1 /* $Id: mdoc_macro.c,v 1.109 2011/04/30 10:18:24 kristaps Exp $ */ 2 /* 3 * Copyright (c) 2008, 2009, 2010 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 <assert.h> 23 #include <ctype.h> 24 #include <stdlib.h> 25 #include <stdio.h> 26 #include <string.h> 27 #include <time.h> 28 29 #include "mdoc.h" 30 #include "mandoc.h" 31 #include "libmdoc.h" 32 #include "libmandoc.h" 33 34 enum rew { /* see rew_dohalt() */ 35 REWIND_NONE, 36 REWIND_THIS, 37 REWIND_MORE, 38 REWIND_FORCE, 39 REWIND_LATER, 40 REWIND_ERROR 41 }; 42 43 static int blk_full(MACRO_PROT_ARGS); 44 static int blk_exp_close(MACRO_PROT_ARGS); 45 static int blk_part_exp(MACRO_PROT_ARGS); 46 static int blk_part_imp(MACRO_PROT_ARGS); 47 static int ctx_synopsis(MACRO_PROT_ARGS); 48 static int in_line_eoln(MACRO_PROT_ARGS); 49 static int in_line_argn(MACRO_PROT_ARGS); 50 static int in_line(MACRO_PROT_ARGS); 51 static int obsolete(MACRO_PROT_ARGS); 52 static int phrase_ta(MACRO_PROT_ARGS); 53 54 static int dword(struct mdoc *, int, int, 55 const char *, enum mdelim); 56 static int append_delims(struct mdoc *, 57 int, int *, char *); 58 static enum mdoct lookup(enum mdoct, const char *); 59 static enum mdoct lookup_raw(const char *); 60 static int make_pending(struct mdoc_node *, enum mdoct, 61 struct mdoc *, int, int); 62 static int phrase(struct mdoc *, int, int, char *); 63 static enum mdoct rew_alt(enum mdoct); 64 static enum rew rew_dohalt(enum mdoct, enum mdoc_type, 65 const struct mdoc_node *); 66 static int rew_elem(struct mdoc *, enum mdoct); 67 static int rew_last(struct mdoc *, 68 const struct mdoc_node *); 69 static int rew_sub(enum mdoc_type, struct mdoc *, 70 enum mdoct, int, int); 71 72 const struct mdoc_macro __mdoc_macros[MDOC_MAX] = { 73 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ap */ 74 { in_line_eoln, MDOC_PROLOGUE }, /* Dd */ 75 { in_line_eoln, MDOC_PROLOGUE }, /* Dt */ 76 { in_line_eoln, MDOC_PROLOGUE }, /* Os */ 77 { blk_full, 0 }, /* Sh */ 78 { blk_full, 0 }, /* Ss */ 79 { in_line_eoln, 0 }, /* Pp */ 80 { blk_part_imp, MDOC_PARSED }, /* D1 */ 81 { blk_part_imp, MDOC_PARSED }, /* Dl */ 82 { blk_full, MDOC_EXPLICIT }, /* Bd */ 83 { blk_exp_close, MDOC_EXPLICIT }, /* Ed */ 84 { blk_full, MDOC_EXPLICIT }, /* Bl */ 85 { blk_exp_close, MDOC_EXPLICIT }, /* El */ 86 { blk_full, MDOC_PARSED }, /* It */ 87 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ad */ 88 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* An */ 89 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ar */ 90 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cd */ 91 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cm */ 92 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Dv */ 93 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Er */ 94 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ev */ 95 { in_line_eoln, 0 }, /* Ex */ 96 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fa */ 97 { in_line_eoln, 0 }, /* Fd */ 98 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fl */ 99 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fn */ 100 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ft */ 101 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ic */ 102 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* In */ 103 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Li */ 104 { blk_full, 0 }, /* Nd */ 105 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Nm */ 106 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Op */ 107 { obsolete, 0 }, /* Ot */ 108 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Pa */ 109 { in_line_eoln, 0 }, /* Rv */ 110 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* St */ 111 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Va */ 112 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Vt */ 113 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Xr */ 114 { in_line_eoln, 0 }, /* %A */ 115 { in_line_eoln, 0 }, /* %B */ 116 { in_line_eoln, 0 }, /* %D */ 117 { in_line_eoln, 0 }, /* %I */ 118 { in_line_eoln, 0 }, /* %J */ 119 { in_line_eoln, 0 }, /* %N */ 120 { in_line_eoln, 0 }, /* %O */ 121 { in_line_eoln, 0 }, /* %P */ 122 { in_line_eoln, 0 }, /* %R */ 123 { in_line_eoln, 0 }, /* %T */ 124 { in_line_eoln, 0 }, /* %V */ 125 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ac */ 126 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Ao */ 127 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Aq */ 128 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* At */ 129 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Bc */ 130 { blk_full, MDOC_EXPLICIT }, /* Bf */ 131 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bo */ 132 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Bq */ 133 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bsx */ 134 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bx */ 135 { in_line_eoln, 0 }, /* Db */ 136 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Dc */ 137 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Do */ 138 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Dq */ 139 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ec */ 140 { blk_exp_close, MDOC_EXPLICIT }, /* Ef */ 141 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Em */ 142 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Eo */ 143 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Fx */ 144 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ms */ 145 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* No */ 146 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Ns */ 147 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Nx */ 148 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ox */ 149 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Pc */ 150 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Pf */ 151 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Po */ 152 