1 /*      $Id: mdoc.c,v 1.196 2011/09/30 00:13:28 schwarze Exp $ */
   2 /*
   3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
   4  * Copyright (c) 2010 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 #ifdef HAVE_CONFIG_H
  19 #include "config.h"
  20 #endif
  21 
  22 #include <sys/types.h>
  23 
  24 #include <assert.h>
  25 #include <stdarg.h>
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29 #include <time.h>
  30 
  31 #include "mdoc.h"
  32 #include "mandoc.h"
  33 #include "libmdoc.h"
  34 #include "libmandoc.h"
  35 
  36 const   char *const __mdoc_macronames[MDOC_MAX] = {              
  37         "Ap",           "Dd",           "Dt",           "Os",
  38         "Sh",           "Ss",           "Pp",           "D1",
  39         "Dl",           "Bd",           "Ed",           "Bl",
  40         "El",           "It",           "Ad",           "An",
  41         "Ar",           "Cd",           "Cm",           "Dv",
  42         "Er",           "Ev",           "Ex",           "Fa",
  43         "Fd",           "Fl",           "Fn",           "Ft",
  44         "Ic",           "In",           "Li",           "Nd",
  45         "Nm",           "Op",           "Ot",           "Pa",
  46         "Rv",           "St",           "Va",           "Vt",
  47         /* LINTED */
  48         "Xr",           "%A",           "%B",           "%D",
  49         /* LINTED */
  50         "%I",           "%J",           "%N",           "%O",
  51         /* LINTED */
  52         "%P",           "%R",           "%T",           "%V",
  53         "Ac",           "Ao",           "Aq",           "At",
  54         "Bc",           "Bf",           "Bo",           "Bq",
  55         "Bsx",          "Bx",           "Db",           "Dc",
  56         "Do",           "Dq",           "Ec",           "Ef",
  57         "Em",           "Eo",           "Fx",           "Ms",
  58         "No",           "Ns",           "Nx",           "Ox",
  59         "Pc",           "Pf",           "Po",           "Pq",
  60         "Qc",           "Ql",           "Qo",           "Qq",
  61         "Re",           "Rs",           "Sc",           "So",
  62         "Sq",           "Sm",           "Sx",           "Sy",
  63         "Tn",           "Ux",           "Xc",           "Xo",
  64         "Fo",           "Fc",           "Oo",           "Oc",
  65         "Bk",           "Ek",           "Bt",           "Hf",
  66         "Fr",           "Ud",           "Lb",           "Lp",
  67         "Lk",           "Mt",           "Brq",          "Bro",
  68         /* LINTED */
  69         "Brc",          "%C",           "Es",           "En",
  70         /* LINTED */
  71         "Dx",           "%Q",           "br",           "sp",
  72         /* LINTED */
  73         "%U",           "Ta"
  74         };
  75 
  76 const   char *const __mdoc_argnames[MDOC_ARG_MAX] = {            
  77         "split",                "nosplit",              "ragged",
  78         "unfilled",             "literal",              "file",          
  79         "offset",               "bullet",               "dash",          
  80         "hyphen",               "item",                 "enum",          
  81         "tag",                  "diag",                 "hang",          
  82         "ohang",                "inset",                "column",        
  83         "width",                "compact",              "std",   
  84         "filled",               "words",                "emphasis",
  85         "symbolic",             "nested",               "centered"
  86         };
  87 
  88 const   char * const *mdoc_macronames = __mdoc_macronames;
  89 const   char * const *mdoc_argnames = __mdoc_argnames;
  90 
  91 static  void              mdoc_node_free(struct mdoc_node *);
  92 static  void              mdoc_node_unlink(struct mdoc *, 
  93                                 struct mdoc_node *);
  94 static  void              mdoc_free1(struct mdoc *);
  95 static  void              mdoc_alloc1(struct mdoc *);
  96 static  struct mdoc_node *node_alloc(struct mdoc *, int, int, 
  97                                 enum mdoct, enum mdoc_type);
  98 static  int               node_append(struct mdoc *, 
  99                                 struct mdoc_node *);
 100 #if 0
 101 static  int               mdoc_preptext(struct mdoc *, int, char *, int);
 102 #endif
 103 static  int               mdoc_ptext(struct mdoc *, int, char *, int);
 104 static  int               mdoc_pmacro(struct mdoc *, int, char *, int);
 105 
 106 const struct mdoc_node *
 107 mdoc_node(const struct mdoc *m)
 108 {
 109 
 110         assert( ! (MDOC_HALT & m->flags));
 111         return(m->first);
 112 }
 113 
 114 
 115 const struct mdoc_meta *
 116 mdoc_meta(const struct mdoc *m)
 117 {
 118 
 119         assert( ! (MDOC_HALT & m->flags));
 120         return(&m->meta);
 121 }
 122 
 123 
 124 /*
 125  * Frees volatile resources (parse tree, meta-data, fields).
