=================================================================== RCS file: /cvs/mandoc/man.c,v retrieving revision 1.46 retrieving revision 1.70 diff -u -p -r1.46 -r1.70 --- mandoc/man.c 2009/11/02 08:40:31 1.46 +++ mandoc/man.c 2010/05/15 22:44:04 1.70 @@ -1,4 +1,4 @@ -/* $Id: man.c,v 1.46 2009/11/02 08:40:31 kristaps Exp $ */ +/* $Id: man.c,v 1.70 2010/05/15 22:44:04 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 @@ -28,9 +32,9 @@ const char *const __man_merrnames[WERRMAX] = { "invalid character", /* WNPRINT */ - "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 +48,9 @@ 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 */ + "document title should be uppercase", /* WTITLECASE */ + "deprecated comment style", /* WBADCOMMENT */ }; const char *const __man_macronames[MAN_MAX] = { @@ -55,28 +61,26 @@ 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", }; 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 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) { @@ -147,7 +151,7 @@ int man_parseln(struct man *m, int ln, char *buf) { - return('.' == *buf ? + return('.' == *buf || '\'' == *buf ? man_pmacro(m, ln, buf) : man_ptext(m, ln, buf)); } @@ -158,13 +162,15 @@ 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) free(man->meta.source); if (man->meta.vol) free(man->meta.vol); + if (man->meta.msec) + free(man->meta.msec); } @@ -177,6 +183,7 @@ man_alloc1(struct man *m) 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; } @@ -204,6 +211,7 @@ man_node_append(struct man *man, struct man_node *p) /* NOTREACHED */ } + assert(p->parent); p->parent->nchild++; if ( ! man_valid_pre(man, p)) @@ -240,7 +248,7 @@ 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; @@ -254,7 +262,7 @@ 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; @@ -267,7 +275,7 @@ 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; @@ -280,7 +288,7 @@ 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; @@ -293,7 +301,7 @@ 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; @@ -305,67 +313,65 @@ man_block_alloc(struct man *m, int line, int pos, int } -static int -pstring(struct man *m, int line, int pos, - const char *p, size_t len) +int +man_word_alloc(struct man *m, int line, int pos, const char *word) { struct man_node *n; - size_t sv; + size_t sv, len; - n = man_node_alloc(line, pos, MAN_TEXT, -1); + len = strlen(word); + + n = man_node_alloc(line, pos, MAN_TEXT, MAN_MAX); n->string = mandoc_malloc(len + 1); - sv = strlcpy(n->string, p, len + 1); + sv = strlcpy(n->string, word, len + 1); /* Prohibit truncation. */ assert(sv < len + 1); if ( ! man_node_append(m, n)) return(0); + m->next = MAN_NEXT_SIBLING; return(1); } -int -man_word_alloc(struct man *m, int line, int pos, const char *word) -{ - - return(pstring(m, line, pos, word, strlen(word))); -} - - -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); } static int man_ptext(struct man *m, int line, char *buf) { - int i, j; + int i; + /* Ignore bogus comments. */ + + if ('\\' == buf[0] && '.' == buf[1] && '\"' == buf[2]) + return(man_pwarn(m, line, 0, WBADCOMMENT)); + /* Literal free-form text whitespace is preserved. */ if (MAN_LITERAL & m->flags) { @@ -374,41 +380,57 @@ man_ptext(struct man *m, int line, char *buf) goto descope; } - /* First de-chunk and allocate words. */ + /* Pump blank lines directly into the backend. */ for (i = 0; ' ' == buf[i]; i++) /* Skip leading whitespace. */ ; - if (0 == buf[i]) { - if ( ! pstring(m, line, 0, &buf[i], 0)) + + if ('\0' == buf[i]) { + /* Allocate a blank entry. */ + if ( ! man_word_alloc(m, line, 0, "")) return(0); goto descope; } - for (j = i; buf[i]; i++) { - if (' ' != buf[i]) - continue; + /* + * Warn if the last un-escaped character is whitespace. Then + * strip away the remaining spaces (tabs stay!). + */ - /* Escaped whitespace. */ - if (i && ' ' == buf[i] && '\\' == buf[i - 1]) - continue; + i = (int)strlen(buf); + assert(i); - buf[i++] = 0; - if ( ! pstring(m, line, j, &buf[j], (size_t)(i - j))) - return(0); + if (' ' == buf[i - 1] || '\t' == buf[i - 1]) { + if (i > 1 && '\\' != buf[i - 2]) + if ( ! man_pwarn(m, line, i - 1, WTSPACE)) + return(0); - for ( ; ' ' == buf[i]; i++) - /* Skip trailing whitespace. */ ; + for (--i; i && ' ' == buf[i]; i--) + /* Spin back to non-space. */ ; - j = i; - if (0 == buf[i]) - break; + /* Jump ahead of escaped whitespace. */ + i += '\\' == buf[i] ? 