version 1.71, 2010/05/15 21:53:11 |
version 1.80, 2010/05/17 00:37:26 |
|
|
#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)) |
|
|
enum rofft { |
enum rofft { |
ROFF_de, |
|
ROFF_dei, |
|
ROFF_am, |
ROFF_am, |
ROFF_ami, |
ROFF_ami, |
|
ROFF_am1, |
|
ROFF_de, |
|
ROFF_dei, |
|
ROFF_de1, |
|
ROFF_if, |
ROFF_ig, |
ROFF_ig, |
ROFF_close, |
ROFF_cblock, |
|
ROFF_ccond, |
|
#if 0 |
|
ROFF_ie, |
|
ROFF_el, |
|
#endif |
ROFF_MAX |
ROFF_MAX |
}; |
}; |
|
|
|
|
void *data; /* privdata for messages */ |
void *data; /* privdata for messages */ |
}; |
}; |
|
|
|
enum roffrule { |
|
ROFFRULE_ALLOW, |
|
ROFFRULE_DENY |
|
}; |
|
|
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; /* custom end-token */ |
|
int line; /* parse line */ |
int line; /* parse line */ |
int col; /* parse col */ |
int col; /* parse col */ |
|
char *end; /* end-rules: custom token */ |
|
int endspan; /* end-rules: next-line or infty */ |
|
enum roffrule rule; |
}; |
}; |
|
|
#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; /* process new macro */ |
roffproc new; /* root of stack (type = ROFF_MAX) */ |
roffproc text; /* process as child text of macro */ |
|
roffproc sub; /* process as child of macro */ |
|
int flags; |
|
#define ROFFMAC_STRUCT (1 << 0) /* always interpret */ |
}; |
}; |
|
|
static enum rofferr roff_ignore(ROFF_ARGS); |
static enum rofferr roff_block(ROFF_ARGS); |
static enum rofferr roff_new_close(ROFF_ARGS); |
static enum rofferr roff_block_text(ROFF_ARGS); |
static enum rofferr roff_new_ig(ROFF_ARGS); |
static enum rofferr roff_block_sub(ROFF_ARGS); |
static enum rofferr roff_sub_ig(ROFF_ARGS); |
static enum rofferr roff_cblock(ROFF_ARGS); |
|
static enum rofferr roff_ccond(ROFF_ARGS); |
|
static enum rofferr roff_if(ROFF_ARGS); |
|
static enum rofferr roff_if_text(ROFF_ARGS); |
|
static enum rofferr roff_if_sub(ROFF_ARGS); |
|
|
const struct roffmac roffs[ROFF_MAX] = { |
const struct roffmac roffs[ROFF_MAX] = { |
{ "de", NULL, roff_ignore }, |
{ "am", roff_block, roff_block_text, roff_block_sub, 0 }, |
{ "dei", NULL, roff_ignore }, |
{ "ami", roff_block, roff_block_text, roff_block_sub, 0 }, |
{ "am", NULL, roff_ignore }, |
{ "am1", roff_block, roff_block_text, roff_block_sub, 0 }, |
{ "ami", NULL, roff_ignore }, |
{ "de", roff_block, roff_block_text, roff_block_sub, 0 }, |
{ "ig", roff_sub_ig, roff_new_ig }, |
{ "dei", roff_block, roff_block_text, roff_block_sub, 0 }, |
{ ".", NULL, roff_new_close }, |
{ "de1", roff_block, roff_block_text, roff_block_sub, 0 }, |
|
{ "if", roff_if, roff_if_text, roff_if_sub, ROFFMAC_STRUCT }, |
|
{ "ig", roff_block, roff_block_text, roff_block_sub, 0 }, |
|
{ ".", roff_cblock, NULL, NULL, 0 }, |
|
{ "\\}", roff_ccond, NULL, NULL, 0 }, |
}; |
}; |
|
|
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 147 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 138 roffnode_push(struct roff *r, enum rofft tok, int line |
|
Line 174 roffnode_push(struct roff *r, enum rofft tok, int line |
|
p->parent = r->last; |
p->parent = r->last; |
p->line = line; |
p->line = line; |
p->col = col; |
p->col = col; |
|
p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY; |
|
|
r->last = p; |
r->last = p; |
return(1); |
return(1); |
Line 187 roff_alloc(const mandocmsg msg, void *data) |
|
Line 224 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; |
|
|
if (NULL != r->last) { |
/* |
/* |
* First, if a scope is open and we're not a macro, pass the |
* If there's a node on the stack, then jump directly |
* text through the macro's filter. If a scope isn't open and |
* into its processing function. |
* we're not a macro, just let it through. |
*/ |
*/ |
|
|
|
if (r->last && ! ROFF_CTL((*bufp)[pos])) { |
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])) |
return(ROFF_CONT); |
return(ROFF_CONT); |
|
|
/* There's nothing on the stack: make us anew. */ |
/* |
|
