-/* $Id: main.c,v 1.110 2010/12/01 10:31:34 kristaps Exp $ */
+/* $Id: main.c,v 1.113 2010/12/01 16:54:25 kristaps Exp $ */
/*
* Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
* Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>
struct curparse {
const char *file; /* Current parse. */
int fd; /* Current parse. */
+ int line; /* Line number in the file. */
enum mandoclevel wlevel; /* Ignore messages below this. */
int wstop; /* Stop after a file with a warning. */
enum intt inttype; /* which parser to use */
+ struct man *pman; /* persistent man parser */
+ struct mdoc *pmdoc; /* persistent mdoc parser */
struct man *man; /* man parser */
struct mdoc *mdoc; /* mdoc parser */
struct roff *roff; /* roff parser (!NULL) */
"generic warning",
+ ".so is fragile, better use ln(1)",
"text should be uppercase",
"sections out of conventional order",
"section name repeats",
"argument count wrong, violates syntax",
"child violates parent syntax",
"argument count wrong, violates syntax",
+ "NOT IMPLEMENTED: .so with absolute path or \"..\"",
"no document body",
"no document prologue",
"utsname system call failed",
"static buffer exhausted",
};
+static void parsebuf(struct curparse *, struct buf, int);
+static void pdesc(struct curparse *);
static void fdesc(struct curparse *);
static void ffile(const char *, struct curparse *);
+static int pfile(const char *, struct curparse *);
static int moptions(enum intt *, char *);
static int mmsg(enum mandocerr, void *,
int, int, const char *);
-static void pset(const char *, int, struct curparse *,
- struct man **, struct mdoc **);
+static void pset(const char *, int, struct curparse *);
static int toptions(struct curparse *, char *);
static void usage(void) __attribute__((noreturn));
static void version(void) __attribute__((noreturn));
if (curp.outfree)
(*curp.outfree)(curp.outdata);
- if (curp.mdoc)
- mdoc_free(curp.mdoc);
- if (curp.man)
- man_free(curp.man);
+ if (curp.pmdoc)
+ mdoc_free(curp.pmdoc);
+ if (curp.pman)
+ man_free(curp.pman);
if (curp.roff)
roff_free(curp.roff);
usage(void)
{
- (void)fprintf(stderr, "usage: %s [-V] [-foption] "
- "[-mformat] [-Ooption] [-Toutput] "
- "[-Werr] [file...]\n", progname);
+ (void)fprintf(stderr, "usage: %s "
+ "[-V] "
+ "[-foption] "
+ "[-mformat] "
+ "[-Ooption] "
+ "[-Toutput] "
+ "[-Werr] "
+ "[file...]\n",
+ progname);
+
exit((int)MANDOCLEVEL_BADARG);
}
-
static void
ffile(const char *file, struct curparse *curp)
{
+ /*
+ * Called once per input file. Get the file ready for reading,
+ * pass it through to the parser-driver, then close it out.
+ * XXX: don't do anything special as this is only called for
+ * files; stdin goes directly to fdesc().
+ */
+
curp->file = file;
+
if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
perror(curp->file);
exit_status = MANDOCLEVEL_SYSERR;
perror(curp->file);
}
+static int
+pfile(const char *file, struct curparse *curp)
+{
+ const char *savefile;
+ int fd, savefd;
+
+ if (-1 == (fd = open(file, O_RDONLY, 0))) {
+ perror(file);
+ exit_status = MANDOCLEVEL_SYSERR;
+ return(0);
+ }
+
+ savefile = curp->file;
+ savefd = curp->fd;
+
+ curp->file = file;
+ curp->fd = fd;
+
+ pdesc(curp);
+
+ curp->file = savefile;
+ curp->fd = savefd;
+
+ if (-1 == close(fd))
+ perror(file);
+
+ return(MANDOCLEVEL_FATAL > exit_status ? 1 : 0);
+}
+
static void
resize_buf(struct buf *buf, size_t initial)
static void
fdesc(struct curparse *curp)
{
- struct buf ln, blk;
- int i, pos, lnn, lnn_start, with_mmap, of;
- enum rofferr re;
- unsigned char c;
- struct man *man;
- struct mdoc *mdoc;
- struct roff *roff;
- man = NULL;
- mdoc = NULL;
- roff = NULL;
+ /*
+ * Called once per file with an opened file descriptor. All
+ * pre-file-parse operations (whether stdin or a file) should go
+ * here.
+ *
+ * This calls down into the nested parser, which drills down and
+ * fully parses a file and all its dependences (i.e., `so'). It
+ * then runs the cleanup validators and pushes to output.
