integrate preconv(1) into mandoc(1);

enhances functionality and reduces code and docs by more than 300 lines

1 files changed, 59 insertions, 368 deletions

diff --git a/preconv.c b/preconv.c
index 1df18b89..ad2bf5e2 100644
--- a/preconv.c
+++ b/preconv.c
@@ -1,6 +1,7 @@
-/*	$Id: preconv.c,v 1.8 2014/08/16 19:00:01 schwarze Exp $ */
+/*	$Id: preconv.c,v 1.9 2014/10/25 01:03:52 schwarze Exp $ */
 /*
  * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2014 Ingo Schwarze <schwarze@openbsd.org>
  *
  * Permission to use, copy, modify, and distribute this software for any
  * purpose with or without fee is hereby granted, provided that the above
@@ -17,117 +18,24 @@
 #include "config.h"
 
 #include <sys/types.h>
-#if HAVE_MMAP
-#include <sys/stat.h>
-#include <sys/mman.h>
-#endif
 
-#include <assert.h>
-#include <fcntl.h>
 #include <stdio.h>
-#include <stdlib.h>
 #include <string.h>
-#include <unistd.h>
+#include "mandoc.h"
+#include "libmandoc.h"
 
-/* 
- * The read_whole_file() and resize_buf() functions are copied from
- * read.c, including all dependency code.
- */
-
-enum	enc {
-	ENC_UTF_8, /* UTF-8 */
-	ENC_US_ASCII, /* US-ASCII */
-	ENC_LATIN_1, /* Latin-1 */
-	ENC__MAX
-};
-
-struct	buf {
-	char		 *buf; /* binary input buffer */
-	size_t	 	  sz; /* size of binary buffer */
-	size_t		  offs; /* starting buffer offset */
-};
-
-struct	encode {
-	const char	 *name;
-	int		(*conv)(const struct buf *);
-};
-
-static	int	 cue_enc(const struct buf *, size_t *, enum enc *);
-static	int	 conv_latin_1(const struct buf *);
-static	int	 conv_us_ascii(const struct buf *);
-static	int	 conv_utf_8(const struct buf *);
-static	int	 read_whole_file(const char *, int, 
-			struct buf *, int *);
-static	void	 resize_buf(struct buf *, size_t);
-static	void	 usage(void);
-
-static	const struct encode encs[ENC__MAX] = {
-	{ "utf-8", conv_utf_8 }, /* ENC_UTF_8 */
-	{ "us-ascii", conv_us_ascii }, /* ENC_US_ASCII */
-	{ "latin-1", conv_latin_1 }, /* ENC_LATIN_1 */
-};
-
-static	const char	 *progname;
-
-static void
-usage(void)
-{
-
-	fprintf(stderr, "usage: %s "
-			"[-D enc] "
-			"[-e ENC] "
-			"[file]\n", progname);
-}
-
-static int
-conv_latin_1(const struct buf *b)
-{
-	size_t		 i;
-	unsigned char	 cu;
-	const char	*cp;
-
-	cp = b->buf + (int)b->offs;
-
-	/*
-	 * Latin-1 falls into the first 256 code-points of Unicode, so
-	 * there's no need for any sort of translation.  Just make the
-	 * 8-bit characters use the Unicode escape.
-	 * Note that binary values 128 < v < 160 are passed through
-	 * unmodified to mandoc.
-	 */
-
-	for (i = b->offs; i < b->sz; i++) {
-		cu = (unsigned char)*cp++;
-		cu < 160U ? putchar(cu) : printf("\\[u%.4X]", cu);
-	}
-
-	return(1);
-}
-
-static int
-conv_us_ascii(const struct buf *b)
-{
-
-	/*
-	 * US-ASCII has no conversion since it falls into the first 128
-	 * bytes of Unicode.
-	 */
-
-	fwrite(b->buf, 1, b->sz, stdout);
-	return(1);
-}
-
-static int
-conv_utf_8(const struct buf *b)
+int
+preconv_encode(struct buf *ib, struct buf *ob, int *filenc)
 {
 	int		 state, be;
 	unsigned int	 accum;
 	size_t		 i;
 	unsigned char	 cu;
-	const char	*cp;
 	const long	 one = 1L;
 