2 : 1; + + buf[i] = '\0'; } - if (j != i && ! pstring(m, line, j, &buf[j], (size_t)(i - j))) + if ( ! man_word_alloc(m, line, 0, buf)) return(0); -descope: + /* + * End-of-sentence check. If the last character is an unescaped + * EOS character, then flag the node as being the end of a + * sentence. The front-end will know how to interpret this. + */ + /* FIXME: chain of close delims. */ + + assert(i); + + if (mandoc_eos(buf, (size_t)i)) + m->last->flags |= MAN_EOS; + +descope: /* * Co-ordinate what happens with having a next-line scope open: * first close out the element scope (if applicable), then close @@ -417,7 +439,7 @@ descope: if (MAN_ELINE & m->flags) { m->flags &= ~MAN_ELINE; - if ( ! man_unscope(m, m->last->parent)) + if ( ! man_unscope(m, m->last->parent, WERRMAX)) return(0); } @@ -425,7 +447,7 @@ descope: return(1); m->flags &= ~MAN_BLINE; - if ( ! man_unscope(m, m->last->parent)) + if ( ! man_unscope(m, m->last->parent, WERRMAX)) return(0); return(man_body_alloc(m, line, 0, m->last->tok)); } @@ -435,8 +457,7 @@ static int macrowarn(struct man *m, int ln, const char *buf) { if ( ! (MAN_IGN_MACRO & m->pflags)) - return(man_verr(m, ln, 0, - "unknown macro: %s%s", + return(man_verr(m, ln, 0, "unknown macro: %s%s", buf, strlen(buf) > 3 ? "..." : "")); return(man_vwarn(m, ln, 0, "unknown macro: %s%s", buf, strlen(buf) > 3 ? "..." : "")); @@ -446,24 +467,28 @@ 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; + enum mant tok; char mac[5]; struct man_node *n; /* Comments and empties are quickly ignored. */ - fl = m->flags; - if ('\0' == buf[1]) return(1); i = 1; - if (' ' == buf[i]) { + /* + * Skip whitespace between the control character and initial + * text. "Whitespace" is both spaces and tabs. + */ + + if (' ' == buf[i] || '\t' == buf[i]) { i++; - while (buf[i] && ' ' == buf[i]) + while (buf[i] && (' ' == buf[i] || '\t' == buf[i])) i++; - if (0 == buf[i]) + if ('\0' == buf[i]) goto out; } @@ -472,7 +497,7 @@ 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; @@ -496,7 +521,7 @@ 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); @@ -507,43 +532,87 @@ man_pmacro(struct man *m, int ln, char *buf) while (buf[i] && ' ' == buf[i]) i++; - /* Remove prior ELINE macro, if applicable. */ + /* + * Trailing whitespace. Note that tabs are allowed to be passed + * into the parser as "text", so we only warn about spaces here. + */ - 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. */ + /* + * Save the fact that we're in the next-line for a block. In + * this way, embedded roff instructions can "remember" state + * when they exit. + */ - assert(man_macros[c].fp); + if (MAN_BLINE & m->flags) + m->flags |= MAN_BPLINE; - if ( ! (*man_macros[c].fp)(m, c, ln, ppos, &i, buf)) + /* Call to handler... */ + + assert(man_macros[tok].fp); + 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. + */ + + if ( ! (MAN_BPLINE & m->flags)) { + m->flags &= ~MAN_ILINE; return(1); + } + m->flags &= ~MAN_BPLINE; + /* + * 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. @@ -557,7 +626,7 @@ out: assert(MAN_BLINE & m->flags); m->flags &= ~MAN_BLINE; - if ( ! man_unscope(m, m->last->parent)) + if ( ! man_unscope(m, m->last->parent, WERRMAX)) return(0); return(man_body_alloc(m, ln, 0, m->last->tok)); @@ -612,4 +681,46 @@ 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; }