1 /*      $Id: man_macro.c,v 1.79 2013/12/25 00:50:05 schwarze Exp $ */
   2 /*
   3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
   4  * Copyright (c) 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
   5  * Copyright (c) 2013 Franco Fichtner <franco@lastsummer.de>
   6  *
   7  * Permission to use, copy, modify, and distribute this software for any
   8  * purpose with or without fee is hereby granted, provided that the above
   9  * copyright notice and this permission notice appear in all copies.
  10  *
  11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  18  */
  19 #ifdef HAVE_CONFIG_H
  20 #include "config.h"
  21 #endif
  22 
  23 #include <assert.h>
  24 #include <ctype.h>
  25 #include <stdlib.h>
  26 #include <string.h>
  27 
  28 #include "man.h"
  29 #include "mandoc.h"
  30 #include "libmandoc.h"
  31 #include "libman.h"
  32 
  33 enum    rew {
  34         REW_REWIND,
  35         REW_NOHALT,
  36         REW_HALT
  37 };
  38 
  39 static  int              blk_close(MACRO_PROT_ARGS);
  40 static  int              blk_exp(MACRO_PROT_ARGS);
  41 static  int              blk_imp(MACRO_PROT_ARGS);
  42 static  int              in_line_eoln(MACRO_PROT_ARGS);
  43 static  int              man_args(struct man *, int, 
  44                                 int *, char *, char **);
  45 
  46 static  int              rew_scope(enum man_type, 
  47                                 struct man *, enum mant);
  48 static  enum rew         rew_dohalt(enum mant, enum man_type, 
  49                                 const struct man_node *);
  50 static  enum rew         rew_block(enum mant, enum man_type, 
  51                                 const struct man_node *);
  52 static  void             rew_warn(struct man *, 
  53                                 struct man_node *, enum mandocerr);
  54 
  55 const   struct man_macro __man_macros[MAN_MAX] = {
  56         { in_line_eoln, MAN_NSCOPED }, /* br */
  57         { in_line_eoln, MAN_BSCOPE }, /* TH */
  58         { blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SH */
  59         { blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SS */
  60         { blk_imp, MAN_BSCOPE | MAN_SCOPED | MAN_FSCOPED }, /* TP */
  61         { blk_imp, MAN_BSCOPE }, /* LP */
  62         { blk_imp, MAN_BSCOPE }, /* PP */
  63         { blk_imp, MAN_BSCOPE }, /* P */
  64         { blk_imp, MAN_BSCOPE }, /* IP */
  65         { blk_imp, MAN_BSCOPE }, /* HP */
  66         { in_line_eoln, MAN_SCOPED }, /* SM */
  67         { in_line_eoln, MAN_SCOPED }, /* SB */
  68         { in_line_eoln, 0 }, /* BI */
  69         { in_line_eoln, 0 }, /* IB */
  70         { in_line_eoln, 0 }, /* BR */
  71         { in_line_eoln, 0 }, /* RB */
  72         { in_line_eoln, MAN_SCOPED }, /* R */
  73         { in_line_eoln, MAN_SCOPED }, /* B */
  74         { in_line_eoln, MAN_SCOPED }, /* I */
  75         { in_line_eoln, 0 }, /* IR */
  76         { in_line_eoln, 0 }, /* RI */
  77         { in_line_eoln, MAN_NSCOPED }, /* na */
  78         { in_line_eoln, MAN_NSCOPED }, /* sp */
  79         { in_line_eoln, MAN_BSCOPE }, /* nf */
  80         { in_line_eoln, MAN_BSCOPE }, /* fi */
  81         { blk_close, 0 }, /* RE */
  82         { blk_exp, MAN_BSCOPE | MAN_EXPLICIT }, /* RS */
  83         { in_line_eoln, 0 }, /* DT */
  84         { in_line_eoln, 0 }, /* UC */
  85         { in_line_eoln, 0 }, /* PD */
  86         { in_line_eoln, 0 }, /* AT */
  87         { in_line_eoln, 0 }, /* in */
  88         { in_line_eoln, 0 }, /* ft */
  89         { in_line_eoln, 0 }, /* OP */
  90         { in_line_eoln, MAN_BSCOPE }, /* EX */
  91         { in_line_eoln, MAN_BSCOPE }, /* EE */
  92         { blk_exp, MAN_BSCOPE | MAN_EXPLICIT }, /* UR */
  93         { blk_close, 0 }, /* UE */
  94 };
  95 
  96 const   struct man_macro * const man_macros = __man_macros;
  97 
  98 
  99 /*
 100  * Warn when "n" is an explicit non-roff macro.
