diff options
author | John Mark Bell <jmb@netsurf-browser.org> | 2008-07-31 14:43:32 +0000 |
---|---|---|
committer | John Mark Bell <jmb@netsurf-browser.org> | 2008-07-31 14:43:32 +0000 |
commit | 6d3950b141b1cdb1f29e929c8c8156562c7ccbc7 (patch) | |
tree | d2bb92960c4c3b54f083f1307241d5fdbac7fdfd /src/charset/codec_iconv.c | |
parent | 2f7a0e956c151c900789a02d8f850bfe79b54071 (diff) | |
download | libhubbub-6d3950b141b1cdb1f29e929c8c8156562c7ccbc7.tar.gz libhubbub-6d3950b141b1cdb1f29e929c8c8156562c7ccbc7.tar.bz2 |
Merged revisions 4631-4838 via svnmerge from
svn://source.netsurf-browser.org/branches/takkaria/hubbub-parserutils
........
r4631 | takkaria | 2008-07-13 12:54:30 +0100 (Sun, 13 Jul 2008) | 2 lines
Initial hatchet job moving to libparserutils (search and replace and a bit of cleaning up). This doesn't compile.
........
r4632 | takkaria | 2008-07-13 15:28:52 +0100 (Sun, 13 Jul 2008) | 2 lines
libparserutilize everything up to the "before attribute name" state. (Not compiling)
........
r4633 | takkaria | 2008-07-13 15:32:14 +0100 (Sun, 13 Jul 2008) | 2 lines
Replace all uses of "current_{comment|chars}" with just "chars".
........
r4634 | takkaria | 2008-07-13 16:12:06 +0100 (Sun, 13 Jul 2008) | 2 lines
Fix lots of compile errors, lpuise "before attribute name" state.
........
r4636 | takkaria | 2008-07-13 17:23:17 +0100 (Sun, 13 Jul 2008) | 2 lines
Finish lpuising the tag states, apart from character references.
........
r4637 | takkaria | 2008-07-13 19:58:52 +0100 (Sun, 13 Jul 2008) | 2 lines
lpuise the comment states.
........
r4638 | takkaria | 2008-07-13 20:04:31 +0100 (Sun, 13 Jul 2008) | 2 lines
Switch to setting hubbub_string::len to 0 instead of hubbub_string::ptr to NULL to indicate an empty buffer, as it was previously.
........
r4639 | takkaria | 2008-07-13 21:02:11 +0100 (Sun, 13 Jul 2008) | 2 lines
"lpu up" about half of the DOCTYPE handling stages.
........
r4640 | takkaria | 2008-07-13 21:23:00 +0100 (Sun, 13 Jul 2008) | 2 lines
Finish off LPUing the doctype modes.
........
r4641 | takkaria | 2008-07-13 21:37:33 +0100 (Sun, 13 Jul 2008) | 2 lines
The tokeniser uses lpu apart from the entity matcher, now.
........
r4643 | takkaria | 2008-07-14 01:20:36 +0100 (Mon, 14 Jul 2008) | 2 lines
Fix up the character reference matching stuff--still not properly dealt with, but compiles futher.
........
r4644 | takkaria | 2008-07-14 01:24:49 +0100 (Mon, 14 Jul 2008) | 2 lines
Get the tokeniser compiling in its LPU'd form.
........
r4645 | takkaria | 2008-07-14 01:26:34 +0100 (Mon, 14 Jul 2008) | 2 lines
Remember to advance the stream position after emitting tokens.
........
r4646 | takkaria | 2008-07-14 01:34:36 +0100 (Mon, 14 Jul 2008) | 2 lines
Nuke the src/input directory and start work on the treebuilder.
........
r4647 | takkaria | 2008-07-14 01:56:27 +0100 (Mon, 14 Jul 2008) | 2 lines
Get hubbub building in its LPU'd form.
........
r4648 | takkaria | 2008-07-14 02:41:03 +0100 (Mon, 14 Jul 2008) | 2 lines
Get the tokeniser2 testrunner working.
........
r4649 | takkaria | 2008-07-14 02:48:55 +0100 (Mon, 14 Jul 2008) | 2 lines
Fix test LDFLAGS so things link properly.
