/*
* Main parse routine for a buffer.
* It assumes encoding and line numbering are already set up.
* It can recurse directly (for invocations of user-defined
* macros, inline equations, and input line traps)
* and indirectly (for .so file inclusion).
*/
static void
mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
{
const struct tbl_span *span;
struct buf ln;
const char *save_file;
char *cp;
size_t pos; /* byte number in the ln buffer */
enum rofferr rr;
int of;
int lnn; /* line number in the real file */
int fd;
unsigned char c;
memset(&ln, 0, sizeof(ln));
lnn = curp->line;
pos = 0;
while (i < blk.sz) {
if (0 == pos && '\0' == blk.buf[i])
break;
if (start) {
curp->line = lnn;
curp->reparse_count = 0;
if (lnn < 3 &&
curp->filenc & MPARSE_UTF8 &&
curp->filenc & MPARSE_LATIN1)
curp->filenc = preconv_cue(&blk, i);
}
while (i < blk.sz && (start || blk.buf[i] != '\0')) {
/*
* When finding an unescaped newline character,
* leave the character loop to process the line.
* Skip a preceding carriage return, if any.
*/
if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
'\n' == blk.buf[i + 1])
++i;
if ('\n' == blk.buf[i]) {
++i;
++lnn;
break;
}
/*
* Make sure we have space for the worst
* case of 11 bytes: "\\[u10ffff]\0"
*/
if (pos + 11 > ln.sz)
resize_buf(&ln, 256);
/*
* Encode 8-bit input.
*/
c = blk.buf[i];
if (c & 0x80) {
if ( ! (curp->filenc && preconv_encode(
&blk, &i, &ln, &pos, &curp->filenc))) {
mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
curp->line, pos, "0x%x", c);
ln.buf[pos++] = '?';
i++;
}
continue;
}
/*
* Exclude control characters.
*/
if (c == 0x7f || (c < 0x20 && c != 0x09)) {
mandoc_vmsg(c == 0x00 || c == 0x04 ||
c > 0x0a ? MANDOCERR_CHAR_BAD :
MANDOCERR_CHAR_UNSUPP,
curp, curp->line, pos, "0x%x", c);
i++;
if (c != '\r')
ln.buf[pos++] = '?';
continue;
}
/* Trailing backslash = a plain char. */
if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
ln.buf[pos++] = blk.buf[i++];
continue;
}
/*
* Found escape and at least one other character.
* When it's a newline character, skip it.
* When there is a carriage return in between,
* skip that one as well.
*/
if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
'\n' == blk.buf[i + 2])
++i;
if ('\n' == blk.buf[i + 1]) {
i += 2;
++lnn;
continue;
}
if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
i += 2;
/* Comment, skip to end of line */
for (; i < blk.sz; ++i) {
if ('\n' == blk.buf[i]) {
++i;
++lnn;
break;
}
}
/* Backout trailing whitespaces */
for (; pos > 0; --pos) {
if (ln.buf[pos - 1] != ' ')
break;
if (pos > 2 && ln.buf[pos - 2] == '\\')
break;
}
break;
}
/* Catch escaped bogus characters. */
c = (unsigned char) blk.buf[i+1];
if ( ! (isascii(c) &&
(isgraph(c) || isblank(c)))) {
mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
curp->line, pos, "0x%x", c);
i += 2;
ln.buf[pos++] = '?';
continue;
}
/* Some other escape sequence, copy & cont. */
ln.buf[pos++] = blk.buf[i++];
ln.buf[pos++] = blk.buf[i++];
}
if (pos >= ln.sz)
resize_buf(&ln, 256);
ln.buf[pos] = '\0';
/*
* A significant amount of complexity is contained by
* the roff preprocessor. It's line-oriented but can be
* expressed on one line, so we need at times to
* readjust our starting point and re-run it. The roff
* preprocessor can also readjust the buffers with new
* data, so we pass them in wholesale.
*/
of = 0;
/*
* Maintain a lookaside buffer of all parsed lines. We
* only do this if mparse_keep() has been invoked (the
* buffer may be accessed with mparse_getkeep()).
