version 1.67, 2010/05/15 18:35:14 |
version 1.78, 2010/05/16 22:28:33 |
|
|
#include <assert.h> |
#include <assert.h> |
#include <stdlib.h> |
#include <stdlib.h> |
#include <string.h> |
#include <string.h> |
|
#include <stdio.h> |
|
|
#include "mandoc.h" |
#include "mandoc.h" |
#include "roff.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 { |
enum rofft { |
ROFF_de, |
ROFF_if, |
ROFF_dei, |
ROFF_ig, |
|
ROFF_cblock, |
|
ROFF_ccond, |
|
#if 0 |
ROFF_am, |
ROFF_am, |
ROFF_ami, |
ROFF_ami, |
ROFF_ig, |
ROFF_de, |
|
ROFF_dei, |
ROFF_close, |
ROFF_close, |
|
#endif |
ROFF_MAX |
ROFF_MAX |
}; |
}; |
|
|
|
|
struct roffnode { |
struct roffnode { |
enum rofft tok; /* type of node */ |
enum rofft tok; /* type of node */ |
struct roffnode *parent; /* up one in stack */ |
struct roffnode *parent; /* up one in stack */ |
|
char *end; /* end-token: custom */ |
int line; /* parse line */ |
int line; /* parse line */ |
int col; /* parse col */ |
int col; /* parse col */ |
|
int endspan; |
}; |
}; |
|
|
#define ROFF_ARGS struct roff *r, /* parse ctx */ \ |
#define ROFF_ARGS struct roff *r, /* parse ctx */ \ |
|
enum rofft tok, /* tok of macro */ \ |
char **bufp, /* input buffer */ \ |
char **bufp, /* input buffer */ \ |
size_t *szp, /* size of input buffer */ \ |
size_t *szp, /* size of input buffer */ \ |
int ln, /* parse line */ \ |
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); |
typedef enum rofferr (*roffproc)(ROFF_ARGS); |
|
|
struct roffmac { |
struct roffmac { |
const char *name; /* macro name */ |
const char *name; /* macro name */ |
roffproc sub; /* child of control black */ |
roffproc proc; |
roffproc new; /* root of stack (type = ROFF_MAX) */ |
roffproc text; |
}; |
}; |
|
|
static enum rofferr roff_ignore(ROFF_ARGS); |
static enum rofferr roff_if(ROFF_ARGS); |
static enum rofferr roff_new_close(ROFF_ARGS); |
static enum rofferr roff_if_text(ROFF_ARGS); |
static enum rofferr roff_new_ig(ROFF_ARGS); |
static enum rofferr roff_ig(ROFF_ARGS); |
static enum rofferr roff_sub_ig(ROFF_ARGS); |
static enum rofferr roff_ig_text(ROFF_ARGS); |
|
static enum rofferr roff_cblock(ROFF_ARGS); |
|
static enum rofferr roff_ccond(ROFF_ARGS); |
|
|
const struct roffmac roffs[ROFF_MAX] = { |
const struct roffmac roffs[ROFF_MAX] = { |
{ "de", NULL, roff_ignore }, |
{ "if", roff_if, roff_if_text }, |
{ "dei", NULL, roff_ignore }, |
{ "ig", roff_ig, roff_ig_text }, |
{ "am", NULL, roff_ignore }, |
{ ".", roff_cblock, NULL }, |
{ "ami", NULL, roff_ignore }, |
{ "\\}", roff_ccond, NULL }, |
{ "ig", roff_sub_ig, roff_new_ig }, |
|
{ ".", NULL, roff_new_close }, |
|
}; |
}; |
|
|
static void roff_alloc1(struct roff *); |
|
static void roff_free1(struct roff *); |
static void roff_free1(struct roff *); |
static enum rofft roff_hash_find(const char *); |
static enum rofft roff_hash_find(const char *); |
|
static void roffnode_cleanscope(struct roff *); |
static int roffnode_push(struct roff *, |
static int roffnode_push(struct roff *, |
enum rofft, int, int); |
enum rofft, int, int); |
static void roffnode_pop(struct roff *); |
static void roffnode_pop(struct roff *); |
Line 113 roffnode_pop(struct roff *r) |
|
Line 135 roffnode_pop(struct roff *r) |
|
{ |
{ |
struct roffnode *p; |
struct roffnode *p; |
|
|
if (NULL == (p = r->last)) |
assert(r->last); |
return; |
p = r->last; |
r->last = p->parent; |
r->last = r->last->parent; |
|
if (p->end) |
|
free(p->end); |
free(p); |
free(p); |
} |
} |
|
|
Line 153 roff_free1(struct roff *r) |
|
Line 177 roff_free1(struct roff *r) |
|
} |
} |
|
|
|
|
static void |
|
roff_alloc1(struct roff *r) |
|
{ |
|
|
|
memset(r, 0, sizeof(struct roff)); |
|
} |
|
|
|
|
|
void |
void |
roff_reset(struct roff *r) |
roff_reset(struct roff *r) |
{ |
{ |
|
|
roff_free1(r); |
roff_free1(r); |