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Pq */ 153 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Qc */ 154 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Ql */ 155 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Qo */ 156 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Qq */ 157 { blk_exp_close, MDOC_EXPLICIT }, /* Re */ 158 { blk_full, MDOC_EXPLICIT }, /* Rs */ 159 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Sc */ 160 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* So */ 161 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Sq */ 162 { in_line_eoln, 0 }, /* Sm */ 163 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sx */ 164 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sy */ 165 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Tn */ 166 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ux */ 167 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Xc */ 168 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Xo */ 169 { blk_full, MDOC_EXPLICIT | MDOC_CALLABLE }, /* Fo */ 170 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Fc */ 171 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Oo */ 172 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Oc */ 173 { blk_full, MDOC_EXPLICIT }, /* Bk */ 174 { blk_exp_close, MDOC_EXPLICIT }, /* Ek */ 175 { in_line_eoln, 0 }, /* Bt */ 176 { in_line_eoln, 0 }, /* Hf */ 177 { obsolete, 0 }, /* Fr */ 178 { in_line_eoln, 0 }, /* Ud */ 179 { in_line, 0 }, /* Lb */ 180 { in_line_eoln, 0 }, /* Lp */ 181 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Lk */ 182 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Mt */ 183 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Brq */ 184 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bro */ 185 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Brc */ 186 { in_line_eoln, 0 }, /* %C */ 187 { obsolete, 0 }, /* Es */ 188 { obsolete, 0 }, /* En */ 189 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Dx */ 190 { in_line_eoln, 0 }, /* %Q */ 191 { in_line_eoln, 0 }, /* br */ 192 { in_line_eoln, 0 }, /* sp */ 193 { in_line_eoln, 0 }, /* %U */ 194 { phrase_ta, MDOC_CALLABLE | MDOC_PARSED }, /* Ta */ 195 }; 196 197 const struct mdoc_macro * const mdoc_macros = __mdoc_macros; 198 199 200 /* 201 * This is called at the end of parsing. It must traverse up the tree, 202 * closing out open [implicit] scopes. Obviously, open explicit scopes 203 * are errors. 204 */ 205 int 206 mdoc_macroend(struct mdoc *m) 207 { 208 struct mdoc_node *n; 209 210 /* Scan for open explicit scopes. */ 211 212 n = MDOC_VALID & m->last->flags ? m->last->parent : m->last; 213 214 for ( ; n; n = n->parent) 215 if (MDOC_BLOCK == n->type && 216 MDOC_EXPLICIT & mdoc_macros[n->tok].flags) 217 mdoc_nmsg(m, n, MANDOCERR_SCOPEEXIT); 218 219 /* Rewind to the first. */ 220 221 return(rew_last(m, m->first)); 222 } 223 224 225 /* 226 * Look up a macro from within a subsequent context. 227 */ 228 static enum mdoct 229 lookup(enum mdoct from, const char *p) 230 { 231 /* FIXME: make -diag lists be un-PARSED. */ 232 233 if ( ! (MDOC_PARSED & mdoc_macros[from].flags)) 234 return(MDOC_MAX); 235 return(lookup_raw(p)); 236 } 237 238 239 /* 240 * Lookup a macro following the initial line macro. 241 */ 242 static enum mdoct 243 lookup_raw(const char *p) 244 { 245 enum mdoct res; 246 247 if (MDOC_MAX == (res = mdoc_hash_find(p))) 248 return(MDOC_MAX); 249 if (MDOC_CALLABLE & mdoc_macros[res].flags) 250 return(res); 251 return(MDOC_MAX); 252 } 253 254 255 static int 256 rew_last(struct mdoc *mdoc, const struct mdoc_node *to) 257 { 258 struct mdoc_node *n, *np; 259 260 assert(to); 261 mdoc->next = MDOC_NEXT_SIBLING; 262 263 /* LINTED */ 264 while (mdoc->last != to) { 265 /* 266 * Save the parent here, because we may delete the 267 * m->last node in the post-validation phase and reset 268 * it to m->last->parent, causing a step in the closing 269 * out to be lost. 270 */ 271 np = mdoc->last->parent; 272 if ( ! mdoc_valid_post(mdoc)) 273 return(0); 274 n = mdoc->last; 275 mdoc->last = np; 276 assert(mdoc->last); 277 mdoc->last->last = n; 278 } 279 280 return(mdoc_valid_post(mdoc)); 281 } 282 283 284 /* 285 * For a block closing macro, return the corresponding opening one. 286 * Otherwise, return the macro itself. 287 */ 288 static enum mdoct 289 rew_alt(enum mdoct tok) 290 { 291 switch (tok) { 292 case (MDOC_Ac): 293 return(MDOC_Ao); 294 case (MDOC_Bc): 295 return(MDOC_Bo); 296 case (MDOC_Brc): 297 return(MDOC_Bro); 298 case (MDOC_Dc): 299 return(MDOC_Do); 300 case (MDOC_Ec): 301 return(MDOC_Eo); 302 case (MDOC_Ed): 303 return(MDOC_Bd); 304 case (MDOC_Ef): 305 return(MDOC_Bf); 306 case (MDOC_Ek): 307 return(MDOC_Bk); 308 case (MDOC_El): 309 return(MDOC_Bl); 310 case (MDOC_Fc): 311 return(MDOC_Fo); 312 case (MDOC_Oc): 313 return(MDOC_Oo); 314 case (MDOC_Pc): 315 return(MDOC_Po); 316 case (MDOC_Qc): 317 return(MDOC_Qo); 318 case (MDOC_Re): 319 return(MDOC_Rs); 320 case (MDOC_Sc): 321 return(MDOC_So); 322 case (MDOC_Xc): 323 return(MDOC_Xo); 324 default: 325 return(tok); 326 } 327 /* NOTREACHED */ 328 } 329 330 331 /* 332 * Rewinding to tok, how do we have to handle *p? 333 * REWIND_NONE: *p would delimit tok, but no tok scope is open 334 * inside *p, so there is no need to rewind anything at all. 335 * REWIND_THIS: *p matches tok, so rewind *p and nothing else. 336 * REWIND_MORE: *p is implicit, rewind it and keep searching for tok. 337 * REWIND_FORCE: *p is explicit, but tok is full, force rewinding *p. 338 * REWIND_LATER: *p is explicit and still open, postpone rewinding. 339 * REWIND_ERROR: No tok block is open at all. 340 */ 341 static enum rew 342 rew_dohalt(enum mdoct tok, enum mdoc_type type, 343 const struct mdoc_node *p) 344 { 345 346 /* 347 * No matching token, no delimiting block, no broken block. 348 * This can happen when full implicit macros are called for 349 * the first time but try to rewind their previous 350 * instance anyway. 