1 /* $Id: roff.c,v 1.142 2011/05/26 11:58:25 kristaps 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
75 char *name; /* key of symbol */
76 char *string; /* current value */
77 struct roffstr *next; /* next in list */
81 struct mparse *parse; /* parse point */
82 struct roffnode *last; /* leaf of stack */
83 enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */
84 int rstackpos; /* position in rstack */
85 struct regset *regs; /* read/writable registers */
86 struct roffstr *first_string; /* user-defined strings & macros */
87 const char *current_string; /* value of last called user macro */
88 struct tbl_node *first_tbl; /* first table parsed */
89 struct tbl_node *last_tbl; /* last table parsed */
90 struct tbl_node *tbl; /* current table being parsed */
91 struct eqn_node *last_eqn; /* last equation parsed */
92 struct eqn_node *first_eqn; /* first equation parsed */
93 struct eqn_node *eqn; /* current equation being parsed */
97 enum rofft tok; /* type of node */
98 struct roffnode *parent; /* up one in stack */
99 int line; /* parse line */
100 int col; /* parse col */
101 char *name; /* node name, e.g. macro name */
102 char *end; /* end-rules: custom token */
103 int endspan; /* end-rules: next-line or infty */
104 enum roffrule rule; /* current evaluation rule */
107 #define ROFF_ARGS struct roff *r, /* parse ctx */ \
108 enum rofft tok, /* tok of macro */ \
109 char **bufp, /* input buffer */ \
110 size_t *szp, /* size of input buffer */ \
111 int ln, /* parse line */ \
112 int ppos, /* original pos in buffer */ \
113 int pos, /* current pos in buffer */ \
114 int *offs /* reset offset of buffer data */
116 typedef enum rofferr (*roffproc)(ROFF_ARGS);
119 const char *name; /* macro name */
120 roffproc proc; /* process new macro */
121 roffproc text; /* process as child text of macro */
122 roffproc sub; /* process as child of macro */
124 #define ROFFMAC_STRUCT (1 << 0) /* always interpret */
125 struct roffmac *next;
129 const char *name; /* predefined input name */
130 const char *str; /* replacement symbol */
133 #define PREDEF(__name, __str) \
134 { (__name), (__str) },
136 static enum rofferr roff_block(ROFF_ARGS);
137 static enum rofferr roff_block_text(ROFF_ARGS);
138 static enum rofferr roff_block_sub(ROFF_ARGS);
139 static enum rofferr roff_cblock(ROFF_ARGS);
140 static enum rofferr roff_ccond(ROFF_ARGS);
141 static enum rofferr roff_cond(ROFF_ARGS);
142 static enum rofferr roff_cond_text(ROFF_ARGS);
143 static enum rofferr roff_cond_sub(ROFF_ARGS);
144 static enum rofferr roff_ds(ROFF_ARGS);
145 static enum roffrule roff_evalcond(const char *, int *);
146 static void roff_freestr(struct roff *);
147 static char *roff_getname(struct roff *, char **, int, int);
148 static const char *roff_getstrn(const struct roff *,
149 const char *, size_t);
150 static enum rofferr roff_line_ignore(ROFF_ARGS);
151 static enum rofferr roff_nr(ROFF_ARGS);
152 static int roff_res(struct roff *,
153 char **, size_t *, int, int);
154 static enum rofferr roff_rm(ROFF_ARGS);
155 static void roff_setstr(struct roff *,
156 const char *, const char *, int);
157 static enum rofferr roff_so(ROFF_ARGS);
158 static enum rofferr roff_TE(ROFF_ARGS);
159 static enum rofferr roff_TS(ROFF_ARGS);
160 static enum rofferr roff_EQ(ROFF_ARGS);
161 static enum rofferr roff_EN(ROFF_ARGS);
162 static enum rofferr roff_T_(ROFF_ARGS);
163 static enum rofferr roff_userdef(ROFF_ARGS);
165 /* See roff_hash_find() */
169 #define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
171 static struct roffmac *hash[HASHWIDTH];
173 static struct roffmac roffs[ROFF_MAX] = {
174 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
175 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
176 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
177 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
178 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
179 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
180 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
181 { "ds", roff_ds, NULL, NULL, 0, NULL },
182 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