*/
if (curp->secondary) {
curp->secondary->buf = mandoc_realloc(
curp->secondary->buf,
curp->secondary->sz + pos + 2);
memcpy(curp->secondary->buf +
curp->secondary->sz,
ln.buf, pos);
curp->secondary->sz += pos;
curp->secondary->buf
[curp->secondary->sz] = '\n';
curp->secondary->sz++;
curp->secondary->buf
[curp->secondary->sz] = '\0';
}
rerun:
rr = roff_parseln(curp->roff, curp->line, &ln, &of);
switch (rr) {
case ROFF_REPARSE:
if (REPARSE_LIMIT >= ++curp->reparse_count)
mparse_buf_r(curp, ln, of, 0);
else
mandoc_msg(MANDOCERR_ROFFLOOP, curp,
curp->line, pos, NULL);
pos = 0;
continue;
case ROFF_APPEND:
pos = strlen(ln.buf);
continue;
case ROFF_RERUN:
goto rerun;
case ROFF_IGN:
pos = 0;
continue;
case ROFF_SO:
if ( ! (curp->options & MPARSE_SO) &&
(i >= blk.sz || blk.buf[i] == '\0')) {
curp->sodest = mandoc_strdup(ln.buf + of);
free(ln.buf);
return;
}
/*
* We remove `so' clauses from our lookaside
* buffer because we're going to descend into
* the file recursively.
*/
if (curp->secondary)
curp->secondary->sz -= pos + 1;
save_file = curp->file;
if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
mparse_readfd(curp, fd, ln.buf + of);
close(fd);
curp->file = save_file;
} else {
curp->file = save_file;
mandoc_vmsg(MANDOCERR_SO_FAIL,
curp, curp->line, pos,
".so %s", ln.buf + of);
ln.sz = mandoc_asprintf(&cp,
".sp\nSee the file %s.\n.sp",
ln.buf + of);
free(ln.buf);
ln.buf = cp;
of = 0;
mparse_buf_r(curp, ln, of, 0);
}
pos = 0;
continue;
default:
break;
}
/*
* If input parsers have not been allocated, do so now.
* We keep these instanced between parsers, but set them
* locally per parse routine since we can use different
* parsers with each one.
*/
if (curp->man == NULL ||
curp->man->macroset == MACROSET_NONE)
choose_parser(curp);
/*
* Lastly, push down into the parsers themselves.
* If libroff returns ROFF_TBL, then add it to the
* currently open parse. Since we only get here if
* there does exist data (see tbl_data.c), we're
* guaranteed that something's been allocated.
* Do the same for ROFF_EQN.
*/
if (rr == ROFF_TBL)
while ((span = roff_span(curp->roff)) != NULL)
roff_addtbl(curp->man, span);
else if (rr == ROFF_EQN)
roff_addeqn(curp->man, roff_eqn(curp->roff));
else if ((curp->man->macroset == MACROSET_MDOC ?
mdoc_parseln(curp->man, curp->line, ln.buf, of) :
man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
break;
/* Temporary buffers typically are not full. */
if (0 == start && '\0' == blk.buf[i])
break;
/* Start the next input line. */
pos = 0;
}
free(ln.buf);
}
/*
* Main parse routine for an opened file. This is called for each
* opened file and simply loops around the full input file, possibly
* nesting (i.e., with `so').
*/
static void
mparse_buf_r(struct mparse *curp, struct buf blk, int start)
{
const struct tbl_span *span;
struct buf ln;
enum rofferr rr;
int i, of, rc;
int pos; /* byte number in the ln buffer */
int lnn; /* line number in the real file */
unsigned char c;
memset(&ln, 0, sizeof(struct buf));
lnn = curp->line;
pos = 0;
for (i = 0; i < (int)blk.sz; ) {
if (0 == pos && '\0' == blk.buf[i])
break;
if (start) {
curp->line = lnn;
curp->reparse_count = 0;
}
while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
/*
* When finding an unescaped newline character,
* leave the character loop to process the line.
* Skip a preceding carriage return, if any.
*/
if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
'\n' == blk.buf[i + 1])
++i;
if ('\n' == blk.buf[i]) {
++i;
++lnn;
break;
}
/*
* Warn about bogus characters. If you're using
* non-ASCII encoding, you're screwing your
* readers. Since I'd rather this not happen,
* I'll be helpful and drop these characters so
* we don't display gibberish. Note to manual
* writers: use special characters.
*/
c = (unsigned char) blk.buf[i];
if ( ! (isascii(c) &&
(isgraph(c) || isblank(c)))) {
mandoc_msg(MANDOCERR_BADCHAR, curp,
curp->line, pos, "ignoring byte");
i++;
continue;
}
/* Trailing backslash = a plain char. */
if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
if (pos >= (int)ln.sz)
resize_buf(&ln, 256);
ln.buf[pos++] = blk.buf[i++];
continue;
}
/*
* Found escape and at least one other character.
* When it's a newline character, skip it.
* When there is a carriage return in between,
* skip that one as well.
