Support cancellation in librt.
[glibc.git] / iconv / iconv_prog.c
blob81edc9603671788c7592c3b20763ab6168dcea0e
1 /* Convert text in given files from the specified from-set to the to-set.
2 Copyright (C) 1998,1999,2000,2001,2002,2003 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, write to the Free
18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
19 02111-1307 USA. */
21 #include <argp.h>
22 #include <assert.h>
23 #include <ctype.h>
24 #include <errno.h>
25 #include <error.h>
26 #include <fcntl.h>
27 #include <iconv.h>
28 #include <langinfo.h>
29 #include <locale.h>
30 #include <search.h>
31 #include <stdbool.h>
32 #include <stdio.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include <unistd.h>
36 #include <libintl.h>
37 #ifdef _POSIX_MAPPED_FILES
38 # include <sys/mman.h>
39 #endif
40 #include <charmap.h>
41 #include <gconv_int.h>
42 #include "iconv_prog.h"
43 #include "iconvconfig.h"
45 /* Get libc version number. */
46 #include "../version.h"
48 #define PACKAGE _libc_intl_domainname
51 /* Name and version of program. */
52 static void print_version (FILE *stream, struct argp_state *state);
53 void (*argp_program_version_hook) (FILE *, struct argp_state *) = print_version;
55 #define OPT_VERBOSE 1000
56 #define OPT_LIST 'l'
58 /* Definitions of arguments for argp functions. */
59 static const struct argp_option options[] =
61 { NULL, 0, NULL, 0, N_("Input/Output format specification:") },
62 { "from-code", 'f', "NAME", 0, N_("encoding of original text") },
63 { "to-code", 't', "NAME", 0, N_("encoding for output") },
64 { NULL, 0, NULL, 0, N_("Information:") },
65 { "list", 'l', NULL, 0, N_("list all known coded character sets") },
66 { NULL, 0, NULL, 0, N_("Output control:") },
67 { NULL, 'c', NULL, 0, N_("omit invalid characters from output") },
68 { "output", 'o', "FILE", 0, N_("output file") },
69 { "silent", 's', NULL, 0, N_("suppress warnings") },
70 { "verbose", OPT_VERBOSE, NULL, 0, N_("print progress information") },
71 { NULL, 0, NULL, 0, NULL }
74 /* Short description of program. */
75 static const char doc[] = N_("\
76 Convert encoding of given files from one encoding to another.");
78 /* Strings for arguments in help texts. */
79 static const char args_doc[] = N_("[FILE...]");
81 /* Prototype for option handler. */
82 static error_t parse_opt (int key, char *arg, struct argp_state *state);
84 /* Function to print some extra text in the help message. */
85 static char *more_help (int key, const char *text, void *input);
87 /* Data structure to communicate with argp functions. */
88 static struct argp argp =
90 options, parse_opt, args_doc, doc, NULL, more_help
93 /* Code sets to convert from and to respectively. An empty string as the
94 default causes the 'iconv_open' function to look up the charset of the
95 currently selected locale and use it. */
96 static const char *from_code = "";
97 static const char *to_code = "";
99 /* File to write output to. If NULL write to stdout. */
100 static const char *output_file;
102 /* Nonzero if verbose ouput is wanted. */
103 int verbose;
105 /* Nonzero if list of all coded character sets is wanted. */
106 static int list;
108 /* If nonzero omit invalid character from output. */
109 int omit_invalid;
111 /* Prototypes for the functions doing the actual work. */
112 static int process_block (iconv_t cd, char *addr, size_t len, FILE *output);
113 static int process_fd (iconv_t cd, int fd, FILE *output);
114 static int process_file (iconv_t cd, FILE *input, FILE *output);
115 static void print_known_names (void) internal_function;
119 main (int argc, char *argv[])
121 int status = EXIT_SUCCESS;
122 int remaining;
123 FILE *output;
124 iconv_t cd;
125 const char *orig_to_code;
126 struct charmap_t *from_charmap = NULL;