351 */ 352 if (MDOC_ROOT == p->type) 353 return(MDOC_BLOCK == type && 354 MDOC_EXPLICIT & mdoc_macros[tok].flags ? 355 REWIND_ERROR : REWIND_NONE); 356 357 /* 358 * When starting to rewind, skip plain text 359 * and nodes that have already been rewound. 360 */ 361 if (MDOC_TEXT == p->type || MDOC_VALID & p->flags) 362 return(REWIND_MORE); 363 364 /* 365 * The easiest case: Found a matching token. 366 * This applies to both blocks and elements. 367 */ 368 tok = rew_alt(tok); 369 if (tok == p->tok) 370 return(p->end ? REWIND_NONE : 371 type == p->type ? REWIND_THIS : REWIND_MORE); 372 373 /* 374 * While elements do require rewinding for themselves, 375 * they never affect rewinding of other nodes. 376 */ 377 if (MDOC_ELEM == p->type) 378 return(REWIND_MORE); 379 380 /* 381 * Blocks delimited by our target token get REWIND_MORE. 382 * Blocks delimiting our target token get REWIND_NONE. 383 */ 384 switch (tok) { 385 case (MDOC_Bl): 386 if (MDOC_It == p->tok) 387 return(REWIND_MORE); 388 break; 389 case (MDOC_It): 390 if (MDOC_BODY == p->type && MDOC_Bl == p->tok) 391 return(REWIND_NONE); 392 break; 393 /* 394 * XXX Badly nested block handling still fails badly 395 * when one block is breaking two blocks of the same type. 396 * This is an incomplete and extremely ugly workaround, 397 * required to let the OpenBSD tree build. 398 */ 399 case (MDOC_Oo): 400 if (MDOC_Op == p->tok) 401 return(REWIND_MORE); 402 break; 403 case (MDOC_Nm): 404 return(REWIND_NONE); 405 case (MDOC_Nd): 406 /* FALLTHROUGH */ 407 case (MDOC_Ss): 408 if (MDOC_BODY == p->type && MDOC_Sh == p->tok) 409 return(REWIND_NONE); 410 /* FALLTHROUGH */ 411 case (MDOC_Sh): 412 if (MDOC_Nd == p->tok || MDOC_Ss == p->tok || 413 MDOC_Sh == p->tok) 414 return(REWIND_MORE); 415 break; 416 default: 417 break; 418 } 419 420 /* 421 * Default block rewinding rules. 422 * In particular, always skip block end markers, 423 * and let all blocks rewind Nm children. 424 */ 425 if (ENDBODY_NOT != p->end || MDOC_Nm == p->tok || 426 (MDOC_BLOCK == p->type && 427 ! (MDOC_EXPLICIT & mdoc_macros[tok].flags))) 428 return(REWIND_MORE); 429 430 /* 431 * By default, closing out full blocks 432 * forces closing of broken explicit blocks, 433 * while closing out partial blocks 434 * allows delayed rewinding by default. 435 */ 436 return (&blk_full == mdoc_macros[tok].fp ? 437 REWIND_FORCE : REWIND_LATER); 438 } 439 440 441 static int 442 rew_elem(struct mdoc *mdoc, enum mdoct tok) 443 { 444 struct mdoc_node *n; 445 446 n = mdoc->last; 447 if (MDOC_ELEM != n->type) 448 n = n->parent; 449 assert(MDOC_ELEM == n->type); 450 assert(tok == n->tok); 451 452 return(rew_last(mdoc, n)); 453 } 454 455 456 /* 457 * We are trying to close a block identified by tok, 458 * but the child block *broken is still open. 459 * Thus, postpone closing the tok block 460 * until the rew_sub call closing *broken. 461 */ 462 static int 463 make_pending(struct mdoc_node *broken, enum mdoct tok, 464 struct mdoc *m, int line, int ppos) 465 { 466 struct mdoc_node *breaker; 467 468 /* 469 * Iterate backwards, searching for the block matching tok, 470 * that is, the block breaking the *broken block. 471 */ 472 for (breaker = broken->parent; breaker; breaker = breaker->parent) { 473 474 /* 475 * If the *broken block had already been broken before 476 * and we encounter its breaker, make the tok block 477 * pending on the inner breaker. 478 * Graphically, "[A breaker=[B broken=[C->B B] tok=A] C]" 479 * becomes "[A broken=[B [C->B B] tok=A] C]" 480 * and finally "[A [B->A [C->B B] A] C]". 481 */ 482 if (breaker == broken->pending) { 483 broken = breaker; 484 continue; 485 } 486 487 if (REWIND_THIS != rew_dohalt(tok, MDOC_BLOCK, breaker)) 488 continue; 489 if (MDOC_BODY == broken->type) 490 broken = broken->parent; 491 492 /* 493 * Found the breaker. 494 * If another, outer breaker is already pending on 495 * the *broken block, we must not clobber the link 496 * to the outer breaker, but make it pending on the 497 * new, now inner breaker. 498 * Graphically, "[A breaker=[B broken=[C->A A] tok=B] C]" 499 * becomes "[A breaker=[B->A broken=[C A] tok=B] C]" 500 * and finally "[A [B->A [C->B A] B] C]". 501 */ 502 if (broken->pending) { 503 struct mdoc_node *taker; 504 505 /* 506 * If the breaker had also been broken before, 507 * it cannot take on the outer breaker itself, 508 * but must hand it on to its own breakers. 509 * Graphically, this is the following situation: 510 * "[A [B breaker=[C->B B] broken=[D->A A] tok=C] D]" 511 * "[A taker=[B->A breaker=[C->B B] [D->C A] C] D]" 512 */ 513 taker = breaker; 514 while (taker->pending) 515 taker = taker->pending; 516 taker->pending = broken->pending; 517 } 518 broken->pending = breaker; 519 mandoc_vmsg(MANDOCERR_SCOPENEST, m->parse, line, ppos, 520 "%s breaks %s", mdoc_macronames[tok], 521 mdoc_macronames[broken->tok]); 522 return(1); 523 } 524 525 /* 526 * Found no matching block for tok. 527 * Are you trying to close a block that is not open? 528 */ 529 return(0); 530 } 531 532 533 static int 534 rew_sub(enum mdoc_type t, struct mdoc *m, 535 enum mdoct tok, int line, int ppos) 536 { 537 struct mdoc_node *n; 538 539 n = m->last; 540 while (n) { 541 switch (rew_dohalt(tok, t, n)) { 542 case (REWIND_NONE): 543 return(1); 544 case (REWIND_THIS): 545 break; 546 case (REWIND_FORCE): 547 mandoc_vmsg(MANDOCERR_SCOPEBROKEN, m->parse, 548 line, ppos, "%s breaks %s", 549 mdoc_macronames[tok], 550 mdoc_macronames[n->tok]); 551 /* FALLTHROUGH */ 552 case (REWIND_MORE): 553 n = n->parent; 554 continue; 555 case (REWIND_LATER): 556 if (make_pending(n, tok, m, line, ppos) || 557 MDOC_BLOCK != t) 558 return(1); 559 /* FALLTHROUGH */ 560 case (REWIND_ERROR): 561 mdoc_pmsg(m, line, ppos, MANDOCERR_NOSCOPE); 562 return(1); 563 } 564 break; 565 } 566 567 assert(n); 568 if ( ! rew_last(m, n)) 569 return(0); 570 571 /* 572 * The current block extends an enclosing block. 