1 /*      $Id: mdoc.c,v 1.206 2013/12/24 19:11:46 schwarze Exp $ */
   2 /*
   3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
   4  * Copyright (c) 2010, 2012, 2013 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 *mdoc)
 108 {
 109 
 110         assert( ! (MDOC_HALT & mdoc->flags));
 111         return(mdoc->first);
 112 }
 113 
 114 
 115 const struct mdoc_meta *
 116 mdoc_meta(const struct mdoc *mdoc)
 117 {
 118 
 119         assert( ! (MDOC_HALT & mdoc->flags));
 120         return(&mdoc->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, char *defos)
 201 {
 202         struct mdoc     *p;
 203 
 204         p = mandoc_calloc(1, sizeof(struct mdoc));
 205 
 206         p->parse = parse;
 207         p->defos = defos;
 208         p->roff = roff;
 209 
 210         mdoc_hash_init();
 211         mdoc_alloc1(p);
 212         return(p);
 213 }
 214 
 215 
 216 /*
 217  * Climb back up the parse tree, validating open scopes.  Mostly calls
 218  * through to macro_end() in macro.c.
 219  */
 220 int
 221 mdoc_endparse(struct mdoc *mdoc)
 222 {
 223 
 224         assert( ! (MDOC_HALT & mdoc->flags));
 225         if (mdoc_macroend(mdoc))
 226                 return(1);
 227         mdoc->flags |= MDOC_HALT;
 228         return(0);
 229 }
 230 
 231 int
 232 mdoc_addeqn(struct mdoc *mdoc, const struct eqn *ep)
 233 {
 234         struct mdoc_node *n;
 235 
 236         assert( ! (MDOC_HALT & mdoc->flags));
 237 
 238         /* No text before an initial macro. */
 239 
 240         if (SEC_NONE == mdoc->lastnamed) {
 241                 mdoc_pmsg(mdoc, ep->ln, ep->pos, MANDOCERR_NOTEXT);
 242                 return(1);
 243         }
 244 
 245         n = node_alloc(mdoc, ep->ln, ep->pos, MDOC_MAX, MDOC_EQN);
 246         n->eqn = ep;
 247 
 248         if ( ! node_append(mdoc, n))
 249                 return(0);
 250 
 251         mdoc->next = MDOC_NEXT_SIBLING;
 252         return(1);
 253 }
 254 
 255 int
 256 mdoc_addspan(struct mdoc *mdoc, const struct tbl_span *sp)
 257 {
 258         struct mdoc_node *n;
 259 
 260         assert( ! (MDOC_HALT & mdoc->flags));
 261 
 262         /* No text before an initial macro. */
 263 
 264         if (SEC_NONE == mdoc->lastnamed) {
 265                 mdoc_pmsg(mdoc, sp->line, 0, MANDOCERR_NOTEXT);
 266                 return(1);
 267         }
 268 
 269         n = node_alloc(mdoc, sp->line, 0, MDOC_MAX, MDOC_TBL);
 270         n->span = sp;
 271 
 272         if ( ! node_append(mdoc, n))
 273                 return(0);
 274 
 275         mdoc->next = MDOC_NEXT_SIBLING;
 276         return(1);
 277 }
 278 
 279 
 280 /*
 281  * Main parse routine.  Parses a single line -- really just hands off to
 282  * the macro (mdoc_pmacro()) or text parser (mdoc_ptext()).
 283  */
 284 int
 285 mdoc_parseln(struct mdoc *mdoc, int ln, char *buf, int offs)
 286 {
 287 
 288         assert( ! (MDOC_HALT & mdoc->flags));
 289 
 290         mdoc->flags |= MDOC_NEWLINE;
 291 
 292         /*
 293          * Let the roff nS register switch SYNOPSIS mode early,
 294          * such that the parser knows at all times
 295          * whether this mode is on or off.
 296          * Note that this mode is also switched by the Sh macro.
 297          */
 298         if (roff_getreg(mdoc->roff, "nS"))
 299                 mdoc->flags |= MDOC_SYNOPSIS;
 300         else
 301                 mdoc->flags &= ~MDOC_SYNOPSIS;
 302 
 303         return(roff_getcontrol(mdoc->roff, buf, &offs) ?