127 struct charmap_t *to_charmap = NULL;
129 /* Set locale via LC_ALL. */
130 setlocale (LC_ALL, "");
132 /* Set the text message domain. */
133 textdomain (_libc_intl_domainname);
135 /* Parse and process arguments. */
136 argp_parse (&argp, argc, argv, 0, &remaining, NULL);
138 /* List all coded character sets if wanted. */
139 if (list)
141 print_known_names ();
142 exit (EXIT_SUCCESS);
145 /* If we have to ignore errors make sure we use the appropriate name for
146 the to-character-set. */
147 orig_to_code = to_code;
148 if (omit_invalid)
150 const char *errhand = strchrnul (to_code, '/');
151 int nslash = 2;
152 char *newp;
153 char *cp;
155 if (*errhand == '/')
157 --nslash;
158 errhand = strchrnul (errhand, '/');
160 if (*errhand == '/')
162 --nslash;
163 errhand = strchr (errhand, '\0');
167 newp = (char *) alloca (errhand - to_code + nslash + 7 + 1);
168 cp = mempcpy (newp, to_code, errhand - to_code);
169 while (nslash-- > 0)
170 *cp++ = '/';
171 if (cp[-1] != '/')
172 *cp++ = ',';
173 memcpy (cp, "IGNORE", sizeof ("IGNORE"));
175 to_code = newp;
178 /* POSIX 1003.2b introduces a silly thing: the arguments to -t anf -f
179 can be file names of charmaps. In this case iconv will have to read
180 those charmaps and use them to do the conversion. But there are
181 holes in the specification. There is nothing said that if -f is a
182 charmap filename that -t must be, too. And vice versa. There is
183 also no word about the symbolic names used. What if they don't
184 match? */
185 if (strchr (from_code, '/') != NULL)
186 /* The from-name might be a charmap file name. Try reading the
187 file. */
188 from_charmap = charmap_read (from_code, /*0, 1*/1, 0, 0);
190 if (strchr (orig_to_code, '/') != NULL)
191 /* The to-name might be a charmap file name. Try reading the
192 file. */
193 to_charmap = charmap_read (orig_to_code, /*0, 1,*/1,0, 0);
196 /* Determine output file. */
197 if (output_file != NULL && strcmp (output_file, "-") != 0)
199 output = fopen (output_file, "w");
200 if (output == NULL)
201 error (EXIT_FAILURE, errno, _("cannot open output file"));
203 else
204 output = stdout;
206 /* At this point we have to handle two cases. The first one is
207 where a charmap is used for the from- or to-charset, or both. We
208 handle this special since it is very different from the sane way of
209 doing things. The other case allows converting using the iconv()
210 function. */
211 if (from_charmap != NULL || to_charmap != NULL)
212 /* Construct the conversion table and do the conversion. */
213 status = charmap_conversion (from_code, from_charmap, to_code, to_charmap,
214 argc, remaining, argv, output);
215 else
217 /* Let's see whether we have these coded character sets. */
218 cd = iconv_open (to_code, from_code);
219 if (cd == (iconv_t) -1)
221 if (errno == EINVAL)
223 /* Try to be nice with the user and tell her which of the
224 two encoding names is wrong. This is possible because
225 all supported encodings can be converted from/to Unicode,
226 in other words, because the graph of encodings is
227 connected. */
228 bool from_wrong =
229 (iconv_open ("UTF-8", from_code) == (iconv_t) -1
230 && errno == EINVAL);
231 bool to_wrong =
232 (iconv_open (to_code, "UTF-8") == (iconv_t) -1
233 && errno == EINVAL);
234 const char *from_pretty =
235 (from_code[0] ? from_code : nl_langinfo (CODESET));
236 const char *to_pretty =
237 (orig_to_code[0] ? orig_to_code : nl_langinfo (CODESET));
239 if (from_wrong)
241 if (to_wrong)
242 error (EXIT_FAILURE, 0,
243 _("\
244 conversion from `%s' and to `%s' are not supported"),
245 from_pretty, to_pretty);
246 else
247 error (EXIT_FAILURE, 0,
248 _("conversion from `%s' is not supported"),
249 from_pretty);
251 else
253 if (to_wrong)
254 error (EXIT_FAILURE, 0,