*/
if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
'\n' == blk.buf[i + 2])
++i;
if ('\n' == blk.buf[i + 1]) {
i += 2;
++lnn;
continue;
}
if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
i += 2;
/* Comment, skip to end of line */
for (; i < (int)blk.sz; ++i) {
if ('\n' == blk.buf[i]) {
++i;
++lnn;
break;
}
}
/* Backout trailing whitespaces */
for (; pos > 0; --pos) {
if (ln.buf[pos - 1] != ' ')
break;
if (pos > 2 && ln.buf[pos - 2] == '\\')
break;
}
break;
}
/* Some other escape sequence, copy & cont. */
if (pos + 1 >= (int)ln.sz)
resize_buf(&ln, 256);
ln.buf[pos++] = blk.buf[i++];
ln.buf[pos++] = blk.buf[i++];
}
if (pos >= (int)ln.sz)
resize_buf(&ln, 256);
ln.buf[pos] = '\0';
/*
* A significant amount of complexity is contained by
* the roff preprocessor. It's line-oriented but can be
* expressed on one line, so we need at times to
* readjust our starting point and re-run it. The roff
* preprocessor can also readjust the buffers with new
* data, so we pass them in wholesale.
*/
of = 0;
/*
* Maintain a lookaside buffer of all parsed lines. We
* only do this if mparse_keep() has been invoked (the
* buffer may be accessed with mparse_getkeep()).
*/
if (curp->secondary) {
curp->secondary->buf =
mandoc_realloc
(curp->secondary->buf,
curp->secondary->sz + pos + 2);
memcpy(curp->secondary->buf +
curp->secondary->sz,
ln.buf, pos);
curp->secondary->sz += pos;
curp->secondary->buf
[curp->secondary->sz] = '\n';
curp->secondary->sz++;
curp->secondary->buf
[curp->secondary->sz] = '\0';
}
rerun:
rr = roff_parseln
(curp->roff, curp->line,
&ln.buf, &ln.sz, of, &of);
switch (rr) {
case (ROFF_REPARSE):
if (REPARSE_LIMIT >= ++curp->reparse_count)
mparse_buf_r(curp, ln, 0);
else
mandoc_msg(MANDOCERR_ROFFLOOP, curp,
curp->line, pos, NULL);
pos = 0;
continue;
case (ROFF_APPEND):
pos = (int)strlen(ln.buf);
continue;
case (ROFF_RERUN):
goto rerun;
case (ROFF_IGN):
pos = 0;
continue;
case (ROFF_ERR):
assert(MANDOCLEVEL_FATAL <= curp->file_status);
break;
case (ROFF_SO):
/*
* We remove `so' clauses from our lookaside
* buffer because we're going to descend into
* the file recursively.
*/
if (curp->secondary)
curp->secondary->sz -= pos + 1;
mparse_readfd_r(curp, -1, ln.buf + of, 1);
if (MANDOCLEVEL_FATAL <= curp->file_status)
break;
pos = 0;
continue;
default:
break;
}
/*
* If we encounter errors in the recursive parse, make
* sure we don't continue parsing.
*/
if (MANDOCLEVEL_FATAL <= curp->file_status)
break;
/*
* If input parsers have not been allocated, do so now.
* We keep these instanced between parsers, but set them
* locally per parse routine since we can use different
* parsers with each one.
*/
if ( ! (curp->man || curp->mdoc))
pset(ln.buf + of, pos - of, curp);
/*
* Lastly, push down into the parsers themselves. One
* of these will have already been set in the pset()
* routine.
* If libroff returns ROFF_TBL, then add it to the
* currently open parse. Since we only get here if
* there does exist data (see tbl_data.c), we're
* guaranteed that something's been allocated.
* Do the same for ROFF_EQN.
*/
rc = -1;
if (ROFF_TBL == rr)
while (NULL != (span = roff_span(curp->roff))) {
rc = curp->man ?
man_addspan(curp->man, span) :
mdoc_addspan(curp->mdoc, span);
if (0 == rc)
break;
}
else if (ROFF_EQN == rr)
rc = curp->mdoc ?
mdoc_addeqn(curp->mdoc,
roff_eqn(curp->roff)) :
man_addeqn(curp->man,
roff_eqn(curp->roff));
else if (curp->man || curp->mdoc)
rc = curp->man ?
man_parseln(curp->man,
curp->line, ln.buf, of) :
mdoc_parseln(curp->mdoc,
curp->line, ln.buf, of);
if (0 == rc) {
assert(MANDOCLEVEL_FATAL <= curp->file_status);
break;
}
/* Temporary buffers typically are not full. */
if (0 == start && '\0' == blk.buf[i])
break;
/* Start the next input line. */
pos = 0;
}
free(ln.buf);
}