=================================================================== RCS file: /cvs/mandoc/preconv.c,v retrieving revision 1.1 retrieving revision 1.12 diff -u -p -r1.1 -r1.12 --- mandoc/preconv.c 2011/05/26 00:30:11 1.1 +++ mandoc/preconv.c 2014/11/14 04:24:04 1.12 @@ -1,6 +1,7 @@ -/* $Id: preconv.c,v 1.1 2011/05/26 00:30:11 kristaps Exp $ */ +/* $Id: preconv.c,v 1.12 2014/11/14 04:24:04 schwarze Exp $ */ /* * Copyright (c) 2011 Kristaps Dzonsons + * Copyright (c) 2014 Ingo Schwarze * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above @@ -14,303 +15,179 @@ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ -#ifdef HAVE_CONFIG_H #include "config.h" -#endif -#include -#include +#include -#include -#include #include -#include #include -#include +#include "mandoc.h" +#include "libmandoc.h" -/* - * The read_whole_file() and resize_buf() functions are copied from - * read.c, including all dependency code (MAP_FILE, etc.). - */ +int +preconv_encode(struct buf *ib, size_t *ii, struct buf *ob, size_t *oi, + int *filenc) +{ + size_t i; + int state; + unsigned int accum; + unsigned char cu; -#ifndef MAP_FILE -#define MAP_FILE 0 -#endif + if ( ! (*filenc & MPARSE_UTF8)) + goto latin; -enum enc { - ENC_UTF_8, /* UTF-8 */ - ENC_US_ASCII, /* US-ASCII */ - ENC_LATIN_1, /* Latin-1 */ - ENC__MAX -}; + state = 0; + accum = 0U; -struct buf { - char *buf; /* binary input buffer */ - size_t sz; /* size of binary buffer */ - size_t offs; /* starting buffer offset */ -}; + for (i = *ii; i < ib->sz; i++) { + cu = ib->buf[i]; + if (state) { + if ( ! (cu & 128) || (cu & 64)) { + /* Bad sequence header. */ + break; + } -struct encode { - const char *name; - int (*conv)(const struct buf *); -}; + /* Accept only legitimate bit patterns. */ -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); + if (cu > 191 || cu < 128) { + /* Bad in-sequence bits. */ + break; + } -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 */ -}; + accum |= (cu & 63) << --state * 6; -static const char *progname; + if (state) + continue; -static void -usage(void) -{ + if (accum < 0x80) + ob->buf[(*oi)++] = accum; + else + *oi += snprintf(ob->buf + *oi, + 11, "\\[u%.4X]", accum); + *ii = 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 + * state from it. + */ + for (state = 0; state < 7; state++) + if ( ! (cu & (1 << (7 - state)))) + break; - fprintf(stderr, "usage: %s " - "[-D enc] " - "[-e ENC] " - "[file]\n", progname); -} + /* Accept only legitimate bit patterns. */ -static int -conv_latin_1(const struct buf *b) -{ - size_t i; - unsigned char c; - 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. - */ - - for (i = b->offs; i < b->sz; i++) { - c = (unsigned char)*cp++; - c < 128 ? putchar(c) : printf("\\[u%.4X]", c); + switch (state--) { + case (4): + if (cu <= 244 && cu >= 240) { + accum = (cu & 7) << 18; + continue; + } + /* Bad 4-sequence start bits. */ + break; + case (3): + if (cu <= 239 && cu >= 224) { + accum = (cu & 15) << 12; + continue; + } + /* Bad 3-sequence start bits. */ + break; + case (2): + if (cu <= 223 && cu >= 194) { + accum = (cu & 31) << 6; + continue; + } + /* Bad 2-sequence start bits. */ + break; + default: + /* Bad sequence bit mask. */ + break; + } + break; + } } - return(1); -} + /* FALLTHROUGH: Invalid or incomplete UTF-8 sequence. */ -static int -conv_us_ascii(const struct buf *b) -{ +latin: + if ( ! (*filenc & MPARSE_LATIN1)) + return(0); - /* - * US-ASCII has no conversion since it falls into the first 128 - * bytes of Unicode. - */ + *oi += snprintf(ob->buf + *oi, 