255 _("conversion to `%s' is not supported"),
256 to_pretty);
257 else
258 error (EXIT_FAILURE, 0,
259 _("conversion from `%s' to `%s' is not supported"),
260 from_pretty, to_pretty);
263 else
264 error (EXIT_FAILURE, errno,
265 _("failed to start conversion processing"));
268 /* Now process the remaining files. Write them to stdout or the file
269 specified with the `-o' parameter. If we have no file given as
270 the parameter process all from stdin. */
271 if (remaining == argc)
273 if (process_file (cd, stdin, output) != 0)
274 status = EXIT_FAILURE;
276 else
279 #ifdef _POSIX_MAPPED_FILES
280 struct stat st;
281 char *addr;
282 #endif
283 int fd;
285 if (verbose)
286 printf ("%s:\n", argv[remaining]);
287 if (strcmp (argv[remaining], "-") == 0)
288 fd = 0;
289 else
291 fd = open (argv[remaining], O_RDONLY);
293 if (fd == -1)
295 error (0, errno, _("cannot open input file `%s'"),
296 argv[remaining]);
297 status = EXIT_FAILURE;
298 continue;
302 #ifdef _POSIX_MAPPED_FILES
303 /* We have possibilities for reading the input file. First try
304 to mmap() it since this will provide the fastest solution. */
305 if (fstat (fd, &st) == 0
306 && ((addr = mmap (NULL, st.st_size, PROT_READ, MAP_PRIVATE,
307 fd, 0)) != MAP_FAILED))
309 /* Yes, we can use mmap(). The descriptor is not needed
310 anymore. */
311 if (close (fd) != 0)
312 error (EXIT_FAILURE, errno,
313 _("error while closing input `%s'"),
314 argv[remaining]);
316 if (process_block (cd, addr, st.st_size, output) < 0)
318 /* Something went wrong. */
319 status = EXIT_FAILURE;
321 /* We don't need the input data anymore. */
322 munmap ((void *) addr, st.st_size);
324 /* We cannot go on with producing output since it might
325 lead to problem because the last output might leave
326 the output stream in an undefined state. */
327 break;
330 /* We don't need the input data anymore. */
331 munmap ((void *) addr, st.st_size);
333 else
334 #endif /* _POSIX_MAPPED_FILES */
336 /* Read the file in pieces. */
337 if (process_fd (cd, fd, output) != 0)
339 /* Something went wrong. */
340 status = EXIT_FAILURE;
342 /* We don't need the input file anymore. */
343 close (fd);
345 /* We cannot go on with producing output since it might
346 lead to problem because the last output might leave
347 the output stream in an undefined state. */
348 break;
351 /* Now close the file. */
352 close (fd);
355 while (++remaining < argc);
358 /* Close the output file now. */
359 if (fclose (output))
360 error (EXIT_FAILURE, errno, _("error while closing output file"));
362 return status;
366 /* Handle program arguments. */
367 static error_t
368 parse_opt (int key, char *arg, struct argp_state *state)
370 switch (key)
372 case 'f':
373 from_code = arg;
374 break;
375 case 't':
376 to_code = arg;
377 break;
378 case 'o':
379 output_file = arg;
380 break;
381 case 's':
382 /* Nothing, for now at least. We are not giving out any information
383 about missing character or so. */
384 break;
385 case 'c':
386 /* Omit invalid characters from output. */
387 omit_invalid = 1;
388 break;
389 case OPT_VERBOSE:
390 verbose = 1;
391 break;
392 case OPT_LIST:
393 list = 1;
394 break;
395 default:
396 return ARGP_ERR_UNKNOWN;
398 return 0;
402 static char *
403 more_help (int key, const char *text, void *input)
405 switch (key)
407 case ARGP_KEY_HELP_EXTRA:
408 /* We print some extra information. */
409 return strdup (gettext ("\
410 Report bugs using the `glibcbug' script to <bugs@gnu.org>.\n"));
411 default:
412 break;
414 return (char *) text;
418 /* Print the version information. */
419 static void
420 print_version (FILE *stream, struct argp_state *state)
422 fprintf (stream, "iconv (GNU %s) %s\n", PACKAGE, VERSION);
423 fprintf (stream, gettext ("\
424 Copyright (C) %s Free Software Foundation, Inc.\n\