573 * Now that the current block ends, close the enclosing block, too. 574 */ 575 while (NULL != (n = n->pending)) { 576 if ( ! rew_last(m, n)) 577 return(0); 578 if (MDOC_HEAD == n->type && 579 ! mdoc_body_alloc(m, n->line, n->pos, n->tok)) 580 return(0); 581 } 582 583 return(1); 584 } 585 586 /* 587 * Allocate a word and check whether it's punctuation or not. 588 * Punctuation consists of those tokens found in mdoc_isdelim(). 589 */ 590 static int 591 dword(struct mdoc *m, int line, 592 int col, const char *p, enum mdelim d) 593 { 594 595 if (DELIM_MAX == d) 596 d = mdoc_isdelim(p); 597 598 if ( ! mdoc_word_alloc(m, line, col, p)) 599 return(0); 600 601 if (DELIM_OPEN == d) 602 m->last->flags |= MDOC_DELIMO; 603 604 /* 605 * Closing delimiters only suppress the preceding space 606 * when they follow something, not when they start a new 607 * block or element, and not when they follow `No'. 608 * 609 * XXX Explicitly special-casing MDOC_No here feels 610 * like a layering violation. Find a better way 611 * and solve this in the code related to `No'! 612 */ 613 614 else if (DELIM_CLOSE == d && m->last->prev && 615 m->last->prev->tok != MDOC_No) 616 m->last->flags |= MDOC_DELIMC; 617 618 return(1); 619 } 620 621 static int 622 append_delims(struct mdoc *m, int line, int *pos, char *buf) 623 { 624 int la; 625 enum margserr ac; 626 char *p; 627 628 if ('\0' == buf[*pos]) 629 return(1); 630 631 for (;;) { 632 la = *pos; 633 ac = mdoc_zargs(m, line, pos, buf, &p); 634 635 if (ARGS_ERROR == ac) 636 return(0); 637 else if (ARGS_EOLN == ac) 638 break; 639 640 dword(m, line, la, p, DELIM_MAX); 641 642 /* 643 * If we encounter end-of-sentence symbols, then trigger 644 * the double-space. 645 * 646 * XXX: it's easy to allow this to propagate outward to 647 * the last symbol, such that `. )' will cause the 648 * correct double-spacing. However, (1) groff isn't 649 * smart enough to do this and (2) it would require 650 * knowing which symbols break this behaviour, for 651 * example, `. ;' shouldn't propagate the double-space. 652 */ 653 if (mandoc_eos(p, strlen(p), 0)) 654 m->last->flags |= MDOC_EOS; 655 } 656 657 return(1); 658 } 659 660 661 /* 662 * Close out block partial/full explicit. 663 */ 664 static int 665 blk_exp_close(MACRO_PROT_ARGS) 666 { 667 struct mdoc_node *body; /* Our own body. */ 668 struct mdoc_node *later; /* A sub-block starting later. */ 669 struct mdoc_node *n; /* For searching backwards. */ 670 671 int j, lastarg, maxargs, flushed, nl; 672 enum margserr ac; 673 enum mdoct atok, ntok; 674 char *p; 675 676 nl = MDOC_NEWLINE & m->flags; 677 678 switch (tok) { 679 case (MDOC_Ec): 680 maxargs = 1; 681 break; 682 default: 683 maxargs = 0; 684 break; 685 } 686 687 /* 688 * Search backwards for beginnings of blocks, 689 * both of our own and of pending sub-blocks. 690 */ 691 atok = rew_alt(tok); 692 body = later = NULL; 693 for (n = m->last; n; n = n->parent) { 694 if (MDOC_VALID & n->flags) 695 continue; 696 697 /* Remember the start of our own body. */ 698 if (MDOC_BODY == n->type && atok == n->tok) { 699 if (ENDBODY_NOT == n->end) 700 body = n; 701 continue; 702 } 703 704 if (MDOC_BLOCK != n->type || MDOC_Nm == n->tok) 705 continue; 706 if (atok == n->tok) { 707 assert(body); 708 709 /* 710 * Found the start of our own block. 711 * When there is no pending sub block, 712 * just proceed to closing out. 713 */ 714 if (NULL == later) 715 break; 716 717 /* 718 * When there is a pending sub block, 719 * postpone closing out the current block 720 * until the rew_sub() closing out the sub-block. 721 */ 722 make_pending(later, tok, m, line, ppos); 723 724 /* 725 * Mark the place where the formatting - but not 726 * the scope - of the current block ends. 727 */ 728 if ( ! mdoc_endbody_alloc(m, line, ppos, 729 atok, body, ENDBODY_SPACE)) 730 return(0); 731 break; 732 } 733 734 /* 735 * When finding an open sub block, remember the last 736 * open explicit block, or, in case there are only 737 * implicit ones, the first open implicit block. 738 */ 739 if (later && 740 MDOC_EXPLICIT & mdoc_macros[later->tok].flags) 741 continue; 742 if (MDOC_CALLABLE & mdoc_macros[n->tok].flags) 743 later = n; 744 } 745 746 if ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)) { 747 /* FIXME: do this in validate */ 748 if (buf[*pos]) 749 mdoc_pmsg(m, line, ppos, MANDOCERR_ARGSLOST); 750 751 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 752 return(0); 753 return(rew_sub(MDOC_BLOCK, m, tok, line, ppos)); 754 } 755 756 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 757 return(0); 758 759 if (NULL == later && maxargs > 0) 760 if ( ! mdoc_tail_alloc(m, line, ppos, rew_alt(tok))) 761 return(0); 762 763 for (flushed = j = 0; ; j++) { 764 lastarg = *pos; 765 766 if (j == maxargs && ! flushed) { 767 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 768 return(0); 769 flushed = 1; 770 } 771 772 ac = mdoc_args(m, line, pos, buf, tok, &p); 773 774 if (ARGS_ERROR == ac) 775 return(0); 776 if (ARGS_PUNCT == ac) 777 break; 778 if (ARGS_EOLN == ac) 779 break; 780 781 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 782 783 if (MDOC_MAX == ntok) { 784 if ( ! dword(m, line, lastarg, p, DELIM_MAX)) 785 return(0); 786 continue; 787 } 788 789 if ( ! flushed) { 790 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 791 return(0); 792 flushed = 1; 793 } 794 if ( ! mdoc_macro(m, ntok, line, lastarg, pos, buf)) 795 return(0); 796 break; 797 } 798 799 if ( ! flushed && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 800 return(0); 801 802 if ( ! nl) 803 return(1); 804 return(append_delims(m, line, pos, buf)); 805 } 806 807 808 static int 809 in_line(MACRO_PROT_ARGS) 810 { 811 int la, scope, cnt, nc, nl; 812 enum margverr av; 813 enum mdoct ntok; 814 enum margserr ac; 815 enum mdelim d; 816 struct mdoc_arg *arg; 817 char *p; 818 