 304                         mdoc_pmacro(mdoc, ln, buf, offs) :
 305                         mdoc_ptext(mdoc, ln, buf, offs));
 306 }
 307 
 308 int
 309 mdoc_macro(MACRO_PROT_ARGS)
 310 {
 311         assert(tok < MDOC_MAX);
 312 
 313         /* If we're in the body, deny prologue calls. */
 314 
 315         if (MDOC_PROLOGUE & mdoc_macros[tok].flags && 
 316                         MDOC_PBODY & mdoc->flags) {
 317                 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_BADBODY);
 318                 return(1);
 319         }
 320 
 321         /* If we're in the prologue, deny "body" macros.  */
 322 
 323         if ( ! (MDOC_PROLOGUE & mdoc_macros[tok].flags) && 
 324                         ! (MDOC_PBODY & mdoc->flags)) {
 325                 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_BADPROLOG);
 326                 if (NULL == mdoc->meta.msec)
 327                         mdoc->meta.msec = mandoc_strdup("1");
 328                 if (NULL == mdoc->meta.title)
 329                         mdoc->meta.title = mandoc_strdup("UNKNOWN");
 330                 if (NULL == mdoc->meta.vol)
 331                         mdoc->meta.vol = mandoc_strdup("LOCAL");
 332                 if (NULL == mdoc->meta.os)
 333                         mdoc->meta.os = mandoc_strdup("LOCAL");
 334                 if (NULL == mdoc->meta.date)
 335                         mdoc->meta.date = mandoc_normdate
 336                                 (mdoc->parse, NULL, line, ppos);
 337                 mdoc->flags |= MDOC_PBODY;
 338         }
 339 
 340         return((*mdoc_macros[tok].fp)(mdoc, tok, line, ppos, pos, buf));
 341 }
 342 
 343 
 344 static int
 345 node_append(struct mdoc *mdoc, struct mdoc_node *p)
 346 {
 347 
 348         assert(mdoc->last);
 349         assert(mdoc->first);
 350         assert(MDOC_ROOT != p->type);
 351 
 352         switch (mdoc->next) {
 353         case (MDOC_NEXT_SIBLING):
 354                 mdoc->last->next = p;
 355                 p->prev = mdoc->last;
 356                 p->parent = mdoc->last->parent;
 357                 break;
 358         case (MDOC_NEXT_CHILD):
 359                 mdoc->last->child = p;
 360                 p->parent = mdoc->last;
 361                 break;
 362         default:
 363                 abort();
 364                 /* NOTREACHED */
 365         }
 366 
 367         p->parent->nchild++;
 368 
 369         /*
 370          * Copy over the normalised-data pointer of our parent.  Not
 371          * everybody has one, but copying a null pointer is fine.
 372          */
 373 
 374         switch (p->type) {
 375         case (MDOC_BODY):
 376                 if (ENDBODY_NOT != p->end)
 377                         break;
 378                 /* FALLTHROUGH */
 379         case (MDOC_TAIL):
 380                 /* FALLTHROUGH */
 381         case (MDOC_HEAD):
 382                 p->norm = p->parent->norm;
 383                 break;
 384         default:
 385                 break;
 386         }
 387 
 388         if ( ! mdoc_valid_pre(mdoc, p))
 389                 return(0);
 390 
 391         switch (p->type) {
 392         case (MDOC_HEAD):
 393                 assert(MDOC_BLOCK == p->parent->type);
 394                 p->parent->head = p;
 395                 break;
 396         case (MDOC_TAIL):
 397                 assert(MDOC_BLOCK == p->parent->type);
 398                 p->parent->tail = p;
 399                 break;
 400         case (MDOC_BODY):
 401                 if (p->end)
 402                         break;
 403                 assert(MDOC_BLOCK == p->parent->type);
 404                 p->parent->body = p;
 405                 break;
 406         default:
 407                 break;
 408         }
 409 
 410         mdoc->last = p;
 411 
