=================================================================== RCS file: /cvs/mandoc/roff.c,v retrieving revision 1.70 retrieving revision 1.74 diff -u -p -r1.70 -r1.74 --- mandoc/roff.c 2010/05/15 20:51:40 1.70 +++ mandoc/roff.c 2010/05/16 10:59:36 1.74 @@ -1,4 +1,4 @@ -/* $Id: roff.c,v 1.70 2010/05/15 20:51:40 kristaps Exp $ */ +/* $Id: roff.c,v 1.74 2010/05/16 10:59:36 kristaps Exp $ */ /* * Copyright (c) 2010 Kristaps Dzonsons * @@ -26,12 +26,15 @@ #include "roff.h" enum rofft { - ROFF_de, - ROFF_dei, +#if 0 ROFF_am, ROFF_ami, + ROFF_de, + ROFF_dei, + ROFF_if, ROFF_ig, ROFF_close, +#endif ROFF_MAX }; @@ -44,15 +47,20 @@ struct roff { struct roffnode { enum rofft tok; /* type of node */ struct roffnode *parent; /* up one in stack */ + char *end; /* end-token: custom */ int line; /* parse line */ int col; /* parse col */ + int flags; +#define ROFF_PARSEONLY (1 << 0) }; #define ROFF_ARGS struct roff *r, /* parse ctx */ \ + enum rofft tok, /* tok of macro */ \ char **bufp, /* input buffer */ \ size_t *szp, /* size of input buffer */ \ int ln, /* parse line */ \ - int ppos /* current pos in buffer */ + int ppos, /* current pos in buffer */ \ + int *offs /* reset offset of buffer data */ typedef enum rofferr (*roffproc)(ROFF_ARGS); @@ -62,18 +70,24 @@ struct roffmac { roffproc new; /* root of stack (type = ROFF_MAX) */ }; -static enum rofferr roff_ignore(ROFF_ARGS); +#if 0 static enum rofferr roff_new_close(ROFF_ARGS); +static enum rofferr roff_new_if(ROFF_ARGS); +static enum rofferr roff_sub_if(ROFF_ARGS); static enum rofferr roff_new_ig(ROFF_ARGS); static enum rofferr roff_sub_ig(ROFF_ARGS); +#endif const struct roffmac roffs[ROFF_MAX] = { - { "de", NULL, roff_ignore }, - { "dei", NULL, roff_ignore }, - { "am", NULL, roff_ignore }, - { "ami", NULL, roff_ignore }, +#if 0 + { "am", roff_sub_ig, roff_new_ig }, + { "ami", roff_sub_ig, roff_new_ig }, + { "de", roff_sub_ig, roff_new_ig }, + { "dei", roff_sub_ig, roff_new_ig }, + { "if", roff_sub_if, roff_new_if }, { "ig", roff_sub_ig, roff_new_ig }, { ".", NULL, roff_new_close }, +#endif }; static void roff_free1(struct roff *); @@ -115,6 +129,8 @@ roffnode_pop(struct roff *r) if (NULL == (p = r->last)) return; r->last = p->parent; + if (p->end) + free(p->end); free(p); } @@ -186,33 +202,45 @@ roff_alloc(const mandocmsg msg, void *data) enum rofferr -roff_parseln(struct roff *r, int ln, char **bufp, size_t *szp) +roff_parseln(struct roff *r, int ln, + char **bufp, size_t *szp, int pos, int *offs) { enum rofft t; - int ppos; + /* If stacked, jump directly into its processing function. */ + if (NULL != r->last) { - /* - * If there's a node on the stack, then jump directly - * into its processing function. - */ t = r->last->tok; assert(roffs[t].sub); - return((*roffs[t].sub)(r, bufp, szp, ln, 0)); - } else if ('.' != (*bufp)[0] && NULL == r->last) - /* Return when in free text without a context. */ + return((*roffs[t].sub)(r, t, bufp, szp, ln, pos, offs)); + } + + /* Return when in free text without a context. */ + + if ('.' != (*bufp)[0] && '\'' != (*bufp)[0] && NULL == r->last) return(ROFF_CONT); /* There's nothing on the stack: make us anew. */ - if (ROFF_MAX == (t = roff_parse(*bufp, &ppos))) + if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) return(ROFF_CONT); assert(roffs[t].new); - return((*roffs[t].new)(r, bufp, szp, ln, ppos)); + return((*roffs[t].new)(r, t, bufp, szp, ln, pos, offs)); } +int +roff_endparse(struct roff *r) +{ + + if (NULL == r->last) + return(1); + return((*r->msg)(MANDOCERR_SCOPEEXIT, r->data, r->last->line, + r->last->col, NULL)); +} + + /* * Parse a roff node's type from the input buffer. This must be in the * form of ".foo xxx" in the usual way. @@ -224,7 +252,7 @@ roff_parse(const char *buf, int *pos) char mac[5]; enum rofft t; - assert('.' == buf[0]); + assert('.' == buf[0] || '\'' == buf[0]); *pos = 1; while (buf[*pos] && (' ' == buf[*pos] || '\t' == buf[*pos])) @@ -254,34 +282,46 @@ roff_parse(const char *buf, int *pos) } +#if 0 /* ARGSUSED */ static enum rofferr -roff_ignore(ROFF_ARGS) -{ - - return(ROFF_IGN); -} - - -/* ARGSUSED */ -static enum rofferr roff_sub_ig(ROFF_ARGS) { - enum rofft t; - int pos; + int i, j; /* Ignore free-text lines. */ - if ('.' != (*bufp)[ppos]) + if ('.' != (*bufp)[ppos] && '\'' != (*bufp)[ppos]) return(ROFF_IGN); - /* Ignore macros unless it's a closing macro. */ + if (r->last->end) { + i = ppos + 1; - t = roff_parse(*bufp, &pos); - if (ROFF_close != t) + while ((*bufp)[i] && ' ' == (*bufp)[i]) + i++; + + for (j = 0; r->last->end[j]; i++, j++) + if ((*bufp)[i] != r->last->end[j]) + return(ROFF_IGN); + + if (r->last->end[j]) + return(ROFF_IGN); + if ((*bufp)[i] && ' ' != (*bufp)[i]) + return(ROFF_IGN); + + while (' ' == (*bufp)[i]) + i++; + + } else if (ROFF_close != roff_parse(*bufp, &i)) return(ROFF_IGN); roffnode_pop(r); + + if ('\0' == (*bufp)[i]) + return(ROFF_IGN); + if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, i, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); } @@ -291,30 +331,120 @@ static enum rofferr roff_new_close(ROFF_ARGS) { - /* if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) return(ROFF_ERR); - */ + return(ROFF_IGN); } -/* ARGSUSED */ static enum rofferr -roff_new_ig(ROFF_ARGS) +roff_sub_if(ROFF_ARGS) { + int i; + enum rofft t; - return(roffnode_push(r, ROFF_ig, ln, ppos) ? - ROFF_IGN : ROFF_ERR); + i = (int)strlen(*bufp); + + if (i > 1 && '}' == (*bufp)[i - 1] && '\\' == (*bufp)[i - 2]) + roffnode_pop(r); + + return(ROFF_IGN); } -int -roff_endparse(struct roff *r) +static enum rofferr +roff_new_if(ROFF_ARGS) { + struct roffnode *n; + enum rofferr re; - if (NULL == r->last) - return(1); - return((*r->msg)(MANDOCERR_SCOPEEXIT, r->data, - r->last->line, r->last->col, NULL)); + /* + * Read ahead past the conditional. + * FIXME: this does not work, as conditionals don't end on + * whitespace, but are parsed according to a formal grammar. + * It's good enough for now, however. + */ + + while ((*bufp)[ppos] && ' ' != (*bufp)[ppos]) + ppos++; + while (' ' == (*bufp)[ppos]) + ppos++; + + if ( ! roffnode_push(r, tok, ln, ppos)) + return(ROFF_ERR); + + n = r->last; + + /* Don't evaluate: just assume NO. */ + + r->last->flags |= ROFF_PARSEONLY; + + if ('\\' == (*bufp)[ppos] && '{' == (*bufp)[ppos + 1]) { + re = roff_parseln(r, ln, bufp, szp, pos); + if (ROFF_ERR == re) + return(re); + if (r->last == n) + roffnode_pop(r, tok, ln, ppos); + return(re); + } + + return(ROFF_IGN); } + + +static enum rofferr +roff_new_ig(ROFF_ARGS) +{ + int i; + + if ( ! roffnode_push(r, tok, ln, ppos)) + return(ROFF_ERR); + + /* + * Other macros (not `ig') using this routine have additional + * crap here that we discard. + */ + + if (ROFF_ig != tok) { + while ((*bufp)[ppos] && ' ' != (*bufp)[ppos]) + ppos++; + while (' ' == (*bufp)[ppos]) + ppos++; + } + + i = (int)ppos; + + while ((*bufp)[i] && ' ' != (*bufp)[i]) + i++; + + if (i == (int)ppos) + return(ROFF_IGN); + + if ((*bufp)[i]) + if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, i, NULL)) + return(ROFF_ERR); + + /* + * If the macro has arguments, the first argument (up to the + * next whitespace) is interpreted as an argument marking the + * macro close. Thus, `.ig foo' will close at `.foo'. + * + * NOTE: the closing macro `.foo' in the above case is not + * allowed to have leading spaces with old groff! Thus `.foo' + * != `. foo'. Oh yeah, everything after the `.foo' is lost. + * Merry fucking Christmas. + */ + + r->last->end = malloc((size_t)(i - ppos) + 1); + if (NULL == r->last->end) { + (*r->msg)(MANDOCERR_MEM, r->data, ln, ppos, NULL); + return(ROFF_ERR); + } + + memcpy(r->last->end, &(*bufp)[ppos], (size_t)(i - ppos)); + r->last->end[i - ppos] = '\0'; + + return(ROFF_IGN); +} +#endif