=================================================================== RCS file: /cvs/mandoc/man.c,v retrieving revision 1.44 retrieving revision 1.54 diff -u -p -r1.44 -r1.54 --- mandoc/man.c 2009/10/27 08:26:12 1.44 +++ mandoc/man.c 2010/03/24 20:10:53 1.54 @@ -1,4 +1,4 @@ -/* $Id: man.c,v 1.44 2009/10/27 08:26:12 kristaps Exp $ */ +/* $Id: man.c,v 1.54 2010/03/24 20:10:53 kristaps Exp $ */ /* * Copyright (c) 2008, 2009 Kristaps Dzonsons * @@ -14,6 +14,10 @@ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + #include #include @@ -24,13 +28,14 @@ #include #include "libman.h" +#include "libmandoc.h" const char *const __man_merrnames[WERRMAX] = { "invalid character", /* WNPRINT */ - "system: malloc error", /* WNMEM */ "invalid manual section", /* WMSEC */ "invalid date format", /* WDATE */ "scope of prior line violated", /* WLNSCOPE */ + "over-zealous prior line scope violation", /* WLNSCOPE2 */ "trailing whitespace", /* WTSPACE */ "unterminated quoted parameter", /* WTQUOTE */ "document has no body", /* WNODATA */ @@ -44,7 +49,8 @@ const char *const __man_merrnames[WERRMAX] = { "scope open on exit", /* WEXITSCOPE */ "no scope context", /* WNOSCOPE */ "literal context already open", /* WOLITERAL */ - "no literal context open" /* WNLITERAL */ + "no literal context open", /* WNLITERAL */ + "invalid nesting of roff declarations", /* WROFFNEST */ }; const char *const __man_macronames[MAN_MAX] = { @@ -55,28 +61,30 @@ const char *const __man_macronames[MAN_MAX] = { "R", "B", "I", "IR", "RI", "na", "i", "sp", "nf", "fi", "r", "RE", - "RS", "DT", "UC", "PD" + "RS", "DT", "UC", "PD", + "Sp", "Vb", "Ve", "de", + "dei", "am", "ami", "ig", + ".", }; const char * const *man_macronames = __man_macronames; static struct man_node *man_node_alloc(int, int, - enum man_type, int); + enum man_type, enum mant); static int man_node_append(struct man *, struct man_node *); +static void man_node_free(struct man_node *); +static void man_node_unlink(struct man *, + struct man_node *); static int man_ptext(struct man *, int, char *); static int man_pmacro(struct man *, int, char *); static void man_free1(struct man *); -static int man_alloc1(struct man *); +static void man_alloc1(struct man *); static int pstring(struct man *, int, int, const char *, size_t); static int macrowarn(struct man *, int, const char *); -#ifdef __linux__ -extern size_t strlcpy(char *, const char *, size_t); -#endif - const struct man_node * man_node(const struct man *m) { @@ -93,12 +101,12 @@ man_meta(const struct man *m) } -int +void man_reset(struct man *man) { man_free1(man); - return(man_alloc1(man)); + man_alloc1(man); } @@ -116,19 +124,16 @@ man_alloc(void *data, int pflags, const struct man_cb { struct man *p; - if (NULL == (p = calloc(1, sizeof(struct man)))) - return(NULL); + p = mandoc_calloc(1, sizeof(struct man)); - if ( ! man_alloc1(p)) { - free(p); - return(NULL); - } + if (cb) + memcpy(&p->cb, cb, sizeof(struct man_cb)); man_hash_init(); - p->data = data; p->pflags = pflags; - (void)memcpy(&p->cb, cb, sizeof(struct man_cb)); + + man_alloc1(p); return(p); } @@ -161,7 +166,7 @@ man_free1(struct man *man) { if (man->first) - man_node_freelist(man->first); + man_node_delete(man, man->first); if (man->meta.title) free(man->meta.title); if (man->meta.source) @@ -171,19 +176,17 @@ man_free1(struct man *man) } -static int +static void man_alloc1(struct man *m) { memset(&m->meta, 0, sizeof(struct man_meta)); m->flags = 0; - m->last = calloc(1, sizeof(struct man_node)); - if (NULL == m->last) - return(0); + m->last = mandoc_calloc(1, sizeof(struct man_node)); m->first = m->last; m->last->type = MAN_ROOT; + m->last->tok = MAN_MAX; m->next = MAN_NEXT_CHILD; - return(1); } @@ -210,6 +213,7 @@ man_node_append(struct man *man, struct man_node *p) /* NOTREACHED */ } + assert(p->parent); p->parent->nchild++; if ( ! man_valid_pre(man, p)) @@ -246,14 +250,11 @@ man_node_append(struct man *man, struct man_node *p) static struct man_node * -man_node_alloc(int line, int pos, enum man_type type, int tok) +man_node_alloc(int line, int pos, enum man_type type, enum mant tok) { struct man_node *p; - p = calloc(1, sizeof(struct