........
r4650 | takkaria | 2008-07-14 16:25:51 +0100 (Mon, 14 Jul 2008) | 2 lines
Get testcases compiling, remove ones now covered by libparserutils.
........
r4651 | takkaria | 2008-07-14 16:37:09 +0100 (Mon, 14 Jul 2008) | 2 lines
Remove more tests covered by libpu.
........
r4652 | takkaria | 2008-07-14 17:53:18 +0100 (Mon, 14 Jul 2008) | 2 lines
Fix up the tokeniser a bit.
........
r4653 | takkaria | 2008-07-14 19:02:15 +0100 (Mon, 14 Jul 2008) | 3 lines
- Remove the buffer_handler stuff from hubbub
- Add the basics of a buffer for attribute values and text.
........
r4654 | takkaria | 2008-07-14 20:00:45 +0100 (Mon, 14 Jul 2008) | 2 lines
Get character references working in attribute values, start trying to make them work in character tokens.
........
r4656 | takkaria | 2008-07-14 23:28:52 +0100 (Mon, 14 Jul 2008) | 2 lines
Get entities working a bit better.
........
r4657 | takkaria | 2008-07-14 23:37:16 +0100 (Mon, 14 Jul 2008) | 2 lines
Get entities working properly. (!)
........
r4658 | takkaria | 2008-07-14 23:56:10 +0100 (Mon, 14 Jul 2008) | 2 lines
Make doctypes work a bit better.
........
r4659 | takkaria | 2008-07-15 00:18:49 +0100 (Tue, 15 Jul 2008) | 2 lines
Get DOCTYPEs working.
........
r4660 | takkaria | 2008-07-15 00:26:36 +0100 (Tue, 15 Jul 2008) | 2 lines
Fix CDATA sections.
........
r4661 | takkaria | 2008-07-15 01:01:16 +0100 (Tue, 15 Jul 2008) | 2 lines
Get comments working again.
........
r4662 | takkaria | 2008-07-15 01:14:19 +0100 (Tue, 15 Jul 2008) | 2 lines
Fix EOF in "after attribute name" state.
........
r4664 | takkaria | 2008-07-15 01:30:27 +0100 (Tue, 15 Jul 2008) | 2 lines
Put the tests in better order, remove one now superceded with libpu.
........
r4665 | takkaria | 2008-07-15 01:46:29 +0100 (Tue, 15 Jul 2008) | 2 lines
Remove a lot of now-redunant clearings of the current stream offset.
........
r4667 | jmb | 2008-07-15 11:56:54 +0100 (Tue, 15 Jul 2008) | 2 lines
Completely purge charset stuff from hubbub. Parserutils handles this now.
........
r4677 | takkaria | 2008-07-15 21:03:42 +0100 (Tue, 15 Jul 2008) | 2 lines
Get more tests passing, handle NUL bytes in data state.
........
r4694 | takkaria | 2008-07-18 17:55:44 +0100 (Fri, 18 Jul 2008) | 3 lines
- Handle CRs correctly in some token states.
- Handle NULs correctly in the CDATA state.
........
r4706 | takkaria | 2008-07-19 14:58:48 +0100 (Sat, 19 Jul 2008) | 2 lines
Improve the tokeniser2 output a bit.
........
r4721 | takkaria | 2008-07-21 20:57:29 +0100 (Mon, 21 Jul 2008) | 2 lines
Get a better framework in place to allow switching to using a buffer mid-collect. This fails a couple of testcases and doesn't implement proper CR or NUL support yet.
........
r4725 | takkaria | 2008-07-23 17:20:07 +0100 (Wed, 23 Jul 2008) | 2 lines
Make comment tokens in tokeniser2 display both expected and actual output.
........
r4726 | takkaria | 2008-07-23 19:10:23 +0100 (Wed, 23 Jul 2008) | 4 lines
- Add FINISH() macro which stops using buffered character collection.
- Make the encoding U+FFFD in UTF-8 a global varabile, for sanity
- Make the bogus comment state deal with NULs correctly.
........
r4730 | takkaria | 2008-07-24 00:35:16 +0100 (Thu, 24 Jul 2008) | 2 lines
Try to get NUL bytes handled as the spec says.
