(INTERNAL_SYSCALL): Add LOAD_REGS_##nr. (LOAD_ARGS_0, LOAD_ARGS_1, LOAD_ARGS_2, LOAD_...
[glibc.git] / iconv / iconv_prog.c
blob5455974f5cbe502f9309fc664cac038921fcf840
1 /* Convert text in given files from the specified from-set to the to-set.
2 Copyright (C) 1998-2003, 2004 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, ret;
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 ret = process_block (cd, addr, st.st_size, output);
318 /* We don't need the input data anymore. */
319 munmap ((void *) addr, st.st_size);
321 if (ret != 0)
323 status = EXIT_FAILURE;
325 if (ret < 0)
326 /* We cannot go on with producing output since it might
327 lead to problem because the last output might leave
328 the output stream in an undefined state. */
329 break;
332 else
333 #endif /* _POSIX_MAPPED_FILES */
335 /* Read the file in pieces. */
336 ret = process_fd (cd, fd, output);
338 /* Now close the file. */
339 close (fd);
341 if (ret != 0)
343 /* Something went wrong. */
344 status = EXIT_FAILURE;
346 if (ret < 0)
347 /* We cannot go on with producing output since it might
348 lead to problem because the last output might leave
349 the output stream in an undefined state. */
350 break;
354 while (++remaining < argc);
357 /* Close the output file now. */
358 if (fclose (output))
359 error (EXIT_FAILURE, errno, _("error while closing output file"));
361 return status;
365 /* Handle program arguments. */
366 static error_t
367 parse_opt (int key, char *arg, struct argp_state *state)
369 switch (key)
371 case 'f':
372 from_code = arg;
373 break;
374 case 't':
375 to_code = arg;
376 break;
377 case 'o':
378 output_file = arg;
379 break;
380 case 's':
381 /* Nothing, for now at least. We are not giving out any information
382 about missing character or so. */
383 break;
384 case 'c':
385 /* Omit invalid characters from output. */
386 omit_invalid = 1;
387 break;
388 case OPT_VERBOSE:
389 verbose = 1;
390 break;
391 case OPT_LIST:
392 list = 1;
393 break;
394 default:
395 return ARGP_ERR_UNKNOWN;
397 return 0;
401 static char *
402 more_help (int key, const char *text, void *input)
404 switch (key)
406 case ARGP_KEY_HELP_EXTRA:
407 /* We print some extra information. */
408 return strdup (gettext ("\
409 Report bugs using the `glibcbug' script to <bugs@gnu.org>.\n"));
410 default:
411 break;
413 return (char *) text;
417 /* Print the version information. */
418 static void
419 print_version (FILE *stream, struct argp_state *state)
421 fprintf (stream, "iconv (GNU %s) %s\n", PACKAGE, VERSION);
422 fprintf (stream, gettext ("\
423 Copyright (C) %s Free Software Foundation, Inc.\n\
424 This is free software; see the source for copying conditions. There is NO\n\
425 warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n\
426 "), "2004");
427 fprintf (stream, gettext ("Written by %s.\n"), "Ulrich Drepper");
431 static int
432 process_block (iconv_t cd, char *addr, size_t len, FILE *output)
434 #define OUTBUF_SIZE 32768
435 const char *start = addr;
436 char outbuf[OUTBUF_SIZE];
437 char *outptr;
438 size_t outlen;
439 size_t n;
440 int ret = 0;
442 while (len > 0)
444 outptr = outbuf;
445 outlen = OUTBUF_SIZE;
446 n = iconv (cd, &addr, &len, &outptr, &outlen);
448 if (n == (size_t) -1 && omit_invalid && errno == EILSEQ)
450 ret = 1;
451 if (len == 0)
452 n = 0;
453 else
454 errno = E2BIG;
457 if (outptr != outbuf)
459 /* We have something to write out. */
460 int errno_save = errno;
462 if (fwrite (outbuf, 1, outptr - outbuf, output)
463 < (size_t) (outptr - outbuf)
464 || ferror (output))
466 /* Error occurred while printing the result. */
467 error (0, 0, _("\
468 conversion stopped due to problem in writing the output"));
469 return -1;
472 errno = errno_save;
475 if (n != (size_t) -1)
477 /* All the input test is processed. For state-dependent
478 character sets we have to flush the state now. */
479 outptr = outbuf;
480 outlen = OUTBUF_SIZE;
481 n = iconv (cd, NULL, NULL, &outptr, &outlen);
483 if (outptr != outbuf)
485 /* We have something to write out. */
486 int errno_save = errno;
488 if (fwrite (outbuf, 1, outptr - outbuf, output)
489 < (size_t) (outptr - outbuf)
490 || ferror (output))
492 /* Error occurred while printing the result. */
493 error (0, 0, _("\
494 conversion stopped due to problem in writing the output"));
495 return -1;
498 errno = errno_save;
501 if (n != (size_t) -1)
502 break;
504 if (omit_invalid && errno == EILSEQ)
506 ret = 1;
507 break;
511 if (errno != E2BIG)
513 /* iconv() ran into a problem. */
514 switch (errno)
516 case EILSEQ:
517 if (! omit_invalid)
518 error (0, 0, _("illegal input sequence at position %ld"),
519 (long int) (addr - start));
520 break;
521 case EINVAL:
522 error (0, 0, _("\
523 incomplete character or shift sequence at end of buffer"));
524 break;
525 case EBADF:
526 error (0, 0, _("internal error (illegal descriptor)"));
527 break;
528 default:
529 error (0, 0, _("unknown iconv() error %d"), errno);
530 break;
533 return -1;
537 return ret;
541 static int
542 process_fd (iconv_t cd, int fd, FILE *output)
544 /* we have a problem with reading from a desriptor since we must not
545 provide the iconv() function an incomplete character or shift