+ */
- memset(&ln, 0, sizeof(struct buf));
+ /* Zero the parse type. */
+
+ curp->mdoc = NULL;
+ curp->man = NULL;
+
+ /* Make sure the mandotory roff parser is initialised. */
+
+ if (NULL == curp->roff) {
+ curp->roff = roff_alloc(&curp->regs, curp, mmsg);
+ assert(curp->roff);
+ }
+
+ /* Fully parse the file. */
+
+ pdesc(curp);
+
+ if (MANDOCLEVEL_FATAL <= exit_status)
+ goto cleanup;
+
+ /* NOTE a parser may not have been assigned, yet. */
+
+ if ( ! (curp->man || curp->mdoc)) {
+ fprintf(stderr, "%s: Not a manual\n", curp->file);
+ exit_status = MANDOCLEVEL_FATAL;
+ goto cleanup;
+ }
+
+ /* Clean up the parse routine ASTs. */
+
+ if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
+ assert(MANDOCLEVEL_FATAL <= exit_status);
+ goto cleanup;
+ }
+
+ if (curp->man && ! man_endparse(curp->man)) {
+ assert(MANDOCLEVEL_FATAL <= exit_status);
+ goto cleanup;
+ }
+
+ assert(curp->roff);
+ if ( ! roff_endparse(curp->roff)) {
+ assert(MANDOCLEVEL_FATAL <= exit_status);
+ goto cleanup;
+ }
/*
- * Two buffers: ln and buf. buf is the input file and may be
- * memory mapped. ln is a line buffer and grows on-demand.
+ * With -Wstop and warnings or errors of at least
+ * the requested level, do not produce output.
+ */
+
+ if (MANDOCLEVEL_OK != exit_status && curp->wstop)
+ goto cleanup;
+
+ /* If unset, allocate output dev now (if applicable). */
+
+ if ( ! (curp->outman && curp->outmdoc)) {
+ switch (curp->outtype) {
+ case (OUTT_XHTML):
+ curp->outdata = xhtml_alloc(curp->outopts);
+ break;
+ case (OUTT_HTML):
+ curp->outdata = html_alloc(curp->outopts);
+ break;
+ case (OUTT_ASCII):
+ curp->outdata = ascii_alloc(curp->outopts);
+ curp->outfree = ascii_free;
+ break;
+ case (OUTT_PDF):
+ curp->outdata = pdf_alloc(curp->outopts);
+ curp->outfree = pspdf_free;
+ break;
+ case (OUTT_PS):
+ curp->outdata = ps_alloc(curp->outopts);
+ curp->outfree = pspdf_free;
+ break;
+ default:
+ break;
+ }
+
+ switch (curp->outtype) {
+ case (OUTT_HTML):
+ /* FALLTHROUGH */
+ case (OUTT_XHTML):
+ curp->outman = html_man;
+ curp->outmdoc = html_mdoc;
+ curp->outfree = html_free;
+ break;
+ case (OUTT_TREE):
+ curp->outman = tree_man;
+ curp->outmdoc = tree_mdoc;
+ break;
+ case (OUTT_PDF):
+ /* FALLTHROUGH */
+ case (OUTT_ASCII):
+ /* FALLTHROUGH */
+ case (OUTT_PS):
+ curp->outman = terminal_man;
+ curp->outmdoc = terminal_mdoc;
+ break;
+ default:
+ break;
+ }
+ }
+
+ /* Execute the out device, if it exists. */
+
+ if (curp->man && curp->outman)
+ (*curp->outman)(curp->outdata, curp->man);
+ if (curp->mdoc && curp->outmdoc)
+ (*curp->outmdoc)(curp->outdata, curp->mdoc);
+
+ cleanup:
+
+ memset(&curp->regs, 0, sizeof(struct regset));
+
+ /* Reset the current-parse compilers. */
+
+ if (curp->mdoc)
+ mdoc_reset(curp->mdoc);
+ if (curp->man)
+ man_reset(curp->man);
+
+ assert(curp->roff);
+ roff_reset(curp->roff);
+
+ return;
+}
+
+static void
+pdesc(struct curparse *curp)
+{
+ struct buf blk;
+ int with_mmap;
+
+ /*
+ * Run for each opened file; may be called more than once for
+ * each full parse sequence if the opened file is nested (i.e.,
+ * from `so'). Simply sucks in the whole file and moves into
+ * the parse phase for the file.
*/
if ( ! read_whole_file(curp, &blk, &with_mmap)) {
return;
}
- if (NULL == curp->roff)
- curp->roff = roff_alloc(&curp->regs, curp, mmsg);
- assert(curp->roff);
- roff = curp->roff;
+ /* Line number is per-file. */
- for (i = 0, lnn = 1; i < (int)blk.sz;) {
- pos = 0;
- lnn_start = lnn;
- while (i < (int)blk.sz) {
+ curp->line = 1;
+
+ parsebuf(curp, blk, 1);
+
+ if (with_mmap)
+ munmap(blk.buf, blk.sz);
+ else
+ free(blk.buf);
+}
+
+static void
+parsebuf(struct curparse *curp, struct buf blk, int start)
+{
+ struct buf ln;
+ int i, of, rc;
+ int pos; /* byte number in the ln buffer */
+ int lnn; /* line number in the real file */
+ unsigned char c;
+
+ /*
+ * 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').
