1 /* $Id: roff.c,v 1.172 2011/10/24 21:41:45 schwarze Exp $ */
3 * Copyright (c) 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2011 Ingo Schwarze <schwarze@openbsd.org>
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.
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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.
29 #include "libmandoc.h"
31 /* Maximum number of nested if-else conditionals. */
32 #define RSTACK_MAX 128
34 /* Maximum number of string expansions per line, to break infinite loops. */
35 #define EXPAND_LIMIT 1000
78 * A single register entity. If "set" is zero, the value of the
79 * register should be the default one, which is per-register.
80 * Registers are assumed to be unsigned ints for now.
83 int set; /* whether set or not */
84 unsigned int u; /* unsigned integer */
88 * An incredibly-simple string buffer.
91 char *p; /* nil-terminated buffer */
92 size_t sz; /* saved strlen(p) */
96 * A key-value roffstr pair as part of a singly-linked list.
101 struct roffkv *next; /* next in list */
105 struct mparse *parse; /* parse point */
106 struct roffnode *last; /* leaf of stack */
107 enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */
108 int rstackpos; /* position in rstack */
109 struct reg regs[REG__MAX];
110 struct roffkv *strtab; /* user-defined strings & macros */
111 struct roffkv *xmbtab; /* multi-byte trans table (`tr') */
112 struct roffstr *xtab; /* single-byte trans table (`tr') */
113 const char *current_string; /* value of last called user macro */
114 struct tbl_node *first_tbl; /* first table parsed */
115 struct tbl_node *last_tbl; /* last table parsed */
116 struct tbl_node *tbl; /* current table being parsed */
117 struct eqn_node *last_eqn; /* last equation parsed */
118 struct eqn_node *first_eqn; /* first equation parsed */
119 struct eqn_node *eqn; /* current equation being parsed */
123 enum rofft tok; /* type of node */
124 struct roffnode *parent; /* up one in stack */
125 int line; /* parse line */
126 int col; /* parse col */
127 char *name; /* node name, e.g. macro name */
128 char *end; /* end-rules: custom token */
129 int endspan; /* end-rules: next-line or infty */
130 enum roffrule rule; /* current evaluation rule */
133 #define ROFF_ARGS struct roff *r, /* parse ctx */ \
134 enum rofft tok, /* tok of macro */ \
135 char **bufp, /* input buffer */ \
136 size_t *szp, /* size of input buffer */ \
137 int ln, /* parse line */ \
138 int ppos, /* original pos in buffer */ \
139 int pos, /* current pos in buffer */ \
140 int *offs /* reset offset of buffer data */
142 typedef enum rofferr (*roffproc)(ROFF_ARGS);
145 const char *name; /* macro name */
146 roffproc proc; /* process new macro */
147 roffproc text; /* process as child text of macro */
148 roffproc sub; /* process as child of macro */
150 #define ROFFMAC_STRUCT (1 << 0) /* always interpret */
151 struct roffmac *next;
155 const char *name; /* predefined input name */
156 const char *str; /* replacement symbol */
159 #define PREDEF(__name, __str) \
160 { (__name), (__str) },
162 static enum rofft roffhash_find(const char *, size_t);
163 static void roffhash_init(void);
164 static void roffnode_cleanscope(struct roff *);
165 static void roffnode_pop(struct roff *);
166 static void roffnode_push(struct roff *, enum rofft,
167 const char *, int, int);
168 static enum rofferr roff_block(ROFF_ARGS);
169 static enum rofferr roff_block_text(ROFF_ARGS);
170 static enum rofferr roff_block_sub(ROFF_ARGS);
171 static enum rofferr roff_cblock(ROFF_ARGS);
172 static enum rofferr roff_ccond(ROFF_ARGS);
173 static enum rofferr roff_cond(ROFF_ARGS);
174 static enum rofferr roff_cond_text(ROFF_ARGS);
175 static enum rofferr roff_cond_sub(ROFF_ARGS);
176 static enum rofferr roff_ds(ROFF_ARGS);
177 static enum roffrule roff_evalcond(const char *, int *);
178 static void roff_free1(struct roff *);
179 static void roff_freestr(struct roffkv *);
180 static char *roff_getname(struct roff *, char **, int, int);
181 static const char *roff_getstrn(const struct roff *,
182 const char *, size_t);