 101  */
 102 static void
 103 rew_warn(struct man *man, struct man_node *n, enum mandocerr er)
 104 {
 105 
 106         if (er == MANDOCERR_MAX || MAN_BLOCK != n->type)
 107                 return;
 108         if (MAN_VALID & n->flags)
 109                 return;
 110         if ( ! (MAN_EXPLICIT & man_macros[n->tok].flags))
 111                 return;
 112 
 113         assert(er < MANDOCERR_FATAL);
 114         man_nmsg(man, n, er);
 115 }
 116 
 117 
 118 /*
 119  * Rewind scope.  If a code "er" != MANDOCERR_MAX has been provided, it
 120  * will be used if an explicit block scope is being closed out.
 121  */
 122 int
 123 man_unscope(struct man *man, const struct man_node *to, 
 124                 enum mandocerr er)
 125 {
 126         struct man_node *n;
 127 
 128         assert(to);
 129 
 130         man->next = MAN_NEXT_SIBLING;
 131 
 132         /* LINTED */
 133         while (man->last != to) {
 134                 /*
 135                  * Save the parent here, because we may delete the
 136                  * man->last node in the post-validation phase and reset
 137                  * it to man->last->parent, causing a step in the closing
 138                  * out to be lost.
 139                  */
 140                 n = man->last->parent;
 141                 rew_warn(man, man->last, er);
 142                 if ( ! man_valid_post(man))
 143                         return(0);
 144                 man->last = n;
 145                 assert(man->last);
 146         }
 147 
 148         rew_warn(man, man->last, er);
 149         if ( ! man_valid_post(man))
 150                 return(0);
 151 
 152         return(1);
 153 }
 154 
 155 
 156 static enum rew
 157 rew_block(enum mant ntok, enum man_type type, const struct man_node *n)
 158 {
 159 
 160         if (MAN_BLOCK == type && ntok == n->parent->tok && 
 161                         MAN_BODY == n->parent->type)
 162                 return(REW_REWIND);
 163         return(ntok == n->tok ? REW_HALT : REW_NOHALT);
 164 }
 165 
 166 
 167 /*
 168  * There are three scope levels: scoped to the root (all), scoped to the
 169  * section (all less sections), and scoped to subsections (all less
 170  * sections and subsections).
 171  */
 172 static enum rew 
 173 rew_dohalt(enum mant tok, enum man_type type, const struct man_node *n)
 174 {
 175         enum rew         c;
 176 
 177         /* We cannot progress beyond the root ever. */
 178         if (MAN_ROOT == n->type)
 179                 return(REW_HALT);
 180 
 181         assert(n->parent);
 182 
 183         /* Normal nodes shouldn't go to the level of the root. */
 184         if (MAN_ROOT == n->parent->type)
 185                 return(REW_REWIND);
 186 
 187         /* Already-validated nodes should be closed out. */
 188         if (MAN_VALID & n->flags)
 189                 return(REW_NOHALT);
 190 
 191         /* First: rewind to ourselves. */
 192         if (type == n->type && tok == n->tok) {
 193                 if (MAN_EXPLICIT & man_macros[n->tok].flags)
 194                         return(REW_HALT);
 195                 else
 196                         return(REW_REWIND);
 197         }
 198 
 199         /* 
 200          * Next follow the implicit scope-smashings as defined by man.7:
 201          * section, sub-section, etc.
 202          */
 203 
 204         switch (tok) {
 205         case (MAN_SH):
 206                 break;
 207         case (MAN_SS):
 208                 /* Rewind to a section, if a block. */
 209                 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
 210                         return(c);
 211                 break;
 212         case (MAN_RS):
 213                 /* Preserve empty paragraphs before RS. */
 214                 if (0 == n->nchild && (MAN_P == n->tok ||
 215                     MAN_PP == n->tok || MAN_LP == n->tok))
 216                         return(REW_HALT);
 217                 /* Rewind to a subsection, if a block. */
 218                 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
 219                         return(c);
 220                 /* Rewind to a section, if a block. */
 221                 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
 222                         return(c);
 223                 break;
 224         default:
 225                 /* Rewind to an offsetter, if a block. */
 226                 if (REW_NOHALT != (c = rew_block(MAN_RS, type, n)))
 227                         return(c);
 228                 /* Rewind to a subsection, if a block. */
 229                 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
 230                         return(c);
 231                 /* Rewind to a section, if a block. */
 232                 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
 233                         return(c);
 234                 break;
 235         }
 236 
 237         return(REW_NOHALT);
 238 }
 239 
 240 
 241 /*
 242  * Rewinding entails ascending the parse tree until a coherent point,
 243  * for example, the `SH' macro will close out any intervening `SS'
 244  * scopes.  When a scope is closed, it must be validated and actioned.