819 nl = MDOC_NEWLINE & m->flags; 820 821 /* 822 * Whether we allow ignored elements (those without content, 823 * usually because of reserved words) to squeak by. 824 */ 825 826 switch (tok) { 827 case (MDOC_An): 828 /* FALLTHROUGH */ 829 case (MDOC_Ar): 830 /* FALLTHROUGH */ 831 case (MDOC_Fl): 832 /* FALLTHROUGH */ 833 case (MDOC_Mt): 834 /* FALLTHROUGH */ 835 case (MDOC_Nm): 836 /* FALLTHROUGH */ 837 case (MDOC_Pa): 838 nc = 1; 839 break; 840 default: 841 nc = 0; 842 break; 843 } 844 845 for (arg = NULL;; ) { 846 la = *pos; 847 av = mdoc_argv(m, line, tok, &arg, pos, buf); 848 849 if (ARGV_WORD == av) { 850 *pos = la; 851 break; 852 } 853 if (ARGV_EOLN == av) 854 break; 855 if (ARGV_ARG == av) 856 continue; 857 858 mdoc_argv_free(arg); 859 return(0); 860 } 861 862 for (cnt = scope = 0;; ) { 863 la = *pos; 864 ac = mdoc_args(m, line, pos, buf, tok, &p); 865 866 if (ARGS_ERROR == ac) 867 return(0); 868 if (ARGS_EOLN == ac) 869 break; 870 if (ARGS_PUNCT == ac) 871 break; 872 873 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 874 875 /* 876 * In this case, we've located a submacro and must 877 * execute it. Close out scope, if open. If no 878 * elements have been generated, either create one (nc) 879 * or raise a warning. 880 */ 881 882 if (MDOC_MAX != ntok) { 883 if (scope && ! rew_elem(m, tok)) 884 return(0); 885 if (nc && 0 == cnt) { 886 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg)) 887 return(0); 888 if ( ! rew_last(m, m->last)) 889 return(0); 890 } else if ( ! nc && 0 == cnt) { 891 mdoc_argv_free(arg); 892 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY); 893 } 894 895 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 896 return(0); 897 if ( ! nl) 898 return(1); 899 return(append_delims(m, line, pos, buf)); 900 } 901 902 /* 903 * Non-quote-enclosed punctuation. Set up our scope, if 904 * a word; rewind the scope, if a delimiter; then append 905 * the word. 906 */ 907 908 d = ARGS_QWORD == ac ? DELIM_NONE : mdoc_isdelim(p); 909 910 if (DELIM_NONE != d) { 911 /* 912 * If we encounter closing punctuation, no word 913 * has been omitted, no scope is open, and we're 914 * allowed to have an empty element, then start 915 * a new scope. `Ar', `Fl', and `Li', only do 916 * this once per invocation. There may be more 917 * of these (all of them?). 918 */ 919 if (0 == cnt && (nc || MDOC_Li == tok) && 920 DELIM_CLOSE == d && ! scope) { 921 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg)) 922 return(0); 923 if (MDOC_Ar == tok || MDOC_Li == tok || 924 MDOC_Fl == tok) 925 cnt++; 926 scope = 1; 927 } 928 /* 929 * Close out our scope, if one is open, before 930 * any punctuation. 931 */ 932 if (scope && ! rew_elem(m, tok)) 933 return(0); 934 scope = 0; 935 } else if ( ! scope) { 936 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg)) 937 return(0); 938 scope = 1; 939 } 940 941 if (DELIM_NONE == d) 942 cnt++; 943 944 if ( ! dword(m, line, la, p, d)) 945 return(0); 946 947 /* 948 * `Fl' macros have their scope re-opened with each new 949 * word so that the `-' can be added to each one without 950 * having to parse out spaces. 951 */ 952 if (scope && MDOC_Fl == tok) { 953 if ( ! rew_elem(m, tok)) 954 return(0); 955 scope = 0; 956 } 957 } 958 959 if (scope && ! rew_elem(m, tok)) 960 return(0); 961 962 /* 963 * If no elements have been collected and we're allowed to have 964 * empties (nc), open a scope and close it out. Otherwise, 965 * raise a warning. 966 */ 967 968 if (nc && 0 == cnt) { 969 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg)) 970 return(0); 971 if ( ! rew_last(m, m->last)) 972 return(0); 973 } else if ( ! nc && 0 == cnt) { 974 mdoc_argv_free(arg); 975 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY); 976 } 977 978 if ( ! nl) 979 return(1); 980 return(append_delims(m, line, pos, buf)); 981 } 982 983 984 static int 985 blk_full(MACRO_PROT_ARGS) 986 { 987 int la, nl; 988 struct mdoc_arg *arg; 989 struct mdoc_node *head; /* save of head macro */ 990 struct mdoc_node *body; /* save of body macro */ 991 struct mdoc_node *n; 992 enum mdoc_type mtt; 993 enum mdoct ntok; 994 enum margserr ac, lac; 995 enum margverr av; 996 char *p; 997 998 nl = MDOC_NEWLINE & m->flags; 999 1000 /* Close out prior implicit scope. */ 1001 1002 if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) { 1003 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 1004 return(0); 1005 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 1006 return(0); 1007 } 1008 1009 /* 1010 * This routine accommodates implicitly- and explicitly-scoped 1011 * macro openings. Implicit ones first close out prior scope 1012 * (seen above). Delay opening the head until necessary to 1013 * allow leading punctuation to print. Special consideration 1014 * for `It -column', which has phrase-part syntax instead of 1015 * regular child nodes. 1016 */ 1017 1018 for (arg = NULL;; ) { 1019 la = *pos; 1020 av = mdoc_argv(m, line, tok, &arg, pos, buf); 1021 1022 if (ARGV_WORD == av) { 1023 *pos = la; 1024 break; 1025 } 1026 1027 if (ARGV_EOLN == av) 1028 break; 1029 if (ARGV_ARG == av) 1030 continue; 1031 1032 mdoc_argv_free(arg); 1033 return(0); 1034 } 1035 1036 if ( ! mdoc_block_alloc(m, line, ppos, tok, arg)) 1037 return(0); 1038 1039 head = body = NULL; 1040 1041 /* 1042 * The `Nd' macro has all arguments in its body: it's a hybrid 1043 * of block partial-explicit and full-implicit. Stupid. 1044 */ 1045 1046 if (MDOC_Nd == tok) { 1047 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1048 return(0); 1049 head = m->last; 1050 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos)) 1051 return(0); 1052 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1053 return(0); 1054 body = m->last; 1055 } 1056 1057 ac = ARGS_ERROR; 1058 1059 for ( ; ; ) { 1060 la = *pos; 1061 /* Initialise last-phrase-type with ARGS_PEND. */ 1062 lac = ARGS_ERROR == ac ? ARGS_PEND : ac; 1063 ac = mdoc_args(m, line, pos, buf, tok, &p); 1064 1065 if (ARGS_PUNCT == ac) 1066 break; 1067 1068 if (ARGS_ERROR == ac) 1069 return(0); 1070 1071 if (ARGS_EOLN == ac) { 1072 if (ARGS_PPHRASE != lac && ARGS_PHRASE != lac) 1073 break; 1074 /* 1075 * This is necessary: if the last token on a 1076 * line is a `Ta' or tab, then we'll get 1077 * ARGS_EOLN, so we must be smart enough to 1078 * reopen our scope if the last parse was a 1079 * phrase or partial phrase. 