........
r4731 | takkaria | 2008-07-24 00:40:59 +0100 (Thu, 24 Jul 2008) | 2 lines
Get CRs working in the data state.
........
r4732 | takkaria | 2008-07-24 00:47:45 +0100 (Thu, 24 Jul 2008) | 2 lines
Set force-quirks correctly when failing to match PUBLIC or SYSTEM in DOCTYPEs.
........
r4773 | takkaria | 2008-07-28 15:34:41 +0100 (Mon, 28 Jul 2008) | 2 lines
Fix up the tokeniser, finally.
........
r4801 | takkaria | 2008-07-29 15:59:31 +0100 (Tue, 29 Jul 2008) | 2 lines
Refactor macros a bit.
........
r4802 | takkaria | 2008-07-29 16:04:17 +0100 (Tue, 29 Jul 2008) | 2 lines
Do s/HUBBUB_TOKENISER_STATE_/STATE_/, for shorter line lengths.
........
r4805 | takkaria | 2008-07-29 16:58:37 +0100 (Tue, 29 Jul 2008) | 4 lines
Start cleaning up the hubbub tokeniser;
- refactor to use new inline emit_character_token() and emit_current_tag() functions; makes code clearer
- check EOF before using the CHAR() macro, so eventually it can be removed.
........
r4806 | takkaria | 2008-07-29 17:45:36 +0100 (Tue, 29 Jul 2008) | 2 lines
More cleanup like the previous commit.
........
r4807 | takkaria | 2008-07-29 19:48:44 +0100 (Tue, 29 Jul 2008) | 2 lines
Rewrite comment-handling code to be just the one function, whilst updating it to handle CRs and NULs properly. (All comments now always use the buffer.)
........
r4820 | takkaria | 2008-07-30 14:14:49 +0100 (Wed, 30 Jul 2008) | 2 lines
Finish off the first sweep of cleaning up and refactoring the tokeniser.
........
r4821 | takkaria | 2008-07-30 15:12:22 +0100 (Wed, 30 Jul 2008) | 2 lines
Add copyright statement.
........
r4822 | takkaria | 2008-07-30 17:23:01 +0100 (Wed, 30 Jul 2008) | 2 lines
Apply changes made to tokeniser2 to tokeniser3.
........
r4829 | takkaria | 2008-07-31 01:59:07 +0100 (Thu, 31 Jul 2008) | 4 lines
- Make the tokeniser save everything into the buffer, at least for now.
- Fix logic errors introduced in refactoring
- Avoid emitting more tokens than we have to (e.g. instead of emitting "<>" and switching back to the data state, just switch back to the data state and let it take care of it)
........
r4830 | takkaria | 2008-07-31 02:03:08 +0100 (Thu, 31 Jul 2008) | 2 lines
Small treebuilder <isindex> fix.
........
r4831 | takkaria | 2008-07-31 02:32:29 +0100 (Thu, 31 Jul 2008) | 2 lines
Stop holding on to pointers to character data across treebuilder calls.
........
r4832 | takkaria | 2008-07-31 02:45:09 +0100 (Thu, 31 Jul 2008) | 18 lines
Merge revisions 4620-4831 from trunk hubbub to libinputstream hubbub, modulo one change to test/Makefile which makes the linker choke when linking tests.
------------------------------------------------------------------------
r4666 | jmb | 2008-07-15 11:52:13 +0100 (Tue, 15 Jul 2008) | 3 lines
Make tree2 perform reference counting.
Fix bits of the treebuilder to perform reference counting correctly in the face of *result not pointing to the same object as the node passed in to the treebuilder client callbacks.
------------------------------------------------------------------------
r4668 | jmb | 2008-07-15 12:37:30 +0100 (Tue, 15 Jul 2008) | 2 lines
Fully document treebuilder callbacks.
------------------------------------------------------------------------
r4675 | takkaria | 2008-07-15 21:01:03 +0100 (Tue, 15 Jul 2008) | 2 lines
Fix memory leak in tokeniser2.