 412         switch (p->type) {
 413         case (MDOC_TBL):
 414                 /* FALLTHROUGH */
 415         case (MDOC_TEXT):
 416                 if ( ! mdoc_valid_post(mdoc))
 417                         return(0);
 418                 break;
 419         default:
 420                 break;
 421         }
 422 
 423         return(1);
 424 }
 425 
 426 
 427 static struct mdoc_node *
 428 node_alloc(struct mdoc *mdoc, int line, int pos, 
 429                 enum mdoct tok, enum mdoc_type type)
 430 {
 431         struct mdoc_node *p;
 432 
 433         p = mandoc_calloc(1, sizeof(struct mdoc_node));
 434         p->sec = mdoc->lastsec;
 435         p->line = line;
 436         p->pos = pos;
 437         p->lastline = line;
 438         p->tok = tok;
 439         p->type = type;
 440 
 441         /* Flag analysis. */
 442 
 443         if (MDOC_SYNOPSIS & mdoc->flags)
 444                 p->flags |= MDOC_SYNPRETTY;
 445         else
 446                 p->flags &= ~MDOC_SYNPRETTY;
 447         if (MDOC_NEWLINE & mdoc->flags)
 448                 p->flags |= MDOC_LINE;
 449         mdoc->flags &= ~MDOC_NEWLINE;
 450 
 451         return(p);
 452 }
 453 
 454 
 455 int
 456 mdoc_tail_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
 457 {
 458         struct mdoc_node *p;
 459 
 460         p = node_alloc(mdoc, line, pos, tok, MDOC_TAIL);
 461         if ( ! node_append(mdoc, p))
 462                 return(0);
 463         mdoc->next = MDOC_NEXT_CHILD;
 464         return(1);
 465 }
 466 
 467 
 468 int
 469 mdoc_head_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
 470 {
 471         struct mdoc_node *p;
 472 
 473         assert(mdoc->first);
 474         assert(mdoc->last);
 475 
 476         p = node_alloc(mdoc, line, pos, tok, MDOC_HEAD);
 477         if ( ! node_append(mdoc, p))
 478                 return(0);
 479         mdoc->next = MDOC_NEXT_CHILD;
 480         return(1);
 481 }
 482 
 483 
 484 int
 485 mdoc_body_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
 486 {
 487         struct mdoc_node *p;
 488 
 489         p = node_alloc(mdoc, line, pos, tok, MDOC_BODY);
 490         if ( ! node_append(mdoc, p))
 491                 return(0);
 492         mdoc->next = MDOC_NEXT_CHILD;
 493         return(1);
 494 }
 495 
 496 
 497 int
 498 mdoc_endbody_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok,
 499                 struct mdoc_node *body, enum mdoc_endbody end)
 500 {
 501         struct mdoc_node *p;
 502 
 503         p = node_alloc(mdoc, line, pos, tok, MDOC_BODY);
 504         p->pending = body;
 505         p->norm = body->norm;
 506         p->end = end;
 507         if ( ! node_append(mdoc, p))
 508                 return(0);
 509         mdoc->next = MDOC_NEXT_SIBLING;
 510         return(1);
 511 }
 512 
 513 
 514 int
 515 mdoc_block_alloc(struct mdoc *mdoc, int line, int pos, 
 516                 enum mdoct tok, struct mdoc_arg *args)
 517 {
 518         struct mdoc_node *p;
 519 
 520         p = node_alloc(mdoc, line, pos, tok, MDOC_BLOCK);
 521         p->args = args;
 522         if (p->args)
 523                 (args->refcnt)++;
 524 
 525         switch (tok) {
 526         case (MDOC_Bd):
 527                 /* FALLTHROUGH */
 528         case (MDOC_Bf):
 529                 /* FALLTHROUGH */
 530         case (MDOC_Bl):
 531                 /* FALLTHROUGH */
 532         case (MDOC_Rs):
 533                 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
 534                 break;
 535         default:
 536                 break;
 537         }
 538 
 539         if ( ! node_append(mdoc, p))
 540                 return(0);
 541         mdoc->next = MDOC_NEXT_CHILD;
 542         return(1);
 543 }
 544 
 545 
 546 int
 547 mdoc_elem_alloc(struct mdoc *mdoc, int line, int pos, 