* If a scope is open, go to the child handler for that macro, |
|
* as it may want to preprocess before doing anything with it. |
|
*/ |
|
|
if (ROFF_MAX == (t = roff_parse(*bufp, &ppos))) |
if (r->last) { |
|
t = r->last->tok; |
|
assert(roffs[t].sub); |
|
return((*roffs[t].sub) |
|
(r, t, bufp, szp, ln, pos, pos, offs)); |
|
} |
|
|
|
/* |
|
* Lastly, as we've no scope open, try to look up and execute |
|
* the new macro. If no macro is found, simply return and let |
|
* the compilers handle it. |
|
*/ |
|
|
|
ppos = pos; |
|
if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) |
return(ROFF_CONT); |
return(ROFF_CONT); |
|
|
assert(roffs[t].new); |
assert(roffs[t].proc); |
return((*roffs[t].new)(r, bufp, szp, ln, ppos)); |
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 225 roff_parse(const char *buf, int *pos) |
|
Line 294 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 257 roff_parse(const char *buf, int *pos) |
|
Line 326 roff_parse(const char *buf, int *pos) |
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_ignore(ROFF_ARGS) |
roff_cblock(ROFF_ARGS) |
{ |
{ |
|
|
|
/* |
|
* A block-close `..' should only be invoked as a child of an |
|
* ignore macro, otherwise raise a warning and just ignore it. |
|
*/ |
|
|
|
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); |
|
|
|
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; |
|
roffnode_pop(r); |
|
} |
|
} |
|
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_sub_ig(ROFF_ARGS) |
roff_ccond(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]) |
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); |
|
} |
|
|
if (r->last->end) { |
if (r->last->endspan > -1) { |
/* |
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
* Allow a macro to break us, if we've defined a special |
return(ROFF_ERR); |
* one for the case. Old groff didn't allow spaces to |
return(ROFF_IGN); |
* buffer the macro, but new groff does. Whee! |
} |
*/ |
|
i = ppos + 1; |
|
while ((*bufp)[i] && ' ' == (*bufp)[i]) |
|
i++; |
|
|
|
if ('\0' == (*bufp)[i]) |
if ((*bufp)[pos]) |
return(ROFF_IGN); |
if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL)) |
|
return(ROFF_ERR); |
|
|
for (j = 0; r->last->end[j]; i++, j++) |
roffnode_pop(r); |
if ((*bufp)[i] != r->last->end[j]) |
roffnode_cleanscope(r); |
return(ROFF_IGN); |
return(ROFF_IGN); |
|
} |
|
|
if (r->last->end[j]) |
|
return(ROFF_IGN); |
|
if ((*bufp)[i] && ' ' != (*bufp)[i]) |
|
return(ROFF_IGN); |
|
|
|
while (' ' == (*bufp)[i]) |
/* ARGSUSED */ |
i++; |
static enum rofferr |
} else if (ROFF_close != roff_parse(*bufp, &i)) |
roff_block(ROFF_ARGS) |
|
{ |
|
int sv; |
|
size_t sz; |
|
|
|
if (ROFF_ig != tok && '\0' == (*bufp)[pos]) { |
|
if ( ! (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL)) |
|
return(ROFF_ERR); |
return(ROFF_IGN); |
return(ROFF_IGN); |
|
} else if (ROFF_ig != tok) { |
|
while ((*bufp)[pos] && ' ' != (*bufp)[pos]) |
|
pos++; |
|
while (' ' == (*bufp)[pos]) |
|
pos++; |
|
} |
|
|
|
if ( ! roffnode_push(r, tok, ln, ppos)) |
|
return(ROFF_ERR); |
|
|
|
if ('\0' == (*bufp)[pos]) |
|
return(ROFF_IGN); |
|
|
|
sv = pos; |
|
while ((*bufp)[pos] && ' ' != (*bufp)[pos] && |
|
'\t' != (*bufp)[pos]) |
|
pos++; |
|
|
/* |
/* |
* Pop off the ignoring context and warn if we're going to lose |
* Note: groff does NOT like escape characters in the input. |
* any of our remaining arguments. |
* Instead of detecting this, we're just going to let it fly and |
|
* to hell with it. |
*/ |
*/ |
|
|
roffnode_pop(r); |
assert(pos > sv); |
|
sz = (size_t)(pos - sv); |
|
|
if ('\0' == (*bufp)[i]) |
if (1 == sz && '.' == (*bufp)[sv]) |
return(ROFF_IGN); |
return(ROFF_IGN); |
if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, i, NULL)) |
|
|
r->last->end = malloc(sz + 1); |
|
|
|
if (NULL == r->last->end) { |
|
(*r->msg)(MANDOCERR_MEM, r->data, ln, pos, NULL); |
return(ROFF_ERR); |
return(ROFF_ERR); |
|
} |
