1 /* $Id: man_macro.c,v 1.33 2011/12/04 00:44:18 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2008, 2009, 2010, 2011 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 "man.h" 23 #include "mandoc.h" 24 #include "libmandoc.h" 25 #include "libman.h" 26 27 enum rew { 28 REW_REWIND, 29 REW_NOHALT, 30 REW_HALT 31 }; 32 33 static int blk_close(MACRO_PROT_ARGS); 34 static int blk_exp(MACRO_PROT_ARGS); 35 static int blk_imp(MACRO_PROT_ARGS); 36 static int in_line_eoln(MACRO_PROT_ARGS); 37 static int man_args(struct man *, int, 38 int *, char *, char **); 39 40 static int rew_scope(enum man_type, 41 struct man *, enum mant); 42 static enum rew rew_dohalt(enum mant, enum man_type, 43 const struct man_node *); 44 static enum rew rew_block(enum mant, enum man_type, 45 const struct man_node *); 46 static void rew_warn(struct man *, 47 struct man_node *, enum mandocerr); 48 49 const struct man_macro __man_macros[MAN_MAX] = { 50 { in_line_eoln, MAN_NSCOPED }, /* br */ 51 { in_line_eoln, MAN_BSCOPE }, /* TH */ 52 { blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SH */ 53 { blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SS */ 54 { blk_imp, MAN_BSCOPE | MAN_SCOPED | MAN_FSCOPED }, /* TP */ 55 { blk_imp, MAN_BSCOPE }, /* LP */ 56 { blk_imp, MAN_BSCOPE }, /* PP */ 57 { blk_imp, MAN_BSCOPE }, /* P */ 58 { blk_imp, MAN_BSCOPE }, /* IP */ 59 { blk_imp, MAN_BSCOPE }, /* HP */ 60 { in_line_eoln, MAN_SCOPED }, /* SM */ 61 { in_line_eoln, MAN_SCOPED }, /* SB */ 62 { in_line_eoln, 0 }, /* BI */ 63 { in_line_eoln, 0 }, /* IB */ 64 { in_line_eoln, 0 }, /* BR */ 65 { in_line_eoln, 0 }, /* RB */ 66 { in_line_eoln, MAN_SCOPED }, /* R */ 67 { in_line_eoln, MAN_SCOPED }, /* B */ 68 { in_line_eoln, MAN_SCOPED }, /* I */ 69 { in_line_eoln, 0 }, /* IR */ 70 { in_line_eoln, 0 }, /* RI */ 71 { in_line_eoln, MAN_NSCOPED }, /* na */ 72 { in_line_eoln, MAN_NSCOPED }, /* sp */ 73 { in_line_eoln, MAN_BSCOPE }, /* nf */ 74 { in_line_eoln, MAN_BSCOPE }, /* fi */ 75 { blk_close, 0 }, /* RE */ 76 { blk_exp, MAN_EXPLICIT }, /* RS */ 77 { in_line_eoln, 0 }, /* DT */ 78 { in_line_eoln, 0 }, /* UC */ 79 { in_line_eoln, 0 }, /* PD */ 80 { in_line_eoln, 0 }, /* AT */ 81 { in_line_eoln, 0 }, /* in */ 82 { in_line_eoln, 0 }, /* ft */ 83 }; 84 85 const struct man_macro * const man_macros = __man_macros; 86 87 88 /* 89 * Warn when "n" is an explicit non-roff macro. 90 */ 91 static void 92 rew_warn(struct man *m, struct man_node *n, enum mandocerr er) 93 { 94 95 if (er == MANDOCERR_MAX || MAN_BLOCK != n->type) 96 return; 97 if (MAN_VALID & n->flags) 98 return; 99 if ( ! (MAN_EXPLICIT & man_macros[n->tok].flags)) 100 return; 101 102 assert(er < MANDOCERR_FATAL); 103 man_nmsg(m, n, er); 104 } 105 106 107 /* 108 * Rewind scope. If a code "er" != MANDOCERR_MAX has been provided, it 109 * will be used if an explicit block scope is being closed out. 110 */ 111 int 112 man_unscope(struct man *m, const struct man_node *to, 113 enum mandocerr er) 114 { 115 struct man_node *n; 116 117 assert(to); 118 119 m->next = MAN_NEXT_SIBLING; 120 121 /* LINTED */ 122 while (m->last != to) { 123 /* 124 * Save the parent here, because we may delete the 125 * m->last node in the post-validation phase and reset 126 * it to m->last->parent, causing a step in the closing 127 * out to be lost. 128 */ 129 n = m->last->parent; 130 rew_warn(m, m->last, er); 131 if ( ! man_valid_post(m)) 132 return(0); 133 m->last = n; 134 assert(m->last); 135 } 136 137 rew_warn(m, m->last, er); 138 if ( ! man_valid_post(m)) 139 return(0); 140 141 return(1); 142 } 143 144 145 static enum rew 146 rew_block(enum mant ntok, enum man_type type, const struct man_node *n) 147 { 148 149 if (MAN_BLOCK == type && ntok == n->parent->tok && 150 MAN_BODY == n->parent->type) 151 return(REW_REWIND); 152 return(ntok == n->tok ? REW_HALT : REW_NOHALT); 153 } 154 155 156 /* 157 * There are three scope levels: scoped to the root (all), scoped to the 158 * section (all less sections), and scoped to subsections (all less 159 * sections and subsections). 