183 static enum rofferr roff_line_ignore(ROFF_ARGS);
184 static enum rofferr roff_nr(ROFF_ARGS);
185 static void roff_openeqn(struct roff *, const char *,
186 int, int, const char *);
187 static enum rofft roff_parse(struct roff *, const char *, int *);
188 static enum rofferr roff_parsetext(char *);
189 static enum rofferr roff_res(struct roff *,
190 char **, size_t *, int, int);
191 static enum rofferr roff_rm(ROFF_ARGS);
192 static void roff_setstr(struct roff *,
193 const char *, const char *, int);
194 static void roff_setstrn(struct roffkv **, const char *,
195 size_t, const char *, size_t, int);
196 static enum rofferr roff_so(ROFF_ARGS);
197 static enum rofferr roff_tr(ROFF_ARGS);
198 static enum rofferr roff_TE(ROFF_ARGS);
199 static enum rofferr roff_TS(ROFF_ARGS);
200 static enum rofferr roff_EQ(ROFF_ARGS);
201 static enum rofferr roff_EN(ROFF_ARGS);
202 static enum rofferr roff_T_(ROFF_ARGS);
203 static enum rofferr roff_userdef(ROFF_ARGS);
205 /* See roffhash_find() */
209 #define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
211 static struct roffmac *hash[HASHWIDTH];
213 static struct roffmac roffs[ROFF_MAX] = {
214 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
215 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
216 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
217 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
218 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
219 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
220 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
221 { "ds", roff_ds, NULL, NULL, 0, NULL },
222 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
223 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
224 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
225 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
226 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
227 { "it", roff_line_ignore, NULL, NULL, 0, NULL },
228 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
229 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
230 { "nr", roff_nr, NULL, NULL, 0, NULL },
231 { "ns", roff_line_ignore, NULL, NULL, 0, NULL },
232 { "ps", roff_line_ignore, NULL, NULL, 0, NULL },
233 { "rm", roff_rm, NULL, NULL, 0, NULL },
234 { "so", roff_so, NULL, NULL, 0, NULL },
235 { "ta", roff_line_ignore, NULL, NULL, 0, NULL },
236 { "tr", roff_tr, NULL, NULL, 0, NULL },
237 { "TS", roff_TS, NULL, NULL, 0, NULL },
238 { "TE", roff_TE, NULL, NULL, 0, NULL },
239 { "T&", roff_T_, NULL, NULL, 0, NULL },
240 { "EQ", roff_EQ, NULL, NULL, 0, NULL },
241 { "EN", roff_EN, NULL, NULL, 0, NULL },
242 { ".", roff_cblock, NULL, NULL, 0, NULL },
243 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
244 { NULL, roff_userdef, NULL, NULL, 0, NULL },
247 /* Array of injected predefined strings. */
248 #define PREDEFS_MAX 38
249 static const struct predef predefs[PREDEFS_MAX] = {
250 #include "predefs.in"
253 /* See roffhash_find() */
254 #define ROFF_HASH(p) (p[0] - ASCII_LO)
262 for (i = 0; i < (int)ROFF_USERDEF; i++) {
263 assert(roffs[i].name[0] >= ASCII_LO);
264 assert(roffs[i].name[0] <= ASCII_HI);
266 buc = ROFF_HASH(roffs[i].name);
268 if (NULL != (n = hash[buc])) {
269 for ( ; n->next; n = n->next)
273 hash[buc] = &roffs[i];
278 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
279 * the nil-terminated string name could be found.
282 roffhash_find(const char *p, size_t s)
288 * libroff has an extremely simple hashtable, for the time
289 * being, which simply keys on the first character, which must
290 * be printable, then walks a chain. It works well enough until
294 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
299 if (NULL == (n = hash[buc]))
301 for ( ; n; n = n->next)
302 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
303 return((enum rofft)(n - roffs));
310 * Pop the current node off of the stack of roff instructions currently
314 roffnode_pop(struct roff *r)
321 r->last = r->last->parent;
329 * Push a roff node onto the instruction stack. This must later be
330 * removed with roffnode_pop().