+ */
+
+ 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;
+
+ while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
if ('\n' == blk.buf[i]) {
++i;
++lnn;
*/
c = (unsigned char) blk.buf[i];
- if ( ! (isascii(c) && (isgraph(c) || isblank(c)))) {
+
+ if ( ! (isascii(c) &&
+ (isgraph(c) || isblank(c)))) {
mmsg(MANDOCERR_BADCHAR, curp,
- lnn_start, pos, "ignoring byte");
+ curp->line, pos, "ignoring byte");
i++;
continue;
}
- /* Trailing backslash is like a plain character. */
+ /* 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 an escape and at least one other character. */
+
+ /* Found escape & at least one other char. */
+
if ('\n' == blk.buf[i + 1]) {
- /* Escaped newlines are skipped over */
i += 2;
+ /* Escaped newlines are skipped over */
++lnn;
continue;
}
+
if ('"' == blk.buf[i + 1]) {
i += 2;
/* Comment, skip to end of line */
break;
}
}
+
/* Backout trailing whitespaces */
for (; pos > 0; --pos) {
if (ln.buf[pos - 1] != ' ')
}
break;
}
- /* Some other escape sequence, copy and continue. */
+
+ /* Some other escape sequence, copy & cont. */
+
if (pos + 1 >= (int)ln.sz)
resize_buf(&ln, 256);
if (pos >= (int)ln.sz)
resize_buf(&ln, 256);
+
ln.buf[pos] = '\0';
/*
*/
of = 0;
- do {
- re = roff_parseln(roff, lnn_start,
- &ln.buf, &ln.sz, of, &of);
- } while (ROFF_RERUN == re);
- if (ROFF_IGN == re) {
+rerun:
+ rc = roff_parseln
+ (curp->roff, curp->line,
+ &ln.buf, &ln.sz, of, &of);
+
+ switch (rc) {
+ case (ROFF_REPARSE):
+ parsebuf(curp, ln, 0);
+ pos = 0;
+ continue;
+ case (ROFF_APPEND):
+ pos = strlen(ln.buf);
continue;
- } else if (ROFF_ERR == re) {
+ case (ROFF_RERUN):
+ goto rerun;
+ case (ROFF_IGN):
+ pos = 0;
+ continue;
+ case (ROFF_ERR):
assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
+ break;
+ case (ROFF_SO):
+ if (pfile(ln.buf + of, curp)) {
+ pos = 0;
+ continue;
+ } else
+ break;
+ case (ROFF_CONT):
+ break;
}
/*
* parsers with each one.
*/
- if ( ! (man || mdoc))
- pset(ln.buf + of, pos - of, curp, &man, &mdoc);
-
- /* Lastly, push down into the parsers themselves. */
-
- if (man && ! man_parseln(man, lnn_start, ln.buf, of)) {
- assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
- }
- if (mdoc && ! mdoc_parseln(mdoc, lnn_start, ln.buf, of)) {
- assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
- }
- }
-
- /* NOTE a parser may not have been assigned, yet. */
-
- if ( ! (man || mdoc)) {
- fprintf(stderr, "%s: Not a manual\n", curp->file);
- exit_status = MANDOCLEVEL_FATAL;
- goto cleanup;
- }
-
- /* Clean up the parse routine ASTs. */
-
- if (mdoc && ! mdoc_endparse(mdoc)) {
- assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
- }
- if (man && ! man_endparse(man)) {
- assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
- }
- if (roff && ! roff_endparse(roff)) {
- assert(MANDOCLEVEL_FATAL <= exit_status);
- goto cleanup;
- }
-
- /*
- * With -Wstop and warnings or errors of at least
- * the requested level, do not produce output.
- */
+ if ( ! (curp->man || curp->mdoc))
+ pset(ln.buf + of, pos - of, curp);
- if (MANDOCLEVEL_OK != exit_status && curp->wstop)
- goto cleanup;
+ /*
+ * Lastly, push down into the parsers themselves. One
+ * of these will have already been set in the pset()
+ * routine.
