=================================================================== RCS file: /cvs/texi2mdoc/util.c,v retrieving revision 1.13 retrieving revision 1.14 diff -u -p -r1.13 -r1.14 --- texi2mdoc/util.c 2015/02/24 14:35:40 1.13 +++ texi2mdoc/util.c 2015/02/25 14:37:17 1.14 @@ -1,4 +1,4 @@ -/* $Id: util.c,v 1.13 2015/02/24 14:35:40 kristaps Exp $ */ +/* $Id: util.c,v 1.14 2015/02/25 14:37:17 kristaps Exp $ */ /* * Copyright (c) 2015 Kristaps Dzonsons * @@ -43,10 +43,7 @@ texifilepop(struct texi *p) assert(p->filepos > 0); f = &p->files[--p->filepos]; - if (TEXISRC_FILE == f->type) - munmap(f->map, f->mapsz); - else - free(f->map); + free(f->map); } static void @@ -206,16 +203,14 @@ texiputchars(struct texi *p, const char *s) /* * This puts all characters onto the output stream but makes sure to * escape mdoc(7) slashes. + * FIXME: useless. */ void -texiputbuf(struct texi *p, const char *buf, size_t start, size_t end) +texiputbuf(struct texi *p, size_t start, size_t end) { - for ( ; start < end; start++) { - texiputchar(p, buf[start]); - if ('\\' == buf[start]) - texiputchar(p, 'e'); - } + for ( ; start < end; start++) + texiputchar(p, BUF(p)[start]); } /* @@ -309,10 +304,10 @@ texivspace(struct texi *p) * in the current input file. */ void -advance(struct texi *p, const char *buf, size_t *pos) +advance(struct texi *p, size_t *pos) { - if ('\n' == buf[*pos]) { + if ('\n' == BUF(p)[*pos]) { p->files[p->filepos - 1].line++; p->files[p->filepos - 1].col = 0; } else @@ -329,15 +324,15 @@ advance(struct texi *p, const char *buf, size_t *pos) * appropriately flush-left punctuation alongside the macro. */ void -texipunctuate(struct texi *p, const char *buf, size_t sz, size_t *pos) +texipunctuate(struct texi *p, size_t *pos) { size_t start, end; if (1 != p->outmacro) return; - for (start = end = *pos; end < sz; end++) { - switch (buf[end]) { + for (start = end = *pos; end < BUFSZ(p); end++) { + switch (BUF(p)[end]) { case (','): case (')'): case ('.'): @@ -353,11 +348,12 @@ texipunctuate(struct texi *p, const char *buf, size_t } if (end == *pos) return; - if (end + 1 == sz || ' ' == buf[end] || '\n' == buf[end]) { + if (end + 1 == BUFSZ(p) || ' ' == BUF(p)[end] || + '\n' == BUF(p)[end]) { for ( ; start < end; start++) { texiputchar(p, ' '); - texiputchar(p, buf[start]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[start]); + advance(p, pos); } } } @@ -368,27 +364,27 @@ texipunctuate(struct texi *p, const char *buf, size_t * doing so. */ static size_t -advancenext(struct texi *p, const char *buf, size_t sz, size_t *pos) +advancenext(struct texi *p, size_t *pos) { if (p->literal) { - while (*pos < sz && ismspace(buf[*pos])) { - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) { + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } return(*pos); } - while (*pos < sz && ismspace(buf[*pos])) { + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) { p->seenws = 1; /* * If it looks like we've printed a double-line, then * output a paragraph. * FIXME: this is stupid. */ - if (*pos && '\n' == buf[*pos] && '\n' == buf[*pos - 1]) + if (*pos && '\n' == BUF(p)[*pos] && '\n' == BUF(p)[*pos - 1]) texivspace(p); - advance(p, buf, pos); + advance(p, pos); } return(*pos); } @@ -399,14 +395,13 @@ advancenext(struct texi *p, const char *buf, size_t sz * the @\n. */ size_t -advanceeoln(struct texi *p, const char *buf, - size_t sz, size_t *pos, int consumenl) +advanceeoln(struct texi *p, size_t *pos, int consumenl) { - while (*pos < sz && '\n' != buf[*pos]) - advance(p, buf, pos); - if (*pos < sz && consumenl) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) + advance(p, pos); + if (*pos < BUFSZ(p) && consumenl) + advance(p, pos); return(*pos); } @@ -415,17 +410,16 @@ advanceeoln(struct texi *p, const char *buf, * current buffer greater than or equal to the current position. */ void -advanceto(struct texi *p, const char *buf, size_t *pos, size_t end) +advanceto(struct texi *p, size_t *pos, size_t end) { assert(*pos <= end); while (*pos < end) - advance(p, buf, pos); + advance(p, pos); } static void -texiexecmacro(struct texi *p, struct teximacro *m, - const char *buf, size_t sz, size_t *pos) +texiexecmacro(struct texi *p, struct teximacro *m, size_t *pos) { size_t valsz, realsz, aasz, asz, ssz, i, j, k, start, end; @@ -433,13 +427,13 @@ texiexecmacro(struct texi *p, struct teximacro *m, char **args; const char *cp; - args = argparse(p, buf, sz, pos, &asz, m->argsz); + args = argparse(p, pos, &asz, m->argsz); if (asz != m->argsz) texiwarn(p, "invalid macro argument length"); aasz = asz < m->argsz ? asz : m->argsz; if (0 == aasz) { - parsemembuf(p, m->value, strlen(m->value)); + texisplice(p, m->value, strlen(m->value), pos); return; } @@ -482,15 +476,10 @@ texiexecmacro(struct texi *p, struct teximacro *m, /* * Argument didn't exist in argument table. - * No need to reallocate here: we just copy the text - * directly from the macro value into the buffer. + * Just ignore it. */ if (k == aasz) { - for ( ; i < end; i++) - val[j++] = m->value[i]; - assert('\\' == m->value[i]); - val[j++] = m->value[i]; - val[j] = '\0'; + i = end; continue; } @@ -508,7 +497,7 @@ texiexecmacro(struct texi *p, struct teximacro *m, i = end; } - parsemembuf(p, val, strlen(val)); + texisplice(p, val, strlen(val), pos); for (i = 0; i < asz; i++) free(args[i]); @@ -522,8 +511,7 @@ texiexecmacro(struct texi *p, struct teximacro *m, * This also will advance the input stream. */ static void -texiword(struct texi *p, const char *buf, - size_t sz, size_t *pos, char extra) +parseword(struct texi *p, size_t *pos, char extra) { if (p->seenws && 0 == p->outmacro && @@ -535,28 +523,28 @@ texiword(struct texi *p, const char *buf, p->seenws = 0; - while (*pos < sz && ! ismspace(buf[*pos])) { - switch (buf[*pos]) { + while (*pos < BUFSZ(p) && ! ismspace(BUF(p)[*pos])) { + switch (BUF(p)[*pos]) { case ('@'): case ('}'): case ('{'): return; } - if ('\0' != extra && buf[*pos] == extra) + if ('\0' != extra && BUF(p)[*pos] == extra) return; - if (*pos < sz - 1 && - '`' == buf[*pos] && - '`' == buf[*pos + 1]) { + if (*pos < BUFSZ(p) - 1 && + '`' == BUF(p)[*pos] && + '`' == BUF(p)[*pos + 1]) { texiputchars(p, "\\(lq"); - advance(p, buf, pos); - } else if (*pos < sz - 1 && - '\'' == buf[*pos] && - '\'' == buf[*pos + 1]) { + advance(p, pos); + } else if (*pos < BUFSZ(p) - 1 && + '\'' == BUF(p)[*pos] && + '\'' == BUF(p)[*pos + 1]) { texiputchars(p, "\\(rq"); - advance(p, buf, pos); + advance(p, pos); } else - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } } @@ -566,39 +554,38 @@ texiword(struct texi *p, const char *buf, * index after the command name. */ enum texicmd -texicmd(struct texi *p, const char *buf, size_t pos, - size_t sz, size_t *end, struct teximacro **macro) +texicmd(struct texi *p, size_t pos, size_t *end, struct teximacro **macro) { size_t i, len, toksz; - assert('@' == buf[pos]); + assert('@' == BUF(p)[pos]); if (NULL != macro) *macro = NULL; - if ((*end = pos) == sz) + if ((*end = pos) == BUFSZ(p)) return(TEXICMD__MAX); - else if ((*end = ++pos) == sz) + else if ((*end = ++pos) == BUFSZ(p)) return(TEXICMD__MAX); /* Alphabetic commands are special. */ - if ( ! isalpha(buf[pos])) { - if ((*end = pos + 1) == sz) + if ( ! isalpha(BUF(p)[pos])) { + if ((*end = pos + 1) == BUFSZ(p)) return(TEXICMD__MAX); for (i = 0; i < TEXICMD__MAX; i++) { if (1 != texitoks[i].len) continue; - if (0 == strncmp(texitoks[i].tok, &buf[pos], 1)) + if (0 == strncmp(texitoks[i].tok, &BUF(p)[pos], 1)) return(i); } - texiwarn(p, "bad command: @%c", buf[pos]); + texiwarn(p, "bad command: @%c", BUF(p)[pos]); return(TEXICMD__MAX); } /* Scan to the end of the possible command name. */ - for (*end = pos; *end < sz && ! ismspace(buf[*end]); (*end)++) - if ((*end > pos && ('@' == buf[*end] || - '{' == buf[*end] || '}' == buf[*end]))) + for (*end = pos; *end < BUFSZ(p) && ! ismspace(BUF(p)[*end]); (*end)++) + if ((*end > pos && ('@' == BUF(p)[*end] || + '{' == BUF(p)[*end] || '}' == BUF(p)[*end]))) break; /* Look for the command. */ @@ -606,7 +593,7 @@ texicmd(struct texi *p, const char *buf, size_t pos, for (i = 0; i < TEXICMD__MAX; i++) { if (len != texitoks[i].len) continue; - if (0 == strncmp(texitoks[i].tok, &buf[pos], len)) + if (0 == strncmp(texitoks[i].tok, &BUF(p)[pos], len)) return(i); } @@ -615,23 +602,23 @@ texicmd(struct texi *p, const char *buf, size_t pos, toksz = strlen(p->indexs[i]); if (len != 5 + toksz) continue; - if (strncmp(&buf[pos], p->indexs[i], toksz)) + if (strncmp(&BUF(p)[pos], p->indexs[i], toksz)) continue; - if (0 == strncmp(&buf[pos + toksz], "index", 5)) + if (0 == strncmp(&BUF(p)[pos + toksz], "index", 5)) return(TEXICMD_USER_INDEX); } for (i = 0; i < p->macrosz; i++) { if (len != strlen(p->macros[i].key)) continue; - if (strncmp(&buf[pos], p->macros[i].key, len)) + if (strncmp(&BUF(p)[pos], p->macros[i].key, len)) continue; if (NULL != macro) *macro = &p->macros[i]; return(TEXICMD__MAX); } - texiwarn(p, "bad command: @%.*s", (int)len, &buf[pos]); + texiwarn(p, "bad command: @%.*s", (int)len, &BUF(p)[pos]); return(TEXICMD__MAX); } @@ -644,48 +631,47 @@ texicmd(struct texi *p, const char *buf, size_t pos, * bracket for the zeroth parse. */ int -parsearg(struct texi *p, const char *buf, - size_t sz, size_t *pos, size_t num) +parsearg(struct texi *p, size_t *pos, size_t num) { size_t end; enum texicmd cmd; struct teximacro *macro; - while (*pos < sz && ismspace(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz || (0 == num && '{' != buf[*pos])) + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p) || (0 == num && '{' != BUF(p)[*pos])) return(0); if (0 == num) - advance(p, buf, pos); + advance(p, pos); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case (','): - advance(p, buf, pos); + advance(p, pos); return(1); case ('}'): - advance(p, buf, pos); + advance(p, pos); return(0); case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, ','); + parseword(p, pos, ','); continue; } - cmd = texicmd(p, buf, *pos, sz, &end, ¯o); - advanceto(p, buf, pos, end); + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (NULL != macro) - texiexecmacro(p, macro, buf, sz, pos); + texiexecmacro(p, macro, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } return(0); } @@ -695,44 +681,44 @@ parsearg(struct texi *p, const char *buf, * This will stop in the event of EOF or if we're not at a bracket. */ void -parsebracket(struct texi *p, const char *buf, size_t sz, size_t *pos) +parsebracket(struct texi *p, size_t *pos) { size_t end; enum texicmd cmd; struct teximacro *macro; - while (*pos < sz && ismspace(buf[*pos])) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) + advance(p, pos); - if (*pos == sz || '{' != buf[*pos]) + if (*pos == BUFSZ(p) || '{' != BUF(p)[*pos]) return; - advance(p, buf, pos); + advance(p, pos); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case ('}'): - advance(p, buf, pos); + advance(p, pos); return; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end, ¯o); - advanceto(p, buf, pos, end); + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (NULL != macro) - texiexecmacro(p, macro, buf, sz, pos); + texiexecmacro(p, macro, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } } @@ -742,88 +728,91 @@ parsebracket(struct texi *p, const char *buf, size_t s * the way. */ void -parseeoln(struct texi *p, const char *buf, size_t sz, size_t *pos) +parseeoln(struct texi *p, size_t *pos) { size_t end; enum texicmd cmd; struct teximacro *macro; - while (*pos < sz && '\n' != buf[*pos]) { - while (*pos < sz && isws(buf[*pos])) { + while (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) { p->seenws = 1; if (p->literal) - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } - switch (buf[*pos]) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end, ¯o); - advanceto(p, buf, pos, end); + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (NULL != macro) - texiexecmacro(p, macro, buf, sz, pos); + texiexecmacro(p, macro, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } + + if (*pos < BUFSZ(p) && '\n' == BUF(p)[*pos]) + advance(p, pos); } /* * Parse a single word or command. * This will return immediately at the EOF. */ -void -parsesingle(struct texi *p, const char *buf, size_t sz, size_t *pos) +static void +parsesingle(struct texi *p, size_t *pos) { size_t end; enum texicmd cmd; struct teximacro *macro; - if ((*pos = advancenext(p, buf, sz, pos)) >= sz) + if ((*pos = advancenext(p, pos)) >= BUFSZ(p)) return; - switch (buf[*pos]) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); return; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); return; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); return; } - cmd = texicmd(p, buf, *pos, sz, &end, ¯o); - advanceto(p, buf, pos, end); + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (NULL != macro) - texiexecmacro(p, macro, buf, sz, pos); + texiexecmacro(p, macro, pos); if (TEXICMD__MAX == cmd) return; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } /* @@ -835,18 +824,18 @@ parsesingle(struct texi *p, const char *buf, size_t sz * line or 1 otherwise. */ int -parselinearg(struct texi *p, const char *buf, size_t sz, size_t *pos) +parselinearg(struct texi *p, size_t *pos) { - while (*pos < sz && isws(buf[*pos])) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) { p->seenws = 1; - advance(p, buf, pos); + advance(p, pos); } - if (*pos < sz && '{' == buf[*pos]) - parsebracket(p, buf, sz, pos); - else if (*pos < sz && '\n' != buf[*pos]) - parsesingle(p, buf, sz, pos); + if (*pos < BUFSZ(p) && '{' == BUF(p)[*pos]) + parsebracket(p, pos); + else if (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) + parsesingle(p, pos); else return(0); @@ -856,41 +845,35 @@ parselinearg(struct texi *p, const char *buf, size_t s /* * Parse til the end of the buffer. */ -void -parseeof(struct texi *p, const char *buf, size_t sz) +static void +parseeof(struct texi *p) { size_t pos; - for (pos = 0; pos < sz; ) - parsesingle(p, buf, sz, &pos); + for (pos = 0; pos < BUFSZ(p); ) + parsesingle(p, &pos); } -/* - * This is like parseeof() except that it's to be invoked on memory - * buffers while parsing a larger scope. - * This is useful for parsing macro sequences. - * The line, column, and name of the calling file context are saved, the - * column and line reset, then all of these restored after parse. - */ void -parsemembuf(struct texi *p, const char *buf, size_t sz) +texisplice(struct texi *p, const char *buf, size_t sz, size_t *pos) { - size_t svln, svcol; - const char *svname; + char *cp; + struct texifile *f; - svln = p->files[p->filepos - 1].line; - svcol = p->files[p->filepos - 1].col; - svname = p->files[p->filepos - 1].name; + assert(p->filepos > 0); + f = &p->files[p->filepos - 1]; - p->files[p->filepos - 1].line = 0; - p->files[p->filepos - 1].col = 0; - p->files[p->filepos - 1].name = ""; + if (f->mapsz + sz > f->mapmaxsz) { + f->mapmaxsz = f->mapsz + sz + 1024; + cp = realloc(f->map, f->mapmaxsz); + if (NULL == cp) + texiabort(p, NULL); + f->map = cp; + } - parseeof(p, buf, sz); - - p->files[p->filepos - 1].line = svln; - p->files[p->filepos - 1].col = svcol; - p->files[p->filepos - 1].name = svname; + memmove(f->map + *pos + sz, f->map + *pos, f->mapsz - *pos); + memcpy(f->map + *pos, buf, sz); + f->mapsz += sz; } /* @@ -899,8 +882,7 @@ parsemembuf(struct texi *p, const char *buf, size_t sz * This will return immediately at EOF. */ void -parseto(struct texi *p, const char *buf, - size_t sz, size_t *pos, const char *endtoken) +parseto(struct texi *p, size_t *pos, const char *endtoken) { size_t end; enum texicmd cmd; @@ -910,50 +892,50 @@ parseto(struct texi *p, const char *buf, endtoksz = strlen(endtoken); assert(endtoksz > 0); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end, ¯o); - advanceto(p, buf, pos, end); + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (TEXICMD_END == cmd) { - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); /* * FIXME: check the full word, not just its * initial substring! */ - if (sz - *pos >= endtoksz && 0 == strncmp - (&buf[*pos], endtoken, endtoksz)) { - advanceeoln(p, buf, sz, pos, 0); + if (BUFSZ(p) - *pos >= endtoksz && 0 == strncmp + (&BUF(p)[*pos], endtoken, endtoksz)) { + advanceeoln(p, pos, 0); break; } if (0 == p->ign) texiwarn(p, "unexpected \"end\""); - advanceeoln(p, buf, sz, pos, 0); + advanceeoln(p, pos, 0); continue; } if (NULL != macro) - texiexecmacro(p, macro, buf, sz, pos); + texiexecmacro(p, macro, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } } @@ -965,7 +947,6 @@ void parsestdin(struct texi *p) { struct texifile *f; - size_t off; ssize_t ssz; assert(0 == p->filepos); @@ -975,18 +956,18 @@ parsestdin(struct texi *p) f->type = TEXISRC_STDIN; f->name = ""; - for (off = 0; ; off += (size_t)ssz) { - if (off == f->mapsz) { - if (f->mapsz == (1U << 31)) + for (f->mapsz = 0; ; f->mapsz += (size_t)ssz) { + if (f->mapsz == f->mapmaxsz) { + if (f->mapmaxsz == (1U << 31)) texierr(p, "stdin buffer too long"); - f->mapsz = f->mapsz > 65536 / 2 ? - 2 * f->mapsz : 65536; - f->map = realloc(f->map, f->mapsz); + f->mapmaxsz = f->mapmaxsz > 65536 / 2 ? + 2 * f->mapmaxsz : 65536; + f->map = realloc(f->map, f->mapmaxsz); if (NULL == f->map) texiabort(p, NULL); } - ssz = read(STDIN_FILENO, - f->map + (int)off, f->mapsz - off); + ssz = read(STDIN_FILENO, f->map + + (int)f->mapsz, f->mapmaxsz - f->mapsz); if (0 == ssz) break; else if (-1 == ssz) @@ -994,7 +975,7 @@ parsestdin(struct texi *p) } p->filepos++; - parseeof(p, f->map, off); + parseeof(p); texifilepop(p); } @@ -1011,6 +992,7 @@ parsefile(struct texi *p, const char *fname, int parse int fd; struct stat st; size_t i; + char *map; if (64 == p->filepos) texierr(p, "too many open files"); @@ -1026,22 +1008,28 @@ parsefile(struct texi *p, const char *fname, int parse texiabort(p, fname); } - f->mapsz = st.st_size; - f->map = mmap(NULL, f->mapsz, + f->mapsz = f->mapmaxsz = st.st_size; + map = mmap(NULL, f->mapsz, PROT_READ, MAP_SHARED, fd, 0); close(fd); - if (MAP_FAILED == f->map) + if (MAP_FAILED == map) texiabort(p, fname); - p->filepos++; if ( ! parse) { for (i = 0; i < f->mapsz; i++) - texiputchar(p, f->map[i]); + texiputchar(p, map[i]); if (p->outcol) texiputchar(p, '\n'); - } else - parseeof(p, f->map, f->mapsz); + munmap(map, f->mapsz); + return; + } + + p->filepos++; + f->map = malloc(f->mapsz); + memcpy(f->map, map, f->mapsz); + munmap(map, f->mapsz); + parseeof(p); texifilepop(p); } @@ -1053,8 +1041,7 @@ parsefile(struct texi *p, const char *fname, int parse * The pointer can point to NULL if the value has been unset. */ static char ** -valuequery(const struct texi *p, - const char *buf, size_t start, size_t end) +valuequery(const struct texi *p, size_t start, size_t end) { size_t i, sz, len; @@ -1066,7 +1053,7 @@ valuequery(const struct texi *p, sz = strlen(p->vals[i].key); if (sz != len) continue; - if (0 == strncmp(p->vals[i].key, &buf[start], len)) + if (0 == strncmp(p->vals[i].key, &BUF(p)[start], len)) return(&p->vals[i].value); } return(NULL); @@ -1077,45 +1064,45 @@ valuequery(const struct texi *p, * pointer to its value via valuequery(). */ static char ** -valuelquery(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuelquery(struct texi *p, size_t *pos) { size_t start, end; char **ret; - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz) + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p)) return(NULL); - for (start = end = *pos; end < sz; end++) - if ('\n' == buf[end]) + for (start = end = *pos; end < BUFSZ(p); end++) + if ('\n' == BUF(p)[end]) break; - advanceto(p, buf, pos, end); - if (*pos < sz) { - assert('\n' == buf[*pos]); - advance(p, buf, pos); + advanceto(p, pos, end); + if (*pos < BUFSZ(p)) { + assert('\n' == BUF(p)[*pos]); + advance(p, pos); } - if (NULL == (ret = valuequery(p, buf, start, end))) + if (NULL == (ret = valuequery(p, start, end))) return(NULL); return(ret); } void -valuelclear(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuelclear(struct texi *p, size_t *pos) { char **ret; - if (NULL == (ret = valuelquery(p, buf, sz, pos))) + if (NULL == (ret = valuelquery(p, pos))) return; free(*ret); *ret = NULL; } const char * -valuellookup(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuellookup(struct texi *p, size_t *pos) { char **ret; - if (NULL == (ret = valuelquery(p, buf, sz, pos))) + if (NULL == (ret = valuelquery(p, pos))) return(NULL); return(*ret); } @@ -1128,25 +1115,25 @@ valuellookup(struct texi *p, const char *buf, size_t s * value had previously been unset. */ const char * -valueblookup(struct texi *p, const char *buf, size_t sz, size_t *pos) +valueblookup(struct texi *p, size_t *pos) { size_t start, end; char **ret; - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz || '{' != buf[*pos]) + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p) || '{' != BUF(p)[*pos]) return(NULL); - advance(p, buf, pos); - for (start = end = *pos; end < sz; end++) - if ('}' == buf[end]) + advance(p, pos); + for (start = end = *pos; end < BUFSZ(p); end++) + if ('}' == BUF(p)[end]) break; - advanceto(p, buf, pos, end); - if (*pos < sz) { - assert('}' == buf[*pos]); - advance(p, buf, pos); + advanceto(p, pos, end); + if (*pos < BUFSZ(p)) { + assert('}' == BUF(p)[*pos]); + advance(p, pos); } - if (NULL == (ret = valuequery(p, buf, start, end))) + if (NULL == (ret = valuequery(p, start, end))) return(NULL); return(*ret); } @@ -1190,19 +1177,18 @@ valueadd(struct texi *p, char *key, char *val) * Ergo, textual: this doesn't interpret the arguments in any way. */ char ** -argparse(struct texi *p, const char *buf, - size_t sz, size_t *pos, size_t *argsz, size_t hint) +argparse(struct texi *p, size_t *pos, size_t *argsz, size_t hint) { char **args; size_t start, end, stack; - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); args = NULL; *argsz = 0; - if ('{' != buf[*pos] && hint) { + if ('{' != BUF(p)[*pos] && hint) { /* * Special case: if we encounter an unbracketed argument * and we're being invoked with non-zero arguments @@ -1214,51 +1200,51 @@ argparse(struct texi *p, const char *buf, if (NULL == args) texiabort(p, NULL); start = *pos; - while (*pos < sz) { - if ('\n' == buf[*pos]) + while (*pos < BUFSZ(p)) { + if ('\n' == BUF(p)[*pos]) break; - advance(p, buf, pos); + advance(p, pos); } args[0] = malloc(*pos - start + 1); - memcpy(args[0], &buf[start], *pos - start); + memcpy(args[0], &BUF(p)[start], *pos - start); args[0][*pos - start] = '\0'; - if (*pos < sz && '\n' == buf[*pos]) - advance(p, buf, pos); + if (*pos < BUFSZ(p) && '\n' == BUF(p)[*pos]) + advance(p, pos); return(args); - } else if ('{' != buf[*pos]) + } else if ('{' != BUF(p)[*pos]) return(args); /* Parse til the closing '}', putting into the array. */ - advance(p, buf, pos); - while (*pos < sz) { - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); + advance(p, pos); + while (*pos < BUFSZ(p)) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); start = *pos; stack = 0; - while (*pos < sz) { + while (*pos < BUFSZ(p)) { /* * According to the manual, commas within * embedded commands are escaped. * We keep track of embedded-ness in the "stack" * state anyway, so this is free. */ - if (',' == buf[*pos] && 0 == stack && 1 != hint) + if (',' == BUF(p)[*pos] && 0 == stack && 1 != hint) break; - else if (0 == stack && '}' == buf[*pos]) + else if (0 == stack && '}' == BUF(p)[*pos]) break; - else if (0 != stack && '}' == buf[*pos]) + else if (0 != stack && '}' == BUF(p)[*pos]) stack--; - else if ('{' == buf[*pos]) + else if ('{' == BUF(p)[*pos]) stack++; - advance(p, buf, pos); + advance(p, pos); } if (stack) texiwarn(p, "unterminated macro " "in macro arguments"); - if ((end = *pos) == sz) + if ((end = *pos) == BUFSZ(p)) break; /* Test for zero-length '{ }'. */ - if (start == end && '}' == buf[*pos] && 0 == *argsz) + if (start == end && '}' == BUF(p)[*pos] && 0 == *argsz) break; /* FIXME: use reallocarray. */ args = realloc @@ -1270,17 +1256,17 @@ argparse(struct texi *p, const char *buf, if (NULL == args[*argsz]) texiabort(p, NULL); memcpy(args[*argsz], - &buf[start], end - start); + &BUF(p)[start], end - start); args[*argsz][end - start] = '\0'; (*argsz)++; - if ('}' == buf[*pos]) + if ('}' == BUF(p)[*pos]) break; - advance(p, buf, pos); + advance(p, pos); } - if (*pos == sz) + if (*pos == BUFSZ(p)) texierr(p, "unterminated arguments"); - assert('}' == buf[*pos]); - advance(p, buf, pos); + assert('}' == BUF(p)[*pos]); + advance(p, pos); return(args); }