333 roffnode_push(struct roff *r, enum rofft tok, const char *name,
338 p = mandoc_calloc(1, sizeof(struct roffnode));
341 p->name = mandoc_strdup(name);
345 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
352 roff_free1(struct roff *r)
358 while (NULL != (t = r->first_tbl)) {
359 r->first_tbl = t->next;
363 r->first_tbl = r->last_tbl = r->tbl = NULL;
365 while (NULL != (e = r->first_eqn)) {
366 r->first_eqn = e->next;
370 r->first_eqn = r->last_eqn = r->eqn = NULL;
375 roff_freestr(r->strtab);
376 roff_freestr(r->xmbtab);
378 r->strtab = r->xmbtab = NULL;
381 for (i = 0; i < 128; i++)
389 roff_reset(struct roff *r)
395 memset(&r->regs, 0, sizeof(struct reg) * REG__MAX);
397 for (i = 0; i < PREDEFS_MAX; i++)
398 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
403 roff_free(struct roff *r)
412 roff_alloc(struct mparse *parse)
417 r = mandoc_calloc(1, sizeof(struct roff));
423 for (i = 0; i < PREDEFS_MAX; i++)
424 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
430 * Pre-filter each and every line for reserved words (one beginning with
431 * `\*', e.g., `\*(ab'). These must be handled before the actual line
433 * This also checks the syntax of regular escapes.
436 roff_res(struct roff *r, char **bufp, size_t *szp, int ln, int pos)
439 const char *stesc; /* start of an escape sequence ('\\') */
440 const char *stnam; /* start of the name, after "[(*" */
441 const char *cp; /* end of the name, e.g. before ']' */
442 const char *res; /* the string to be substituted */
443 int i, maxl, expand_count;
451 while (NULL != (cp = strchr(cp, '\\'))) {
455 * The second character must be an asterisk.
456 * If it isn't, skip it anyway: It is escaped,
457 * so it can't start another escape sequence.
465 esc = mandoc_escape(&cp, NULL, NULL);
466 if (ESCAPE_ERROR != esc)
470 (MANDOCERR_BADESCAPE, r->parse,
471 ln, (int)(stesc - *bufp), NULL);
478 * The third character decides the length
479 * of the name of the string.
480 * Save a pointer to the name.
500 /* Advance to the end of the name. */
502 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
505 (MANDOCERR_BADESCAPE,
507 (int)(stesc - *bufp), NULL);
510 if (0 == maxl && ']' == *cp)
515 * Retrieve the replacement string; if it is
516 * undefined, resume searching for escapes.
519 res = roff_getstrn(r, stnam, (size_t)i);
523 (MANDOCERR_BADESCAPE, r->parse,
524 ln, (int)(stesc - *bufp), NULL);
528 /* Replace the escape sequence by the string. */
532 nsz = *szp + strlen(res) + 1;
533 n = mandoc_malloc(nsz);
535 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
536 strlcat(n, res, nsz);
537 strlcat(n, cp + (maxl ? 0 : 1), nsz);
544 if (EXPAND_LIMIT >= ++expand_count)
547 /* Just leave the string unexpanded. */
548 mandoc_msg(MANDOCERR_ROFFLOOP, r->parse, ln, pos, NULL);
555 * Process text streams: convert all breakable hyphens into ASCII_HYPH.
558 roff_parsetext(char *p)
567 sz = strcspn(p, "-\\");
574 /* Skip over escapes. */
577 ((const char **)&p, NULL, NULL);
578 if (ESCAPE_ERROR == esc)
581 } else if (p == start) {
586 if (isalpha((unsigned char)p[-1]) &&
587 isalpha((unsigned char)p[1]))
596 roff_parseln(struct roff *r, int ln, char **bufp,
597 size_t *szp, int pos, int *offs)
604 * Run the reserved-word filter only if we have some reserved
608 e = roff_res(r, bufp, szp, ln, pos);
611 assert(ROFF_CONT == e);
614 ctl = mandoc_getcontrol(*bufp, &pos);
617 * First, if a scope is open and we're not a macro, pass the
618 * text through the macro's filter. If a scope isn't open and
619 * we're not a macro, just let it through.