 126  */
 127 static void
 128 mdoc_free1(struct mdoc *mdoc)
 129 {
 130 
 131         if (mdoc->first)
 132                 mdoc_node_delete(mdoc, mdoc->first);
 133         if (mdoc->meta.title)
 134                 free(mdoc->meta.title);
 135         if (mdoc->meta.os)
 136                 free(mdoc->meta.os);
 137         if (mdoc->meta.name)
 138                 free(mdoc->meta.name);
 139         if (mdoc->meta.arch)
 140                 free(mdoc->meta.arch);
 141         if (mdoc->meta.vol)
 142                 free(mdoc->meta.vol);
 143         if (mdoc->meta.msec)
 144                 free(mdoc->meta.msec);
 145         if (mdoc->meta.date)
 146                 free(mdoc->meta.date);
 147 }
 148 
 149 
 150 /*
 151  * Allocate all volatile resources (parse tree, meta-data, fields).
 152  */
 153 static void
 154 mdoc_alloc1(struct mdoc *mdoc)
 155 {
 156 
 157         memset(&mdoc->meta, 0, sizeof(struct mdoc_meta));
 158         mdoc->flags = 0;
 159         mdoc->lastnamed = mdoc->lastsec = SEC_NONE;
 160         mdoc->last = mandoc_calloc(1, sizeof(struct mdoc_node));
 161         mdoc->first = mdoc->last;
 162         mdoc->last->type = MDOC_ROOT;
 163         mdoc->last->tok = MDOC_MAX;
 164         mdoc->next = MDOC_NEXT_CHILD;
 165 }
 166 
 167 
 168 /*
 169  * Free up volatile resources (see mdoc_free1()) then re-initialises the
 170  * data with mdoc_alloc1().  After invocation, parse data has been reset
 171  * and the parser is ready for re-invocation on a new tree; however,
 172  * cross-parse non-volatile data is kept intact.
 173  */
 174 void
 175 mdoc_reset(struct mdoc *mdoc)
 176 {
 177 
 178         mdoc_free1(mdoc);
 179         mdoc_alloc1(mdoc);
 180 }
 181 
 182 
 183 /*
 184  * Completely free up all volatile and non-volatile parse resources.
 185  * After invocation, the pointer is no longer usable.
 186  */
 187 void
 188 mdoc_free(struct mdoc *mdoc)
 189 {
 190 
 191         mdoc_free1(mdoc);
 192         free(mdoc);
 193 }
 194 
 195 
 196 /*
 197  * Allocate volatile and non-volatile parse resources.  
 198  */
 199 struct mdoc *
 200 mdoc_alloc(struct roff *roff, struct mparse *parse)
 201 {
 202         struct mdoc     *p;
 203 
 204         p = mandoc_calloc(1, sizeof(struct mdoc));
 205 
 206         p->parse = parse;
 207         p->roff = roff;
 208 
 209         mdoc_hash_init();
 210         mdoc_alloc1(p);
 211         return(p);
 212 }
 213 
 214 
 215 /*
 216  * Climb back up the parse tree, validating open scopes.  Mostly calls
 217  * through to macro_end() in macro.c.