425 This is free software; see the source for copying conditions. There is NO\n\
426 warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n\
427 "), "2003");
428 fprintf (stream, gettext ("Written by %s.\n"), "Ulrich Drepper");
432 static int
433 process_block (iconv_t cd, char *addr, size_t len, FILE *output)
435 #define OUTBUF_SIZE 32768
436 const char *start = addr;
437 char outbuf[OUTBUF_SIZE];
438 char *outptr;
439 size_t outlen;
440 size_t n;
442 while (len > 0)
444 outptr = outbuf;
445 outlen = OUTBUF_SIZE;
446 n = iconv (cd, &addr, &len, &outptr, &outlen);
448 if (outptr != outbuf)
450 /* We have something to write out. */
451 int errno_save = errno;
453 if (fwrite (outbuf, 1, outptr - outbuf, output)
454 < (size_t) (outptr - outbuf)
455 || ferror (output))
457 /* Error occurred while printing the result. */
458 error (0, 0, _("\
459 conversion stopped due to problem in writing the output"));
460 return -1;
463 errno = errno_save;
466 if (n != (size_t) -1)
468 /* All the input test is processed. For state-dependent
469 character sets we have to flush the state now. */
470 outptr = outbuf;
471 outlen = OUTBUF_SIZE;
472 (void) iconv (cd, NULL, NULL, &outptr, &outlen);
474 if (outptr != outbuf)
476 /* We have something to write out. */
477 int errno_save = errno;
479 if (fwrite (outbuf, 1, outptr - outbuf, output)
480 < (size_t) (outptr - outbuf)
481 || ferror (output))
483 /* Error occurred while printing the result. */
484 error (0, 0, _("\
485 conversion stopped due to problem in writing the output"));
486 return -1;
489 errno = errno_save;
492 break;
495 if (errno != E2BIG)
497 /* iconv() ran into a problem. */
498 switch (errno)
500 case EILSEQ:
501 if (! omit_invalid)
502 error (0, 0, _("illegal input sequence at position %ld"),
503 (long int) (addr - start));
504 break;
505 case EINVAL:
506 error (0, 0, _("\
507 incomplete character or shift sequence at end of buffer"));
508 break;
509 case EBADF:
510 error (0, 0, _("internal error (illegal descriptor)"));
511 break;
512 default:
513 error (0, 0, _("unknown iconv() error %d"), errno);
514 break;
517 return -1;
521 return 0;
525 static int
526 process_fd (iconv_t cd, int fd, FILE *output)
528 /* we have a problem with reading from a desriptor since we must not
529 provide the iconv() function an incomplete character or shift
530 sequence at the end of the buffer. Since we have to deal with
531 arbitrary encodings we must read the whole text in a buffer and
532 process it in one step. */
533 static char *inbuf = NULL;
534 static size_t maxlen = 0;
535 char *inptr = NULL;
536 size_t actlen = 0;
538 while (actlen < maxlen)
540 ssize_t n = read (fd, inptr, maxlen - actlen);
542 if (n == 0)
543 /* No more text to read. */
544 break;
546 if (n == -1)
548 /* Error while reading. */
549 error (0, errno, _("error while reading the input"));
550 return -1;
553 inptr += n;
554 actlen += n;
557 if (actlen == maxlen)
558 while (1)
560 ssize_t n;
561 char *new_inbuf;
563 /* Increase the buffer. */
564 new_inbuf = (char *) realloc (inbuf, maxlen + 32768);
565 if (new_inbuf == NULL)
567 error (0, errno, _("unable to allocate buffer for input"));
568 return -1;
570 inbuf = new_inbuf;
571 maxlen += 32768;
572 inptr = inbuf + actlen;
576 n = read (fd, inptr, maxlen - actlen);
578 if (n == 0)
579 /* No more text to read. */
580 break;
582 if (n == -1)
584 /* Error while reading. */
585 error (0, errno, _("error while reading the input"));
586 return -1;
589 inptr += n;
590 actlen += n;
592 while (actlen < maxlen);
594 if (n == 0)
595 /* Break again so we leave both loops. */
596 break;
599 /* Now we have all the input in the buffer. Process it in one run. */
600 return process_block (cd, inbuf, actlen, output);
604 static int
605 process_file (iconv_t cd, FILE *input, FILE *output)