620 * Finally, if there's an equation scope open, divert it into it
621 * no matter our state.
624 if (r->last && ! ctl) {
626 assert(roffs[t].text);
628 (r, t, bufp, szp, ln, pos, pos, offs);
629 assert(ROFF_IGN == e || ROFF_CONT == e);
633 return(eqn_read(&r->eqn, ln, *bufp, pos, offs));
635 return(tbl_read(r->tbl, ln, *bufp, pos));
636 return(roff_parsetext(*bufp + pos));
639 return(eqn_read(&r->eqn, ln, *bufp, pos, offs));
641 return(tbl_read(r->tbl, ln, *bufp, pos));
642 return(roff_parsetext(*bufp + pos));
644 return(eqn_read(&r->eqn, ln, *bufp, ppos, offs));
647 * If a scope is open, go to the child handler for that macro,
648 * as it may want to preprocess before doing anything with it.
649 * Don't do so if an equation is open.
654 assert(roffs[t].sub);
655 return((*roffs[t].sub)
657 ln, ppos, pos, offs));
661 * Lastly, as we've no scope open, try to look up and execute
662 * the new macro. If no macro is found, simply return and let
663 * the compilers handle it.
666 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
669 assert(roffs[t].proc);
670 return((*roffs[t].proc)
672 ln, ppos, pos, offs));
677 roff_endparse(struct roff *r)
681 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
682 r->last->line, r->last->col, NULL);
685 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
686 r->eqn->eqn.ln, r->eqn->eqn.pos, NULL);
691 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
692 r->tbl->line, r->tbl->pos, NULL);
698 * Parse a roff node's type from the input buffer. This must be in the
699 * form of ".foo xxx" in the usual way.
702 roff_parse(struct roff *r, const char *buf, int *pos)
708 if ('\0' == buf[*pos] || '"' == buf[*pos] ||
709 '\t' == buf[*pos] || ' ' == buf[*pos])
713 * We stop the macro parse at an escape, tab, space, or nil.
714 * However, `\}' is also a valid macro, so make sure we don't
715 * clobber it by seeing the `\' as the end of token.
719 maclen = strcspn(mac + 1, " \\\t\0") + 1;
721 t = (r->current_string = roff_getstrn(r, mac, maclen))
722 ? ROFF_USERDEF : roffhash_find(mac, maclen);
726 while (buf[*pos] && ' ' == buf[*pos])
734 roff_cblock(ROFF_ARGS)
738 * A block-close `..' should only be invoked as a child of an
739 * ignore macro, otherwise raise a warning and just ignore it.
742 if (NULL == r->last) {
743 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
747 switch (r->last->tok) {
755 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
762 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
767 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
770 roffnode_cleanscope(r);
777 roffnode_cleanscope(struct roff *r)
781 if (--r->last->endspan < 0)
790 roff_ccond(ROFF_ARGS)
793 if (NULL == r->last) {
794 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
798 switch (r->last->tok) {
806 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
810 if (r->last->endspan > -1) {
811 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
816 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
819 roffnode_cleanscope(r);
826 roff_block(ROFF_ARGS)
834 if (ROFF_ig != tok) {
835 if ('\0' == (*bufp)[pos]) {
836 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
841 * Re-write `de1', since we don't really care about
842 * groff's strange compatibility mode, into `de'.
850 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos,
853 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
856 while (isspace((unsigned char)(*bufp)[pos]))
857 (*bufp)[pos++] = '\0';
860 roffnode_push(r, tok, name, ln, ppos);
863 * At the beginning of a `de' macro, clear the existing string
864 * with the same name, if there is one. New content will be
865 * added from roff_block_text() in multiline mode.