 218  */
 219 int
 220 mdoc_endparse(struct mdoc *m)
 221 {
 222 
 223         assert( ! (MDOC_HALT & m->flags));
 224         if (mdoc_macroend(m))
 225                 return(1);
 226         m->flags |= MDOC_HALT;
 227         return(0);
 228 }
 229 
 230 int
 231 mdoc_addeqn(struct mdoc *m, const struct eqn *ep)
 232 {
 233         struct mdoc_node *n;
 234 
 235         assert( ! (MDOC_HALT & m->flags));
 236 
 237         /* No text before an initial macro. */
 238 
 239         if (SEC_NONE == m->lastnamed) {
 240                 mdoc_pmsg(m, ep->ln, ep->pos, MANDOCERR_NOTEXT);
 241                 return(1);
 242         }
 243 
 244         n = node_alloc(m, ep->ln, ep->pos, MDOC_MAX, MDOC_EQN);
 245         n->eqn = ep;
 246 
 247         if ( ! node_append(m, n))
 248                 return(0);
 249 
 250         m->next = MDOC_NEXT_SIBLING;
 251         return(1);
 252 }
 253 
 254 int
 255 mdoc_addspan(struct mdoc *m, const struct tbl_span *sp)
 256 {
 257         struct mdoc_node *n;
 258 
 259         assert( ! (MDOC_HALT & m->flags));
 260 
 261         /* No text before an initial macro. */
 262 
 263         if (SEC_NONE == m->lastnamed) {
 264                 mdoc_pmsg(m, sp->line, 0, MANDOCERR_NOTEXT);
 265                 return(1);
 266         }
 267 
 268         n = node_alloc(m, sp->line, 0, MDOC_MAX, MDOC_TBL);
 269         n->span = sp;
 270 
 271         if ( ! node_append(m, n))
 272                 return(0);
 273 
 274         m->next = MDOC_NEXT_SIBLING;
 275         return(1);
 276 }
 277 
 278 
 279 /*
 280  * Main parse routine.  Parses a single line -- really just hands off to
 281  * the macro (mdoc_pmacro()) or text parser (mdoc_ptext()).
 282  */
 283 int
 284 mdoc_parseln(struct mdoc *m, int ln, char *buf, int offs)
 285 {
 286 
 287         assert( ! (MDOC_HALT & m->flags));
 288 
 289         m->flags |= MDOC_NEWLINE;
 290 
 291         /*
 292          * Let the roff nS register switch SYNOPSIS mode early,
 293          * such that the parser knows at all times
 294          * whether this mode is on or off.
 295          * Note that this mode is also switched by the Sh macro.
 296          */
 297         if (roff_regisset(m->roff, REG_nS)) {
 298                 if (roff_regget(m->roff, REG_nS))
 299                         m->flags |= MDOC_SYNOPSIS;
 300                 else
 301                         m->flags &= ~MDOC_SYNOPSIS;
 302         }
 303 
 304         return(mandoc_getcontrol(buf, &offs) ?
 305                         mdoc_pmacro(m, ln, buf, offs) :
 306                         mdoc_ptext(m, ln, buf, offs));
 307 }
 308 
 309 int
 310 mdoc_macro(MACRO_PROT_ARGS)
 311 {
 312         assert(tok < MDOC_MAX);
 313 
 314         /* If we're in the body, deny prologue calls. */
 315 
 316         if (MDOC_PROLOGUE & mdoc_macros[tok].flags && 
 317                         MDOC_PBODY & m->flags) {
 318                 mdoc_pmsg(m, line, ppos, MANDOCERR_BADBODY);
 319                 return(1);
 320         }
 321 
 322         /* If we're in the prologue, deny "body" macros.  */
 323 
 324         if ( ! (MDOC_PROLOGUE & mdoc_macros[tok].flags) && 
 325                         ! (MDOC_PBODY & m->flags)) {
 326                 mdoc_pmsg(m, line, ppos, MANDOCERR_BADPROLOG);
 327                 if (NULL == m->meta.msec)
 328                         m->meta.msec = mandoc_strdup("1");
 329                 if (NULL == m->meta.title)
 330                         m->meta.title = mandoc_strdup("UNKNOWN");
 331                 if (NULL == m->meta.vol)
 332                         m->meta.vol = mandoc_strdup("LOCAL");
 333                 if (NULL == m->meta.os)
 334                         m->meta.os = mandoc_strdup("LOCAL");
 335                 if (NULL == m->meta.date)
 336                         m->meta.date = mandoc_normdate
 337                                 (m->parse, NULL, line, ppos);
 338                 m->flags |= MDOC_PBODY;
 339         }
 340 
 341         return((*mdoc_macros[tok].fp)(m, tok, line, ppos, pos, buf));
 342 }
 343 
 344 
 345 static int
 346 node_append(struct mdoc *mdoc, struct mdoc_node *p)
 347 {
 348 
 349         assert(mdoc->last);
 350         assert(mdoc->first);
 351         assert(MDOC_ROOT != p->type);
 352 
 353         switch (mdoc->next) {
 354         case (MDOC_NEXT_SIBLING):
 355                 mdoc->last->next = p;
 356                 p->prev = mdoc->last;
 357                 p->parent = mdoc->last->parent;
 358                 break;
 359         case (MDOC_NEXT_CHILD):
 360                 mdoc->last->child = p;
 361                 p->parent = mdoc->last;
 362                 break;
 363         default:
 364                 abort();
 365                 /* NOTREACHED */
 366         }
 367 
 368         p->parent->nchild++;
 369 
 370         /*
 371          * Copy over the normalised-data pointer of our parent.  Not
 372          * everybody has one, but copying a null pointer is fine.
