459: * guaranteed that something's been allocated.
460: * Do the same for ROFF_EQN.
461: */
462:
463: rc = -1;
464:
465: if (ROFF_TBL == rr)
466: while (NULL != (span = roff_span(curp->roff))) {
467: rc = curp->man ?
468: man_addspan(curp->man, span) :
469: mdoc_addspan(curp->mdoc, span);
470: if (0 == rc)
471: break;
472: }
473: else if (ROFF_EQN == rr)
474: rc = curp->mdoc ?
475: mdoc_addeqn(curp->mdoc,
476: roff_eqn(curp->roff)) :
477: man_addeqn(curp->man,
478: roff_eqn(curp->roff));
479: else if (curp->man || curp->mdoc)
480: rc = curp->man ?
481: man_parseln(curp->man,
482: curp->line, ln.buf, of) :
483: mdoc_parseln(curp->mdoc,
484: curp->line, ln.buf, of);
485:
486: if (0 == rc) {
487: assert(MANDOCLEVEL_FATAL <= curp->file_status);
488: break;
489: }
490:
491: /* Temporary buffers typically are not full. */
492:
493: if (0 == start && '\0' == blk.buf[i])
494: break;
495:
496: /* Start the next input line. */
497:
498: pos = 0;
499: }
500:
501: free(ln.buf);
502: }
503:
504: static void
505: pdesc(struct mparse *curp, const char *file, int fd)
506: {
507: struct buf blk;
508: int with_mmap;
509:
510: /*
511: * Run for each opened file; may be called more than once for
512: * each full parse sequence if the opened file is nested (i.e.,
513: * from `so'). Simply sucks in the whole file and moves into
514: * the parse phase for the file.
515: */
516:
517: if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
518: curp->file_status = MANDOCLEVEL_SYSERR;
519: return;
520: }
521:
522: /* Line number is per-file. */
523:
524: curp->line = 1;
525:
526: mparse_buf_r(curp, blk, 1);
527:
528: if (with_mmap)
529: munmap(blk.buf, blk.sz);
530: else
531: free(blk.buf);
532: }
533:
534: static int
535: read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
536: {
537: struct stat st;
538: size_t off;
539: ssize_t ssz;
540:
541: if (-1 == fstat(fd, &st)) {
542: perror(file);
543: return(0);
544: }
545:
546: /*
547: * If we're a regular file, try just reading in the whole entry
548: * via mmap(). This is faster than reading it into blocks, and
549: * since each file is only a few bytes to begin with, I'm not
550: * concerned that this is going to tank any machines.
551: */
552:
553: if (S_ISREG(st.st_mode)) {
554: if (st.st_size >= (1U << 31)) {
555: fprintf(stderr, "%s: input too large\n", file);
556: return(0);
557: }
558: *with_mmap = 1;
559: fb->sz = (size_t)st.st_size;
560: fb->buf = mmap(NULL, fb->sz, PROT_READ,
561: MAP_FILE|MAP_SHARED, fd, 0);
562: if (fb->buf != MAP_FAILED)
563: return(1);
564: }
565:
566: /*
567: * If this isn't a regular file (like, say, stdin), then we must
568: * go the old way and just read things in bit by bit.
569: */
570:
571: *with_mmap = 0;
572: off = 0;
573: fb->sz = 0;
574: fb->buf = NULL;
575: for (;;) {
576: if (off == fb->sz) {
577: if (fb->sz == (1U << 31)) {
578: fprintf(stderr, "%s: input too large\n", file);
579: break;
580: }
581: resize_buf(fb, 65536);
582: }
583: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
584: if (ssz == 0) {
585: fb->sz = off;
586: return(1);
587: }
588: if (ssz == -1) {
589: perror(file);
590: break;
591: }
592: off += (size_t)ssz;
593: }
594:
595: free(fb->buf);
596: fb->buf = NULL;
597: return(0);
598: }
599:
600: static void
601: mparse_end(struct mparse *curp)
602: {
603:
604: if (MANDOCLEVEL_FATAL <= curp->file_status)
605: return;
606:
607: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
608: assert(MANDOCLEVEL_FATAL <= curp->file_status);
609: return;
610: }
611:
612: if (curp->man && ! man_endparse(curp->man)) {
613: assert(MANDOCLEVEL_FATAL <= curp->file_status);
614: return;
615: }
616:
617: if ( ! (curp->man || curp->mdoc)) {
1.6 kristaps 618: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
1.1 kristaps 619: curp->file_status = MANDOCLEVEL_FATAL;
1.6 kristaps 620: return;
1.1 kristaps 621: }
622:
623: roff_endparse(curp->roff);
624: }
625:
626: static void
627: mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
628: {
629: const char *svfile;
630:
631: if (-1 == fd)
632: if (-1 == (fd = open(file, O_RDONLY, 0))) {
633: perror(file);
634: curp->file_status = MANDOCLEVEL_SYSERR;
635: return;
636: }
637:
1.3 kristaps 638: svfile = curp->file;
639: curp->file = file;
1.1 kristaps 640:
641: pdesc(curp, file, fd);
642:
643: if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
644: mparse_end(curp);
645:
646: if (STDIN_FILENO != fd && -1 == close(fd))
647: perror(file);
648:
1.3 kristaps 649: curp->file = svfile;
1.1 kristaps 650: }
651:
652: enum mandoclevel
653: mparse_readfd(struct mparse *curp, int fd, const char *file)
654: {
655:
656: mparse_readfd_r(curp, fd, file, 0);
657: return(curp->file_status);
658: }
659:
660: struct mparse *
1.4 kristaps 661: mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
1.1 kristaps 662: {
663: struct mparse *curp;
664:
665: curp = mandoc_calloc(1, sizeof(struct mparse));
666:
1.3 kristaps 667: curp->wlevel = wlevel;
1.1 kristaps 668: curp->mmsg = mmsg;
669: curp->arg = arg;
670: curp->inttype = inttype;
671:
1.3 kristaps 672: curp->roff = roff_alloc(&curp->regs, curp);
1.1 kristaps 673: return(curp);
674: }
675:
676: void
677: mparse_reset(struct mparse *curp)
678: {
679:
680: memset(&curp->regs, 0, sizeof(struct regset));
681:
682: roff_reset(curp->roff);
683:
684: if (curp->mdoc)
685: mdoc_reset(curp->mdoc);
686: if (curp->man)
687: man_reset(curp->man);
688:
689: curp->file_status = MANDOCLEVEL_OK;
690: curp->mdoc = NULL;
691: curp->man = NULL;
692: }
693:
694: void
695: mparse_free(struct mparse *curp)
696: {
697:
698: if (curp->pmdoc)
699: mdoc_free(curp->pmdoc);
700: if (curp->pman)
701: man_free(curp->pman);
702: if (curp->roff)
703: roff_free(curp->roff);
704:
705: free(curp);
706: }
707:
708: void
709: mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
710: {
711:
712: *mdoc = curp->mdoc;
713: *man = curp->man;
1.3 kristaps 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:
1.8 ! kristaps 743: if (m->mmsg)
! 744: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 745:
746: if (m->file_status < level)
747: m->file_status = level;
1.7 kristaps 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]);
1.1 kristaps 761: }
CVSweb