869 roff_setstr(r, name, "", 0);
871 if ('\0' == (*bufp)[pos])
874 /* If present, process the custom end-of-line marker. */
877 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
881 * Note: groff does NOT like escape characters in the input.
882 * Instead of detecting this, we're just going to let it fly and
887 sz = (size_t)(pos - sv);
889 if (1 == sz && '.' == (*bufp)[sv])
892 r->last->end = mandoc_malloc(sz + 1);
894 memcpy(r->last->end, *bufp + sv, sz);
895 r->last->end[(int)sz] = '\0';
898 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
906 roff_block_sub(ROFF_ARGS)
912 * First check whether a custom macro exists at this level. If
913 * it does, then check against it. This is some of groff's
914 * stranger behaviours. If we encountered a custom end-scope
915 * tag and that tag also happens to be a "real" macro, then we
916 * need to try interpreting it again as a real macro. If it's
917 * not, then return ignore. Else continue.
921 for (i = pos, j = 0; r->last->end[j]; j++, i++)
922 if ((*bufp)[i] != r->last->end[j])
925 if ('\0' == r->last->end[j] &&
926 ('\0' == (*bufp)[i] ||
928 '\t' == (*bufp)[i])) {
930 roffnode_cleanscope(r);
932 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
936 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
943 * If we have no custom end-query or lookup failed, then try
944 * pulling it out of the hashtable.
947 t = roff_parse(r, *bufp, &pos);
950 * Macros other than block-end are only significant
951 * in `de' blocks; elsewhere, simply throw them away.
953 if (ROFF_cblock != t) {
955 roff_setstr(r, r->last->name, *bufp + ppos, 1);
959 assert(roffs[t].proc);
960 return((*roffs[t].proc)(r, t, bufp, szp,
961 ln, ppos, pos, offs));
967 roff_block_text(ROFF_ARGS)
971 roff_setstr(r, r->last->name, *bufp + pos, 1);
979 roff_cond_sub(ROFF_ARGS)
986 roffnode_cleanscope(r);
989 * If the macro is unknown, first check if it contains a closing
990 * delimiter `\}'. If it does, close out our scope and return
991 * the currently-scoped rule (ignore or continue). Else, drop
992 * into the currently-scoped rule.
995 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
997 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
1003 * Make the \} go away.
1004 * This is a little haphazard, as it's not quite
1005 * clear how nroff does this.
1006 * If we're at the end of line, then just chop
1007 * off the \} and resize the buffer.
1008 * If we aren't, then conver it to spaces.
1011 if ('\0' == *(ep + 1)) {
1015 *(ep - 1) = *ep = ' ';
1017 roff_ccond(r, ROFF_ccond, bufp, szp,
1018 ln, pos, pos + 2, offs);
1021 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
1025 * A denied conditional must evaluate its children if and only
1026 * if they're either structurally required (such as loops and
1027 * conditionals) or a closing macro.
1030 if (ROFFRULE_DENY == rr)
1031 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
1032 if (ROFF_ccond != t)
1035 assert(roffs[t].proc);
1036 return((*roffs[t].proc)(r, t, bufp, szp,
1037 ln, ppos, pos, offs));
1042 roff_cond_text(ROFF_ARGS)
1048 roffnode_cleanscope(r);
1051 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
1056 roff_ccond(r, ROFF_ccond, bufp, szp,
1057 ln, pos, pos + 2, offs);
1059 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
1062 static enum roffrule
1063 roff_evalcond(const char *v, int *pos)
1069 return(ROFFRULE_ALLOW);
1076 return(ROFFRULE_DENY);
1081 while (v[*pos] && ' ' != v[*pos])
1083 return(ROFFRULE_DENY);
1088 roff_line_ignore(ROFF_ARGS)
1092 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos, "it");
1099 roff_cond(ROFF_ARGS)
1105 * An `.el' has no conditional body: it will consume the value
1106 * of the current rstack entry set in prior `ie' calls or
1109 * If we're not an `el', however, then evaluate the conditional.
1112 rule = ROFF_el == tok ?