 373          */
 374 
 375         switch (p->type) {
 376         case (MDOC_BODY):
 377                 /* FALLTHROUGH */
 378         case (MDOC_TAIL):
 379                 /* FALLTHROUGH */
 380         case (MDOC_HEAD):
 381                 p->norm = p->parent->norm;
 382                 break;
 383         default:
 384                 break;
 385         }
 386 
 387         if ( ! mdoc_valid_pre(mdoc, p))
 388                 return(0);
 389 
 390         switch (p->type) {
 391         case (MDOC_HEAD):
 392                 assert(MDOC_BLOCK == p->parent->type);
 393                 p->parent->head = p;
 394                 break;
 395         case (MDOC_TAIL):
 396                 assert(MDOC_BLOCK == p->parent->type);
 397                 p->parent->tail = p;
 398                 break;
 399         case (MDOC_BODY):
 400                 if (p->end)
 401                         break;
 402                 assert(MDOC_BLOCK == p->parent->type);
 403                 p->parent->body = p;
 404                 break;
 405         default:
 406                 break;
 407         }
 408 
 409         mdoc->last = p;
 410 
 411         switch (p->type) {
 412         case (MDOC_TBL):
 413                 /* FALLTHROUGH */
 414         case (MDOC_TEXT):
 415                 if ( ! mdoc_valid_post(mdoc))
 416                         return(0);
 417                 break;
 418         default:
 419                 break;
 420         }
 421 
 422         return(1);
 423 }
 424 
 425 
 426 static struct mdoc_node *
 427 node_alloc(struct mdoc *m, int line, int pos, 
 428                 enum mdoct tok, enum mdoc_type type)
 429 {
 430         struct mdoc_node *p;
 431 
 432         p = mandoc_calloc(1, sizeof(struct mdoc_node));
 433         p->sec = m->lastsec;
 434         p->line = line;
 435         p->pos = pos;
 436         p->tok = tok;
 437         p->type = type;
 438 
 439         /* Flag analysis. */
 440 
 441         if (MDOC_SYNOPSIS & m->flags)
 442                 p->flags |= MDOC_SYNPRETTY;
 443         else
 444                 p->flags &= ~MDOC_SYNPRETTY;
 445         if (MDOC_NEWLINE & m->flags)
 446                 p->flags |= MDOC_LINE;
 447         m->flags &= ~MDOC_NEWLINE;
 448 
 449         return(p);
 450 }
 451 
 452 
 453 int
 454 mdoc_tail_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
 455 {
 456         struct mdoc_node *p;
 457 
 458         p = node_alloc(m, line, pos, tok, MDOC_TAIL);
 459         if ( ! node_append(m, p))
 460                 return(0);
 461         m->next = MDOC_NEXT_CHILD;
 462         return(1);
 463 }
 464 
 465 
 466 int
 467 mdoc_head_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
 468 {
 469         struct mdoc_node *p;
 470 
 471         assert(m->first);
 472         assert(m->last);
 473 
 474         p = node_alloc(m, line, pos, tok, MDOC_HEAD);
 475         if ( ! node_append(m, p))
 476                 return(0);
 477         m->next = MDOC_NEXT_CHILD;
 478         return(1);
 479 }
 480 
 481 
 482 int
 483 mdoc_body_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
 484 {
 485         struct mdoc_node *p;
 486 
 487         p = node_alloc(m, line, pos, tok, MDOC_BODY);
 488         if ( ! node_append(m, p))
 489                 return(0);
 490         m->next = MDOC_NEXT_CHILD;
 491         return(1);
 492 }
 493 
 494 
 495 int
 496 mdoc_endbody_alloc(struct mdoc *m, int line, int pos, enum mdoct tok,
 497                 struct mdoc_node *body, enum mdoc_endbody end)
 498 {
 499         struct mdoc_node *p;
 500 
 501         p = node_alloc(m, line, pos, tok, MDOC_BODY);
 502         p->pending = body;
 503         p->end = end;
 504         if ( ! node_append(m, p))
 505                 return(0);
 506         m->next = MDOC_NEXT_SIBLING;
 507         return(1);
 508 }
 509 
 510 
 511 int
 512 mdoc_block_alloc(struct mdoc *m, int line, int pos, 
