summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorIngo Schwarze <schwarze@cvs.openbsd.org>2010-05-15 17:32:26 +0000
committerIngo Schwarze <schwarze@cvs.openbsd.org>2010-05-15 17:32:26 +0000
commitc2c62a88f6f9d56daa2265e1476dbf15d196f99c (patch)
tree89b797a223b6f03bb47b9c12e5505b0ee1238587
parentcf7f7beef09cac77f6c9235e61abbbacd7f4b06b (diff)
Bugfxing and cleanup by Joerg Sonnenberger, edited by kristaps@:
* Flag errors and warnings in global variables. * Exit with error code if either is not null on exit. * Push down the buffers to the functions using them. * Handle the stdin/file argument in one function. * Use mmap when possible or else read the whole file into a single buffer. * Push buffer resizing into its own function.
-rw-r--r--usr.bin/mandoc/main.c365
1 files changed, 209 insertions, 156 deletions
diff --git a/usr.bin/mandoc/main.c b/usr.bin/mandoc/main.c
index 1d8a79c58bd..071bbcb7cd9 100644
--- a/usr.bin/mandoc/main.c
+++ b/usr.bin/mandoc/main.c
@@ -1,4 +1,4 @@
-/* $Id: main.c,v 1.26 2010/05/15 15:58:56 schwarze Exp $ */
+/* $Id: main.c,v 1.27 2010/05/15 17:32:25 schwarze Exp $ */
/*
* Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
*
@@ -14,6 +14,8 @@
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
+#include <sys/types.h>
+#include <sys/mman.h>
#include <sys/stat.h>
#include <assert.h>
@@ -64,11 +66,11 @@ struct curparse {
#define FL_NIGN_ESCAPE (1 << 1) /* Don't ignore bad escapes. */
#define FL_NIGN_MACRO (1 << 2) /* Don't ignore bad macros. */
#define FL_IGN_ERRORS (1 << 4) /* Ignore failed parse. */
+#define FL_STRICT FL_NIGN_ESCAPE | \
+ FL_NIGN_MACRO
enum intt inttype; /* Input parsers... */
struct man *man;
- struct man *lastman;
struct mdoc *mdoc;
- struct mdoc *lastmdoc;
enum outt outtype; /* Output devices... */
out_mdoc outmdoc;
out_man outman;
@@ -77,34 +79,29 @@ struct curparse {
char outopts[BUFSIZ];
};
-#define FL_STRICT FL_NIGN_ESCAPE | \
- FL_NIGN_MACRO
-
+static void fdesc(struct curparse *);
+static void ffile(const char *, struct curparse *);
static int foptions(int *, char *);
-static int toptions(struct curparse *, char *);
-static int moptions(enum intt *, char *);
-static int woptions(int *, char *);
+static struct man *man_init(struct curparse *);
+static struct mdoc *mdoc_init(struct curparse *);
static int merr(void *, int, int, const char *);
+static int moptions(enum intt *, char *);
static int mwarn(void *, int, int, const char *);
-static int ffile(struct buf *, struct buf *,
- const char *, struct curparse *);
-static int fdesc(struct buf *, struct buf *,
- struct curparse *);
static int pset(const char *, int, struct curparse *,
struct man **, struct mdoc **);
-static struct man *man_init(struct curparse *);
-static struct mdoc *mdoc_init(struct curparse *);
-static void version(void) __attribute__((noreturn));
+static int toptions(struct curparse *, char *);
static void usage(void) __attribute__((noreturn));
+static void version(void) __attribute__((noreturn));
+static int woptions(int *, char *);
static const char *progname;
-
+static int with_error;