11, + "\\[u%.4X]", (unsigned char)ib->buf[(*ii)++]); - fwrite(b->buf, 1, b->sz, stdout); + *filenc &= ~MPARSE_UTF8; return(1); } -static int -conv_utf_8(const struct buf *b) +int +preconv_cue(const struct buf *b, size_t offset) { + const char *ln, *eoln, *eoph; + size_t sz, phsz; - return(1); -} + ln = b->buf + offset; + sz = b->sz - offset; -static void -resize_buf(struct buf *buf, size_t initial) -{ + /* Look for the end-of-line. */ - buf->sz = buf->sz > initial / 2 ? - 2 * buf->sz : initial; + if (NULL == (eoln = memchr(ln, '\n', sz))) + eoln = ln + sz; - buf->buf = realloc(buf->buf, buf->sz); - if (NULL == buf->buf) { - perror(NULL); - exit(EXIT_FAILURE); - } -} + /* Check if we have the correct header/trailer. */ -static int -read_whole_file(const char *f, int fd, - struct buf *fb, int *with_mmap) -{ - struct stat st; - size_t off; - ssize_t ssz; + if ((sz = (size_t)(eoln - ln)) < 10 || + memcmp(ln, ".\\\" -*-", 7) || memcmp(eoln - 3, "-*-", 3)) + return(MPARSE_UTF8 | MPARSE_LATIN1); - if (-1 == fstat(fd, &st)) { - perror(f); - return(0); - } + /* Move after the header and adjust for the trailer. */ - /* - * 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. - */ + ln += 7; + sz -= 10; - if (S_ISREG(st.st_mode) && st.st_size >= (1U << 31)) { - fprintf(stderr, "%s: input too large\n", f); - 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_FILE|MAP_SHARED, 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 && 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); + while (sz > 0) { + while (sz > 0 && ' ' == *ln) { + ln++; + sz--; } - if (ssz == -1) { - perror(f); + if (0 == sz) break; - } - off += (size_t)ssz; - } - free(fb->buf); - fb->buf = NULL; - return(0); -} + /* Find the end-of-phrase marker (or eoln). */ -int -main(int argc, char *argv[]) -{ - int i, ch, map, fd, rc; - struct buf buf; - const char *fn; - enum enc enc, def; - extern int optind; - extern char *optarg; + if (NULL == (eoph = memchr(ln, ';', sz))) + eoph = eoln - 3; + else + eoph++; - progname = strrchr(argv[0], '/'); - if (progname == NULL) - progname = argv[0]; - else - ++progname; + /* Only account for the "coding" phrase. */ - fn = ""; - fd = STDIN_FILENO; - rc = EXIT_FAILURE; - enc = def = ENC__MAX; - map = 0; - - memset(&buf, 0, sizeof(struct buf)); - - while (-1 != (ch = getopt(argc, argv, "D:e:rdvh"))) - switch (ch) { - case ('D'): - /* FALLTHROUGH */ - case ('e'): - for (i = 0; i < ENC__MAX; i++) { - if (strcasecmp(optarg, encs[i].name)) - continue; - break; - } - if (i < 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); + if ((phsz = eoph - ln) < 7 || + strncasecmp(ln, "coding:", 7)) { + sz -= phsz; + ln += phsz; + continue; } - argc -= optind; - argv += optind; - - /* - * Open and read the first argument on the command-line. - * If we don't have one, we default to stdin. - */ + sz -= 7; + ln += 7; - if (argc > 0) { - fn = *argv; - fd = open(fn, O_RDONLY, 0); - if (-1 == fd) { - perror(fn); - return(EXIT_FAILURE); + while (sz > 0 && ' ' == *ln) { + ln++; + sz--; } - } + if (0 == sz) + return(0); - if ( ! read_whole_file(fn, fd, &buf, &map)) - goto out; + /* Check us against known encodings. */ - if (ENC__MAX == enc) { - /* TODO: search for BOM. */ + 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); } - - /* - * 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)(&buf)) - goto out; - - rc = EXIT_SUCCESS; -out: - if (map) - munmap(buf.buf, buf.sz); - else - free(buf.buf); - - if (fd > STDIN_FILENO) - close(fd); - - return(rc); + return(MPARSE_UTF8 | MPARSE_LATIN1); }