183 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
184 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
185 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
186 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
187 { "it", roff_line_ignore, NULL, NULL, 0, NULL },
188 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
189 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
190 { "nr", roff_nr, NULL, NULL, 0, NULL },
191 { "ns", roff_line_ignore, NULL, NULL, 0, NULL },
192 { "ps", roff_line_ignore, NULL, NULL, 0, NULL },
193 { "rm", roff_rm, NULL, NULL, 0, NULL },
194 { "so", roff_so, NULL, NULL, 0, NULL },
195 { "ta", roff_line_ignore, NULL, NULL, 0, NULL },
196 { "tr", roff_line_ignore, NULL, NULL, 0, NULL },
197 { "TS", roff_TS, NULL, NULL, 0, NULL },
198 { "TE", roff_TE, NULL, NULL, 0, NULL },
199 { "T&", roff_T_, NULL, NULL, 0, NULL },
200 { "EQ", roff_EQ, NULL, NULL, 0, NULL },
201 { "EN", roff_EN, NULL, NULL, 0, NULL },
202 { ".", roff_cblock, NULL, NULL, 0, NULL },
203 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
204 { NULL, roff_userdef, NULL, NULL, 0, NULL },
207 /* Array of injected predefined strings. */
208 #define PREDEFS_MAX 38
209 static const struct predef predefs[PREDEFS_MAX] = {
210 #include "predefs.in"
213 static void roff_free1(struct roff *);
214 static enum rofft roff_hash_find(const char *, size_t);
215 static void roff_hash_init(void);
216 static void roffnode_cleanscope(struct roff *);
217 static void roffnode_push(struct roff *, enum rofft,
218 const char *, int, int);
219 static void roffnode_pop(struct roff *);
220 static enum rofft roff_parse(struct roff *, const char *, int *);
222 /* See roff_hash_find() */
223 #define ROFF_HASH(p) (p[0] - ASCII_LO)
231 for (i = 0; i < (int)ROFF_USERDEF; i++) {
232 assert(roffs[i].name[0] >= ASCII_LO);
233 assert(roffs[i].name[0] <= ASCII_HI);
235 buc = ROFF_HASH(roffs[i].name);
237 if (NULL != (n = hash[buc])) {
238 for ( ; n->next; n = n->next)
242 hash[buc] = &roffs[i];
247 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
248 * the nil-terminated string name could be found.
251 roff_hash_find(const char *p, size_t s)
257 * libroff has an extremely simple hashtable, for the time
258 * being, which simply keys on the first character, which must
259 * be printable, then walks a chain. It works well enough until
263 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
268 if (NULL == (n = hash[buc]))
270 for ( ; n; n = n->next)
271 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
272 return((enum rofft)(n - roffs));
279 * Pop the current node off of the stack of roff instructions currently
283 roffnode_pop(struct roff *r)
290 r->last = r->last->parent;
298 * Push a roff node onto the instruction stack. This must later be
299 * removed with roffnode_pop().
302 roffnode_push(struct roff *r, enum rofft tok, const char *name,
307 p = mandoc_calloc(1, sizeof(struct roffnode));
310 p->name = mandoc_strdup(name);
314 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
321 roff_free1(struct roff *r)
326 while (NULL != (t = r->first_tbl)) {
327 r->first_tbl = t->next;
331 r->first_tbl = r->last_tbl = r->tbl = NULL;
333 while (NULL != (e = r->first_eqn)) {
334 r->first_eqn = e->next;
338 r->first_eqn = r->last_eqn = r->eqn = NULL;
348 roff_reset(struct roff *r)
356 roff_free(struct roff *r)
365 roff_alloc(struct regset *regs, struct mparse *parse)
370 r = mandoc_calloc(1, sizeof(struct roff));
377 for (i = 0; i < PREDEFS_MAX; i++)
378 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
385 * Pre-filter each and every line for reserved words (one beginning with
386 * `\*', e.g., `\*(ab'). These must be handled before the actual line
390 roff_res(struct roff *r, char **bufp, size_t *szp, int ln, int pos)
392 const char *stesc; /* start of an escape sequence ('\\') */
393 const char *stnam; /* start of the name, after "[(*" */
394 const char *cp; /* end of the name, e.g. before ']' */
395 const char *res; /* the string to be substituted */
400 /* Search for a leading backslash and save a pointer to it. */
403 while (NULL != (cp = strchr(cp, '\\'))) {
407 * The second character must be an asterisk.