+static int with_warning;
int
main(int argc, char *argv[])
{
- int c, rc;
- struct buf ln, blk;
+ int c;
struct curparse curp;
progname = strrchr(argv[0], '/');
@@ -152,52 +149,26 @@ main(int argc, char *argv[])
argc -= optind;
argv += optind;
- memset(&ln, 0, sizeof(struct buf));
- memset(&blk, 0, sizeof(struct buf));
-
- rc = 1;
-
if (NULL == *argv) {
curp.file = "<stdin>";
curp.fd = STDIN_FILENO;
- c = fdesc(&blk, &ln, &curp);
- if ( ! (FL_IGN_ERRORS & curp.fflags))
- rc = 1 == c ? 1 : 0;
- else
- rc = -1 == c ? 0 : 1;
+ fdesc(&curp);
}
- while (rc && *argv) {
- c = ffile(&blk, &ln, *argv, &curp);
- if ( ! (FL_IGN_ERRORS & curp.fflags))
- rc = 1 == c ? 1 : 0;
- else
- rc = -1 == c ? 0 : 1;
-
- argv++;
- if (*argv && rc) {
- if (curp.lastman)
- man_reset(curp.lastman);
- if (curp.lastmdoc)
- mdoc_reset(curp.lastmdoc);
- curp.lastman = NULL;
- curp.lastmdoc = NULL;
- }
+ while (*argv) {
+ ffile(*argv, &curp);
+
+ if (with_error && !(curp.fflags & FL_IGN_ERRORS))
+ break;
+ ++argv;
}
- if (blk.buf)
- free(blk.buf);
- if (ln.buf)
- free(ln.buf);
if (curp.outfree)
(*curp.outfree)(curp.outdata);
- if (curp.mdoc)
- mdoc_free(curp.mdoc);
- if (curp.man)
- man_free(curp.man);
- return(rc ? EXIT_SUCCESS : EXIT_FAILURE);
+ return((with_warning || with_error) ?
+ EXIT_FAILURE : EXIT_SUCCESS);
}
@@ -267,40 +238,129 @@ mdoc_init(struct curparse *curp)
}
-static int
-ffile(struct buf *blk, struct buf *ln,
- const char *file, struct curparse *curp)
+static void
+ffile(const char *file, struct curparse *curp)
{
- int c;
curp->file = file;
if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
perror(curp->file);
- return(-1);
+ with_error = 1;
+ return;
}
- c = fdesc(blk, ln, curp);
+ fdesc(curp);
if (-1 == close(curp->fd))
perror(curp->file);
+}
+
+
+static int
+resize_buf(struct buf *buf, size_t initial)
+{
+ void *tmp;
+ size_t sz;
- return(c);
+ if (buf->sz == 0)
+ sz = initial;
+ else
+ sz = 2 * buf->sz;
+ tmp = realloc(buf->buf, sz);
+ if (NULL == tmp) {
+ perror(NULL);
+ return(0);
+ }
+ buf->buf = tmp;
+ buf->sz = sz;
+ return(1);
}
static int
-fdesc(struct buf *blk, struct buf *ln, struct curparse *curp)
+read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
{
- size_t sz;
- ssize_t ssz;
struct stat st;
- int j, i, pos, lnn, comment;
+ size_t off;
+ ssize_t ssz;
+
+ if (-1 == fstat(curp->fd, &st)) {
+ perror(curp->file);
+ with_error = 1;
+ return(0);
+ }
+
+ /*
+ * If we're a regular file, try just reading in the whole entry
+ * via mmap(). This is faster than reading it into blocks, and
+ * since each file is only a few bytes to begin with, I'm not
+ * concerned that this is going to tank any machines.
+ */
+
+ if (S_ISREG(st.st_mode)) {
+ if (st.st_size >= (1U << 31)) {
+ fprintf(stderr, "%s: input too large\n",
+ curp->file);
+ with_error = 1;
+ return(0);
+ }
+ *with_mmap = 1;
+ fb->sz = st.st_size;
+ fb->buf = mmap(NULL, fb->sz, PROT_READ,
+ MAP_FILE, curp->fd, 0);
+ if (fb->buf != MAP_FAILED)
+ return(1);
+ }
+
+ /*
+ * If this isn't a regular file (like, say, stdin), then we must
+ * go the old way and just read things in bit by bit.