1114 ROFFRULE_DENY : r->rstack[r->rstackpos--]) :
1115 roff_evalcond(*bufp, &pos);
1118 while (' ' == (*bufp)[pos])
1122 * Roff is weird. If we have just white-space after the
1123 * conditional, it's considered the BODY and we exit without
1124 * really doing anything. Warn about this. It's probably
1128 if ('\0' == (*bufp)[pos] && sv != pos) {
1129 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
1133 roffnode_push(r, tok, NULL, ln, ppos);
1135 r->last->rule = rule;
1138 * An if-else will put the NEGATION of the current evaluated
1139 * conditional into the stack of rules.
1142 if (ROFF_ie == tok) {
1143 if (r->rstackpos == RSTACK_MAX - 1) {
1144 mandoc_msg(MANDOCERR_MEM,
1145 r->parse, ln, ppos, NULL);
1148 r->rstack[++r->rstackpos] =
1149 ROFFRULE_DENY == r->last->rule ?
1150 ROFFRULE_ALLOW : ROFFRULE_DENY;
1153 /* If the parent has false as its rule, then so do we. */
1155 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
1156 r->last->rule = ROFFRULE_DENY;
1159 * Determine scope. If we're invoked with "\{" trailing the
1160 * conditional, then we're in a multiline scope. Else our scope
1161 * expires on the next line.
1164 r->last->endspan = 1;
1166 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1167 r->last->endspan = -1;
1172 * If there are no arguments on the line, the next-line scope is
1176 if ('\0' == (*bufp)[pos])
1179 /* Otherwise re-run the roff parser after recalculating. */
1190 char *name, *string;
1193 * A symbol is named by the first word following the macro
1194 * invocation up to a space. Its value is anything after the
1195 * name's trailing whitespace and optional double-quote. Thus,
1199 * will have `bar " ' as its value.
1202 string = *bufp + pos;
1203 name = roff_getname(r, &string, ln, pos);
1207 /* Read past initial double-quote. */
1211 /* The rest is the value. */
1212 roff_setstr(r, name, string, 0);
1217 roff_regisset(const struct roff *r, enum regs reg)
1220 return(r->regs[(int)reg].set);
1224 roff_regget(const struct roff *r, enum regs reg)
1227 return(r->regs[(int)reg].u);
1231 roff_regunset(struct roff *r, enum regs reg)
1234 r->regs[(int)reg].set = 0;
1246 key = roff_getname(r, &val, ln, pos);
1248 if (0 == strcmp(key, "nS")) {
1249 r->regs[(int)REG_nS].set = 1;
1250 if ((iv = mandoc_strntoi(val, strlen(val), 10)) >= 0)
1251 r->regs[(int)REG_nS].u = (unsigned)iv;
1253 r->regs[(int)REG_nS].u = 0u;
1267 while ('\0' != *cp) {
1268 name = roff_getname(r, &cp, ln, (int)(cp - *bufp));
1270 roff_setstr(r, name, NULL, 0);
1281 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1294 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1296 tbl_restart(ppos, ln, r->tbl);
1303 roff_closeeqn(struct roff *r)
1306 return(r->eqn && ROFF_EQN == eqn_end(&r->eqn) ? 1 : 0);
1311 roff_openeqn(struct roff *r, const char *name, int line,
1312 int offs, const char *buf)
1317 assert(NULL == r->eqn);
1318 e = eqn_alloc(name, offs, line, r->parse);
1321 r->last_eqn->next = e;
1323 r->first_eqn = r->last_eqn = e;
1325 r->eqn = r->last_eqn = e;
1329 eqn_read(&r->eqn, line, buf, offs, &poff);
1338 roff_openeqn(r, *bufp + pos, ln, ppos, NULL);
1347 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1358 mandoc_msg(MANDOCERR_SCOPEBROKEN, r->parse, ln, ppos, NULL);
1362 t = tbl_alloc(ppos, ln, r->parse);
1365 r->last_tbl->next = t;
1367 r->first_tbl = r->last_tbl = t;
1369 r->tbl = r->last_tbl = t;
1377 const char *p, *first, *second;
