=================================================================== RCS file: /cvs/mandoc/preconv.c,v retrieving revision 1.2 retrieving revision 1.7 diff -u -p -r1.2 -r1.7 --- mandoc/preconv.c 2011/05/26 12:01:14 1.2 +++ mandoc/preconv.c 2014/08/10 23:54:41 1.7 @@ -1,4 +1,4 @@ -/* $Id: preconv.c,v 1.2 2011/05/26 12:01:14 kristaps Exp $ */ +/* $Id: preconv.c,v 1.7 2014/08/10 23:54:41 schwarze Exp $ */ /* * Copyright (c) 2011 Kristaps Dzonsons * @@ -14,12 +14,13 @@ * 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 +#ifdef HAVE_MMAP #include #include +#endif #include #include @@ -30,13 +31,9 @@ /* * The read_whole_file() and resize_buf() functions are copied from - * read.c, including all dependency code (MAP_FILE, etc.). + * read.c, including all dependency code. */ -#ifndef MAP_FILE -#define MAP_FILE 0 -#endif - enum enc { ENC_UTF_8, /* UTF-8 */ ENC_US_ASCII, /* US-ASCII */ @@ -55,6 +52,7 @@ struct encode { 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 *); @@ -94,11 +92,13 @@ conv_latin_1(const struct buf *b) * 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 < 128U ? putchar(cu) : printf("\\[u%.4X]", cu); + cu < 160U ? putchar(cu) : printf("\\[u%.4X]", cu); } return(1); @@ -134,7 +134,7 @@ conv_utf_8(const struct buf *b) /* Quick test for big-endian value. */ - if ( ! (*((char *)(&one)))) + if ( ! (*((const char *)(&one)))) be = 1; for (i = b->offs; i < b->sz; i++) { @@ -241,10 +241,11 @@ 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; +#ifdef HAVE_MMAP + struct stat st; if (-1 == fstat(fd, &st)) { perror(f); return(0); @@ -265,11 +266,11 @@ read_whole_file(const char *f, int fd, 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); + 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 @@ -306,6 +307,93 @@ read_whole_file(const char *f, int fd, return(0); } +static int +cue_enc(const struct buf *b, size_t *offs, enum enc *enc) +{ + const char *ln, *eoln, *eoph; + size_t sz, phsz, nsz; + int i; + + ln = b->buf + (int)*offs; + sz = b->sz - *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); + + /* Check if we have the correct header/trailer. */ + + if ((sz = (size_t)(eoln - ln)) < 10 || + memcmp(ln, ".\\\" -*-", 7) || + memcmp(eoln - 3, "-*-", 3)) + return(0); + + /* Move after the header and adjust for the trailer. */ + + ln += 7; + sz -= 10; + + while (sz > 0) { + while (sz > 0 && ' ' == *ln) { + ln++; + sz--; + } + if (0 == sz) + break; + + /* Find the end-of-phrase marker (or eoln). */ + + if (NULL == (eoph = memchr(ln, ';', sz))) + eoph = eoln - 3; + else + eoph++; + + /* Only account for the "coding" phrase. */ + + if ((phsz = (size_t)(eoph - ln)) < 7 || + strncasecmp(ln, "coding:", 7)) { + sz -= phsz; + ln += phsz; + continue; + } + + sz -= 7; + ln += 7; + + while (sz > 0 && ' ' == *ln) { + ln++; + sz--; + } + if (0 == sz) + break; + + /* 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[]) { @@ -313,7 +401,8 @@ main(int argc, char *argv[]) struct buf b; const char *fn; enum enc enc, def; - const char bom[3] = { 0xEF, 0xBB, 0xBF }; + unsigned char bom[3] = { 0xEF, 0xBB, 0xBF }; + size_t offs; extern int optind; extern char *optarg; @@ -336,12 +425,12 @@ main(int argc, char *argv[]) case ('D'): /* FALLTHROUGH */ case ('e'): - for (i = 0; i < ENC__MAX; i++) { + for (i = 0; i < (int)ENC__MAX; i++) { if (strcasecmp(optarg, encs[i].name)) continue; break; } - if (i < ENC__MAX) { + if (i < (int)ENC__MAX) { if ('D' == ch) def = (enum enc)i; else @@ -394,6 +483,15 @@ main(int argc, char *argv[]) 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, @@ -403,14 +501,18 @@ main(int argc, char *argv[]) if (ENC__MAX == enc) enc = ENC__MAX == def ? ENC_LATIN_1 : def; - if ( ! (*encs[(int)enc].conv)(&b)) + if ( ! (*encs[(int)enc].conv)(&b)) { + fprintf(stderr, "%s: Bad encoding\n", fn); goto out; + } rc = EXIT_SUCCESS; out: +#ifdef HAVE_MMAP if (map) munmap(b.buf, b.sz); else +#endif free(b.buf); if (fd > STDIN_FILENO)