roff_alloc1(r); |
|
} |
} |
|
|
|
|
Line 196 roff_alloc(const mandocmsg msg, void *data) |
|
Line 211 roff_alloc(const mandocmsg msg, void *data) |
|
|
|
|
|
enum rofferr |
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; |
enum rofft t; |
int ppos; |
int ppos, i, j, wtf; |
|
|
if (NULL != r->last) { |
if (r->last && ! ROFF_CTL((*bufp)[pos])) { |
/* |
/* |
* If there's a node on the stack, then jump directly |
* If a scope is open and we're not a macro, pass it |
* into its processing function. |
* through our text detector and continue as quickly as |
|
* possible. |
*/ |
*/ |
t = r->last->tok; |
t = r->last->tok; |
assert(roffs[t].sub); |
assert(roffs[t].text); |
return((*roffs[t].sub)(r, bufp, szp, ln, 0)); |
return((*roffs[t].text) |
} else if ('.' != (*bufp)[0] && NULL == r->last) |
(r, t, bufp, szp, ln, pos, pos, offs)); |
/* Return when in free text without a context. */ |
} else if ( ! ROFF_CTL((*bufp)[pos])) |
|
/* |
|
* Don't do anything if we're free-form text. |
|
*/ |
return(ROFF_CONT); |
return(ROFF_CONT); |
|
|
/* There's nothing on the stack: make us anew. */ |
/* A macro-ish line with a possibly-open macro context. */ |
|
|
if (ROFF_MAX == (t = roff_parse(*bufp, &ppos))) |
wtf = 0; |
return(ROFF_CONT); |
|
|
|
assert(roffs[t].new); |
if (r->last && r->last->end) { |
return((*roffs[t].new)(r, bufp, szp, ln, ppos)); |
/* |
|
* We have a scope open that has a custom end-macro |
|
* handler. Try to match it against the input. |
|
*/ |
|
i = pos + 1; |
|
while (' ' == (*bufp)[i] || '\t' == (*bufp)[i]) |
|
i++; |
|
|
|
for (j = 0; r->last->end[j]; j++, i++) |
|
if ((*bufp)[i] != r->last->end[j]) |
|
break; |
|
|
|
if ('\0' == r->last->end[j] && |
|
('\0' == (*bufp)[i] || |
|
' ' == (*bufp)[i] || |
|
'\t' == (*bufp)[i])) { |
|
roffnode_pop(r); |
|
roffnode_cleanscope(r); |
|
wtf = 1; |
|
} |
|
} |
|
|
|
ppos = pos; |
|
if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) { |
|
/* |
|
* This is some of groff's stranger behaviours. If we |
|
* encountered a custom end-scope tag and that tag also |
|
* happens to be a "real" macro, then we need to try |
|
* interpreting it again as a real macro. If it's not, |
|
* then return ignore. Else continue. |
|
*/ |
|
if (wtf) |
|
return(ROFF_IGN); |
|
else if (NULL == r->last) |
|
return(ROFF_CONT); |
|
|
|
/* FIXME: this assumes that we ignore!? */ |
|
return(ROFF_IGN); |
|
} |
|
|
|
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 |
* Parse a roff node's type from the input buffer. This must be in the |
* form of ".foo xxx" in the usual way. |
* form of ".foo xxx" in the usual way. |
Line 234 roff_parse(const char *buf, int *pos) |
|
Line 306 roff_parse(const char *buf, int *pos) |
|
char mac[5]; |
char mac[5]; |
enum rofft t; |
enum rofft t; |
|
|
assert('.' == buf[0]); |
assert(ROFF_CTL(buf[*pos])); |
*pos = 1; |
(*pos)++; |
|
|
while (buf[*pos] && (' ' == buf[*pos] || '\t' == buf[*pos])) |
while (buf[*pos] && (' ' == buf[*pos] || '\t' == buf[*pos])) |
(*pos)++; |
(*pos)++; |
Line 266 roff_parse(const char *buf, int *pos) |
|
Line 338 roff_parse(const char *buf, int *pos) |
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_ignore(ROFF_ARGS) |
roff_cblock(ROFF_ARGS) |
{ |
{ |
|
|
|
if (NULL == r->last) { |
|
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
|
return(ROFF_ERR); |
|
return(ROFF_IGN); |
|
} |
|
|
|
if (ROFF_ig != r->last->tok) { |
|
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 ignore block"); |
|
roffnode_pop(r); |
|
roffnode_cleanscope(r); |
return(ROFF_IGN); |
return(ROFF_IGN); |
|
|
} |
} |
|
|
|
|
|
static void |
|
roffnode_cleanscope(struct roff *r) |
|
{ |
|
|
|
while (r->last) { |
|
if (--r->last->endspan < 0) |
|
break; |
|
ROFF_MDEBUG(r, "closing implicit scope"); |
|