408 * If it isn't, skip it anyway: It is escaped,
409 * so it can't start another escape sequence.
418 * The third character decides the length
419 * of the name of the string.
420 * Save a pointer to the name.
440 /* Advance to the end of the name. */
442 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
444 return(1); /* Error. */
445 if (0 == maxl && ']' == *cp)
450 * Retrieve the replacement string; if it is
451 * undefined, resume searching for escapes.
454 res = roff_getstrn(r, stnam, (size_t)i);
457 /* TODO: keep track of the correct position. */
458 mandoc_msg(MANDOCERR_BADESCAPE, r->parse, ln, pos, NULL);
462 /* Replace the escape sequence by the string. */
464 nsz = *szp + strlen(res) + 1;
465 n = mandoc_malloc(nsz);
467 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
468 strlcat(n, res, nsz);
469 strlcat(n, cp + (maxl ? 0 : 1), nsz);
483 roff_parseln(struct roff *r, int ln, char **bufp,
484 size_t *szp, int pos, int *offs)
491 * Run the reserved-word filter only if we have some reserved
495 if (r->first_string && ! roff_res(r, bufp, szp, ln, pos))
496 return(ROFF_REPARSE);
499 ctl = mandoc_getcontrol(*bufp, &pos);
502 * First, if a scope is open and we're not a macro, pass the
503 * text through the macro's filter. If a scope isn't open and
504 * we're not a macro, just let it through.
505 * Finally, if there's an equation scope open, divert it into it
506 * no matter our state.
509 if (r->last && ! ctl) {
511 assert(roffs[t].text);
513 (r, t, bufp, szp, ln, pos, pos, offs);
514 assert(ROFF_IGN == e || ROFF_CONT == e);
518 return(eqn_read(&r->eqn, ln, *bufp, pos));
520 return(tbl_read(r->tbl, ln, *bufp, pos));
524 return(eqn_read(&r->eqn, ln, *bufp, pos));
526 return(tbl_read(r->tbl, ln, *bufp, pos));
529 return(eqn_read(&r->eqn, ln, *bufp, ppos));
532 * If a scope is open, go to the child handler for that macro,
533 * as it may want to preprocess before doing anything with it.
534 * Don't do so if an equation is open.
539 assert(roffs[t].sub);
540 return((*roffs[t].sub)
542 ln, ppos, pos, offs));
546 * Lastly, as we've no scope open, try to look up and execute
547 * the new macro. If no macro is found, simply return and let
548 * the compilers handle it.
551 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
554 assert(roffs[t].proc);
555 return((*roffs[t].proc)
557 ln, ppos, pos, offs));
562 roff_endparse(struct roff *r)
566 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
567 r->last->line, r->last->col, NULL);
570 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
571 r->eqn->eqn.line, r->eqn->eqn.pos, NULL);
577 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
578 r->tbl->line, r->tbl->pos, NULL);
585 * Parse a roff node's type from the input buffer. This must be in the
586 * form of ".foo xxx" in the usual way.
589 roff_parse(struct roff *r, const char *buf, int *pos)
595 if ('\0' == buf[*pos] || '"' == buf[*pos])
599 maclen = strcspn(mac, " \\\t\0");
601 t = (r->current_string = roff_getstrn(r, mac, maclen))
602 ? ROFF_USERDEF : roff_hash_find(mac, maclen);
606 while (buf[*pos] && ' ' == buf[*pos])
614 roff_cblock(ROFF_ARGS)
618 * A block-close `..' should only be invoked as a child of an
619 * ignore macro, otherwise raise a warning and just ignore it.