1379 enum mandoc_esc esc;
1384 mandoc_msg(MANDOCERR_ARGCOUNT, r->parse, ln, ppos, NULL);
1388 while ('\0' != *p) {
1392 if ('\\' == *first) {
1393 esc = mandoc_escape(&p, NULL, NULL);
1394 if (ESCAPE_ERROR == esc) {
1396 (MANDOCERR_BADESCAPE, r->parse,
1397 ln, (int)(p - *bufp), NULL);
1400 fsz = (size_t)(p - first);
1404 if ('\\' == *second) {
1405 esc = mandoc_escape(&p, NULL, NULL);
1406 if (ESCAPE_ERROR == esc) {
1408 (MANDOCERR_BADESCAPE, r->parse,
1409 ln, (int)(p - *bufp), NULL);
1412 ssz = (size_t)(p - second);
1413 } else if ('\0' == *second) {
1414 mandoc_msg(MANDOCERR_ARGCOUNT, r->parse,
1415 ln, (int)(p - *bufp), NULL);
1421 roff_setstrn(&r->xmbtab, first,
1422 fsz, second, ssz, 0);
1426 if (NULL == r->xtab)
1427 r->xtab = mandoc_calloc
1428 (128, sizeof(struct roffstr));
1430 free(r->xtab[(int)*first].p);
1431 r->xtab[(int)*first].p = mandoc_strndup(second, ssz);
1432 r->xtab[(int)*first].sz = ssz;
1444 mandoc_msg(MANDOCERR_SO, r->parse, ln, ppos, NULL);
1447 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1448 * opening anything that's not in our cwd or anything beneath
1449 * it. Thus, explicitly disallow traversing up the file-system
1450 * or using absolute paths.
1454 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
1455 mandoc_msg(MANDOCERR_SOPATH, r->parse, ln, pos, NULL);
1465 roff_userdef(ROFF_ARGS)
1472 * Collect pointers to macro argument strings
1473 * and null-terminate them.
1476 for (i = 0; i < 9; i++)
1477 arg[i] = '\0' == *cp ? "" :
1478 mandoc_getarg(r->parse, &cp, ln, &pos);
1481 * Expand macro arguments.
1484 n1 = cp = mandoc_strdup(r->current_string);
1485 while (NULL != (cp = strstr(cp, "\\$"))) {
1487 if (0 > i || 8 < i) {
1488 /* Not an argument invocation. */
1493 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1494 n2 = mandoc_malloc(*szp);
1496 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1497 strlcat(n2, arg[i], *szp);
1498 strlcat(n2, cp + 3, *szp);
1500 cp = n2 + (cp - n1);
1506 * Replace the macro invocation
1507 * by the expanded macro.
1512 *szp = strlen(*bufp) + 1;
1514 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1515 ROFF_REPARSE : ROFF_APPEND);
1519 roff_getname(struct roff *r, char **cpp, int ln, int pos)
1527 /* Read until end of name. */
1528 for (cp = name; '\0' != *cp && ' ' != *cp; cp++) {
1534 mandoc_msg(MANDOCERR_NAMESC, r->parse, ln, pos, NULL);
1539 /* Nil-terminate name. */
1543 /* Read past spaces. */
1552 * Store *string into the user-defined string called *name.
1553 * In multiline mode, append to an existing entry and append '\n';
1554 * else replace the existing entry, if there is one.
1555 * To clear an existing entry, call with (*r, *name, NULL, 0).
1558 roff_setstr(struct roff *r, const char *name, const char *string,
1562 roff_setstrn(&r->strtab, name, strlen(name), string,
1563 string ? strlen(string) : 0, multiline);
1567 roff_setstrn(struct roffkv **r, const char *name, size_t namesz,
1568 const char *string, size_t stringsz, int multiline)
1573 size_t oldch, newch;
1575 /* Search for an existing string with the same name. */
1578 while (n && strcmp(name, n->key.p))
1582 /* Create a new string table entry. */
1583 n = mandoc_malloc(sizeof(struct roffkv));
1584 n->key.p = mandoc_strndup(name, namesz);
1590 } else if (0 == multiline) {
1591 /* In multiline mode, append; else replace. */
1601 * One additional byte for the '\n' in multiline mode,
1602 * and one for the terminating '\0'.
