=================================================================== RCS file: /cvs/mandoc/man.c,v retrieving revision 1.53 retrieving revision 1.62 diff -u -p -r1.53 -r1.62 --- mandoc/man.c 2010/03/23 21:50:43 1.53 +++ mandoc/man.c 2010/05/08 07:30:19 1.62 @@ -1,4 +1,4 @@ -/* $Id: man.c,v 1.53 2010/03/23 21:50:43 kristaps Exp $ */ +/* $Id: man.c,v 1.62 2010/05/08 07:30:19 kristaps Exp $ */ /* * Copyright (c) 2008, 2009 Kristaps Dzonsons * @@ -49,7 +49,11 @@ 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 */ + "scope in roff instructions broken", /* WROFFSCOPE */ + "document title should be uppercase", /* WTITLECASE */ + "deprecated comment style", /* WBADCOMMENT */ }; const char *const __man_macronames[MAN_MAX] = { @@ -61,7 +65,9 @@ const char *const __man_macronames[MAN_MAX] = { "RI", "na", "i", "sp", "nf", "fi", "r", "RE", "RS", "DT", "UC", "PD", - "Sp", "Vb", "Ve", + "Sp", "Vb", "Ve", "de", + "dei", "am", "ami", "ig", + ".", }; const char * const *man_macronames = __man_macronames; @@ -70,12 +76,13 @@ static struct man_node *man_node_alloc(int, 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 *); @@ -149,7 +156,7 @@ 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)); } @@ -160,7 +167,7 @@ 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) @@ -179,6 +186,7 @@ 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; } @@ -206,6 +214,7 @@ man_node_append(struct man *man, struct man_node *p) /* NOTREACHED */ } + assert(p->parent); p->parent->nchild++; if ( ! man_valid_pre(man, p)) @@ -307,68 +316,65 @@ man_block_alloc(struct man *m, int line, int pos, enum } -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; + 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; - char sv; + 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) { @@ -377,61 +383,32 @@ 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]) { - /* 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)) + /* 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. */ - /* Escaped whitespace. */ - if (i && ' ' == buf[i] && '\\' == buf[i - 1]) - continue; + i = (int)strlen(buf); + assert(i); - 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 (' ' == buf[i - 1] || '\t' == buf[i - 1]) + if (1 == i || ('\\' != buf[i - 2])) if ( ! man_pwarn(m, line, i - 1, WTSPACE)) return(0); - for ( ; ' ' == buf[i]; i++) - /* Skip trailing whitespace. */ ; - - j = 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; - } - - if (j != i && ! pstring(m, line, j, &buf[j], (size_t)(i - j))) + if ( ! man_word_alloc(m, line, 0, buf)) return(0); descope: - /* * Co-ordinate what happens with having a next-line scope open: * first close out the element scope (if applicable), then close @@ -440,7 +417,7 @@ 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); } @@ -448,7 +425,7 @@ 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)); } @@ -458,8 +435,7 @@ 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 ? "..." : "")); @@ -469,23 +445,26 @@ macrowarn(struct man *m, int ln, const char *buf) int man_pmacro(struct man *m, int ln, char *buf) { - int i, j, 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]) goto out; @@ -531,7 +510,10 @@ man_pmacro(struct man *m, int ln, char *buf) while (buf[i] && ' ' == buf[i]) i++; - /* Trailing whitespace? */ + /* + * Trailing whitespace. Note that tabs are allowed to be passed + * into the parser as "text", so we only warn about spaces here. + */ if ('\0' == buf[i] && ' ' == buf[i - 1]) if ( ! man_pwarn(m, ln, i - 1, WTSPACE)) @@ -541,6 +523,9 @@ man_pmacro(struct man *m, int ln, char *buf) * 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. + * NOTE: we don't allow roff blocks (NOCLOSE) to be embedded + * here because that would stipulate blocks as children of + * elements! */ if ( ! (MAN_NSCOPED & man_macros[tok].flags) && @@ -568,15 +553,22 @@ man_pmacro(struct man *m, int ln, char *buf) if ( ! man_nwarn(m, n, WLNSCOPE)) return(0); - man_node_unlink(m, n); - 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[tok].fp); + if (MAN_BLINE & m->flags) + m->flags |= MAN_BPLINE; + /* Call to handler... */ + + assert(man_macros[tok].fp); if ( ! (*man_macros[tok].fp)(m, tok, ln, ppos, &i, buf)) goto err; @@ -586,10 +578,11 @@ out: * above-parsed macro, so return. */ - if ( ! (MAN_BLINE & fl)) { + 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 @@ -614,7 +607,7 @@ 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)); @@ -672,26 +665,43 @@ man_err(struct man *m, int line, int pos, int iserr, e } -void +/* + * 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) { - if (n->prev) { + /* Adjust siblings. */ + + if (n->prev) n->prev->next = n->next; - if (m->last == n) { - assert(NULL == 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 { - n->parent->child = n->next; - if (m->last == n) { - assert(NULL == n->next); + } else { m->last = n->parent; m->next = MAN_NEXT_CHILD; } } - if (n->next) - n->next->prev = n->prev; + if (m && m->first == n) + m->first = NULL; }