=================================================================== RCS file: /cvs/mandoc/roff.c,v retrieving revision 1.68 retrieving revision 1.75 diff -u -p -r1.68 -r1.75 --- mandoc/roff.c 2010/05/15 18:43:59 1.68 +++ mandoc/roff.c 2010/05/16 13:49:23 1.75 @@ -1,4 +1,4 @@ -/* $Id: roff.c,v 1.68 2010/05/15 18:43:59 kristaps Exp $ */ +/* $Id: roff.c,v 1.75 2010/05/16 13:49:23 kristaps Exp $ */ /* * Copyright (c) 2010 Kristaps Dzonsons * @@ -21,17 +21,33 @@ #include #include #include +#include #include "mandoc.h" #include "roff.h" +#define ROFF_CTL(c) \ + ('.' == (c) || '\'' == (c)) +#ifdef ROFF_DEBUG +#define ROFF_MDEBUG(p, str) \ + fprintf(stderr, "%s: %s (%d:%d)\n", (str), \ + roffs[(p)->last->tok].name, \ + (p)->last->line, (p)->last->col) +#else +#define ROFF_MDEBUG(p, str) while (/* CONSTCOND */ 0) +#endif + enum rofft { - ROFF_de, - ROFF_dei, + ROFF_if, + ROFF_ccond, +#if 0 ROFF_am, ROFF_ami, + ROFF_de, + ROFF_dei, ROFF_ig, ROFF_close, +#endif ROFF_MAX }; @@ -44,39 +60,49 @@ 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 endspan; }; #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, /* original pos in buffer */ \ + int pos, /* current pos in buffer */ \ + int *offs /* reset offset of buffer data */ typedef enum rofferr (*roffproc)(ROFF_ARGS); struct roffmac { const char *name; /* macro name */ - roffproc sub; /* child of control black */ - roffproc new; /* root of stack (type = ROFF_MAX) */ + roffproc proc; }; -static enum rofferr roff_ignore(ROFF_ARGS); +static enum rofferr roff_if(ROFF_ARGS); +static enum rofferr roff_ccond(ROFF_ARGS); +#if 0 static enum rofferr roff_new_close(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", roff_if }, + { "\\}", roff_ccond }, +#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 }, { "ig", roff_sub_ig, roff_new_ig }, { ".", NULL, roff_new_close }, +#endif }; -static void roff_alloc1(struct roff *); static void roff_free1(struct roff *); static enum rofft roff_hash_find(const char *); static int roffnode_push(struct roff *, @@ -113,9 +139,11 @@ roffnode_pop(struct roff *r) { struct roffnode *p; - if (NULL == (p = r->last)) - return; - r->last = p->parent; + assert(r->last); + p = r->last; + r->last = r->last->parent; + if (p->end) + free(p->end); free(p); } @@ -153,20 +181,11 @@ roff_free1(struct roff *r) } -static void -roff_alloc1(struct roff *r) -{ - - /* Do nothing for now. */ -} - - void roff_reset(struct roff *r) { roff_free1(r); - roff_alloc1(r); } @@ -196,33 +215,48 @@ 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 (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 when in free text without a context. */ + + if (r->last && ! ROFF_CTL((*bufp)[pos])) { + /* XXX: this assumes we're just discarding. */ + while (r->last) { + if (r->last->endspan-- < 0) + break; + ROFF_MDEBUG(r, "closing implicit scope"); + roffnode_pop(r); + } + return(ROFF_IGN); + } else if ( ! ROFF_CTL((*bufp)[pos])) return(ROFF_CONT); /* There's nothing on the stack: make us anew. */ - if (ROFF_MAX == (t = roff_parse(*bufp, &ppos))) + ppos = pos; + if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) return(ROFF_CONT); - assert(roffs[t].new); - return((*roffs[t].new)(r, bufp, szp, ln, ppos)); + assert(roffs[t].proc); + return((*roffs[t].proc)(r, t, bufp, szp, ln, ppos, 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. @@ -234,8 +268,8 @@ roff_parse(const char *buf, int *pos) char mac[5]; enum rofft t; - assert('.' == buf[0]); - *pos = 1; + assert(ROFF_CTL(buf[*pos])); + (*pos)++; while (buf[*pos] && (' ' == buf[*pos] || '\t' == buf[*pos])) (*pos)++; @@ -264,67 +298,181 @@ roff_parse(const char *buf, int *pos) } +#if 0 /* ARGSUSED */ static enum rofferr -roff_ignore(ROFF_ARGS) +roff_sub_ig(ROFF_ARGS) { + int i, j; + /* Ignore free-text lines. */ + + if ('.' != (*bufp)[ppos] && '\'' != (*bufp)[ppos]) + return(ROFF_IGN); + + if (r->last->end) { + i = ppos + 1; + + 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); } /* ARGSUSED */ static enum rofferr -roff_sub_ig(ROFF_ARGS) +roff_new_close(ROFF_ARGS) { - enum rofft t; - int pos; - /* Ignore free-text lines. */ + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); - if ('.' != (*bufp)[ppos]) - return(ROFF_IGN); + return(ROFF_IGN); +} +#endif - /* Ignore macros unless it's a closing macro. */ - t = roff_parse(*bufp, &pos); - if (ROFF_close != t) +/* ARGSUSED */ +static enum rofferr +roff_ccond(ROFF_ARGS) +{ + + if (NULL == r->last || ROFF_if != r->last->tok || r->last->endspan > -1) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); return(ROFF_IGN); + } + ROFF_MDEBUG(r, "closing explicit scope"); roffnode_pop(r); + + while (r->last) { + if (--r->last->endspan < 0) + break; + + ROFF_MDEBUG(r, "closing implicit scope"); + roffnode_pop(r); + } + return(ROFF_IGN); } /* ARGSUSED */ static enum rofferr -roff_new_close(ROFF_ARGS) +roff_if(ROFF_ARGS) { /* - if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, 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. + */ + + if ( ! roffnode_push(r, tok, ln, ppos)) return(ROFF_ERR); - */ - return(ROFF_CONT); + + while ((*bufp)[pos] && ' ' != (*bufp)[pos]) + pos++; + while (' ' == (*bufp)[pos]) + pos++; + + /* Don't evaluate: just assume NO. */ + + r->last->endspan = 1; + + if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) { + ROFF_MDEBUG(r, "opening explicit scope"); + r->last->endspan = -1; + pos += 2; + } else + ROFF_MDEBUG(r, "opening implicit scope"); + + if ('\0' == (*bufp)[pos]) + return(ROFF_IGN); + + *offs = pos; + return(ROFF_RERUN); } -/* ARGSUSED */ +#if 0 static enum rofferr roff_new_ig(ROFF_ARGS) { + int i; - return(roffnode_push(r, ROFF_ig, ln, ppos) ? - ROFF_IGN : ROFF_ERR); -} + 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++; + } -int -roff_endparse(struct roff *r) -{ + i = (int)ppos; - if (NULL == r->last) - return(1); - return((*r->msg)(MANDOCERR_SCOPEEXIT, r->data, - r->last->line, r->last->col, NULL)); + 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