man_node)); - if (NULL == p) - return(NULL); - + p = mandoc_calloc(1, sizeof(struct man_node)); p->line = line; p->pos = pos; p->type = type; @@ -263,13 +264,11 @@ man_node_alloc(int line, int pos, enum man_type type, int -man_elem_alloc(struct man *m, int line, int pos, int tok) +man_elem_alloc(struct man *m, int line, int pos, enum mant tok) { struct man_node *p; p = man_node_alloc(line, pos, MAN_ELEM, tok); - if (NULL == p) - return(0); if ( ! man_node_append(m, p)) return(0); m->next = MAN_NEXT_CHILD; @@ -278,13 +277,11 @@ man_elem_alloc(struct man *m, int line, int pos, int t int -man_head_alloc(struct man *m, int line, int pos, int tok) +man_head_alloc(struct man *m, int line, int pos, enum mant tok) { struct man_node *p; p = man_node_alloc(line, pos, MAN_HEAD, tok); - if (NULL == p) - return(0); if ( ! man_node_append(m, p)) return(0); m->next = MAN_NEXT_CHILD; @@ -293,13 +290,11 @@ man_head_alloc(struct man *m, int line, int pos, int t int -man_body_alloc(struct man *m, int line, int pos, int tok) +man_body_alloc(struct man *m, int line, int pos, enum mant tok) { struct man_node *p; p = man_node_alloc(line, pos, MAN_BODY, tok); - if (NULL == p) - return(0); if ( ! man_node_append(m, p)) return(0); m->next = MAN_NEXT_CHILD; @@ -308,13 +303,11 @@ man_body_alloc(struct man *m, int line, int pos, int t int -man_block_alloc(struct man *m, int line, int pos, int tok) +man_block_alloc(struct man *m, int line, int pos, enum mant tok) { struct man_node *p; p = man_node_alloc(line, pos, MAN_BLOCK, tok); - if (NULL == p) - return(0); if ( ! man_node_append(m, p)) return(0); m->next = MAN_NEXT_CHILD; @@ -329,16 +322,8 @@ pstring(struct man *m, int line, int pos, struct man_node *n; size_t sv; - n = man_node_alloc(line, pos, MAN_TEXT, -1); - if (NULL == n) - return(0); - - n->string = malloc(len + 1); - if (NULL == n->string) { - free(n); - return(0); - } - + n = man_node_alloc(line, pos, MAN_TEXT, MAN_MAX); + n->string = mandoc_malloc(len + 1); sv = strlcpy(n->string, p, len + 1); /* Prohibit truncation. */ @@ -359,30 +344,29 @@ man_word_alloc(struct man *m, int line, int pos, const } -void +/* + * Free all of the resources held by a node. This does NOT unlink a + * node from its context; for that, see man_node_unlink(). + */ +static void man_node_free(struct man_node *p) { if (p->string) free(p->string); - if (p->parent) - p->parent->nchild--; free(p); } void -man_node_freelist(struct man_node *p) +man_node_delete(struct man *m, struct man_node *p) { - struct man_node *n; - if (p->child) - man_node_freelist(p->child); - assert(0 == p->nchild); - n = p->next; + while (p->child) + man_node_delete(m, p->child); + + man_node_unlink(m, p); man_node_free(p); - if (n) - man_node_freelist(n); } @@ -390,6 +374,7 @@ static int man_ptext(struct man *m, int line, char *buf) { int i, j; + char sv; /* Literal free-form text whitespace is preserved. */ @@ -403,7 +388,12 @@ man_ptext(struct man *m, int line, char *buf) for (i = 0; ' ' == buf[i]; i++) /* Skip leading whitespace. */ ; - if (0 == buf[i]) { + + if ('\0' == buf[i]) { + /* Trailing whitespace? */ + if (i && ' ' == buf[i - 1]) + if ( ! man_pwarn(m, line, i - 1, WTSPACE)) + return(0); if ( ! pstring(m, line, 0, &buf[i], 0)) return(0); goto descope; @@ -417,15 +407,30 @@ man_ptext(struct man *m, int line, char *buf) if (i && ' ' == buf[i] && '\\' == buf[i - 1]) continue; - buf[i++] = 0; + sv = buf[i]; + buf[i++] = '\0'; + if ( ! pstring(m, line, j, &buf[j], (size_t)(i - j))) return(0); + /* Trailing whitespace? Check at overwritten byte. */ + + if (' ' == sv && '\0' == buf[i]) + if ( ! man_pwarn(m, line, i - 1, WTSPACE)) + return(0); + for ( ; ' ' == buf[i]; i++) /* Skip trailing whitespace. */ ; j = i; - if (0 == buf[i]) + + /* Trailing whitespace? */ + + if (' ' == buf[i - 1] && '\0' == buf[i]) + if ( ! man_pwarn(m, line, i - 1, WTSPACE)) + return(0); + + if ('\0' == buf[i]) break; } @@ -471,7 +476,8 @@ macrowarn(struct man *m, int ln, const char *buf) int man_pmacro(struct man *m, int ln, char *buf) { - int i, j, c, ppos, fl; + int i, j, ppos, fl; + enum mant tok; char mac[5]; struct man_node *n; @@ -479,8 +485,8 @@ man_pmacro(struct man *m, int ln, char *buf) fl = m->flags; - if (0 == buf[1]) - goto out; + if ('\0' == buf[1]) + return(1); i = 1; @@ -488,7 +494,7 @@ man_pmacro(struct man *m, int ln, char *buf) i++; while (buf[i] && ' ' == buf[i]) i++; - if (0 == buf[i]) + if ('\0' == buf[i]) goto out; } @@ -497,7 +503,7 @@ man_pmacro(struct man *m, int ln, char *buf) /* Copy the first word into a nil-terminated buffer. */ for (j = 0; j < 4; j++, i++) { - if (0 == (mac[j] = buf[i])) + if ('\0' == (mac[j] = buf[i])) break; else if (' ' == buf[i]) break; @@ -509,7 +515,7 @@ man_pmacro(struct man *m, int ln, char *buf) return(man_perr(m, ln, i, WNPRINT)); } - mac[j] = 0; + mac[j] = '\0'; if (j == 4 || j < 1) { if ( ! (MAN_IGN_MACRO & m->pflags)) { @@ -521,7 +527,7 @@ man_pmacro(struct man *m, int ln, char *buf) return(1); } - if (MAN_MAX == (c = man_hash_find(mac))) { + if (MAN_MAX == (tok = man_hash_find(mac))) { if ( ! macrowarn(m, ln, mac)) goto err; return(1); @@ -532,43 +538,78 @@ man_pmacro(struct man *m, int ln, char *buf) while (buf[i] && ' ' == buf[i]) i++; - /* Remove prior ELINE macro, if applicable. */ + /* Trailing whitespace? */ - if (m->flags & MAN_ELINE) { + if ('\0' == buf[i] && ' ' == buf[i - 1]) + if ( ! man_pwarn(m, ln, i - 1, WTSPACE)) + goto err; + + /* + * Remove prior ELINE macro, as it's being clobbering by a new + * macro. Note that NSCOPED macros do not close out ELINE + * macros---they don't print text---so we let those slip by. + */ + + if ( ! (MAN_NSCOPED & man_macros[tok].flags) && + m->flags & MAN_ELINE) { + assert(MAN_TEXT != m->last->type); + + /* + * This occurs in the following construction: + * .B + * .br + * .B + * .br + * I hate man macros. + * Flat-out disallow this madness. + */ + if (MAN_NSCOPED & man_macros[m->last->tok].flags) + return(man_perr(m, ln, ppos, WLNSCOPE)); + n = m->last; + + assert(n); assert(NULL == n->child); assert(0 == n->nchild); + if ( ! man_nwarn(m, n, WLNSCOPE)) return(0); - if (n->prev) { - assert(n != n->parent->child); - assert(n == n->prev->next); - n->prev->next = NULL; - m->last = n->prev; - m->next = MAN_NEXT_SIBLING; - } else { - assert(n == n->parent->child); - n->parent->child = NULL; - m->last = n->parent; - m->next = MAN_NEXT_CHILD; - } - - man_node_free(n); + man_node_delete(m, n); m->flags &= ~MAN_ELINE; } /* Begin recursive parse sequence. */ - assert(man_macros[c].fp); + assert(man_macros[tok].fp); - if ( ! (*man_macros[c].fp)(m, c, ln, ppos, &i, buf)) + if ( ! (*man_macros[tok].fp)(m, tok, ln, ppos, &i, buf)) goto err; out: - if ( ! (MAN_BLINE & fl)) + /* + * We weren't in a block-line scope when entering the + * above-parsed macro, so return. + * + * FIXME: this prohibits the nesting of blocks (e.g., `de' and + * family) within BLINE or ELINE systems. This is annoying. + */ + + if ( ! (MAN_BLINE & fl)) { + m->flags &= ~MAN_ILINE; return(1); + } + /* + * If we're in a block scope, then allow this macro to slip by + * without closing scope around it. + */ + + if (MAN_ILINE & m->flags) { + m->flags &= ~MAN_ILINE; + return(1); + } + /* * If we've opened a new next-line element scope, then return * now, as the next line will close out the block scope. @@ -637,4 +678,46 @@ man_err(struct man *m, int line, int pos, int iserr, e return(man_verr(m, line, pos, p)); return(man_vwarn(m, line, pos, p)); +} + + +/* + * Unlink a node from its context. If "m" is provided, the last parse + * point will also be adjusted accordingly. + */ +static void +man_node_unlink(struct man *m, struct man_node *n) +{ + + /* Adjust siblings. */ + + if (n->prev) + n->prev->next = n->next; + if (n->next) + n->next->prev = n->prev; + + /* Adjust parent. */ + + if (n->parent) { + n->parent->nchild--; + if (n->parent->child == n) + n->parent->child = n->prev ? n->prev : n->next; + } + + /* Adjust parse point, if applicable. */ + + if (m && m->last == n) { + /*XXX: this can occur when bailing from validation. */ + /*assert(NULL == n->next);*/ + if (n->prev) { + m->last = n->prev; + m->next = MAN_NEXT_SIBLING; + } else { + m->last = n->parent; + m->next = MAN_NEXT_CHILD; + } + } + + if (m && m->first == n) + m->first = NULL; }