546 sequence at the end of the buffer. Since we have to deal with
547 arbitrary encodings we must read the whole text in a buffer and
548 process it in one step. */
549 static char *inbuf = NULL;
550 static size_t maxlen = 0;
551 char *inptr = NULL;
552 size_t actlen = 0;
554 while (actlen < maxlen)
556 ssize_t n = read (fd, inptr, maxlen - actlen);
558 if (n == 0)
559 /* No more text to read. */
560 break;
562 if (n == -1)
564 /* Error while reading. */
565 error (0, errno, _("error while reading the input"));
566 return -1;
569 inptr += n;
570 actlen += n;
573 if (actlen == maxlen)
574 while (1)
576 ssize_t n;
577 char *new_inbuf;
579 /* Increase the buffer. */
580 new_inbuf = (char *) realloc (inbuf, maxlen + 32768);
581 if (new_inbuf == NULL)
583 error (0, errno, _("unable to allocate buffer for input"));
584 return -1;
586 inbuf = new_inbuf;
587 maxlen += 32768;
588 inptr = inbuf + actlen;
592 n = read (fd, inptr, maxlen - actlen);
594 if (n == 0)
595 /* No more text to read. */
596 break;
598 if (n == -1)
600 /* Error while reading. */
601 error (0, errno, _("error while reading the input"));
602 return -1;
605 inptr += n;
606 actlen += n;
608 while (actlen < maxlen);
610 if (n == 0)
611 /* Break again so we leave both loops. */
612 break;
615 /* Now we have all the input in the buffer. Process it in one run. */
616 return process_block (cd, inbuf, actlen, output);
620 static int
621 process_file (iconv_t cd, FILE *input, FILE *output)
623 /* This should be safe since we use this function only for `stdin' and
624 we haven't read anything so far. */
625 return process_fd (cd, fileno (input), output);
629 /* Print all known character sets/encodings. */
630 static void *printlist;
631 static size_t column;
632 static int not_first;
634 static void
635 insert_print_list (const void *nodep, VISIT value, int level)
637 if (value == leaf || value == postorder)
639 const struct gconv_alias *s = *(const struct gconv_alias **) nodep;
640 tsearch (s->fromname, &printlist, (__compar_fn_t) strverscmp);
644 static void
645 do_print_human (const void *nodep, VISIT value, int level)
647 if (value == leaf || value == postorder)
649 const char *s = *(const char **) nodep;
650 size_t len = strlen (s);
651 size_t cnt;
653 while (len > 0 && s[len - 1] == '/')
654 --len;
656 for (cnt = 0; cnt < len; ++cnt)
657 if (isalnum (s[cnt]))
658 break;
659 if (cnt == len)
660 return;
662 if (not_first)
664 putchar (',');
665 ++column;
667 if (column > 2 && column + len > 77)
669 fputs ("\n ", stdout);
670 column = 2;
672 else
674 putchar (' ');
675 ++column;
678 else
679 not_first = 1;
681 fwrite (s, len, 1, stdout);
682 column += len;
686 static void
687 do_print (const void *nodep, VISIT value, int level)
689 if (value == leaf || value == postorder)
691 const char *s = *(const char **) nodep;
693 puts (s);
697 static void
698 internal_function
699 add_known_names (struct gconv_module *node)
701 if (node->left != NULL)
702 add_known_names (node->left);
703 if (node->right != NULL)
704 add_known_names (node->right);
707 if (strcmp (node->from_string, "INTERNAL"))
708 tsearch (node->from_string, &printlist,
709 (__compar_fn_t) strverscmp);
710 if (strcmp (node->to_string, "INTERNAL") != 0)
711 tsearch (node->to_string, &printlist, (__compar_fn_t) strverscmp);
713 node = node->same;
715 while (node != NULL);
719 static void
720 insert_cache (void)
722 const struct gconvcache_header *header;
723 const char *strtab;
724 const struct hash_entry *hashtab;
725 size_t cnt;
727 header = (const struct gconvcache_header *) __gconv_get_cache ();
728 strtab = (char *) header + header->string_offset;
729 hashtab = (struct hash_entry *) ((char *) header + header->hash_offset);
731 for (cnt = 0; cnt < header->hash_size; ++cnt)
732 if (hashtab[cnt].string_offset != 0)
734 const char *str = strtab + hashtab[cnt].string_offset;
736 if (strcmp (str, "INTERNAL") != 0)
737 tsearch (str, &printlist, (__compar_fn_t) strverscmp);
742 static void
743 internal_function
744 print_known_names (void)
746 iconv_t h;
747 void *cache;
749 /* We must initialize the internal databases first. */
750 h = iconv_open ("L1", "L1");
751 iconv_close (h);
753 /* See whether we have a cache. */
754 cache = __gconv_get_cache ();
755 if (cache != NULL)
756 /* Yep, use only this information. */
757 insert_cache ();
758 else
760 struct gconv_module *modules;
762 /* No, then use the information read from the gconv-modules file.
763 First add the aliases. */
764 twalk (__gconv_get_alias_db (), insert_print_list);
766 /* Add the from- and to-names from the known modules. */
767 modules = __gconv_get_modules_db ();
768 if (modules != NULL)
769 add_known_names (modules);
772 fputs (_("\
773 The following list contain all the coded character sets known. This does\n\
774 not necessarily mean that all combinations of these names can be used for\n\
775 the FROM and TO command line parameters. One coded character set can be\n\
776 listed with several different names (aliases).\n\n "), stdout);
778 /* Now print the collected names. */
779 column = 2;
780 if (isatty (fileno (stdout)))
782 twalk (printlist, do_print_human);
784 if (column != 0)
785 puts ("");
787 else
788 twalk (printlist, do_print);