 245  */
 246 static int
 247 rew_scope(enum man_type type, struct man *man, enum mant tok)
 248 {
 249         struct man_node *n;
 250         enum rew         c;
 251 
 252         /* LINTED */
 253         for (n = man->last; n; n = n->parent) {
 254                 /* 
 255                  * Whether we should stop immediately (REW_HALT), stop
 256                  * and rewind until this point (REW_REWIND), or keep
 257                  * rewinding (REW_NOHALT).
 258                  */
 259                 c = rew_dohalt(tok, type, n);
 260                 if (REW_HALT == c)
 261                         return(1);
 262                 if (REW_REWIND == c)
 263                         break;
 264         }
 265 
 266         /* 
 267          * Rewind until the current point.  Warn if we're a roff
 268          * instruction that's mowing over explicit scopes.
 269          */
 270         assert(n);
 271 
 272         return(man_unscope(man, n, MANDOCERR_MAX));
 273 }
 274 
 275 
 276 /*
 277  * Close out a generic explicit macro.
 278  */
 279 /* ARGSUSED */
 280 int
 281 blk_close(MACRO_PROT_ARGS)
 282 {
 283         enum mant                ntok;
 284         const struct man_node   *nn;
 285 
 286         switch (tok) {
 287         case (MAN_RE):
 288                 ntok = MAN_RS;
 289                 break;
 290         case (MAN_UE):
 291                 ntok = MAN_UR;
 292                 break;
 293         default:
 294                 abort();
 295                 /* NOTREACHED */
 296         }
 297 
 298         for (nn = man->last->parent; nn; nn = nn->parent)
 299                 if (ntok == nn->tok && MAN_BLOCK == nn->type)
 300                         break;
 301 
 302         if (NULL == nn) {
 303                 man_pmsg(man, line, ppos, MANDOCERR_NOSCOPE);
 304                 if ( ! rew_scope(MAN_BLOCK, man, MAN_PP))
 305                         return(0);
 306         } else 
 307                 man_unscope(man, nn, MANDOCERR_MAX);
 308 
 309         return(1);
 310 }
 311 
 312 
 313 /* ARGSUSED */
 314 int
 315 blk_exp(MACRO_PROT_ARGS)
 316 {
 317         struct man_node *n;
 318         int              la;
 319         char            *p;
 320 
 321         /* Close out prior implicit scopes. */
 322 
 323         if ( ! rew_scope(MAN_BLOCK, man, tok))
 324                 return(0);
 325 
 326         if ( ! man_block_alloc(man, line, ppos, tok))
 327                 return(0);
 328         if ( ! man_head_alloc(man, line, ppos, tok))
 329                 return(0);
 330 
 331         for (;;) {
 332                 la = *pos;
 333                 if ( ! man_args(man, line, pos, buf, &p))
 334                         break;
 335                 if ( ! man_word_alloc(man, line, la, p))
 336                         return(0);
 337         }
 338 
 339         assert(man);
 340         assert(tok != MAN_MAX);
 341 
 342         for (n = man->last; n; n = n->parent) {
 343                 if (n->tok != tok)
 344                         continue;
 345                 assert(MAN_HEAD == n->type);
 346                 man_unscope(man, n, MANDOCERR_MAX);
 347                 break;
 348         }
 349 
 350         return(man_body_alloc(man, line, ppos, tok));
 351 }
 352 
 353 
 354 
 355 /*
 356  * Parse an implicit-block macro.  These contain a MAN_HEAD and a
 357  * MAN_BODY contained within a MAN_BLOCK.  Rules for closing out other
 358  * scopes, such as `SH' closing out an `SS', are defined in the rew
 359  * routines.