160 */ 161 static enum rew 162 rew_dohalt(enum mant tok, enum man_type type, const struct man_node *n) 163 { 164 enum rew c; 165 166 /* We cannot progress beyond the root ever. */ 167 if (MAN_ROOT == n->type) 168 return(REW_HALT); 169 170 assert(n->parent); 171 172 /* Normal nodes shouldn't go to the level of the root. */ 173 if (MAN_ROOT == n->parent->type) 174 return(REW_REWIND); 175 176 /* Already-validated nodes should be closed out. */ 177 if (MAN_VALID & n->flags) 178 return(REW_NOHALT); 179 180 /* First: rewind to ourselves. */ 181 if (type == n->type && tok == n->tok) 182 return(REW_REWIND); 183 184 /* 185 * Next follow the implicit scope-smashings as defined by man.7: 186 * section, sub-section, etc. 187 */ 188 189 switch (tok) { 190 case (MAN_SH): 191 break; 192 case (MAN_SS): 193 /* Rewind to a section, if a block. */ 194 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 195 return(c); 196 break; 197 case (MAN_RS): 198 /* Rewind to a subsection, if a block. */ 199 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n))) 200 return(c); 201 /* Rewind to a section, if a block. */ 202 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 203 return(c); 204 break; 205 default: 206 /* Rewind to an offsetter, if a block. */ 207 if (REW_NOHALT != (c = rew_block(MAN_RS, type, n))) 208 return(c); 209 /* Rewind to a subsection, if a block. */ 210 if (REW_NOHALT != (c = rew_block(MAN_SS, type, n))) 211 return(c); 212 /* Rewind to a section, if a block. */ 213 if (REW_NOHALT != (c = rew_block(MAN_SH, type, n))) 214 return(c); 215 break; 216 } 217 218 return(REW_NOHALT); 219 } 220 221 222 /* 223 * Rewinding entails ascending the parse tree until a coherent point, 224 * for example, the `SH' macro will close out any intervening `SS' 225 * scopes. When a scope is closed, it must be validated and actioned. 226 */ 227 static int 228 rew_scope(enum man_type type, struct man *m, enum mant tok) 229 { 230 struct man_node *n; 231 enum rew c; 232 233 /* LINTED */ 234 for (n = m->last; n; n = n->parent) { 235 /* 236 * Whether we should stop immediately (REW_HALT), stop 237 * and rewind until this point (REW_REWIND), or keep 238 * rewinding (REW_NOHALT). 239 */ 240 c = rew_dohalt(tok, type, n); 241 if (REW_HALT == c) 242 return(1); 243 if (REW_REWIND == c) 244 break; 245 } 246 247 /* 248 * Rewind until the current point. Warn if we're a roff 249 * instruction that's mowing over explicit scopes. 250 */ 251 assert(n); 252 253 return(man_unscope(m, n, MANDOCERR_MAX)); 254 } 255 256 257 /* 258 * Close out a generic explicit macro. 259 */ 260 /* ARGSUSED */ 261 int 262 blk_close(MACRO_PROT_ARGS) 263 { 264 enum mant ntok; 265 const struct man_node *nn; 266 267 switch (tok) { 268 case (MAN_RE): 269 ntok = MAN_RS; 270 break; 271 default: 272 abort(); 273 /* NOTREACHED */ 274 } 275 276 for (nn = m->last->parent; nn; nn = nn->parent) 277 if (ntok == nn->tok) 278 break; 279 280 if (NULL == nn) 281 man_pmsg(m, line, ppos, MANDOCERR_NOSCOPE); 282 283 if ( ! rew_scope(MAN_BODY, m, ntok)) 284 return(0); 285 if ( ! rew_scope(MAN_BLOCK, m, ntok)) 286 return(0); 287 288 return(1); 289 } 290 291 292 /* ARGSUSED */ 293 int 294 blk_exp(MACRO_PROT_ARGS) 295 { 296 int la; 297 char *p; 298 299 /* 300 * Close out prior scopes. "Regular" explicit macros cannot be 301 * nested, but we allow roff macros to be placed just about 302 * anywhere. 