 513                 enum mdoct tok, struct mdoc_arg *args)
 514 {
 515         struct mdoc_node *p;
 516 
 517         p = node_alloc(m, line, pos, tok, MDOC_BLOCK);
 518         p->args = args;
 519         if (p->args)
 520                 (args->refcnt)++;
 521 
 522         switch (tok) {
 523         case (MDOC_Bd):
 524                 /* FALLTHROUGH */
 525         case (MDOC_Bf):
 526                 /* FALLTHROUGH */
 527         case (MDOC_Bl):
 528                 /* FALLTHROUGH */
 529         case (MDOC_Rs):
 530                 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
 531                 break;
 532         default:
 533                 break;
 534         }
 535 
 536         if ( ! node_append(m, p))
 537                 return(0);
 538         m->next = MDOC_NEXT_CHILD;
 539         return(1);
 540 }
 541 
 542 
 543 int
 544 mdoc_elem_alloc(struct mdoc *m, int line, int pos, 
 545                 enum mdoct tok, struct mdoc_arg *args)
 546 {
 547         struct mdoc_node *p;
 548 
 549         p = node_alloc(m, line, pos, tok, MDOC_ELEM);
 550         p->args = args;
 551         if (p->args)
 552                 (args->refcnt)++;
 553 
 554         switch (tok) {
 555         case (MDOC_An):
 556                 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
 557                 break;
 558         default:
 559                 break;
 560         }
 561 
 562         if ( ! node_append(m, p))
 563                 return(0);
 564         m->next = MDOC_NEXT_CHILD;
 565         return(1);
 566 }
 567 
 568 int
 569 mdoc_word_alloc(struct mdoc *m, int line, int pos, const char *p)
 570 {
 571         struct mdoc_node *n;
 572 
 573         n = node_alloc(m, line, pos, MDOC_MAX, MDOC_TEXT);
 574         n->string = roff_strdup(m->roff, p);
 575 
 576         if ( ! node_append(m, n))
 577                 return(0);
 578 
 579         m->next = MDOC_NEXT_SIBLING;
 580         return(1);
 581 }
 582 
 583 
 584 static void
 585 mdoc_node_free(struct mdoc_node *p)
 586 {
 587 
 588         if (MDOC_BLOCK == p->type || MDOC_ELEM == p->type)
 589                 free(p->norm);
 590         if (p->string)
 591                 free(p->string);
 592         if (p->args)
 593                 mdoc_argv_free(p->args);
 594         free(p);
 595 }
 596 
 597 
 598 static void
 599 mdoc_node_unlink(struct mdoc *m, struct mdoc_node *n)
 600 {
 601 
 602         /* Adjust siblings. */
 603 
 604         if (n->prev)
 605                 n->prev->next = n->next;
 606         if (n->next)
 607                 n->next->prev = n->prev;
 608 
 609         /* Adjust parent. */
 610 
 611         if (n->parent) {
 612                 n->parent->nchild--;
 613                 if (n->parent->child == n)
 614                         n->parent->child = n->prev ? n->prev : n->next;
 615                 if (n->parent->last == n)
 616                         n->parent->last = n->prev ? n->prev : NULL;
 617         }
 618 
 619         /* Adjust parse point, if applicable. */
 620 
 621         if (m && m->last == n) {
 622                 if (n->prev) {
 623                         m->last = n->prev;
 624                         m->next = MDOC_NEXT_SIBLING;
 625                 } else {
 626                         m->last = n->parent;
 627                         m->next = MDOC_NEXT_CHILD;
 628                 }
 629         }
 630 
 631         if (m && m->first == n)
 632                 m->first = NULL;
 633 }
 634 
 635 
 636 void
 637 mdoc_node_delete(struct mdoc *m, struct mdoc_node *p)
 638 {
 639 
 640         while (p->child) {
 641                 assert(p->nchild);
 642                 mdoc_node_delete(m, p->child);
 643         }
 644         assert(0 == p->nchild);
 645 
 646         mdoc_node_unlink(m, p);
 647         mdoc_node_free(p);
 648 }
 649 
 650 #if 0
 651 /*
 652  * Pre-treat a text line.