+ */
+
+ *with_mmap = 0;
+ off = 0;
+ fb->sz = 0;
+ fb->buf = NULL;
+ for (;;) {
+ if (off == fb->sz) {
+ if (fb->sz == (1U << 31)) {
+ fprintf(stderr, "%s: input too large\n",
+ curp->file);
+ break;
+ }
+ if (! resize_buf(fb, 65536))
+ break;
+ }
+ ssz = read(curp->fd, fb->buf + off, fb->sz - off);
+ if (ssz == 0) {
+ fb->sz = off;
+ return(1);
+ }
+ if (ssz == -1) {
+ perror(curp->file);
+ break;
+ }
+ off += ssz;
+ }
+
+ free(fb->buf);
+ fb->buf = NULL;
+ with_error = 1;
+ return(0);
+}
+
+
+static void
+fdesc(struct curparse *curp)
+{
+ struct buf ln, blk;
+ int j, i, pos, lnn, comment, with_mmap;
struct man *man;
struct mdoc *mdoc;
- sz = BUFSIZ;
man = NULL;
mdoc = NULL;
+ memset(&ln, 0, sizeof(struct buf));
/*
* Two buffers: ln and buf. buf is the input buffer optimised
@@ -308,114 +368,87 @@ fdesc(struct buf *blk, struct buf *ln, struct curparse *curp)
* growable, hence passed in by ptr-ptr.
*/
- if (-1 == fstat(curp->fd, &st))
- perror(curp->file);
- else if ((size_t)st.st_blksize > sz)
- sz = st.st_blksize;
-
- if (sz > blk->sz) {
- blk->buf = realloc(blk->buf, sz);
- if (NULL == blk->buf) {
- perror(NULL);
- exit(EXIT_FAILURE);
- }
- blk->sz = sz;
- }
+ if (!read_whole_file(curp, &blk, &with_mmap))
+ return;
/* Fill buf with file blocksize. */
- for (lnn = pos = comment = 0; ; ) {
- if (-1 == (ssz = read(curp->fd, blk->buf, sz))) {
- perror(curp->file);
- return(-1);
- } else if (0 == ssz)
- break;
-
- /* Parse the read block into partial or full lines. */
-
- for (i = 0; i < (int)ssz; i++) {
- if (pos >= (int)ln->sz) {
- ln->sz += 256; /* Step-size. */
- ln->buf = realloc(ln->buf, ln->sz);
- if (NULL == ln->buf) {
- perror(NULL);
- return(EXIT_FAILURE);
- }
- }
+ for (i = lnn = pos = comment = 0; i < (int)blk.sz; ++i) {
+ if (pos >= (int)ln.sz) {
+ if (! resize_buf(&ln, 256))
+ goto bailout;
+ }
- if ('\n' != blk->buf[i]) {
- if (comment)
- continue;
- ln->buf[pos++] = blk->buf[i];
+ if ('\n' != blk.buf[i]) {
+ if (comment)
+ continue;
+ ln.buf[pos++] = blk.buf[i];
- /* Handle in-line `\"' comments. */
+ /* Handle in-line `\"' comments. */
- if (1 == pos || '\"' != ln->buf[pos - 1])
- continue;
+ if (1 == pos || '\"' != ln.buf[pos - 1])
+ continue;
- for (j = pos - 2; j >= 0; j--)
- if ('\\' != ln->buf[j])
- break;
+ for (j = pos - 2; j >= 0; j--)
+ if ('\\' != ln.buf[j])
+ break;
- if ( ! ((pos - 2 - j) % 2))
- continue;
+ if ( ! ((pos - 2 - j) % 2))
+ continue;
- comment = 1;
- pos -= 2;
- for (; pos > 0; --pos) {
- if (ln->buf[pos - 1] != ' ')
- break;
- if (pos > 2 && ln->buf[pos - 2] == '\\')
- break;
- }
+ comment = 1;
+ pos -= 2;
+ for (; pos > 0; --pos) {
+ if (ln.buf[pos - 1] != ' ')
+ break;
+ if (pos > 2 && ln.buf[pos - 2] == '\\')
+ break;
+ }
+ continue;
+ }
+
+ /* Handle escaped `\\n' newlines. */
+
+ if (pos > 0 && 0 == comment && '\\' == ln.buf[pos - 1]) {
+ for (j = pos - 1; j >= 0; j--)
+ if ('\\' != ln.buf[j])
+ break;
+ if ( ! ((pos - j) % 2)) {
+ pos--;
+ lnn++;
continue;