+ */
- /* If unset, allocate output dev now (if applicable). */
+ 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 ( ! (curp->outman && curp->outmdoc)) {
- switch (curp->outtype) {
- case (OUTT_XHTML):
- curp->outdata = xhtml_alloc(curp->outopts);
- break;
- case (OUTT_HTML):
- curp->outdata = html_alloc(curp->outopts);
- break;
- case (OUTT_ASCII):
- curp->outdata = ascii_alloc(curp->outopts);
- curp->outfree = ascii_free;
- break;
- case (OUTT_PDF):
- curp->outdata = pdf_alloc(curp->outopts);
- curp->outfree = pspdf_free;
- break;
- case (OUTT_PS):
- curp->outdata = ps_alloc(curp->outopts);
- curp->outfree = pspdf_free;
- break;
- default:
- break;
+ if ( ! rc) {
+ assert(MANDOCLEVEL_FATAL <= exit_status);
+ break;
+ }
}
- switch (curp->outtype) {
- case (OUTT_HTML):
- /* FALLTHROUGH */
- case (OUTT_XHTML):
- curp->outman = html_man;
- curp->outmdoc = html_mdoc;
- curp->outfree = html_free;
- break;
- case (OUTT_TREE):
- curp->outman = tree_man;
- curp->outmdoc = tree_mdoc;
- break;
- case (OUTT_PDF):
- /* FALLTHROUGH */
- case (OUTT_ASCII):
- /* FALLTHROUGH */
- case (OUTT_PS):
- curp->outman = terminal_man;
- curp->outmdoc = terminal_mdoc;
- break;
- default:
- break;
- }
- }
+ /* Temporary buffers typically are not full. */
- /* Execute the out device, if it exists. */
+ if (0 == start && '\0' == blk.buf[i])
+ break;
- if (man && curp->outman)
- (*curp->outman)(curp->outdata, man);
- if (mdoc && curp->outmdoc)
- (*curp->outmdoc)(curp->outdata, mdoc);
+ /* Start the next input line. */
- cleanup:
- memset(&curp->regs, 0, sizeof(struct regset));
- if (mdoc)
- mdoc_reset(mdoc);
- if (man)
- man_reset(man);
- if (roff)
- roff_reset(roff);
- if (ln.buf)
- free(ln.buf);
- if (with_mmap)
- munmap(blk.buf, blk.sz);
- else
- free(blk.buf);
+ pos = 0;
+ }
- return;
+ free(ln.buf);
}
-
static void
-pset(const char *buf, int pos, struct curparse *curp,
- struct man **man, struct mdoc **mdoc)
+pset(const char *buf, int pos, struct curparse *curp)
{
int i;
* explicitly. If passed as -mandoc, then try to guess from the
* line: either skip dot-lines, use -mdoc when finding `.Dt', or
* default to -man, which is more lenient.
+ *
+ * Separate out pmdoc/pman from mdoc/man: the first persists
+ * through all parsers, while the latter is used per-parse.
*/
if ('.' == buf[0] || '\'' == buf[0]) {
switch (curp->inttype) {
case (INTT_MDOC):
- if (NULL == curp->mdoc)
- curp->mdoc = mdoc_alloc(&curp->regs, curp, mmsg);
- assert(curp->mdoc);
- *mdoc = curp->mdoc;
+ if (NULL == curp->pmdoc)
+ curp->pmdoc = mdoc_alloc
+ (&curp->regs, curp, mmsg);
+ assert(curp->pmdoc);
+ curp->mdoc = curp->pmdoc;
return;
case (INTT_MAN):
- if (NULL == curp->man)
- curp->man = man_alloc(&curp->regs, curp, mmsg);
- assert(curp->man);
- *man = curp->man;
+ if (NULL == curp->pman)
+ curp->pman = man_alloc
+ (&curp->regs, curp, mmsg);
+ assert(curp->pman);
+ curp->man = curp->pman;
return;
default:
break;
}
if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
- if (NULL == curp->mdoc)
- curp->mdoc = mdoc_alloc(&curp->regs, curp, mmsg);
- assert(curp->mdoc);
- *mdoc = curp->mdoc;
+ if (NULL == curp->pmdoc)
+ curp->pmdoc = mdoc_alloc
+ (&curp->regs, curp, mmsg);
+ assert(curp->pmdoc);
+ curp->mdoc = curp->pmdoc;
return;
}
- if (NULL == curp->man)
- curp->man = man_alloc(&curp->regs, curp, mmsg);
- assert(curp->man);
- *man = curp->man;
+ if (NULL == curp->pman)
+ curp->pman = man_alloc(&curp->regs, curp, mmsg);
+ assert(curp->pman);
+ curp->man = curp->pman;
}
-
static int
moptions(enum intt *tflags, char *arg)
{
return(1);
}
-
static int
toptions(struct curparse *curp, char *arg)
{
return(1);
}
-
static int
woptions(struct curparse *curp, char *arg)
{
return(1);
}
-
static int
mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg)
{
git.ckatri.com / mandoc.git / blobdiff