 653  * Text lines can consist of equations, which must be handled apart from
 654  * the regular text.
 655  * Thus, use this function to step through a line checking if it has any
 656  * equations embedded in it.
 657  * This must handle multiple equations AND equations that do not end at
 658  * the end-of-line, i.e., will re-enter in the next roff parse.
 659  */
 660 static int
 661 mdoc_preptext(struct mdoc *m, int line, char *buf, int offs)
 662 {
 663         char            *start, *end;
 664         char             delim;
 665 
 666         while ('\0' != buf[offs]) {
 667                 /* Mark starting position if eqn is set. */
 668                 start = NULL;
 669                 if ('\0' != (delim = roff_eqndelim(m->roff)))
 670                         if (NULL != (start = strchr(buf + offs, delim)))
 671                                 *start++ = '\0';
 672 
 673                 /* Parse text as normal. */
 674                 if ( ! mdoc_ptext(m, line, buf, offs))
 675                         return(0);
 676 
 677                 /* Continue only if an equation exists. */
 678                 if (NULL == start)
 679                         break;
 680 
 681                 /* Read past the end of the equation. */
 682                 offs += start - (buf + offs);
 683                 assert(start == &buf[offs]);
 684                 if (NULL != (end = strchr(buf + offs, delim))) {
 685                         *end++ = '\0';
 686                         while (' ' == *end)
 687                                 end++;
 688                 }
 689 
 690                 /* Parse the equation itself. */
 691                 roff_openeqn(m->roff, NULL, line, offs, buf);
 692 
 693                 /* Process a finished equation? */
 694                 if (roff_closeeqn(m->roff))
 695                         if ( ! mdoc_addeqn(m, roff_eqn(m->roff)))
 696                                 return(0);
 697                 offs += (end - (buf + offs));
 698         } 
 699 
 700         return(1);
 701 }
 702 #endif
 703 
 704 /*
 705  * Parse free-form text, that is, a line that does not begin with the
 706  * control character.
 707  */
 708 static int
 709 mdoc_ptext(struct mdoc *m, int line, char *buf, int offs)
 710 {
 711         char             *c, *ws, *end;
 712         struct mdoc_node *n;
 713 
 714         /* No text before an initial macro. */
 715 
 716         if (SEC_NONE == m->lastnamed) {
 717                 mdoc_pmsg(m, line, offs, MANDOCERR_NOTEXT);
 718                 return(1);
 719         }
 720 
 721         assert(m->last);
 722         n = m->last;
 723 
 724         /*
 725          * Divert directly to list processing if we're encountering a
 726          * columnar MDOC_BLOCK with or without a prior MDOC_BLOCK entry
 727          * (a MDOC_BODY means it's already open, in which case we should
 728          * process within its context in the normal way).
 729          */
 730 
 731         if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
 732                         LIST_column == n->norm->Bl.type) {
 733                 /* `Bl' is open without any children. */
 734                 m->flags |= MDOC_FREECOL;
 735                 return(mdoc_macro(m, MDOC_It, line, offs, &offs, buf));
 736         }
 737 
 738         if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
 739                         NULL != n->parent &&
 740                         MDOC_Bl == n->parent->tok &&
 741                         LIST_column == n->parent->norm->Bl.type) {
 742                 /* `Bl' has block-level `It' children. */
 743                 m->flags |= MDOC_FREECOL;
 744                 return(mdoc_macro(m, MDOC_It, line, offs, &offs, buf));
 745         }
 746 
 747         /*
 748          * Search for the beginning of unescaped trailing whitespace (ws)
 749          * and for the first character not to be output (end).