 548                 enum mdoct tok, struct mdoc_arg *args)
 549 {
 550         struct mdoc_node *p;
 551 
 552         p = node_alloc(mdoc, line, pos, tok, MDOC_ELEM);
 553         p->args = args;
 554         if (p->args)
 555                 (args->refcnt)++;
 556 
 557         switch (tok) {
 558         case (MDOC_An):
 559                 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
 560                 break;
 561         default:
 562                 break;
 563         }
 564 
 565         if ( ! node_append(mdoc, p))
 566                 return(0);
 567         mdoc->next = MDOC_NEXT_CHILD;
 568         return(1);
 569 }
 570 
 571 int
 572 mdoc_word_alloc(struct mdoc *mdoc, int line, int pos, const char *p)
 573 {
 574         struct mdoc_node *n;
 575 
 576         n = node_alloc(mdoc, line, pos, MDOC_MAX, MDOC_TEXT);
 577         n->string = roff_strdup(mdoc->roff, p);
 578 
 579         if ( ! node_append(mdoc, n))
 580                 return(0);
 581 
 582         mdoc->next = MDOC_NEXT_SIBLING;
 583         return(1);
 584 }
 585 
 586 void
 587 mdoc_word_append(struct mdoc *mdoc, const char *p)
 588 {
 589         struct mdoc_node        *n;
 590         char                    *addstr, *newstr;
 591 
 592         n = mdoc->last;
 593         addstr = roff_strdup(mdoc->roff, p);
 594         if (-1 == asprintf(&newstr, "%s %s", n->string, addstr)) {
 595                 perror(NULL);
 596                 exit((int)MANDOCLEVEL_SYSERR);
 597         }
 598         free(addstr);
 599         free(n->string);
 600         n->string = newstr;
 601         mdoc->next = MDOC_NEXT_SIBLING;
 602 }
 603 
 604 static void
 605 mdoc_node_free(struct mdoc_node *p)
 606 {
 607 
 608         if (MDOC_BLOCK == p->type || MDOC_ELEM == p->type)
 609                 free(p->norm);
 610         if (p->string)
 611                 free(p->string);
 612         if (p->args)
 613                 mdoc_argv_free(p->args);
 614         free(p);
 615 }
 616 
 617 
 618 static void
 619 mdoc_node_unlink(struct mdoc *mdoc, struct mdoc_node *n)
 620 {
 621 
 622         /* Adjust siblings. */
 623 
 624         if (n->prev)
 625                 n->prev->next = n->next;
 626         if (n->next)
 627                 n->next->prev = n->prev;
 628 
 629         /* Adjust parent. */
 630 
 631         if (n->parent) {
 632                 n->parent->nchild--;
 633                 if (n->parent->child == n)
 634                         n->parent->child = n->prev ? n->prev : n->next;
 635                 if (n->parent->last == n)
 636                         n->parent->last = n->prev ? n->prev : NULL;
 637         }
 638 
 639         /* Adjust parse point, if applicable. */
 640 
 641         if (mdoc && mdoc->last == n) {
 642                 if (n->prev) {
 643                         mdoc->last = n->prev;
 644                         mdoc->next = MDOC_NEXT_SIBLING;
 645                 } else {
 646                         mdoc->last = n->parent;
 647                         mdoc->next = MDOC_NEXT_CHILD;
 648                 }
 649         }
 650 
 651         if (mdoc && mdoc->first == n)
 652                 mdoc->first = NULL;
 653 }
 654 
 655 
 656 void
 657 mdoc_node_delete(struct mdoc *mdoc, struct mdoc_node *p)
 658 {
 659 
 660         while (p->child) {
 661                 assert(p->nchild);
 662                 mdoc_node_delete(mdoc, p->child);
 663         }
 664         assert(0 == p->nchild);
 665 
 666         mdoc_node_unlink(mdoc, p);
 667         mdoc_node_free(p);
 668 }
 669 
 670 int
 671 mdoc_node_relink(struct mdoc *mdoc, struct mdoc_node *p)
 672 {
 673 
 674         mdoc_node_unlink(mdoc, p);
 675         return(node_append(mdoc, p));
 676 }
 677 
 678 #if 0
 679 /*
 680  * Pre-treat a text line.