- }
-
- /* Handle escaped `\\n' newlines. */
-
- if (pos > 0 && 0 == comment &&
- '\\' == ln->buf[pos - 1]) {
- for (j = pos - 1; j >= 0; j--)
- if ('\\' != ln->buf[j])
- break;
- if ( ! ((pos - j) % 2)) {
- pos--;
- lnn++;
- continue;
- }
}
+ }
- ln->buf[pos] = 0;
- lnn++;
+ ln.buf[pos] = 0;
+ lnn++;
- /* If unset, assign parser in pset(). */
+ /* If unset, assign parser in pset(). */
- if ( ! (man || mdoc) && ! pset(ln->buf,
- pos, curp, &man, &mdoc))
- return(-1);
+ if ( ! (man || mdoc) && ! pset(ln.buf, pos, curp, &man, &mdoc))
+ goto bailout;
- pos = comment = 0;
+ pos = comment = 0;
- /* Pass down into parsers. */
+ /* Pass down into parsers. */
- if (man && ! man_parseln(man, lnn, ln->buf))
- return(0);
- if (mdoc && ! mdoc_parseln(mdoc, lnn, ln->buf))
- return(0);
- }
+ if (man && ! man_parseln(man, lnn, ln.buf))
+ goto bailout;
+ if (mdoc && ! mdoc_parseln(mdoc, lnn, ln.buf))
+ goto bailout;
}
/* NOTE a parser may not have been assigned, yet. */
if ( ! (man || mdoc)) {
fprintf(stderr, "%s: Not a manual\n", curp->file);
- return(0);
+ goto bailout;
}
if (mdoc && ! mdoc_endparse(mdoc))
- return(0);
+ goto bailout;
if (man && ! man_endparse(man))
- return(0);
+ goto bailout;
/* If unset, allocate output dev now (if applicable). */
@@ -455,7 +488,26 @@ fdesc(struct buf *blk, struct buf *ln, struct curparse *curp)
if (mdoc && curp->outmdoc)
(*curp->outmdoc)(curp->outdata, mdoc);
- return(1);
+ cleanup:
+ if (curp->mdoc) {
+ mdoc_free(curp->mdoc);
+ curp->mdoc = NULL;
+ }
+ if (curp->man) {
+ man_free(curp->man);
+ curp->man = NULL;
+ }
+ if (ln.buf)
+ free(ln.buf);
+ if (with_mmap)
+ munmap(blk.buf, blk.sz);
+ else
+ free(blk.buf);
+ return;
+
+ bailout:
+ with_error = 1;
+ goto cleanup;
}
@@ -487,14 +539,12 @@ pset(const char *buf, int pos, struct curparse *curp,
curp->mdoc = mdoc_init(curp);
if (NULL == (*mdoc = curp->mdoc))
return(0);
- curp->lastmdoc = *mdoc;
return(1);
case (INTT_MAN):
if (NULL == curp->man)
curp->man = man_init(curp);
if (NULL == (*man = curp->man))
return(0);
- curp->lastman = *man;
return(1);
default:
break;
@@ -505,7 +555,6 @@ pset(const char *buf, int pos, struct curparse *curp,
curp->mdoc = mdoc_init(curp);
if (NULL == (*mdoc = curp->mdoc))
return(0);
- curp->lastmdoc = *mdoc;
return(1);
}
@@ -513,7 +562,6 @@ pset(const char *buf, int pos, struct curparse *curp,
curp->man = man_init(curp);
if (NULL == (*man = curp->man))
return(0);
- curp->lastman = *man;
return(1);
}
@@ -648,6 +696,8 @@ merr(void *arg, int line, int col, const char *msg)
(void)fprintf(stderr, "%s:%d:%d: error: %s\n",
curp->file, line, col + 1, msg);
+ with_error = 1;
+
return(0);
}
@@ -665,9 +715,12 @@ mwarn(void *arg, int line, int col, const char *msg)
(void)fprintf(stderr, "%s:%d:%d: warning: %s\n",
curp->file, line, col + 1, msg);
- if ( ! (curp->wflags & WARN_WERR))
- return(1);
-
- return(0);
+ with_warning = 1;
+ if (curp->wflags & WARN_WERR) {
+ with_error = 1;
+ return(0);
+ }
+
+ return(1);
}