1080 */ 1081 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 1082 return(0); 1083 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1084 return(0); 1085 body = m->last; 1086 break; 1087 } 1088 1089 /* 1090 * Emit leading punctuation (i.e., punctuation before 1091 * the MDOC_HEAD) for non-phrase types. 1092 */ 1093 1094 if (NULL == head && 1095 ARGS_PEND != ac && 1096 ARGS_PHRASE != ac && 1097 ARGS_PPHRASE != ac && 1098 ARGS_QWORD != ac && 1099 DELIM_OPEN == mdoc_isdelim(p)) { 1100 if ( ! dword(m, line, la, p, DELIM_OPEN)) 1101 return(0); 1102 continue; 1103 } 1104 1105 /* Open a head if one hasn't been opened. */ 1106 1107 if (NULL == head) { 1108 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1109 return(0); 1110 head = m->last; 1111 } 1112 1113 if (ARGS_PHRASE == ac || 1114 ARGS_PEND == ac || 1115 ARGS_PPHRASE == ac) { 1116 /* 1117 * If we haven't opened a body yet, rewind the 1118 * head; if we have, rewind that instead. 1119 */ 1120 1121 mtt = body ? MDOC_BODY : MDOC_HEAD; 1122 if ( ! rew_sub(mtt, m, tok, line, ppos)) 1123 return(0); 1124 1125 /* Then allocate our body context. */ 1126 1127 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1128 return(0); 1129 body = m->last; 1130 1131 /* 1132 * Process phrases: set whether we're in a 1133 * partial-phrase (this effects line handling) 1134 * then call down into the phrase parser. 1135 */ 1136 1137 if (ARGS_PPHRASE == ac) 1138 m->flags |= MDOC_PPHRASE; 1139 if (ARGS_PEND == ac && ARGS_PPHRASE == lac) 1140 m->flags |= MDOC_PPHRASE; 1141 1142 if ( ! phrase(m, line, la, buf)) 1143 return(0); 1144 1145 m->flags &= ~MDOC_PPHRASE; 1146 continue; 1147 } 1148 1149 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 1150 1151 if (MDOC_MAX == ntok) { 1152 if ( ! dword(m, line, la, p, DELIM_MAX)) 1153 return(0); 1154 continue; 1155 } 1156 1157 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 1158 return(0); 1159 break; 1160 } 1161 1162 if (NULL == head) { 1163 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1164 return(0); 1165 head = m->last; 1166 } 1167 1168 if (nl && ! append_delims(m, line, pos, buf)) 1169 return(0); 1170 1171 /* If we've already opened our body, exit now. */ 1172 1173 if (NULL != body) 1174 goto out; 1175 1176 /* 1177 * If there is an open (i.e., unvalidated) sub-block requiring 1178 * explicit close-out, postpone switching the current block from 1179 * head to body until the rew_sub() call closing out that 1180 * sub-block. 1181 */ 1182 for (n = m->last; n && n != head; n = n->parent) { 1183 if (MDOC_BLOCK == n->type && 1184 MDOC_EXPLICIT & mdoc_macros[n->tok].flags && 1185 ! (MDOC_VALID & n->flags)) { 1186 n->pending = head; 1187 return(1); 1188 } 1189 } 1190 1191 /* Close out scopes to remain in a consistent state. */ 1192 1193 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos)) 1194 return(0); 1195 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1196 return(0); 1197 1198 out: 1199 if ( ! (MDOC_FREECOL & m->flags)) 1200 return(1); 1201 1202 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 1203 return(0); 1204 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 1205 return(0); 1206 1207 m->flags &= ~MDOC_FREECOL; 1208 return(1); 1209 } 1210 1211 1212 static int 1213 blk_part_imp(MACRO_PROT_ARGS) 1214 { 1215 int la, nl; 1216 enum mdoct ntok; 1217 enum margserr ac; 1218 char *p; 1219 struct mdoc_node *blk; /* saved block context */ 1220 struct mdoc_node *body; /* saved body context */ 1221 struct mdoc_node *n; 1222 1223 nl = MDOC_NEWLINE & m->flags; 1224 1225 /* 1226 * A macro that spans to the end of the line. This is generally 1227 * (but not necessarily) called as the first macro. The block 1228 * has a head as the immediate child, which is always empty, 1229 * followed by zero or more opening punctuation nodes, then the 1230 * body (which may be empty, depending on the macro), then zero 1231 * or more closing punctuation nodes. 1232 */ 1233 1234 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL)) 1235 return(0); 1236 1237 blk = m->last; 1238 1239 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1240 return(0); 1241 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos)) 1242 return(0); 1243 1244 /* 1245 * Open the body scope "on-demand", that is, after we've 1246 * processed all our the leading delimiters (open parenthesis, 1247 * etc.). 1248 */ 1249 1250 for (body = NULL; ; ) { 1251 la = *pos; 1252 ac = mdoc_args(m, line, pos, buf, tok, &p); 1253 1254 if (ARGS_ERROR == ac) 1255 return(0); 1256 if (ARGS_EOLN == ac) 1257 break; 1258 if (ARGS_PUNCT == ac) 1259 break; 1260 1261 if (NULL == body && ARGS_QWORD != ac && 1262 DELIM_OPEN == mdoc_isdelim(p)) { 1263 if ( ! dword(m, line, la, p, DELIM_OPEN)) 1264 return(0); 1265 continue; 1266 } 1267 1268 if (NULL == body) { 1269 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1270 return(0); 1271 body = m->last; 1272 } 1273 1274 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 1275 1276 if (MDOC_MAX == ntok) { 1277 if ( ! dword(m, line, la, p, DELIM_MAX)) 1278 return(0); 1279 continue; 1280 } 1281 1282 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 1283 return(0); 1284 break; 1285 } 1286 1287 /* Clean-ups to leave in a consistent state. */ 1288 1289 if (NULL == body) { 1290 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1291 return(0); 1292 body = m->last; 1293 } 1294 1295 for (n = body->child; n && n->next; n = n->next) 1296 /* Do nothing. */ ; 1297 1298 /* 1299 * End of sentence spacing: if the last node is a text node and 1300 * has a trailing period, then mark it as being end-of-sentence. 