622 if (NULL == r->last) {
623 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
627 switch (r->last->tok) {
635 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
642 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
647 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
650 roffnode_cleanscope(r);
657 roffnode_cleanscope(struct roff *r)
661 if (--r->last->endspan < 0)
670 roff_ccond(ROFF_ARGS)
673 if (NULL == r->last) {
674 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
678 switch (r->last->tok) {
686 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
690 if (r->last->endspan > -1) {
691 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
696 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
699 roffnode_cleanscope(r);
706 roff_block(ROFF_ARGS)
714 if (ROFF_ig != tok) {
715 if ('\0' == (*bufp)[pos]) {
716 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
721 * Re-write `de1', since we don't really care about
722 * groff's strange compatibility mode, into `de'.
730 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos,
733 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
736 while (isspace((unsigned char)(*bufp)[pos]))
737 (*bufp)[pos++] = '\0';
740 roffnode_push(r, tok, name, ln, ppos);
743 * At the beginning of a `de' macro, clear the existing string
744 * with the same name, if there is one. New content will be
745 * added from roff_block_text() in multiline mode.
749 roff_setstr(r, name, "", 0);
751 if ('\0' == (*bufp)[pos])
754 /* If present, process the custom end-of-line marker. */
757 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
761 * Note: groff does NOT like escape characters in the input.
762 * Instead of detecting this, we're just going to let it fly and
767 sz = (size_t)(pos - sv);
769 if (1 == sz && '.' == (*bufp)[sv])
772 r->last->end = mandoc_malloc(sz + 1);
774 memcpy(r->last->end, *bufp + sv, sz);
775 r->last->end[(int)sz] = '\0';
778 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
786 roff_block_sub(ROFF_ARGS)
792 * First check whether a custom macro exists at this level. If
793 * it does, then check against it. This is some of groff's
794 * stranger behaviours. If we encountered a custom end-scope
795 * tag and that tag also happens to be a "real" macro, then we
796 * need to try interpreting it again as a real macro. If it's
797 * not, then return ignore. Else continue.
801 for (i = pos, j = 0; r->last->end[j]; j++, i++)
802 if ((*bufp)[i] != r->last->end[j])
805 if ('\0' == r->last->end[j] &&
806 ('\0' == (*bufp)[i] ||
808 '\t' == (*bufp)[i])) {
810 roffnode_cleanscope(r);
812 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
816 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
823 * If we have no custom end-query or lookup failed, then try
824 * pulling it out of the hashtable.
827 t = roff_parse(r, *bufp, &pos);
830 * Macros other than block-end are only significant
831 * in `de' blocks; elsewhere, simply throw them away.
833 if (ROFF_cblock != t) {
835 roff_setstr(r, r->last->name, *bufp + ppos, 1);
839 assert(roffs[t].proc);
840 return((*roffs[t].proc)(r, t, bufp, szp,
841 ln, ppos, pos, offs));
847 roff_block_text(ROFF_ARGS)
851 roff_setstr(r, r->last->name, *bufp + pos, 1);
859 roff_cond_sub(ROFF_ARGS)
866 roffnode_cleanscope(r);
869 * If the macro is unknown, first check if it contains a closing
870 * delimiter `\}'. If it does, close out our scope and return
871 * the currently-scoped rule (ignore or continue). Else, drop
872 * into the currently-scoped rule.
875 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
877 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
882 roff_ccond(r, ROFF_ccond, bufp, szp,
883 ln, pos, pos + 2, offs);
886 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
890 * A denied conditional must evaluate its children if and only
891 * if they're either structurally required (such as loops and
892 * conditionals) or a closing macro.