 750          */
 751 
 752         /* FIXME: replace with strcspn(). */
 753         ws = NULL;
 754         for (c = end = buf + offs; *c; c++) {
 755                 switch (*c) {
 756                 case ' ':
 757                         if (NULL == ws)
 758                                 ws = c;
 759                         continue;
 760                 case '\t':
 761                         /*
 762                          * Always warn about trailing tabs,
 763                          * even outside literal context,
 764                          * where they should be put on the next line.
 765                          */
 766                         if (NULL == ws)
 767                                 ws = c;
 768                         /*
 769                          * Strip trailing tabs in literal context only;
 770                          * outside, they affect the next line.
 771                          */
 772                         if (MDOC_LITERAL & m->flags)
 773                                 continue;
 774                         break;
 775                 case '\\':
 776                         /* Skip the escaped character, too, if any. */
 777                         if (c[1])
 778                                 c++;
 779                         /* FALLTHROUGH */
 780                 default:
 781                         ws = NULL;
 782                         break;
 783                 }
 784                 end = c + 1;
 785         }
 786         *end = '\0';
 787 
 788         if (ws)
 789                 mdoc_pmsg(m, line, (int)(ws-buf), MANDOCERR_EOLNSPACE);
 790 
 791         if ('\0' == buf[offs] && ! (MDOC_LITERAL & m->flags)) {
 792                 mdoc_pmsg(m, line, (int)(c-buf), MANDOCERR_NOBLANKLN);
 793 
 794                 /*
 795                  * Insert a `sp' in the case of a blank line.  Technically,
 796                  * blank lines aren't allowed, but enough manuals assume this
 797                  * behaviour that we want to work around it.
 798                  */
 799                 if ( ! mdoc_elem_alloc(m, line, offs, MDOC_sp, NULL))
 800                         return(0);
 801 
 802                 m->next = MDOC_NEXT_SIBLING;
 803                 return(1);
 804         }
 805 
 806         if ( ! mdoc_word_alloc(m, line, offs, buf+offs))
 807                 return(0);
 808 
 809         if (MDOC_LITERAL & m->flags)
 810                 return(1);
 811 
 812         /*
 813          * End-of-sentence check.  If the last character is an unescaped
 814          * EOS character, then flag the node as being the end of a
 815          * sentence.  The front-end will know how to interpret this.
 816          */
 817 
 818         assert(buf < end);
 819 
 820         if (mandoc_eos(buf+offs, (size_t)(end-buf-offs), 0))
 821                 m->last->flags |= MDOC_EOS;
 822 
 823         return(1);
 824 }
 825 
 826 
 827 /*
 828  * Parse a macro line, that is, a line beginning with the control
 829  * character.
 830  */
 831 static int
 832 mdoc_pmacro(struct mdoc *m, int ln, char *buf, int offs)
 833 {
 834         enum mdoct        tok;
 835         int               i, sv;
 836         char              mac[5];
 837         struct mdoc_node *n;
 838 
 839         /* Empty post-control lines are ignored. */
 840 
 841         if ('"' == buf[offs]) {
 842                 mdoc_pmsg(m, ln, offs, MANDOCERR_BADCOMMENT);
 843                 return(1);
 844         } else if ('\0' == buf[offs])
 845                 return(1);
 846 
 847         sv = offs;
 848 
 849         /* 
 850          * Copy the first word into a nil-terminated buffer.
 851          * Stop copying when a tab, space, or eoln is encountered.