-	cp = b->buf + (int)b->offs;
+	if ( ! (*filenc & MPARSE_UTF8))
+		goto latin;
+
 	state = 0;
 	accum = 0U;
 	be = 0;
@@ -137,23 +45,26 @@ conv_utf_8(const struct buf *b)
 	if ( ! (*((const char *)(&one))))
 		be = 1;
 
-	for (i = b->offs; i < b->sz; i++) {
-		cu = (unsigned char)*cp++;
+	for (i = ib->offs; i < ib->sz; i++) {
+		cu = ib->buf[i];
 		if (state) {
 			if ( ! (cu & 128) || (cu & 64)) {
 				/* Bad sequence header. */
-				return(0);
+				break;
 			}
 
 			/* Accept only legitimate bit patterns. */
 
 			if (cu > 191 || cu < 128) {
 				/* Bad in-sequence bits. */
-				return(0);
+				break;
 			}
 
 			accum |= (cu & 63) << --state * 6;
 
+			if (state)
+				continue;
+
 			/*
 			 * Accum is held in little-endian order as
 			 * stipulated by the UTF-8 sequence coding.  We
@@ -161,18 +72,21 @@ conv_utf_8(const struct buf *b)
 			 * architecture requires it.
 			 */
 
-			if (0 == state && be) 
+			if (be)
 				accum = (accum >> 24) | 
 					((accum << 8) & 0x00FF0000) |
 					((accum >> 8) & 0x0000FF00) |
 					(accum << 24);
 
-			if (0 == state) {
-				accum < 128U ? putchar(accum) : 
-					printf("\\[u%.4X]", accum);
-				accum = 0U;
-			}
-		} else if (cu & (1 << 7)) {
+			if (accum < 0x80)
+				ob->buf[ob->offs++] = accum;
+			else
+				ob->offs += snprintf(ob->buf + ob->offs,
+				    11, "\\[u%.4X]", accum);
+			ib->offs = i + 1;
+			*filenc &= ~MPARSE_LATIN1;
+			return(1);
+		} else {
 			/*
 			 * Entering a UTF-8 state:  if we encounter a
 			 * UTF-8 bitmask, calculate the expected UTF-8
@@ -184,154 +98,69 @@ conv_utf_8(const struct buf *b)
 
 			/* Accept only legitimate bit patterns. */
 
-			switch (state) {
+			switch (state--) {
 			case (4):
 				if (cu <= 244 && cu >= 240) {
 					accum = (cu & 7) << 18;
-					break;
+					continue;
 				}
 				/* Bad 4-sequence start bits. */
-				return(0);
+				break;
 			case (3):
 				if (cu <= 239 && cu >= 224) {
 					accum = (cu & 15) << 12;
-					break;
+					continue;
 				}
 				/* Bad 3-sequence start bits. */
-				return(0);
+				break;
 			case (2):
 				if (cu <= 223 && cu >= 194) {
 					accum = (cu & 31) << 6;
-					break;
+					continue;
 				}
 				/* Bad 2-sequence start bits. */
-				return(0);
+				break;
 			default:
 				/* Bad sequence bit mask. */
-				return(0);
+				break;
 			}
-			state--;
-		} else
-			putchar(cu);
-	}
-
-	if (0 != state) {
-		/* Bad trailing bits. */
-		return(0);
-	}
-
-	return(1);
-}
-
-static void
-resize_buf(struct buf *buf, size_t initial)
-{
-
-	buf->sz = buf->sz > initial / 2 ? 
-		2 * buf->sz : initial;
-
-	buf->buf = realloc(buf->buf, buf->sz);
-	if (NULL == buf->buf) {
-		perror(NULL);
-		exit(EXIT_FAILURE);
-	}
-}
-
-static int
-read_whole_file(const char *f, int fd, 
-		struct buf *fb, int *with_mmap)
-{
-	size_t		 off;
-	ssize_t		 ssz;
-
-#if HAVE_MMAP
-	struct stat	 st;
-	if (-1 == fstat(fd, &st)) {
-		perror(f);
-		return(0);
+			break;
+		}
 	}
 
