=================================================================== RCS file: /cvs/mandoc/roff.c,v retrieving revision 1.74 retrieving revision 1.76 diff -u -p -r1.74 -r1.76 --- mandoc/roff.c 2010/05/16 10:59:36 1.74 +++ mandoc/roff.c 2010/05/16 14:47:19 1.76 @@ -1,4 +1,4 @@ -/* $Id: roff.c,v 1.74 2010/05/16 10:59:36 kristaps Exp $ */ +/* $Id: roff.c,v 1.76 2010/05/16 14:47:19 kristaps Exp $ */ /* * Copyright (c) 2010 Kristaps Dzonsons * @@ -21,18 +21,32 @@ #include #include #include +#include #include "mandoc.h" #include "roff.h" +#define ROFF_CTL(c) \ + ('.' == (c) || '\'' == (c)) +#if 0 +#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_if, + ROFF_ig, + ROFF_cblock, + ROFF_ccond, #if 0 ROFF_am, ROFF_ami, ROFF_de, ROFF_dei, - ROFF_if, - ROFF_ig, ROFF_close, #endif ROFF_MAX @@ -50,8 +64,7 @@ struct roffnode { char *end; /* end-token: custom */ int line; /* parse line */ int col; /* parse col */ - int flags; -#define ROFF_PARSEONLY (1 << 0) + int endspan; }; #define ROFF_ARGS struct roff *r, /* parse ctx */ \ @@ -59,39 +72,38 @@ struct roffnode { 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; }; -#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 +static enum rofferr roff_if(ROFF_ARGS); +static enum rofferr roff_ig(ROFF_ARGS); +static enum rofferr roff_cblock(ROFF_ARGS); +static enum rofferr roff_ccond(ROFF_ARGS); const struct roffmac roffs[ROFF_MAX] = { + { "if", roff_if }, + { "ig", roff_ig }, + { ".", roff_cblock }, + { "\\}", 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 }, - { "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 *); static enum rofft roff_hash_find(const char *); +static void roffnode_cleanscope(struct roff *); static int roffnode_push(struct roff *, enum rofft, int, int); static void roffnode_pop(struct roff *); @@ -126,9 +138,9 @@ 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); @@ -206,27 +218,28 @@ 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) { - t = r->last->tok; - assert(roffs[t].sub); - 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) + if (r->last && ! ROFF_CTL((*bufp)[pos])) { + if (ROFF_ig == r->last->tok) + return(ROFF_IGN); + roffnode_cleanscope(r); + /* FIXME: this assumes we're discarding! */ + 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, &pos))) + ppos = pos; + if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) { + if (r->last && ROFF_ig == r->last->tok) + return(ROFF_IGN); return(ROFF_CONT); + } - assert(roffs[t].new); - return((*roffs[t].new)(r, t, bufp, szp, ln, pos, offs)); + assert(roffs[t].proc); + return((*roffs[t].proc)(r, t, bufp, szp, ln, ppos, pos, offs)); } @@ -252,8 +265,8 @@ roff_parse(const char *buf, int *pos) char mac[5]; enum rofft t; - assert('.' == buf[0] || '\'' == buf[0]); - *pos = 1; + assert(ROFF_CTL(buf[*pos])); + (*pos)++; while (buf[*pos] && (' ' == buf[*pos] || '\t' == buf[*pos])) (*pos)++; @@ -282,82 +295,102 @@ roff_parse(const char *buf, int *pos) } -#if 0 /* ARGSUSED */ static enum rofferr -roff_sub_ig(ROFF_ARGS) +roff_cblock(ROFF_ARGS) { - int i, j; - /* Ignore free-text lines. */ + if (NULL == r->last) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); + } - if ('.' != (*bufp)[ppos] && '\'' != (*bufp)[ppos]) + if (ROFF_ig != r->last->tok) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); return(ROFF_IGN); + } - if (r->last->end) { - i = ppos + 1; + if ((*bufp)[pos]) + if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL)) + return(ROFF_ERR); - while ((*bufp)[i] && ' ' == (*bufp)[i]) - i++; + ROFF_MDEBUG(r, "closing ignore block"); + roffnode_pop(r); + roffnode_cleanscope(r); + return(ROFF_IGN); - 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++; +static void +roffnode_cleanscope(struct roff *r) +{ - } 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); + while (r->last) { + if (--r->last->endspan < 0) + break; + ROFF_MDEBUG(r, "closing implicit scope"); + roffnode_pop(r); + } } /* ARGSUSED */ static enum rofferr -roff_new_close(ROFF_ARGS) +roff_ccond(ROFF_ARGS) { - if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) - return(ROFF_ERR); + if (NULL == r->last) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); + } + if (ROFF_if != r->last->tok) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); + } + + if (r->last->endspan > -1) { + if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); + } + + if ((*bufp)[pos]) + if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL)) + return(ROFF_ERR); + + ROFF_MDEBUG(r, "closing explicit scope"); + roffnode_pop(r); + roffnode_cleanscope(r); return(ROFF_IGN); } +/* ARGSUSED */ static enum rofferr -roff_sub_if(ROFF_ARGS) +roff_ig(ROFF_ARGS) { - int i; - enum rofft t; - i = (int)strlen(*bufp); + if ( ! roffnode_push(r, tok, ln, ppos)) + return(ROFF_ERR); - if (i > 1 && '}' == (*bufp)[i - 1] && '\\' == (*bufp)[i - 2]) - roffnode_pop(r); + ROFF_MDEBUG(r, "opening ignore block"); + /* FIXME: warn about end of line. */ + return(ROFF_IGN); } +/* ARGSUSED */ static enum rofferr -roff_new_if(ROFF_ARGS) +roff_if(ROFF_ARGS) { - struct roffnode *n; - enum rofferr re; /* * Read ahead past the conditional. @@ -366,85 +399,28 @@ roff_new_if(ROFF_ARGS) * 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; + while ((*bufp)[pos] && ' ' != (*bufp)[pos]) + pos++; + while (' ' == (*bufp)[pos]) + pos++; /* Don't evaluate: just assume NO. */ - r->last->flags |= ROFF_PARSEONLY; + r->last->endspan = 1; - 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); - } + 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"); - 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) + if ('\0' == (*bufp)[pos]) 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); + *offs = pos; + return(ROFF_RERUN); } -#endif