1604 newch = stringsz + (multiline ? 2u : 1u);
1606 if (NULL == n->val.p) {
1607 n->val.p = mandoc_malloc(newch);
1612 n->val.p = mandoc_realloc(n->val.p, oldch + newch);
1615 /* Skip existing content in the destination buffer. */
1616 c = n->val.p + (int)oldch;
1618 /* Append new content to the destination buffer. */
1620 while (i < (int)stringsz) {
1622 * Rudimentary roff copy mode:
1623 * Handle escaped backslashes.
1625 if ('\\' == string[i] && '\\' == string[i + 1])
1630 /* Append terminating bytes. */
1635 n->val.sz = (int)(c - n->val.p);
1639 roff_getstrn(const struct roff *r, const char *name, size_t len)
1641 const struct roffkv *n;
1643 for (n = r->strtab; n; n = n->next)
1644 if (0 == strncmp(name, n->key.p, len) &&
1645 '\0' == n->key.p[(int)len])
1652 roff_freestr(struct roffkv *r)
1654 struct roffkv *n, *nn;
1656 for (n = r; n; n = nn) {
1664 const struct tbl_span *
1665 roff_span(const struct roff *r)
1668 return(r->tbl ? tbl_span(r->tbl) : NULL);
1672 roff_eqn(const struct roff *r)
1675 return(r->last_eqn ? &r->last_eqn->eqn : NULL);
1679 * Duplicate an input string, making the appropriate character
1680 * conversations (as stipulated by `tr') along the way.
1681 * Returns a heap-allocated string with all the replacements made.
1684 roff_strdup(const struct roff *r, const char *p)
1686 const struct roffkv *cp;
1690 enum mandoc_esc esc;
1692 if (NULL == r->xmbtab && NULL == r->xtab)
1693 return(mandoc_strdup(p));
1694 else if ('\0' == *p)
1695 return(mandoc_strdup(""));
1698 * Step through each character looking for term matches
1699 * (remember that a `tr' can be invoked with an escape, which is
1700 * a glyph but the escape is multi-character).
1701 * We only do this if the character hash has been initialised
1702 * and the string is >0 length.
1708 while ('\0' != *p) {
1709 if ('\\' != *p && r->xtab && r->xtab[(int)*p].p) {
1710 sz = r->xtab[(int)*p].sz;
1711 res = mandoc_realloc(res, ssz + sz + 1);
1712 memcpy(res + ssz, r->xtab[(int)*p].p, sz);
1716 } else if ('\\' != *p) {
1717 res = mandoc_realloc(res, ssz + 2);
1722 /* Search for term matches. */
1723 for (cp = r->xmbtab; cp; cp = cp->next)
1724 if (0 == strncmp(p, cp->key.p, cp->key.sz))
1729 * A match has been found.
1730 * Append the match to the array and move
1731 * forward by its keysize.
1733 res = mandoc_realloc
1734 (res, ssz + cp->val.sz + 1);
1735 memcpy(res + ssz, cp->val.p, cp->val.sz);
1737 p += (int)cp->key.sz;
1742 * Handle escapes carefully: we need to copy
1743 * over just the escape itself, or else we might
1744 * do replacements within the escape itself.
1745 * Make sure to pass along the bogus string.
1748 esc = mandoc_escape(&p, NULL, NULL);
1749 if (ESCAPE_ERROR == esc) {
1751 res = mandoc_realloc(res, ssz + sz + 1);
1752 memcpy(res + ssz, pp, sz);
1756 * We bail out on bad escapes.
1757 * No need to warn: we already did so when
1758 * roff_res() was called.
1761 res = mandoc_realloc(res, ssz + sz + 1);
1762 memcpy(res + ssz, pp, sz);
1766 res[(int)ssz] = '\0';