 681  * Text lines can consist of equations, which must be handled apart from
 682  * the regular text.
 683  * Thus, use this function to step through a line checking if it has any
 684  * equations embedded in it.
 685  * This must handle multiple equations AND equations that do not end at
 686  * the end-of-line, i.e., will re-enter in the next roff parse.
 687  */
 688 static int
 689 mdoc_preptext(struct mdoc *mdoc, int line, char *buf, int offs)
 690 {
 691         char            *start, *end;
 692         char             delim;
 693 
 694         while ('\0' != buf[offs]) {
 695                 /* Mark starting position if eqn is set. */
 696                 start = NULL;
 697                 if ('\0' != (delim = roff_eqndelim(mdoc->roff)))
 698                         if (NULL != (start = strchr(buf + offs, delim)))
 699                                 *start++ = '\0';
 700 
 701                 /* Parse text as normal. */
 702                 if ( ! mdoc_ptext(mdoc, line, buf, offs))
 703                         return(0);
 704 
 705                 /* Continue only if an equation exists. */
 706                 if (NULL == start)
 707                         break;
 708 
 709                 /* Read past the end of the equation. */
 710                 offs += start - (buf + offs);
 711                 assert(start == &buf[offs]);
 712                 if (NULL != (end = strchr(buf + offs, delim))) {
 713                         *end++ = '\0';
 714                         while (' ' == *end)
 715                                 end++;
 716                 }
 717 
 718                 /* Parse the equation itself. */
 719                 roff_openeqn(mdoc->roff, NULL, line, offs, buf);
 720 
 721                 /* Process a finished equation? */
 722                 if (roff_closeeqn(mdoc->roff))
 723                         if ( ! mdoc_addeqn(mdoc, roff_eqn(mdoc->roff)))
 724                                 return(0);
 725                 offs += (end - (buf + offs));
 726         } 
 727 
 728         return(1);
 729 }
 730 #endif
 731 
 732 /*
 733  * Parse free-form text, that is, a line that does not begin with the
 734  * control character.
 735  */
 736 static int
 737 mdoc_ptext(struct mdoc *mdoc, int line, char *buf, int offs)
 738 {
 739         char             *c, *ws, *end;
 740         struct mdoc_node *n;
 741 
 742         /* No text before an initial macro. */
 743 
 744         if (SEC_NONE == mdoc->lastnamed) {
 745                 mdoc_pmsg(mdoc, line, offs, MANDOCERR_NOTEXT);
 746                 return(1);
 747         }
 748 
 749         assert(mdoc->last);
 750         n = mdoc->last;
 751 
 752         /*
 753          * Divert directly to list processing if we're encountering a
 754          * columnar MDOC_BLOCK with or without a prior MDOC_BLOCK entry
 755          * (a MDOC_BODY means it's already open, in which case we should
 756          * process within its context in the normal way).
 757          */
 758 
 759         if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
 760                         LIST_column == n->norm->Bl.type) {
 761                 /* `Bl' is open without any children. */
 762                 mdoc->flags |= MDOC_FREECOL;
 763                 return(mdoc_macro(mdoc, MDOC_It, line, offs, &offs, buf));
 764         }
 765 
 766         if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
 767                         NULL != n->parent &&
 768                         MDOC_Bl == n->parent->tok &&
 769                         LIST_column == n->parent->norm->Bl.type) {
 770                 /* `Bl' has block-level `It' children. */
 771                 mdoc->flags |= MDOC_FREECOL;
 772                 return(mdoc_macro(mdoc, MDOC_It, line, offs, &offs, buf));
 773         }
 774 
 775         /*
 776          * Search for the beginning of unescaped trailing whitespace (ws)
 777          * and for the first character not to be output (end).
