1 /* $Id: man_macro.c,v 1.28 2011/04/21 22:59:54 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv> 4 * 5 * Permission to use, copy, modify, and distribute this software for any 6 * purpose with or without fee is hereby granted, provided that the above 7 * copyright notice and this permission notice appear in all copies. 8 * 9 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 10 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 11 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 12 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 13 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 14 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 15 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 16 */ 17 #include <assert.h> 18 #include <ctype.h> 19 #include <stdlib.h> 20 #include <string.h> 21 22 #include "mandoc.h" 23 #include "libman.h" 24 25 enum rew { 26 REW_REWIND, 27 REW_NOHALT, 28 REW_HALT 29 }; 30 31 static int blk_close(MACRO_PROT_ARGS); 32 static int blk_exp(MACRO_PROT_ARGS); 33 static int blk_imp(MACRO_PROT_ARGS); 34 static int in_line_eoln(MACRO_PROT_ARGS); 35 36 static int rew_scope(enum man_type, 37 struct man *, enum mant); 38 static enum rew rew_dohalt(enum mant, enum man_type, 39 const struct man_node *); 40 static enum rew rew_block(enum mant, enum man_type, 41 const struct man_node *); 42 static void rew_warn(struct man *, 43 struct man_node *, enum mandocerr); 44 45 const struct man_macro __man_macros[MAN_MAX] = { 46 { in_line_eoln, MAN_NSCOPED }, /* br */ 47 { in_line_eoln, 0 }, /* TH */ 48 { blk_imp, MAN_SCOPED }, /* SH */ 49 { blk_imp, MAN_SCOPED }, /* SS */ 50 { blk_imp, MAN_SCOPED | MAN_FSCOPED }, /* TP */ 51 { blk_imp, 0 }, /* LP */ 52 { blk_imp, 0 }, /* PP */ 53 { blk_imp, 0 }, /* P */ 54 { blk_imp, 0 }, /* IP */ 55 { blk_imp, 0 }, /* HP */ 56 { in_line_eoln, MAN_SCOPED }, /* SM */ 57 { in_line_eoln, MAN_SCOPED }, /* SB */ 58 { in_line_eoln, 0 }, /* BI */ 59 { in_line_eoln, 0 }, /* IB */ 60 { in_line_eoln, 0 }, /* BR */ 61 { in_line_eoln, 0 }, /* RB */ 62 { in_line_eoln, MAN_SCOPED }, /* R */ 63 { in_line_eoln, MAN_SCOPED }, /* B */ 64 { in_line_eoln, MAN_SCOPED }, /* I */ 65 { in_line_eoln, 0 }, /* IR */ 66 { in_line_eoln, 0 }, /* RI */ 67 { in_line_eoln, MAN_NSCOPED }, /* na */ 68 { in_line_eoln, MAN_NSCOPED }, /* sp */ 69 { in_line_eoln, 0 }, /* nf */ 70 { in_line_eoln, 0 }, /* fi */ 71 { blk_close, 0 }, /* RE */ 72 { blk_exp, MAN_EXPLICIT }, /* RS */ 73 { in_line_eoln, 0 }, /* DT */ 74 { in_line_eoln, 0 }, /* UC */ 75 { in_line_eoln, 0 }, /* PD */ 76 { in_line_eoln, 0 }, /* AT */ 77 { in_line_eoln, 0 }, /* in */ 78 { in_line_eoln, 0 }, /* ft */ 79 }; 80 81 const struct man_macro * const man_macros = __man_macros; 82 83 84 /* 85 * Warn when "n" is an explicit non-roff macro. 86 */ 87 static void 88 rew_warn(struct man *m, struct man_node *n, enum mandocerr er) 89 { 90 91 if (er == MANDOCERR_MAX || MAN_BLOCK != n->type) 92 return; 93 if (MAN_VALID & n->flags) 94 return; 95 if ( ! (MAN_EXPLICIT & man_macros[n->tok].flags)) 96 return; 97 98 assert(er < MANDOCERR_FATAL); 99 man_nmsg(m, n, er); 100 } 101 102 103 /* 104 * Rewind scope. If a code "er" != MANDOCERR_MAX has been provided, it 105 * will be used if an explicit block scope is being closed out. 106 */ 107 int 108 man_unscope(struct man *m, const struct man_node *to, 109 enum mandocerr er) 110 { 111 struct man_node *n; 112 113 assert(to); 114 115 /* LINTED */ 116 while (m->last != to) { 117 /* 118 * Save the parent here, because we may delete the 119 * m->last node in the post-validation phase and reset 120 * it to m->last->parent, causing a step in the closing 121 * out to be lost. 122 */ 123 n = m->last->parent; 124 rew_warn(m, m->last, er); 125 if ( ! man_valid_post(m)) 126 return(0); 127 m->last = n; 128 assert(m->last); 129 } 130 131 rew_warn(m, m->last, er); 132 if ( ! man_valid_post(m)) 133 return(0); 134 135 m->next = MAN_ROOT == m->last->type ? 