-	/*
-	 * 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.
-	 */
+	/* FALLTHROUGH: Invalid or incomplete UTF-8 sequence. */
 
-	if (S_ISREG(st.st_mode) && st.st_size >= (1U << 31)) {
-		fprintf(stderr, "%s: input too large\n", f);
+latin:
+	if ( ! (*filenc & MPARSE_LATIN1))
 		return(0);
-	} 
-	
-	if (S_ISREG(st.st_mode)) {
-		*with_mmap = 1;
-		fb->sz = (size_t)st.st_size;
-		fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
-		if (fb->buf != MAP_FAILED)
-			return(1);
-	}
-#endif
-
-	/*
-	 * 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 && fb->sz == (1U << 31)) {
-			fprintf(stderr, "%s: input too large\n", f);
-			break;
-		} 
-		
-		if (off == fb->sz)
-			resize_buf(fb, 65536);
 
-		ssz = read(fd, fb->buf + (int)off, fb->sz - off);
-		if (ssz == 0) {
-			fb->sz = off;
-			return(1);
-		}
-		if (ssz == -1) {
-			perror(f);
-			break;
-		}
-		off += (size_t)ssz;
-	}
+	ob->offs += snprintf(ob->buf + ob->offs, 11,
+	    "\\[u%.4X]", (unsigned char)ib->buf[ib->offs++]);
 
-	free(fb->buf);
-	fb->buf = NULL;
-	return(0);
+	*filenc &= ~MPARSE_UTF8;
+	return(1);
 }
 
-static int
-cue_enc(const struct buf *b, size_t *offs, enum enc *enc)
+int
+preconv_cue(const struct buf *b)
 {
 	const char	*ln, *eoln, *eoph;
-	size_t		 sz, phsz, nsz;
-	int		 i;
+	size_t		 sz, phsz;
 
-	ln = b->buf + (int)*offs;
-	sz = b->sz - *offs;
+	ln = b->buf + b->offs;
+	sz = b->sz - b->offs;
 
 	/* Look for the end-of-line. */
 
 	if (NULL == (eoln = memchr(ln, '\n', sz)))
-		return(-1);
-
-	/* Set next-line marker. */
-
-	*offs = (size_t)((eoln + 1) - b->buf);
+		eoln = ln + sz;
 
 	/* Check if we have the correct header/trailer. */
 
 	if ((sz = (size_t)(eoln - ln)) < 10 || 
 			memcmp(ln, ".\\\" -*-", 7) ||
 			memcmp(eoln - 3, "-*-", 3))
-		return(0);
+		return(MPARSE_UTF8 | MPARSE_LATIN1);
 
 	/* Move after the header and adjust for the trailer. */
 
@@ -355,8 +184,8 @@ cue_enc(const struct buf *b, size_t *offs, enum enc *enc)
 
 		/* Only account for the "coding" phrase. */
 
-		if ((phsz = (size_t)(eoph - ln)) < 7 ||
-				strncasecmp(ln, "coding:", 7)) {
+		if ((phsz = eoph - ln) < 7 ||
+		    strncasecmp(ln, "coding:", 7)) {
 			sz -= phsz;
 			ln += phsz;
 			continue;
@@ -370,153 +199,15 @@ cue_enc(const struct buf *b, size_t *offs, enum enc *enc)
 			sz--;
 		}
 		if (0 == sz)
-			break;
+			return(0);
 