|
|
|
memcpy(r->last->end, *bufp + sv, sz); |
|
r->last->end[(int)sz] = '\0'; |
|
|
|
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_close(ROFF_ARGS) |
roff_if_sub(ROFF_ARGS) |
{ |
{ |
|
enum rofft t; |
|
enum roffrule rr; |
|
|
|
ppos = pos; |
|
rr = r->last->rule; |
|
roffnode_cleanscope(r); |
|
|
|
if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) |
|
return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT); |
|
|
/* |
/* |
if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) |
* A denied conditional must evaluate its children if and only |
return(ROFF_ERR); |
* if they're either structurally required (such as loops and |
*/ |
* conditionals) or a closing macro. |
return(ROFF_IGN); |
*/ |
|
if (ROFFRULE_DENY == rr) |
|
if ( ! (ROFFMAC_STRUCT & roffs[t].flags)) |
|
if (ROFF_ccond != t) |
|
return(ROFF_IGN); |
|
|
|
assert(roffs[t].proc); |
|
return((*roffs[t].proc) |
|
(r, t, bufp, szp, ln, ppos, pos, offs)); |
} |
} |
|
|
|
|
/* ARGSUSED */ |
/* ARGSUSED */ |
static enum rofferr |
static enum rofferr |
roff_new_ig(ROFF_ARGS) |
roff_block_sub(ROFF_ARGS) |
{ |
{ |
int i; |
enum rofft t; |
|
int i, j; |
|
|
if ( ! roffnode_push(r, ROFF_ig, ln, ppos)) |
/* |
return(ROFF_ERR); |
* First check whether a custom macro exists at this level. If |
|
* it does, then check against it. 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. |
|
*/ |
|
|
i = (int)ppos; |
if (r->last->end) { |
while ((*bufp)[i] && ' ' != (*bufp)[i]) |
i = pos + 1; |
i++; |
while (' ' == (*bufp)[i] || '\t' == (*bufp)[i]) |
|
i++; |
|
|
if (i == (int)ppos) |
for (j = 0; r->last->end[j]; j++, i++) |
return(ROFF_IGN); |
if ((*bufp)[i] != r->last->end[j]) |
if ((*bufp)[i]) |
break; |
if ( ! (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, i, NULL)) |
|
return(ROFF_ERR); |
|
|
|
|
if ('\0' == r->last->end[j] && |
|
('\0' == (*bufp)[i] || |
|
' ' == (*bufp)[i] || |
|
'\t' == (*bufp)[i])) { |
|
roffnode_pop(r); |
|
roffnode_cleanscope(r); |
|
|
|
if (ROFF_MAX != roff_parse(*bufp, &pos)) |
|
return(ROFF_RERUN); |
|
return(ROFF_IGN); |
|
} |
|
} |
|
|
/* |
/* |
* If `.ig' has arguments, the first argument (up to the next |
* If we have no custom end-query or lookup failed, then try |
* whitespace) is interpreted as an argument marking the macro |
* pulling it out of the hashtable. |
* 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); |
ppos = pos; |
if (NULL == r->last->end) { |
t = roff_parse(*bufp, &pos); |
(*r->msg)(MANDOCERR_MEM, r->data, ln, ppos, NULL); |
|
return(ROFF_ERR); |
/* If we're not a comment-end, then throw it away. */ |
|
if (ROFF_cblock != t) |
|
return(ROFF_IGN); |
|
|
|
assert(roffs[t].proc); |
|
return((*roffs[t].proc)(r, t, bufp, |
|
szp, ln, ppos, pos, offs)); |
|
} |
|
|
|
|
|
/* ARGSUSED */ |
|
static enum rofferr |
|
roff_block_text(ROFF_ARGS) |
|
{ |
|
|
|
return(ROFF_IGN); |
|
} |
|
|
|
|
|
/* ARGSUSED */ |
|
static enum rofferr |
|
roff_if_text(ROFF_ARGS) |
|
{ |
|
char *ep, *st; |
|
|
|
st = &(*bufp)[pos]; |
|
if (NULL == (ep = strstr(st, "\\}"))) { |
|
roffnode_cleanscope(r); |
|
return(ROFF_IGN); |
} |
} |
|
|
memcpy(r->last->end, &(*bufp)[ppos], (size_t)i - ppos); |
if (ep > st && '\\' != *(ep - 1)) |
r->last->end[(size_t)i - ppos] = '\0'; |
roffnode_pop(r); |
|
|
|
roffnode_cleanscope(r); |
return(ROFF_IGN); |
return(ROFF_IGN); |
} |
} |
|
|
|
|
int |
/* ARGSUSED */ |
roff_endparse(struct roff *r) |
static enum rofferr |
|
roff_if(ROFF_ARGS) |
{ |
{ |
|
int sv; |
|
|
if (NULL == r->last) |
/* |
return(1); |
* Read ahead past the conditional. |
return((*r->msg)(MANDOCERR_SCOPEEXIT, r->data, |
* FIXME: this does not work, as conditionals don't end on |
r->last->line, r->last->col, NULL)); |
* 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]) { |
|
r->last->endspan = -1; |
|
pos += 2; |
|
} |
|
|
|
/* |
|
* 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); |
} |
} |