 360  */
 361 /* ARGSUSED */
 362 int
 363 blk_imp(MACRO_PROT_ARGS)
 364 {
 365         int              la;
 366         char            *p;
 367         struct man_node *n;
 368 
 369         /* Close out prior scopes. */
 370 
 371         if ( ! rew_scope(MAN_BODY, man, tok))
 372                 return(0);
 373         if ( ! rew_scope(MAN_BLOCK, man, tok))
 374                 return(0);
 375 
 376         /* Allocate new block & head scope. */
 377 
 378         if ( ! man_block_alloc(man, line, ppos, tok))
 379                 return(0);
 380         if ( ! man_head_alloc(man, line, ppos, tok))
 381                 return(0);
 382 
 383         n = man->last;
 384 
 385         /* Add line arguments. */
 386 
 387         for (;;) {
 388                 la = *pos;
 389                 if ( ! man_args(man, line, pos, buf, &p))
 390                         break;
 391                 if ( ! man_word_alloc(man, line, la, p))
 392                         return(0);
 393         }
 394 
 395         /* Close out head and open body (unless MAN_SCOPE). */
 396 
 397         if (MAN_SCOPED & man_macros[tok].flags) {
 398                 /* If we're forcing scope (`TP'), keep it open. */
 399                 if (MAN_FSCOPED & man_macros[tok].flags) {
 400                         man->flags |= MAN_BLINE;
 401                         return(1);
 402                 } else if (n == man->last) {
 403                         man->flags |= MAN_BLINE;
 404                         return(1);
 405                 }
 406         }
 407 
 408         if ( ! rew_scope(MAN_HEAD, man, tok))
 409                 return(0);
 410         return(man_body_alloc(man, line, ppos, tok));
 411 }
 412 
 413 
 414 /* ARGSUSED */
 415 int
 416 in_line_eoln(MACRO_PROT_ARGS)
 417 {
 418         int              la;
 419         char            *p;
 420         struct man_node *n;
 421 
 422         if ( ! man_elem_alloc(man, line, ppos, tok))
 423                 return(0);
 424 
 425         n = man->last;
 426 
 427         for (;;) {
 428                 la = *pos;
 429                 if ( ! man_args(man, line, pos, buf, &p))
 430                         break;
 431                 if ( ! man_word_alloc(man, line, la, p))
 432                         return(0);
 433         }
 434 
 435         /*
 436          * Append MAN_EOS in case the last snipped argument
 437          * ends with a dot, e.g. `.IR syslog (3).'
 438          */
 439 
 440         if (n != man->last &&
 441             mandoc_eos(man->last->string, strlen(man->last->string), 0))
 442                 man->last->flags |= MAN_EOS;
 443 
 444         /*
 445          * If no arguments are specified and this is MAN_SCOPED (i.e.,
 446          * next-line scoped), then set our mode to indicate that we're
 447          * waiting for terms to load into our context.
 448          */
 449 
 450         if (n == man->last && MAN_SCOPED & man_macros[tok].flags) {
 451                 assert( ! (MAN_NSCOPED & man_macros[tok].flags));
 452                 man->flags |= MAN_ELINE;
 453                 return(1);
 454         } 
 455 
 456         /* Set ignorable context, if applicable. */
 457 
 458         if (MAN_NSCOPED & man_macros[tok].flags) {
 459                 assert( ! (MAN_SCOPED & man_macros[tok].flags));
 460                 man->flags |= MAN_ILINE;
 461         }
 462 
 463         assert(MAN_ROOT != man->last->type);
 464         man->next = MAN_NEXT_SIBLING;
 465         
 466         /*
 467          * Rewind our element scope.  Note that when TH is pruned, we'll
 468          * be back at the root, so make sure that we don't clobber as
 469          * its sibling.
 470          */
 471 
 472         for ( ; man->last; man->last = man->last->parent) {
 473                 if (man->last == n)
 474                         break;
 475                 if (man->last->type == MAN_ROOT)
 476                         break;
 477                 if ( ! man_valid_post(man))
 478                         return(0);
 479         }
 480 
 481         assert(man->last);
 482 
 483         /*
 484          * Same here regarding whether we're back at the root. 
 485          */
 486 
 487         if (man->last->type != MAN_ROOT && ! man_valid_post(man))
 488                 return(0);
 489 
 490         return(1);
 491 }
 492 
 493 
 494 int
 495 man_macroend(struct man *man)
 496 {
 497 
 498         return(man_unscope(man, man->first, MANDOCERR_SCOPEEXIT));
 499 }
 500 
 501 static int
 502 man_args(struct man *man, int line, int *pos, char *buf, char **v)
 503 {
 504         char     *start;
 505 
 506         assert(*pos);
 507         *v = start = buf + *pos;
 508         assert(' ' != *start);
 509 
 510         if ('\0' == *start)
 511                 return(0);
 512 
 513         *v = mandoc_getarg(man->parse, v, line, pos);
 514         return(1);
 515 }