 		/* Check us against known encodings. */
 
-		for (i = 0; i < (int)ENC__MAX; i++) {
-			nsz = strlen(encs[i].name);
-			if (phsz < nsz)
-				continue;
-			if (strncasecmp(ln, encs[i].name, nsz))
-				continue;
-
-			*enc = (enum enc)i;
-			return(1);
-		}
-
-		/* Unknown encoding. */
-
-		*enc = ENC__MAX;
-		return(1);
-	}
-
-	return(0);
-}
-
-int
-main(int argc, char *argv[])
-{
-	int	 	 i, ch, map, fd, rc;
-	struct buf	 b;
-	const char	*fn;
-	enum enc	 enc, def;
-	unsigned char 	 bom[3] = { 0xEF, 0xBB, 0xBF };
-	size_t		 offs;
-	extern int	 optind;
-	extern char	*optarg;
-
-	progname = strrchr(argv[0], '/');
-	if (progname == NULL)
-		progname = argv[0];
-	else
-		++progname;
-
-	fn = "<stdin>";
-	fd = STDIN_FILENO;
-	rc = EXIT_FAILURE;
-	enc = def = ENC__MAX;
-	map = 0;
-
-	memset(&b, 0, sizeof(struct buf));
-
-	while (-1 != (ch = getopt(argc, argv, "D:e:rdvh")))
-		switch (ch) {
-		case ('D'):
-			/* FALLTHROUGH */
-		case ('e'):
-			for (i = 0; i < (int)ENC__MAX; i++) {
-				if (strcasecmp(optarg, encs[i].name))
-					continue;
-				break;
-			}
-			if (i < (int)ENC__MAX) {
-				if ('D' == ch)
-					def = (enum enc)i;
-				else
-					enc = (enum enc)i;
-				break;
-			}
-
-			fprintf(stderr, "%s: Bad encoding\n", optarg);
-			return(EXIT_FAILURE);
-		case ('r'):
-			/* FALLTHROUGH */
-		case ('d'):
-			/* FALLTHROUGH */
-		case ('v'):
-			/* Compatibility with GNU preconv. */
-			break;
-		case ('h'):
-			/* Compatibility with GNU preconv. */
-			/* FALLTHROUGH */
-		default:
-			usage();
-			return(EXIT_FAILURE);
-		}
-
-	argc -= optind;
-	argv += optind;
-	
-	/* 
-	 * Open and read the first argument on the command-line.
-	 * If we don't have one, we default to stdin.
-	 */
-
-	if (argc > 0) {
-		fn = *argv;
-		fd = open(fn, O_RDONLY, 0);
-		if (-1 == fd) {
-			perror(fn);
-			return(EXIT_FAILURE);
-		}
-	}
-
-	if ( ! read_whole_file(fn, fd, &b, &map))
-		goto out;
-
-	/* Try to read the UTF-8 BOM. */
-
-	if (ENC__MAX == enc)
-		if (b.sz > 3 && 0 == memcmp(b.buf, bom, 3)) {
-			b.offs = 3;
-			enc = ENC_UTF_8;
-		}
-
-	/* Try reading from the "-*-" cue. */
-
-	if (ENC__MAX == enc) {
-		offs = b.offs;
-		ch = cue_enc(&b, &offs, &enc);
-		if (0 == ch)
-			ch = cue_enc(&b, &offs, &enc);
-	}
-
-	/*
-	 * No encoding has been detected.
-	 * Thus, we either fall into our default encoder, if specified,
-	 * or use Latin-1 if all else fails.
-	 */
-
-	if (ENC__MAX == enc) 
-		enc = ENC__MAX == def ? ENC_LATIN_1 : def;
-
-	if ( ! (*encs[(int)enc].conv)(&b)) {
-		fprintf(stderr, "%s: Bad encoding\n", fn);
-		goto out;
+		if (phsz > 4 && !strncasecmp(ln, "utf-8", 5))
+			return(MPARSE_UTF8);
+		if (phsz > 10 && !strncasecmp(ln, "iso-latin-1", 11))
+			return(MPARSE_LATIN1);
+		return(0);
 	}
-
-	rc = EXIT_SUCCESS;
-out:
-#if HAVE_MMAP
-	if (map)
-		munmap(b.buf, b.sz);
-	else 
-#endif
-		free(b.buf);
-
-	if (fd > STDIN_FILENO)
-		close(fd);
-
-	return(rc);
+	return(MPARSE_UTF8 | MPARSE_LATIN1);
 }