895 if (ROFFRULE_DENY == rr)
896 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
900 assert(roffs[t].proc);
901 return((*roffs[t].proc)(r, t, bufp, szp,
902 ln, ppos, pos, offs));
907 roff_cond_text(ROFF_ARGS)
913 roffnode_cleanscope(r);
916 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
921 roff_ccond(r, ROFF_ccond, bufp, szp,
922 ln, pos, pos + 2, offs);
924 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
928 roff_evalcond(const char *v, int *pos)
934 return(ROFFRULE_ALLOW);
941 return(ROFFRULE_DENY);
946 while (v[*pos] && ' ' != v[*pos])
948 return(ROFFRULE_DENY);
953 roff_line_ignore(ROFF_ARGS)
957 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos, "it");
970 * An `.el' has no conditional body: it will consume the value
971 * of the current rstack entry set in prior `ie' calls or
974 * If we're not an `el', however, then evaluate the conditional.
977 rule = ROFF_el == tok ?
979 ROFFRULE_DENY : r->rstack[r->rstackpos--]) :
980 roff_evalcond(*bufp, &pos);
983 while (' ' == (*bufp)[pos])
987 * Roff is weird. If we have just white-space after the
988 * conditional, it's considered the BODY and we exit without
989 * really doing anything. Warn about this. It's probably
993 if ('\0' == (*bufp)[pos] && sv != pos) {
994 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
998 roffnode_push(r, tok, NULL, ln, ppos);
1000 r->last->rule = rule;
1003 * An if-else will put the NEGATION of the current evaluated
1004 * conditional into the stack of rules.
1007 if (ROFF_ie == tok) {
1008 if (r->rstackpos == RSTACK_MAX - 1) {
1009 mandoc_msg(MANDOCERR_MEM,
1010 r->parse, ln, ppos, NULL);
1013 r->rstack[++r->rstackpos] =
1014 ROFFRULE_DENY == r->last->rule ?
1015 ROFFRULE_ALLOW : ROFFRULE_DENY;
1018 /* If the parent has false as its rule, then so do we. */
1020 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
1021 r->last->rule = ROFFRULE_DENY;
1024 * Determine scope. If we're invoked with "\{" trailing the
1025 * conditional, then we're in a multiline scope. Else our scope
1026 * expires on the next line.
1029 r->last->endspan = 1;
1031 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1032 r->last->endspan = -1;
1037 * If there are no arguments on the line, the next-line scope is
1041 if ('\0' == (*bufp)[pos])
1044 /* Otherwise re-run the roff parser after recalculating. */
1055 char *name, *string;
1058 * A symbol is named by the first word following the macro
1059 * invocation up to a space. Its value is anything after the
1060 * name's trailing whitespace and optional double-quote. Thus,
1064 * will have `bar " ' as its value.
1067 string = *bufp + pos;
1068 name = roff_getname(r, &string, ln, pos);
1072 /* Read past initial double-quote. */
1076 /* The rest is the value. */
1077 roff_setstr(r, name, string, 0);
1092 key = roff_getname(r, &val, ln, pos);
1095 if (0 == strcmp(key, "nS")) {
1096 rg[(int)REG_nS].set = 1;
1097 if ((iv = mandoc_strntou(val, strlen(val), 10)) >= 0)
1098 rg[REG_nS].v.u = (unsigned)iv;
1100 rg[(int)REG_nS].v.u = 0u;
1114 while ('\0' != *cp) {
1115 name = roff_getname(r, &cp, ln, (int)(cp - *bufp));
1117 roff_setstr(r, name, NULL, 0);
1128 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1142 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1144 tbl_restart(ppos, ln, r->tbl);
1155 assert(NULL == r->eqn);
1156 e = eqn_alloc(ppos, ln);
1159 r->last_eqn->next = e;
1161 r->first_eqn = r->last_eqn = e;
1163 r->eqn = r->last_eqn = e;
1172 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1183 mandoc_msg(MANDOCERR_SCOPEBROKEN, r->parse, ln, ppos, NULL);
1187 t = tbl_alloc(ppos, ln, r->parse);
1190 r->last_tbl->next = t;
1192 r->first_tbl = r->last_tbl = t;
1194 r->tbl = r->last_tbl = t;
1204 mandoc_msg(MANDOCERR_SO, r->parse, ln, ppos, NULL);
1207 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1208 * opening anything that's not in our cwd or anything beneath
1209 * it. Thus, explicitly disallow traversing up the file-system
1210 * or using absolute paths.