607 /* This should be safe since we use this function only for `stdin' and
608 we haven't read anything so far. */
609 return process_fd (cd, fileno (input), output);
613 /* Print all known character sets/encodings. */
614 static void *printlist;
615 static size_t column;
616 static int not_first;
618 static void
619 insert_print_list (const void *nodep, VISIT value, int level)
621 if (value == leaf || value == postorder)
623 const struct gconv_alias *s = *(const struct gconv_alias **) nodep;
624 tsearch (s->fromname, &printlist, (__compar_fn_t) strverscmp);
628 static void
629 do_print_human (const void *nodep, VISIT value, int level)
631 if (value == leaf || value == postorder)
633 const char *s = *(const char **) nodep;
634 size_t len = strlen (s);
635 size_t cnt;
637 while (len > 0 && s[len - 1] == '/')
638 --len;
640 for (cnt = 0; cnt < len; ++cnt)
641 if (isalnum (s[cnt]))
642 break;
643 if (cnt == len)
644 return;
646 if (not_first)
648 putchar (',');
649 ++column;
651 if (column > 2 && column + len > 77)
653 fputs ("\n ", stdout);
654 column = 2;
656 else
658 putchar (' ');
659 ++column;
662 else
663 not_first = 1;
665 fwrite (s, len, 1, stdout);
666 column += len;
670 static void
671 do_print (const void *nodep, VISIT value, int level)
673 if (value == leaf || value == postorder)
675 const char *s = *(const char **) nodep;
677 puts (s);
681 static void
682 internal_function
683 add_known_names (struct gconv_module *node)
685 if (node->left != NULL)
686 add_known_names (node->left);
687 if (node->right != NULL)
688 add_known_names (node->right);
691 if (strcmp (node->from_string, "INTERNAL"))
692 tsearch (node->from_string, &printlist,
693 (__compar_fn_t) strverscmp);
694 if (strcmp (node->to_string, "INTERNAL") != 0)
695 tsearch (node->to_string, &printlist, (__compar_fn_t) strverscmp);
697 node = node->same;
699 while (node != NULL);
703 static void
704 insert_cache (void)
706 const struct gconvcache_header *header;
707 const char *strtab;
708 const struct hash_entry *hashtab;
709 size_t cnt;
711 header = (const struct gconvcache_header *) __gconv_get_cache ();
712 strtab = (char *) header + header->string_offset;
713 hashtab = (struct hash_entry *) ((char *) header + header->hash_offset);
715 for (cnt = 0; cnt < header->hash_size; ++cnt)
716 if (hashtab[cnt].string_offset != 0)
718 const char *str = strtab + hashtab[cnt].string_offset;
720 if (strcmp (str, "INTERNAL") != 0)
721 tsearch (str, &printlist, (__compar_fn_t) strverscmp);
726 static void
727 internal_function
728 print_known_names (void)
730 iconv_t h;
731 void *cache;
733 /* We must initialize the internal databases first. */
734 h = iconv_open ("L1", "L1");
735 iconv_close (h);
737 /* See whether we have a cache. */
738 cache = __gconv_get_cache ();
739 if (cache != NULL)
740 /* Yep, use only this information. */
741 insert_cache ();
742 else
744 struct gconv_module *modules;
746 /* No, then use the information read from the gconv-modules file.
747 First add the aliases. */
748 twalk (__gconv_get_alias_db (), insert_print_list);
750 /* Add the from- and to-names from the known modules. */
751 modules = __gconv_get_modules_db ();
752 if (modules != NULL)
753 add_known_names (modules);
756 fputs (_("\
757 The following list contain all the coded character sets known. This does\n\
758 not necessarily mean that all combinations of these names can be used for\n\
759 the FROM and TO command line parameters. One coded character set can be\n\
760 listed with several different names (aliases).\n\n "), stdout);
762 /* Now print the collected names. */
763 column = 2;
764 if (isatty (fileno (stdout)))
766 twalk (printlist, do_print_human);
768 if (column != 0)
769 puts ("");
771 else
772 twalk (printlist, do_print);