 778          */
 779 
 780         /* FIXME: replace with strcspn(). */
 781         ws = NULL;
 782         for (c = end = buf + offs; *c; c++) {
 783                 switch (*c) {
 784                 case ' ':
 785                         if (NULL == ws)
 786                                 ws = c;
 787                         continue;
 788                 case '\t':
 789                         /*
 790                          * Always warn about trailing tabs,
 791                          * even outside literal context,
 792                          * where they should be put on the next line.
 793                          */
 794                         if (NULL == ws)
 795                                 ws = c;
 796                         /*
 797                          * Strip trailing tabs in literal context only;
 798                          * outside, they affect the next line.
 799                          */
 800                         if (MDOC_LITERAL & mdoc->flags)
 801                                 continue;
 802                         break;
 803                 case '\\':
 804                         /* Skip the escaped character, too, if any. */
 805                         if (c[1])
 806                                 c++;
 807                         /* FALLTHROUGH */
 808                 default:
 809                         ws = NULL;
 810                         break;
 811                 }
 812                 end = c + 1;
 813         }
 814         *end = '\0';
 815 
 816         if (ws)
 817                 mdoc_pmsg(mdoc, line, (int)(ws-buf), MANDOCERR_EOLNSPACE);
 818 
 819         if ('\0' == buf[offs] && ! (MDOC_LITERAL & mdoc->flags)) {
 820                 mdoc_pmsg(mdoc, line, (int)(c-buf), MANDOCERR_NOBLANKLN);
 821 
 822                 /*
 823                  * Insert a `sp' in the case of a blank line.  Technically,
 824                  * blank lines aren't allowed, but enough manuals assume this
 825                  * behaviour that we want to work around it.
 826                  */
 827                 if ( ! mdoc_elem_alloc(mdoc, line, offs, MDOC_sp, NULL))
 828                         return(0);
 829 
 830                 mdoc->next = MDOC_NEXT_SIBLING;
 831 
 832                 return(mdoc_valid_post(mdoc));
 833         }
 834 
 835         if ( ! mdoc_word_alloc(mdoc, line, offs, buf+offs))
 836                 return(0);
 837 
 838         if (MDOC_LITERAL & mdoc->flags)
 839                 return(1);
 840 
 841         /*
 842          * End-of-sentence check.  If the last character is an unescaped
 843          * EOS character, then flag the node as being the end of a
 844          * sentence.  The front-end will know how to interpret this.
 845          */
 846 
 847         assert(buf < end);
 848 
 849         if (mandoc_eos(buf+offs, (size_t)(end-buf-offs), 0))
 850                 mdoc->last->flags |= MDOC_EOS;
 851 
 852         return(1);
 853 }
 854 
 855 
 856 /*
 857  * Parse a macro line, that is, a line beginning with the control
 858  * character.
 859  */
 860 static int
 861 mdoc_pmacro(struct mdoc *mdoc, int ln, char *buf, int offs)
 862 {
 863         enum mdoct        tok;
 864         int               i, sv;
 865         char              mac[5];
 866         struct mdoc_node *n;
 867 
 868         /* Empty post-control lines are ignored. */
 869 
 870         if ('"' == buf[offs]) {
 871                 mdoc_pmsg(mdoc, ln, offs, MANDOCERR_BADCOMMENT);
 872                 return(1);
 873         } else if ('\0' == buf[offs])
 874                 return(1);
 875 
 876         sv = offs;
 877 
 878         /* 
 879          * Copy the first word into a nil-terminated buffer.
 880          * Stop copying when a tab, space, or eoln is encountered.
 881          */
 882 
 883         i = 0;
 884         while (i < 4 && '\0' != buf[offs] && 
 885                         ' ' != buf[offs] && '\t' != buf[offs])
 886                 mac[i++] = buf[offs++];
 887 
 888         mac[i] = '\0';
 889 
 890         tok = (i > 1 || i < 4) ? mdoc_hash_find(mac) : MDOC_MAX;
 891 
 892         if (MDOC_MAX == tok) {
 893                 mandoc_vmsg(MANDOCERR_MACRO, mdoc->parse, 
 894                                 ln, sv, "%s", buf + sv - 1);
 895                 return(1);
 896         }
 897 
 898         /* Disregard the first trailing tab, if applicable. */
 899 
 900         if ('\t' == buf[offs])
 901                 offs++;
 902 
 903         /* Jump to the next non-whitespace word. */
 904 
 905         while (buf[offs] && ' ' == buf[offs])
 906                 offs++;
 907 
 908         /* 
 909          * Trailing whitespace.  Note that tabs are allowed to be passed
 910          * into the parser as "text", so we only warn about spaces here.