1301 */ 1302 1303 if (n && MDOC_TEXT == n->type && n->string) 1304 if (mandoc_eos(n->string, strlen(n->string), 1)) 1305 n->flags |= MDOC_EOS; 1306 1307 /* Up-propagate the end-of-space flag. */ 1308 1309 if (n && (MDOC_EOS & n->flags)) { 1310 body->flags |= MDOC_EOS; 1311 body->parent->flags |= MDOC_EOS; 1312 } 1313 1314 /* 1315 * If there is an open sub-block requiring explicit close-out, 1316 * postpone closing out the current block 1317 * until the rew_sub() call closing out the sub-block. 1318 */ 1319 for (n = m->last; n && n != body && n != blk->parent; n = n->parent) { 1320 if (MDOC_BLOCK == n->type && 1321 MDOC_EXPLICIT & mdoc_macros[n->tok].flags && 1322 ! (MDOC_VALID & n->flags)) { 1323 make_pending(n, tok, m, line, ppos); 1324 if ( ! mdoc_endbody_alloc(m, line, ppos, 1325 tok, body, ENDBODY_NOSPACE)) 1326 return(0); 1327 return(1); 1328 } 1329 } 1330 1331 /* 1332 * If we can't rewind to our body, then our scope has already 1333 * been closed by another macro (like `Oc' closing `Op'). This 1334 * is ugly behaviour nodding its head to OpenBSD's overwhelming 1335 * crufty use of `Op' breakage. 1336 */ 1337 if (n != body) 1338 mandoc_vmsg(MANDOCERR_SCOPENEST, m->parse, line, ppos, 1339 "%s broken", mdoc_macronames[tok]); 1340 1341 if (n && ! rew_sub(MDOC_BODY, m, tok, line, ppos)) 1342 return(0); 1343 1344 /* Standard appending of delimiters. */ 1345 1346 if (nl && ! append_delims(m, line, pos, buf)) 1347 return(0); 1348 1349 /* Rewind scope, if applicable. */ 1350 1351 if (n && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos)) 1352 return(0); 1353 1354 return(1); 1355 } 1356 1357 1358 static int 1359 blk_part_exp(MACRO_PROT_ARGS) 1360 { 1361 int la, nl; 1362 enum margserr ac; 1363 struct mdoc_node *head; /* keep track of head */ 1364 struct mdoc_node *body; /* keep track of body */ 1365 char *p; 1366 enum mdoct ntok; 1367 1368 nl = MDOC_NEWLINE & m->flags; 1369 1370 /* 1371 * The opening of an explicit macro having zero or more leading 1372 * punctuation nodes; a head with optional single element (the 1373 * case of `Eo'); and a body that may be empty. 1374 */ 1375 1376 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL)) 1377 return(0); 1378 1379 for (head = body = NULL; ; ) { 1380 la = *pos; 1381 ac = mdoc_args(m, line, pos, buf, tok, &p); 1382 1383 if (ARGS_ERROR == ac) 1384 return(0); 1385 if (ARGS_PUNCT == ac) 1386 break; 1387 if (ARGS_EOLN == ac) 1388 break; 1389 1390 /* Flush out leading punctuation. */ 1391 1392 if (NULL == head && ARGS_QWORD != ac && 1393 DELIM_OPEN == mdoc_isdelim(p)) { 1394 assert(NULL == body); 1395 if ( ! dword(m, line, la, p, DELIM_OPEN)) 1396 return(0); 1397 continue; 1398 } 1399 1400 if (NULL == head) { 1401 assert(NULL == body); 1402 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1403 return(0); 1404 head = m->last; 1405 } 1406 1407 /* 1408 * `Eo' gobbles any data into the head, but most other 1409 * macros just immediately close out and begin the body. 1410 */ 1411 1412 if (NULL == body) { 1413 assert(head); 1414 /* No check whether it's a macro! */ 1415 if (MDOC_Eo == tok) 1416 if ( ! dword(m, line, la, p, DELIM_MAX)) 1417 return(0); 1418 1419 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos)) 1420 return(0); 1421 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1422 return(0); 1423 body = m->last; 1424 1425 if (MDOC_Eo == tok) 1426 continue; 1427 } 1428 1429 assert(NULL != head && NULL != body); 1430 1431 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 1432 1433 if (MDOC_MAX == ntok) { 1434 if ( ! dword(m, line, la, p, DELIM_MAX)) 1435 return(0); 1436 continue; 1437 } 1438 1439 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 1440 return(0); 1441 break; 1442 } 1443 1444 /* Clean-up to leave in a consistent state. */ 1445 1446 if (NULL == head) { 1447 if ( ! mdoc_head_alloc(m, line, ppos, tok)) 1448 return(0); 1449 head = m->last; 1450 } 1451 1452 if (NULL == body) { 1453 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos)) 1454 return(0); 1455 if ( ! mdoc_body_alloc(m, line, ppos, tok)) 1456 return(0); 1457 body = m->last; 1458 } 1459 1460 /* Standard appending of delimiters. */ 1461 1462 if ( ! nl) 1463 return(1); 1464 return(append_delims(m, line, pos, buf)); 1465 } 1466 1467 1468 /* ARGSUSED */ 1469 static int 1470 in_line_argn(MACRO_PROT_ARGS) 1471 { 1472 int la, flushed, j, maxargs, nl; 1473 enum margserr ac; 1474 enum margverr av; 1475 struct mdoc_arg *arg; 1476 char *p; 1477 enum mdoct ntok; 1478 1479 nl = MDOC_NEWLINE & m->flags; 1480 1481 /* 1482 * A line macro that has a fixed number of arguments (maxargs). 1483 * Only open the scope once the first non-leading-punctuation is 1484 * found (unless MDOC_IGNDELIM is noted, like in `Pf'), then 1485 * keep it open until the maximum number of arguments are 1486 * exhausted. 1487 */ 1488 1489 switch (tok) { 1490 case (MDOC_Ap): 1491 /* FALLTHROUGH */ 1492 case (MDOC_No): 1493 /* FALLTHROUGH */ 1494 case (MDOC_Ns): 1495 /* FALLTHROUGH */ 1496 case (MDOC_Ux): 1497 maxargs = 0; 1498 break; 1499 case (MDOC_Bx): 1500 /* FALLTHROUGH */ 1501 case (MDOC_Xr): 1502 maxargs = 2; 1503 break; 1504 default: 1505 maxargs = 1; 1506 break; 1507 } 1508 1509 for (arg = NULL; ; ) { 1510 la = *pos; 1511 av = mdoc_argv(m, line, tok, &arg, pos, buf); 1512 1513 if (ARGV_WORD == av) { 1514 *pos = la; 1515 break; 1516 } 1517 1518 if (ARGV_EOLN == av) 1519 break; 1520 if (ARGV_ARG == av) 1521 continue; 1522 1523 mdoc_argv_free(arg); 1524 return(0); 1525 } 1526 1527 for (flushed = j = 0; ; ) { 1528 la = *pos; 1529 ac = mdoc_args(m, line, pos, buf, tok, &p); 1530 1531 if (ARGS_ERROR == ac) 1532 return(0); 1533 if (ARGS_PUNCT == ac) 1534 break; 1535 if (ARGS_EOLN == ac) 1536 break; 1537 1538 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) && 