roffnode_pop(r); |
|
} |
|
} |
|
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_sub_ig(ROFF_ARGS) |
roff_ccond(ROFF_ARGS) |
{ |
{ |
enum rofft t; |
|
int pos; |
|
|
|
/* 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]) |
if (ROFF_if != r->last->tok) { |
|
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
|
return(ROFF_ERR); |
return(ROFF_IGN); |
return(ROFF_IGN); |
|
} |
|
|
/* Ignore macros unless it's a closing macro. */ |
if (r->last->endspan > -1) { |
|
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
t = roff_parse(*bufp, &pos); |
return(ROFF_ERR); |
if (ROFF_close != t) |
|
return(ROFF_IGN); |
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_pop(r); |
|
roffnode_cleanscope(r); |
return(ROFF_IGN); |
return(ROFF_IGN); |
} |
} |
|
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_new_close(ROFF_ARGS) |
roff_ig(ROFF_ARGS) |
{ |
{ |
|
int sv; |
|
size_t sz; |
|
|
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
if ( ! roffnode_push(r, tok, ln, ppos)) |
return(ROFF_ERR); |
return(ROFF_ERR); |
|
|
|
if ('\0' == (*bufp)[pos]) { |
|
ROFF_MDEBUG(r, "opening ignore block"); |
|
return(ROFF_IGN); |
|
} |
|
|
|
sv = pos; |
|
while ((*bufp)[pos] && ' ' != (*bufp)[pos] && |
|
'\t' != (*bufp)[pos]) |
|
pos++; |
|
|
|
/* |
|
* Note: groff does NOT like escape characters in the input. |
|
* Instead of detecting this, we're just going to let it fly and |
|
* to hell with it. |
|
*/ |
|
|
|
assert(pos > sv); |
|
sz = (size_t)(pos - sv); |
|
|
|
r->last->end = malloc(sz + 1); |
|
|
|
if (NULL == r->last->end) { |
|
(*r->msg)(MANDOCERR_MEM, r->data, ln, pos, NULL); |
|
return(ROFF_ERR); |
|
} |
|
|
|
memcpy(r->last->end, *bufp + sv, sz); |
|
r->last->end[(int)sz] = '\0'; |
|
|
|
ROFF_MDEBUG(r, "opening explicit ignore block"); |
|
|
|
if ((*bufp)[pos]) |
|
if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL)) |
|
return(ROFF_ERR); |
|
|
return(ROFF_IGN); |
return(ROFF_IGN); |
} |
} |
|
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_new_ig(ROFF_ARGS) |
roff_ig_text(ROFF_ARGS) |
{ |
{ |
|
|
return(roffnode_push(r, ROFF_ig, ln, ppos) ? |
return(ROFF_IGN); |
ROFF_IGN : ROFF_ERR); |
|
} |
} |
|
|
|
|
int |
/* ARGSUSED */ |
roff_endparse(struct roff *r) |
static enum rofferr |
|
roff_if_text(ROFF_ARGS) |
{ |
{ |
|
char *ep, *st; |
|
|
if (NULL == r->last) |
st = &(*bufp)[pos]; |
return(1); |
if (NULL == (ep = strstr(st, "\\}"))) { |
return((*r->msg)(MANDOCERR_SCOPEEXIT, r->data, |
roffnode_cleanscope(r); |
r->last->line, r->last->col, NULL)); |
return(ROFF_IGN); |
|
} |
|
|
|
if (ep > st && '\\' != *(ep - 1)) { |
|
ROFF_MDEBUG(r, "closing explicit scope (in-line)"); |
|
roffnode_pop(r); |
|
} |
|
|
|
roffnode_cleanscope(r); |
|
return(ROFF_IGN); |
|
} |
|
|
|
|
|
/* ARGSUSED */ |
|
static enum rofferr |
|
roff_if(ROFF_ARGS) |
|
{ |
|
int sv; |
|
|
|
/* |
|
* 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)[pos] && ' ' != (*bufp)[pos]) |
|
pos++; |
|
|
|
sv = pos; |
|
while (' ' == (*bufp)[pos]) |
|
pos++; |
|
|
|
/* |
|
* Roff is weird. If we have just white-space after the |
|
* conditional, it's considered the BODY and we exit without |
|
* really doing anything. Warn about this. It's probably |
|
* wrong. |
|
*/ |
|
|
|
if ('\0' == (*bufp)[pos] && sv != pos) { |
|
if ( ! (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL)) |
|
return(ROFF_ERR); |
|
return(ROFF_IGN); |
|
} |
|
|
|
if ( ! roffnode_push(r, tok, ln, ppos)) |
|
return(ROFF_ERR); |
|
|
|
/* 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 there are no arguments on the line, the next-line scope is |
|
* assumed. |
|
*/ |
|
|
|
if ('\0' == (*bufp)[pos]) |
|
return(ROFF_IGN); |
|
|
|
/* Otherwise re-run the roff parser after recalculating. */ |
|
|
|
*offs = pos; |
|
return(ROFF_RERUN); |
} |
} |