 852          */
 853 
 854         i = 0;
 855         while (i < 4 && '\0' != buf[offs] && 
 856                         ' ' != buf[offs] && '\t' != buf[offs])
 857                 mac[i++] = buf[offs++];
 858 
 859         mac[i] = '\0';
 860 
 861         tok = (i > 1 || i < 4) ? mdoc_hash_find(mac) : MDOC_MAX;
 862 
 863         if (MDOC_MAX == tok) {
 864                 mandoc_vmsg(MANDOCERR_MACRO, m->parse, 
 865                                 ln, sv, "%s", buf + sv - 1);
 866                 return(1);
 867         }
 868 
 869         /* Disregard the first trailing tab, if applicable. */
 870 
 871         if ('\t' == buf[offs])
 872                 offs++;
 873 
 874         /* Jump to the next non-whitespace word. */
 875 
 876         while (buf[offs] && ' ' == buf[offs])
 877                 offs++;
 878 
 879         /* 
 880          * Trailing whitespace.  Note that tabs are allowed to be passed
 881          * into the parser as "text", so we only warn about spaces here.
 882          */
 883 
 884         if ('\0' == buf[offs] && ' ' == buf[offs - 1])
 885                 mdoc_pmsg(m, ln, offs - 1, MANDOCERR_EOLNSPACE);
 886 
 887         /*
 888          * If an initial macro or a list invocation, divert directly
 889          * into macro processing.
 890          */
 891 
 892         if (NULL == m->last || MDOC_It == tok || MDOC_El == tok) {
 893                 if ( ! mdoc_macro(m, tok, ln, sv, &offs, buf)) 
 894                         goto err;
 895                 return(1);
 896         }
 897 
 898         n = m->last;
 899         assert(m->last);
 900 
 901         /*
 902          * If the first macro of a `Bl -column', open an `It' block
 903          * context around the parsed macro.
 904          */
 905 
 906         if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
 907                         LIST_column == n->norm->Bl.type) {
 908                 m->flags |= MDOC_FREECOL;
 909                 if ( ! mdoc_macro(m, MDOC_It, ln, sv, &sv, buf))
 910                         goto err;
 911                 return(1);
 912         }
 913 
 914         /*
 915          * If we're following a block-level `It' within a `Bl -column'
 916          * context (perhaps opened in the above block or in ptext()),
 917          * then open an `It' block context around the parsed macro.
 918          */
 919 
 920         if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
 921                         NULL != n->parent &&
 922                         MDOC_Bl == n->parent->tok &&
 923                         LIST_column == n->parent->norm->Bl.type) {
 924                 m->flags |= MDOC_FREECOL;
 925                 if ( ! mdoc_macro(m, MDOC_It, ln, sv, &sv, buf)) 
 926                         goto err;
 927                 return(1);
 928         }
 929 
 930         /* Normal processing of a macro. */
 931 
 932         if ( ! mdoc_macro(m, tok, ln, sv, &offs, buf)) 
 933                 goto err;
 934 
 935         return(1);
 936 
 937 err:    /* Error out. */
 938 
 939         m->flags |= MDOC_HALT;
 940         return(0);
 941 }
 942 
 943 enum mdelim
 944 mdoc_isdelim(const char *p)
 945 {
 946 
 947         if ('\0' == p[0])
 948                 return(DELIM_NONE);
 949 
 950         if ('\0' == p[1])
 951                 switch (p[0]) {
 952                 case('('):
 953                         /* FALLTHROUGH */
 954                 case('['):
 955                         return(DELIM_OPEN);
 956                 case('|'):
 957                         return(DELIM_MIDDLE);
 958                 case('.'):
 959                         /* FALLTHROUGH */
 960                 case(','):
 961                         /* FALLTHROUGH */
 962                 case(';'):
 963                         /* FALLTHROUGH */
 964                 case(':'):
 965                         /* FALLTHROUGH */
 966                 case('?'):
 967                         /* FALLTHROUGH */
 968                 case('!'):
 969                         /* FALLTHROUGH */
 970                 case(')'):
 971                         /* FALLTHROUGH */
 972                 case(']'):
 973                         return(DELIM_CLOSE);
 974                 default:
 975                         return(DELIM_NONE);
 976                 }
 977 
 978         if ('\\' != p[0])
 979                 return(DELIM_NONE);
 980 
 981         if (0 == strcmp(p + 1, "."))
 982                 return(DELIM_CLOSE);
 983         if (0 == strcmp(p + 1, "*(Ba"))
 984                 return(DELIM_MIDDLE);
 985 
 986         return(DELIM_NONE);
 987 }