1539 ARGS_QWORD != ac && 0 == j && 1540 DELIM_OPEN == mdoc_isdelim(p)) { 1541 if ( ! dword(m, line, la, p, DELIM_OPEN)) 1542 return(0); 1543 continue; 1544 } else if (0 == j) 1545 if ( ! mdoc_elem_alloc(m, line, la, tok, arg)) 1546 return(0); 1547 1548 if (j == maxargs && ! flushed) { 1549 if ( ! rew_elem(m, tok)) 1550 return(0); 1551 flushed = 1; 1552 } 1553 1554 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 1555 1556 if (MDOC_MAX != ntok) { 1557 if ( ! flushed && ! rew_elem(m, tok)) 1558 return(0); 1559 flushed = 1; 1560 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 1561 return(0); 1562 j++; 1563 break; 1564 } 1565 1566 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) && 1567 ARGS_QWORD != ac && 1568 ! flushed && 1569 DELIM_NONE != mdoc_isdelim(p)) { 1570 if ( ! rew_elem(m, tok)) 1571 return(0); 1572 flushed = 1; 1573 } 1574 1575 /* 1576 * XXX: this is a hack to work around groff's ugliness 1577 * as regards `Xr' and extraneous arguments. It should 1578 * ideally be deprecated behaviour, but because this is 1579 * code is no here, it's unlikely to be removed. 1580 */ 1581 1582 #ifdef __OpenBSD__ 1583 if (MDOC_Xr == tok && j == maxargs) { 1584 if ( ! mdoc_elem_alloc(m, line, la, MDOC_Ns, NULL)) 1585 return(0); 1586 if ( ! rew_elem(m, MDOC_Ns)) 1587 return(0); 1588 } 1589 #endif 1590 1591 if ( ! dword(m, line, la, p, DELIM_MAX)) 1592 return(0); 1593 j++; 1594 } 1595 1596 if (0 == j && ! mdoc_elem_alloc(m, line, la, tok, arg)) 1597 return(0); 1598 1599 /* Close out in a consistent state. */ 1600 1601 if ( ! flushed && ! rew_elem(m, tok)) 1602 return(0); 1603 if ( ! nl) 1604 return(1); 1605 return(append_delims(m, line, pos, buf)); 1606 } 1607 1608 1609 static int 1610 in_line_eoln(MACRO_PROT_ARGS) 1611 { 1612 int la; 1613 enum margserr ac; 1614 enum margverr av; 1615 struct mdoc_arg *arg; 1616 char *p; 1617 enum mdoct ntok; 1618 1619 assert( ! (MDOC_PARSED & mdoc_macros[tok].flags)); 1620 1621 if (tok == MDOC_Pp) 1622 rew_sub(MDOC_BLOCK, m, MDOC_Nm, line, ppos); 1623 1624 /* Parse macro arguments. */ 1625 1626 for (arg = NULL; ; ) { 1627 la = *pos; 1628 av = mdoc_argv(m, line, tok, &arg, pos, buf); 1629 1630 if (ARGV_WORD == av) { 1631 *pos = la; 1632 break; 1633 } 1634 if (ARGV_EOLN == av) 1635 break; 1636 if (ARGV_ARG == av) 1637 continue; 1638 1639 mdoc_argv_free(arg); 1640 return(0); 1641 } 1642 1643 /* Open element scope. */ 1644 1645 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg)) 1646 return(0); 1647 1648 /* Parse argument terms. */ 1649 1650 for (;;) { 1651 la = *pos; 1652 ac = mdoc_args(m, line, pos, buf, tok, &p); 1653 1654 if (ARGS_ERROR == ac) 1655 return(0); 1656 if (ARGS_EOLN == ac) 1657 break; 1658 1659 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p); 1660 1661 if (MDOC_MAX == ntok) { 1662 if ( ! dword(m, line, la, p, DELIM_MAX)) 1663 return(0); 1664 continue; 1665 } 1666 1667 if ( ! rew_elem(m, tok)) 1668 return(0); 1669 return(mdoc_macro(m, ntok, line, la, pos, buf)); 1670 } 1671 1672 /* Close out (no delimiters). */ 1673 1674 return(rew_elem(m, tok)); 1675 } 1676 1677 1678 /* ARGSUSED */ 1679 static int 1680 ctx_synopsis(MACRO_PROT_ARGS) 1681 { 1682 int nl; 1683 1684 nl = MDOC_NEWLINE & m->flags; 1685 1686 /* If we're not in the SYNOPSIS, go straight to in-line. */ 1687 if ( ! (MDOC_SYNOPSIS & m->flags)) 1688 return(in_line(m, tok, line, ppos, pos, buf)); 1689 1690 /* If we're a nested call, same place. */ 1691 if ( ! nl) 1692 return(in_line(m, tok, line, ppos, pos, buf)); 1693 1694 /* 1695 * XXX: this will open a block scope; however, if later we end 1696 * up formatting the block scope, then child nodes will inherit 1697 * the formatting. Be careful. 1698 */ 1699 if (MDOC_Nm == tok) 1700 return(blk_full(m, tok, line, ppos, pos, buf)); 1701 assert(MDOC_Vt == tok); 1702 return(blk_part_imp(m, tok, line, ppos, pos, buf)); 1703 } 1704 1705 1706 /* ARGSUSED */ 1707 static int 1708 obsolete(MACRO_PROT_ARGS) 1709 { 1710 1711 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROOBS); 1712 return(1); 1713 } 1714 1715 1716 /* 1717 * Phrases occur within `Bl -column' entries, separated by `Ta' or tabs. 1718 * They're unusual because they're basically free-form text until a 1719 * macro is encountered. 1720 */ 1721 static int 1722 phrase(struct mdoc *m, int line, int ppos, char *buf) 1723 { 1724 int la, pos; 1725 enum margserr ac; 1726 enum mdoct ntok; 1727 char *p; 1728 1729 for (pos = ppos; ; ) { 1730 la = pos; 1731 1732 ac = mdoc_zargs(m, line, &pos, buf, &p); 1733 1734 if (ARGS_ERROR == ac) 1735 return(0); 1736 if (ARGS_EOLN == ac) 1737 break; 1738 1739 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p); 1740 1741 if (MDOC_MAX == ntok) { 1742 if ( ! dword(m, line, la, p, DELIM_MAX)) 1743 return(0); 1744 continue; 1745 } 1746 1747 if ( ! mdoc_macro(m, ntok, line, la, &pos, buf)) 1748 return(0); 1749 return(append_delims(m, line, &pos, buf)); 1750 } 1751 1752 return(1); 1753 } 1754 1755 1756 /* ARGSUSED */ 1757 static int 1758 phrase_ta(MACRO_PROT_ARGS) 1759 { 1760 int la; 1761 enum mdoct ntok; 1762 enum margserr ac; 1763 char *p; 1764 1765 /* 1766 * FIXME: this is overly restrictive: if the `Ta' is unexpected, 1767 * it should simply error out with ARGSLOST. 1768 */ 1769 1770 if ( ! rew_sub(MDOC_BODY, m, MDOC_It, line, ppos)) 1771 return(0); 1772 if ( ! mdoc_body_alloc(m, line, ppos, MDOC_It)) 1773 return(0); 1774 1775 for (;;) { 1776 la = *pos; 1777 ac = mdoc_zargs(m, line, pos, buf, &p); 1778 1779 if (ARGS_ERROR == ac) 1780 return(0); 1781 if (ARGS_EOLN == ac) 1782 break; 1783 1784 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p); 1785 1786 if (MDOC_MAX == ntok) { 1787 if ( ! dword(m, line, la, p, DELIM_MAX)) 1788 return(0); 1789 continue; 1790 } 1791 1792 if ( ! mdoc_macro(m, ntok, line, la, pos, buf)) 1793 return(0); 1794 return(append_delims(m, line, pos, buf)); 1795 } 1796 1797 return(1); 1798 } 1799