136 MAN_NEXT_CHILD : MAN_NEXT_SIBLING; 137 138 return(1); 139 } 140 141 142 static enum rew 143 rew_block(enum mant ntok, enum man_type type, const struct man_node *n) 144 { 145 146 if (MAN_BLOCK == type && ntok == n->parent->tok && 147 MAN_BODY == n->parent->type) 148 return(REW_REWIND); 149 return(ntok == n->tok ? REW_HALT : REW_NOHALT); 150 } 151 152 153 /* 154 * There are three scope levels: scoped to the root (all), scoped to the 155 * section (all less sections), and scoped to subsections (all less 156 * sections and subsections). 157 */ 158 static enum rew 159 rew_dohalt(enum mant tok, enum man_type type, const struct man_node *n) 160 { 161 enum rew c; 162 163 /* We cannot progress beyond the root ever. */ 164 if (MAN_ROOT == n->type) 165 return(REW_HALT); 166 167 assert(n->parent); 168 169 /* Normal nodes shouldn't go to the level of the root. */ 170 if (MAN_ROOT == n->parent->type) 171 return(REW_REWIND); 172 173 /* Already-validated nodes should be closed out. */ 174 if (MAN_VALID & n->flags) 175 return(REW_NOHALT); 176 177 /* First: rewind to ourselves. */ 178 if (type == n->type && tok == n->tok) 179 return(REW_REWIND); 180 181 /* 182 * Next follow the implicit scope-smashings as defined by man.7: 183 * section, sub-section, etc. 184 */ 185 186 switch (tok) { 187 case (MAN_SH): 188 break; 189 case (MAN_SS): 190 /* Rewind to a section, if a block. */ 191 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 192 return(c); 193 break; 194 case (MAN_RS): 195 /* Rewind to a subsection, if a block. */ 196 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n))) 197 return(c); 198 /* Rewind to a section, if a block. */ 199 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 200 return(c); 201 break; 202 default: 203 /* Rewind to an offsetter, if a block. */ 204 if (REW_NOHALT != (c = rew_block(MAN_RS, type, n))) 205 return(c); 206 /* Rewind to a subsection, if a block. */ 207 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n))) 208 return(c); 209 /* Rewind to a section, if a block. */ 210 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 211 return(c); 212 break; 213 } 214 215 return(REW_NOHALT); 216 } 217 218 219 /* 220 * Rewinding entails ascending the parse tree until a coherent point, 221 * for example, the `SH' macro will close out any intervening `SS' 222 * scopes. When a scope is closed, it must be validated and actioned. 223 */ 224 static int 225 rew_scope(enum man_type type, struct man *m, enum mant tok) 226 { 227 struct man_node *n; 228 enum rew c; 229 230 /* LINTED */ 231 for (n = m->last; n; n = n->parent) { 232 /* 233 * Whether we should stop immediately (REW_HALT), stop 234 * and rewind until this point (REW_REWIND), or keep 235 * rewinding (REW_NOHALT). 236 */ 237 c = rew_dohalt(tok, type, n); 238 if (REW_HALT == c) 239 return(1); 240 if (REW_REWIND == c) 241 break; 242 } 243 244 /* 245 * Rewind until the current point. Warn if we're a roff 246 * instruction that's mowing over explicit scopes. 247 */ 248 assert(n); 249 250 return(man_unscope(m, n, MANDOCERR_MAX)); 251 } 252 253 254 /* 255 * Close out a generic explicit macro. 256 */ 257 /* ARGSUSED */ 258 int 259 blk_close(MACRO_PROT_ARGS) 260 { 261 enum mant ntok; 262 const struct man_node *nn; 263 264 switch (tok) { 265 case (MAN_RE): 266 ntok = MAN_RS; 267 break; 268 default: 269 abort(); 270 /* NOTREACHED */ 271 } 272 273 for (nn = m->last->parent; nn; nn = nn->parent) 274 if (ntok == nn->tok) 275 break; 276 277 if (NULL == nn) 278 man_pmsg(m, line, ppos, MANDOCERR_NOSCOPE); 279 280 if ( ! rew_scope(MAN_BODY, m, ntok)) 281 return(0); 282 if ( ! rew_scope(MAN_BLOCK, m, ntok)) 283 return(0); 284 285 return(1); 286 } 287 288 289 /* ARGSUSED */ 290 int 291 blk_exp(MACRO_PROT_ARGS) 292 { 293 int w, la; 294 char *p; 295 296 /* 297 * Close out prior scopes. "Regular" explicit macros cannot be 298 * nested, but we allow roff macros to be placed just about 299 * anywhere. 