303 */ 304 305 if ( ! man_block_alloc(m, line, ppos, tok)) 306 return(0); 307 if ( ! man_head_alloc(m, line, ppos, tok)) 308 return(0); 309 310 for (;;) { 311 la = *pos; 312 if ( ! man_args(m, line, pos, buf, &p)) 313 break; 314 if ( ! man_word_alloc(m, line, la, p)) 315 return(0); 316 } 317 318 assert(m); 319 assert(tok != MAN_MAX); 320 321 if ( ! rew_scope(MAN_HEAD, m, tok)) 322 return(0); 323 return(man_body_alloc(m, line, ppos, tok)); 324 } 325 326 327 328 /* 329 * Parse an implicit-block macro. These contain a MAN_HEAD and a 330 * MAN_BODY contained within a MAN_BLOCK. Rules for closing out other 331 * scopes, such as `SH' closing out an `SS', are defined in the rew 332 * routines. 333 */ 334 /* ARGSUSED */ 335 int 336 blk_imp(MACRO_PROT_ARGS) 337 { 338 int la; 339 char *p; 340 struct man_node *n; 341 342 /* Close out prior scopes. */ 343 344 if ( ! rew_scope(MAN_BODY, m, tok)) 345 return(0); 346 if ( ! rew_scope(MAN_BLOCK, m, tok)) 347 return(0); 348 349 /* Allocate new block & head scope. */ 350 351 if ( ! man_block_alloc(m, line, ppos, tok)) 352 return(0); 353 if ( ! man_head_alloc(m, line, ppos, tok)) 354 return(0); 355 356 n = m->last; 357 358 /* Add line arguments. */ 359 360 for (;;) { 361 la = *pos; 362 if ( ! man_args(m, line, pos, buf, &p)) 363 break; 364 if ( ! man_word_alloc(m, line, la, p)) 365 return(0); 366 } 367 368 /* Close out head and open body (unless MAN_SCOPE). */ 369 370 if (MAN_SCOPED & man_macros[tok].flags) { 371 /* If we're forcing scope (`TP'), keep it open. */ 372 if (MAN_FSCOPED & man_macros[tok].flags) { 373 m->flags |= MAN_BLINE; 374 return(1); 375 } else if (n == m->last) { 376 m->flags |= MAN_BLINE; 377 return(1); 378 } 379 } 380 381 if ( ! rew_scope(MAN_HEAD, m, tok)) 382 return(0); 383 return(man_body_alloc(m, line, ppos, tok)); 384 } 385 386 387 /* ARGSUSED */ 388 int 389 in_line_eoln(MACRO_PROT_ARGS) 390 { 391 int la; 392 char *p; 393 struct man_node *n; 394 395 if ( ! man_elem_alloc(m, line, ppos, tok)) 396 return(0); 397 398 n = m->last; 399 400 for (;;) { 401 la = *pos; 402 if ( ! man_args(m, line, pos, buf, &p)) 403 break; 404 if ( ! man_word_alloc(m, line, la, p)) 405 return(0); 406 } 407 408 /* 409 * If no arguments are specified and this is MAN_SCOPED (i.e., 410 * next-line scoped), then set our mode to indicate that we're 411 * waiting for terms to load into our context. 412 */ 413 414 if (n == m->last && MAN_SCOPED & man_macros[tok].flags) { 415 assert( ! (MAN_NSCOPED & man_macros[tok].flags)); 416 m->flags |= MAN_ELINE; 417 return(1); 418 } 419 420 /* Set ignorable context, if applicable. */ 421 422 if (MAN_NSCOPED & man_macros[tok].flags) { 423 assert( ! (MAN_SCOPED & man_macros[tok].flags)); 424 m->flags |= MAN_ILINE; 425 } 426 427 assert(MAN_ROOT != m->last->type); 428 m->next = MAN_NEXT_SIBLING; 429 430 /* 431 * Rewind our element scope. Note that when TH is pruned, we'll 432 * be back at the root, so make sure that we don't clobber as 433 * its sibling. 434 */ 435 436 for ( ; m->last; m->last = m->last->parent) { 437 if (m->last == n) 438 break; 439 if (m->last->type == MAN_ROOT) 440 break; 441 if ( ! man_valid_post(m)) 442 return(0); 443 } 444 445 assert(m->last); 446 447 /* 448 * Same here regarding whether we're back at the root. 449 */ 450 451 if (m->last->type != MAN_ROOT && ! man_valid_post(m)) 452 return(0); 453 454 return(1); 455 } 456 457 458 int 459 man_macroend(struct man *m) 460 { 461 462 return(man_unscope(m, m->first, MANDOCERR_SCOPEEXIT)); 463 } 464 465 static int 466 man_args(struct man *m, int line, int *pos, char *buf, char **v) 467 { 468 char *start; 469 470 assert(*pos); 471 *v = start = buf + *pos; 472 assert(' ' != *start); 473 474 if ('\0' == *start) 475 return(0); 476 477 *v = mandoc_getarg(m->parse, v, line, pos); 478 return(1); 479 } 480