1214 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
1215 mandoc_msg(MANDOCERR_SOPATH, r->parse, ln, pos, NULL);
1225 roff_userdef(ROFF_ARGS)
1232 * Collect pointers to macro argument strings
1233 * and null-terminate them.
1236 for (i = 0; i < 9; i++)
1237 arg[i] = '\0' == *cp ? "" :
1238 mandoc_getarg(r->parse, &cp, ln, &pos);
1241 * Expand macro arguments.
1244 n1 = cp = mandoc_strdup(r->current_string);
1245 while (NULL != (cp = strstr(cp, "\\$"))) {
1247 if (0 > i || 8 < i) {
1248 /* Not an argument invocation. */
1253 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1254 n2 = mandoc_malloc(*szp);
1256 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1257 strlcat(n2, arg[i], *szp);
1258 strlcat(n2, cp + 3, *szp);
1260 cp = n2 + (cp - n1);
1266 * Replace the macro invocation
1267 * by the expanded macro.
1272 *szp = strlen(*bufp) + 1;
1274 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1275 ROFF_REPARSE : ROFF_APPEND);
1279 roff_getname(struct roff *r, char **cpp, int ln, int pos)
1287 /* Read until end of name. */
1288 for (cp = name; '\0' != *cp && ' ' != *cp; cp++) {
1294 mandoc_msg(MANDOCERR_NAMESC, r->parse, ln, pos, NULL);
1299 /* Nil-terminate name. */
1303 /* Read past spaces. */
1312 * Store *string into the user-defined string called *name.
1313 * In multiline mode, append to an existing entry and append '\n';
1314 * else replace the existing entry, if there is one.
1315 * To clear an existing entry, call with (*r, *name, NULL, 0).
1318 roff_setstr(struct roff *r, const char *name, const char *string,
1323 size_t oldch, newch;
1325 /* Search for an existing string with the same name. */
1326 n = r->first_string;
1327 while (n && strcmp(name, n->name))
1331 /* Create a new string table entry. */
1332 n = mandoc_malloc(sizeof(struct roffstr));
1333 n->name = mandoc_strdup(name);
1335 n->next = r->first_string;
1336 r->first_string = n;
1337 } else if (0 == multiline) {
1338 /* In multiline mode, append; else replace. */
1347 * One additional byte for the '\n' in multiline mode,
1348 * and one for the terminating '\0'.
1350 newch = strlen(string) + (multiline ? 2u : 1u);
1351 if (NULL == n->string) {
1352 n->string = mandoc_malloc(newch);
1356 oldch = strlen(n->string);
1357 n->string = mandoc_realloc(n->string, oldch + newch);
1360 /* Skip existing content in the destination buffer. */
1361 c = n->string + (int)oldch;
1363 /* Append new content to the destination buffer. */
1366 * Rudimentary roff copy mode:
1367 * Handle escaped backslashes.
1369 if ('\\' == *string && '\\' == *(string + 1))
1374 /* Append terminating bytes. */
1381 roff_getstrn(const struct roff *r, const char *name, size_t len)
1383 const struct roffstr *n;
1385 n = r->first_string;
1386 while (n && (strncmp(name, n->name, len) || '\0' != n->name[(int)len]))
1389 return(n ? n->string : NULL);
1393 roff_freestr(struct roff *r)
1395 struct roffstr *n, *nn;
1397 for (n = r->first_string; n; n = nn) {
1404 r->first_string = NULL;
1407 const struct tbl_span *
1408 roff_span(const struct roff *r)
1411 return(r->tbl ? tbl_span(r->tbl) : NULL);
1415 roff_eqn(const struct roff *r)
1418 return(r->last_eqn ? &r->last_eqn->eqn : NULL);