 911          */
 912 
 913         if ('\0' == buf[offs] && ' ' == buf[offs - 1])
 914                 mdoc_pmsg(mdoc, ln, offs - 1, MANDOCERR_EOLNSPACE);
 915 
 916         /*
 917          * If an initial macro or a list invocation, divert directly
 918          * into macro processing.
 919          */
 920 
 921         if (NULL == mdoc->last || MDOC_It == tok || MDOC_El == tok) {
 922                 if ( ! mdoc_macro(mdoc, tok, ln, sv, &offs, buf)) 
 923                         goto err;
 924                 return(1);
 925         }
 926 
 927         n = mdoc->last;
 928         assert(mdoc->last);
 929 
 930         /*
 931          * If the first macro of a `Bl -column', open an `It' block
 932          * context around the parsed macro.
 933          */
 934 
 935         if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
 936                         LIST_column == n->norm->Bl.type) {
 937                 mdoc->flags |= MDOC_FREECOL;
 938                 if ( ! mdoc_macro(mdoc, MDOC_It, ln, sv, &sv, buf))
 939                         goto err;
 940                 return(1);
 941         }
 942 
 943         /*
 944          * If we're following a block-level `It' within a `Bl -column'
 945          * context (perhaps opened in the above block or in ptext()),
 946          * then open an `It' block context around the parsed macro.
 947          */
 948 
 949         if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
 950                         NULL != n->parent &&
 951                         MDOC_Bl == n->parent->tok &&
 952                         LIST_column == n->parent->norm->Bl.type) {
 953                 mdoc->flags |= MDOC_FREECOL;
 954                 if ( ! mdoc_macro(mdoc, MDOC_It, ln, sv, &sv, buf)) 
 955                         goto err;
 956                 return(1);
 957         }
 958 
 959         /* Normal processing of a macro. */
 960 
 961         if ( ! mdoc_macro(mdoc, tok, ln, sv, &offs, buf)) 
 962                 goto err;
 963 
 964         return(1);
 965 
 966 err:    /* Error out. */
 967 
 968         mdoc->flags |= MDOC_HALT;
 969         return(0);
 970 }
 971 
 972 enum mdelim
 973 mdoc_isdelim(const char *p)
 974 {
 975 
 976         if ('\0' == p[0])
 977                 return(DELIM_NONE);
 978 
 979         if ('\0' == p[1])
 980                 switch (p[0]) {
 981                 case('('):
 982                         /* FALLTHROUGH */
 983                 case('['):
 984                         return(DELIM_OPEN);
 985                 case('|'):
 986                         return(DELIM_MIDDLE);
 987                 case('.'):
 988                         /* FALLTHROUGH */
 989                 case(','):
 990                         /* FALLTHROUGH */
 991                 case(';'):
 992                         /* FALLTHROUGH */
 993                 case(':'):
 994                         /* FALLTHROUGH */
 995                 case('?'):
 996                         /* FALLTHROUGH */
 997                 case('!'):
 998                         /* FALLTHROUGH */
 999                 case(')'):
1000                         /* FALLTHROUGH */
1001                 case(']'):
1002                         return(DELIM_CLOSE);
1003                 default:
1004                         return(DELIM_NONE);
1005                 }
1006 
1007         if ('\\' != p[0])
1008                 return(DELIM_NONE);
1009 
1010         if (0 == strcmp(p + 1, "."))
1011                 return(DELIM_CLOSE);
1012         if (0 == strcmp(p + 1, "fR|\\fP"))
1013                 return(DELIM_MIDDLE);
1014 
1015         return(DELIM_NONE);
1016 }