300 */ 301 302 if ( ! rew_scope(MAN_BODY, m, tok)) 303 return(0); 304 if ( ! rew_scope(MAN_BLOCK, m, tok)) 305 return(0); 306 307 if ( ! man_block_alloc(m, line, ppos, tok)) 308 return(0); 309 if ( ! man_head_alloc(m, line, ppos, tok)) 310 return(0); 311 312 for (;;) { 313 la = *pos; 314 w = man_args(m, line, pos, buf, &p); 315 316 if (-1 == w) 317 return(0); 318 if (0 == w) 319 break; 320 321 if ( ! man_word_alloc(m, line, la, p)) 322 return(0); 323 } 324 325 assert(m); 326 assert(tok != MAN_MAX); 327 328 if ( ! rew_scope(MAN_HEAD, m, tok)) 329 return(0); 330 return(man_body_alloc(m, line, ppos, tok)); 331 } 332 333 334 335 /* 336 * Parse an implicit-block macro. These contain a MAN_HEAD and a 337 * MAN_BODY contained within a MAN_BLOCK. Rules for closing out other 338 * scopes, such as `SH' closing out an `SS', are defined in the rew 339 * routines. 340 */ 341 /* ARGSUSED */ 342 int 343 blk_imp(MACRO_PROT_ARGS) 344 { 345 int w, la; 346 char *p; 347 struct man_node *n; 348 349 /* Close out prior scopes. */ 350 351 if ( ! rew_scope(MAN_BODY, m, tok)) 352 return(0); 353 if ( ! rew_scope(MAN_BLOCK, m, tok)) 354 return(0); 355 356 /* Allocate new block & head scope. */ 357 358 if ( ! man_block_alloc(m, line, ppos, tok)) 359 return(0); 360 if ( ! man_head_alloc(m, line, ppos, tok)) 361 return(0); 362 363 n = m->last; 364 365 /* Add line arguments. */ 366 367 for (;;) { 368 la = *pos; 369 w = man_args(m, line, pos, buf, &p); 370 371 if (-1 == w) 372 return(0); 373 if (0 == w) 374 break; 375 376 if ( ! man_word_alloc(m, line, la, p)) 377 return(0); 378 } 379 380 /* Close out head and open body (unless MAN_SCOPE). */ 381 382 if (MAN_SCOPED & man_macros[tok].flags) { 383 /* If we're forcing scope (`TP'), keep it open. */ 384 if (MAN_FSCOPED & man_macros[tok].flags) { 385 m->flags |= MAN_BLINE; 386 return(1); 387 } else if (n == m->last) { 388 m->flags |= MAN_BLINE; 389 return(1); 390 } 391 } 392 393 if ( ! rew_scope(MAN_HEAD, m, tok)) 394 return(0); 395 return(man_body_alloc(m, line, ppos, tok)); 396 } 397 398 399 /* ARGSUSED */ 400 int 401 in_line_eoln(MACRO_PROT_ARGS) 402 { 403 int w, la; 404 char *p; 405 struct man_node *n; 406 407 if ( ! man_elem_alloc(m, line, ppos, tok)) 408 return(0); 409 410 n = m->last; 411 412 for (;;) { 413 la = *pos; 414 w = man_args(m, line, pos, buf, &p); 415 416 if (-1 == w) 417 return(0); 418 if (0 == w) 419 break; 420 if ( ! man_word_alloc(m, line, la, p)) 421 return(0); 422 } 423 424 /* 425 * If no arguments are specified and this is MAN_SCOPED (i.e., 426 * next-line scoped), then set our mode to indicate that we're 427 * waiting for terms to load into our context. 428 */ 429 430 if (n == m->last && MAN_SCOPED & man_macros[tok].flags) { 431 assert( ! (MAN_NSCOPED & man_macros[tok].flags)); 432 m->flags |= MAN_ELINE; 433 return(1); 434 } 435 436 /* Set ignorable context, if applicable. */ 437 438 if (MAN_NSCOPED & man_macros[tok].flags) { 439 assert( ! (MAN_SCOPED & man_macros[tok].flags)); 440 m->flags |= MAN_ILINE; 441 } 442 443 /* 444 * Rewind our element scope. Note that when TH is pruned, we'll 445 * be back at the root, so make sure that we don't clobber as 446 * its sibling. 447 */ 448 449 for ( ; m->last; m->last = m->last->parent) { 450 if (m->last == n) 451 break; 452 if (m->last->type == MAN_ROOT) 453 break; 454 if ( ! man_valid_post(m)) 455 return(0); 456 } 457 458 assert(m->last); 459 460 /* 461 * Same here regarding whether we're back at the root. 462 */ 463 464 if (m->last->type != MAN_ROOT && ! man_valid_post(m)) 465 return(0); 466 467 m->next = MAN_ROOT == m->last->type ? 468 MAN_NEXT_CHILD : MAN_NEXT_SIBLING; 469 470 return(1); 471 } 472 473 474 int 475 man_macroend(struct man *m) 476 { 477 478 return(man_unscope(m, m->first, MANDOCERR_SCOPEEXIT)); 479 } 480 481