------------------------------------------------------------------------
........
r4834 | jmb | 2008-07-31 09:57:51 +0100 (Thu, 31 Jul 2008) | 2 lines
Fix infinite loop in charset detector
........
r4835 | jmb | 2008-07-31 13:01:24 +0100 (Thu, 31 Jul 2008) | 2 lines
Actually store namespaces on formatting list. Otherwise we read uninitialised memory. Add some semblance of filling allocations with junk to myrealloc().
........
r4836 | jmb | 2008-07-31 13:06:07 +0100 (Thu, 31 Jul 2008) | 2 lines
Lose debug again
........
r4837 | jmb | 2008-07-31 15:09:19 +0100 (Thu, 31 Jul 2008) | 2 lines
Lose obsolete testdata (this is now part of lpu)
........
svn path=/trunk/hubbub/; revision=4839
Diffstat (limited to 'src/charset/codec_iconv.c')
-rw-r--r-- | src/charset/codec_iconv.c | 837 |
1 files changed, 0 insertions, 837 deletions
diff --git a/src/charset/codec_iconv.c b/src/charset/codec_iconv.c deleted file mode 100644 index 097e82a..0000000 --- a/src/charset/codec_iconv.c +++ /dev/null @@ -1,837 +0,0 @@ -/* - * This file is part of Hubbub. - * Licensed under the MIT License, - * http://www.opensource.org/licenses/mit-license.php - * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org> - */ - -/* This codec is hideously slow. Only use it as a last resort */ - -#include <errno.h> -#include <stdlib.h> -#include <string.h> - -#include <iconv.h> - -/* These two are for htonl / ntohl */ -#include <arpa/inet.h> -#include <netinet/in.h> - -#include "charset/aliases.h" -#include "utils/utils.h" - -#include "codec_impl.h" - -/** - * A note on endianness: - * - * UCS4 is big-endian by default. Therefore, this codec reads and writes - * big-endian values. This is fine, and causes no problems. However, to - * make life easier for client-supplied filter code, character values passed - * to a filter and those read back from a filter are in host-endian. - * Therefore, we need to convert from big-endian to host-endian when passing - * characters to a filter and perform the reverse translation when reading - * characters back. - */ - -/** - * Iconv-based charset codec - */ -typedef struct hubbub_iconv_codec { - hubbub_charsetcodec base; /**< Base class */ - - iconv_t read_cd; /**< Iconv handle for reading */ -#define INVAL_BUFSIZE (32) - uint8_t inval_buf[INVAL_BUFSIZE]; /**< Buffer for fixing up - * incomplete input - * sequences */ - size_t inval_len; /**< Number of bytes in inval_buf */ - -#define READ_BUFSIZE (8) - uint32_t read_buf[READ_BUFSIZE]; /**< Buffer for partial - * output sequences (decode) - */ - size_t read_len; /**< Number of characters in - * read_buf */ - - iconv_t write_cd; /**< Iconv handle for writing */ -#define WRITE_BUFSIZE (8) - uint32_t write_buf[WRITE_BUFSIZE]; /**< Buffer for partial - * output sequences (encode) - */ - size_t write_len; /**< Number of characters in - * write_buf */ -} hubbub_iconv_codec; - - -static bool hubbub_iconv_codec_handles_charset(const char *charset); -static hubbub_charsetcodec *hubbub_iconv_codec_create(const char *charset, - hubbub_alloc alloc, void *pw); -static void hubbub_iconv_codec_destroy (hubbub_charsetcodec *codec); -static hubbub_error hubbub_iconv_codec_encode(hubbub_charsetcodec *codec, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen); -static hubbub_error hubbub_iconv_codec_decode(hubbub_charsetcodec *codec, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen); -static hubbub_error hubbub_iconv_codec_reset(hubbub_charsetcodec *codec); -static hubbub_error hubbub_iconv_codec_filter_decoded_char( - hubbub_iconv_codec *c, uint32_t ucs4, uint8_t **dest, - size_t *destlen); -static bool hubbub_iconv_codec_is_unicode(hubbub_iconv_codec *c); -static hubbub_error hubbub_iconv_codec_read_char(hubbub_iconv_codec *c, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen); -static hubbub_error hubbub_iconv_codec_write_char(hubbub_iconv_codec *c, - uint32_t ucs4, uint8_t **dest, size_t *destlen); - -/** - * Determine whether this codec handles a specific charset - * - * \param charset Charset to test - * \return true if handleable, false otherwise - */ -bool hubbub_iconv_codec_handles_charset(const char *charset) -{ - iconv_t cd; - bool ret; - - cd = iconv_open("UCS-4", charset); - - ret = (cd != (iconv_t) -1); - - if (ret) - iconv_close(cd); - - return ret; -} - -/** - * Create an iconv-based codec - * - * \param charset The charset to read from / write to - * \param alloc Memory (de)allocation function - * \param pw Pointer to client-specific private data (may be NULL) - * \return Pointer to codec, or NULL on failure - */ -hubbub_charsetcodec *hubbub_iconv_codec_create(const char *charset, - hubbub_alloc alloc, void *pw) -{ - hubbub_iconv_codec *codec; - - codec = alloc(NULL, sizeof(hubbub_iconv_codec), pw); - if (codec == NULL) - return NULL; - - codec->read_cd = iconv_open("UCS-4", charset); - if (codec->read_cd == (iconv_t) -1) { - alloc(codec, 0, pw); - return NULL; - } - - codec->write_cd = iconv_open(charset, "UCS-4"); - if (codec->write_cd == (iconv_t) -1) { - iconv_close(codec->read_cd); - alloc(codec, 0, pw); - return NULL; - } - - codec->inval_buf[0] = '\0'; - codec->inval_len = 0; - - codec->read_buf[0] = 0; - codec->read_len = 0; - - codec->write_buf[0] = 0; - codec->write_len = 0; - - /* Finally, populate vtable */ - codec->base.handler.destroy = hubbub_iconv_codec_destroy; - codec->base.handler.encode = hubbub_iconv_codec_encode; - codec->base.handler.decode = hubbub_iconv_codec_decode; - codec->base.handler.reset = hubbub_iconv_codec_reset; - - return (hubbub_charsetcodec *) codec; -} - -/** - * Destroy an iconv-based codec - * - * \param codec The codec to destroy - */ -void hubbub_iconv_codec_destroy (hubbub_charsetcodec *codec) -{ - hubbub_iconv_codec *c = (hubbub_iconv_codec *) codec; - - iconv_close(c->read_cd); - iconv_close(c->write_cd); - - return; -} - -/** - * Encode a chunk of UCS4 data into an iconv-based codec's charset - * - * \param codec The codec to use - * \param source Pointer to pointer to source data - * \param sourcelen Pointer to length (in bytes) of source data - * \param dest Pointer to pointer to output buffer - * \param destlen Pointer to length (in bytes) of output buffer - * \return HUBBUB_OK on success, - * HUBBUB_NOMEM if output buffer is too small, - * HUBBUB_INVALID if a character cannot be represented and the - * codec's error handling mode is set to STRICT, - * <any_other_error> as a result of the failure of the - * client-provided filter function. - * - * On exit, ::source will point immediately _after_ the last input character - * read. Any remaining output for the character will be buffered by the - * codec for writing on the next call. This buffered data is post-filtering, - * so will not be refiltered on the next call. - * - * In the case of the filter function failing, ::source will point _at_ the - * last input character read; nothing will be written or buffered for the - * failed character. It is up to the client to fix the cause of the failure - * and retry the encoding process. - * - * Note that, if failure occurs whilst attempting to write any output - * buffered by the last call, then ::source and ::sourcelen will remain - * unchanged (as nothing more has been read). - * - * There is no way to determine the output character which caused a - * failure (as it may be one in a filter-injected replacement sequence). - * It is, however, possible to determine which source character caused it - * (this being the character immediately before the location pointed to by - * ::source on exit). - * - * [I.e. the process of filtering results in a potential one-to-many mapping - * between source characters and output characters, and identification of - * individual output characters is impossible.] - * - * ::sourcelen will be reduced appropriately on exit. - * - * ::dest will point immediately _after_ the last character written. - * - * ::destlen will be reduced appropriately on exit. - */ -hubbub_error hubbub_iconv_codec_encode(hubbub_charsetcodec *codec, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen) -{ - hubbub_iconv_codec *c = (hubbub_iconv_codec *) codec; - uint32_t ucs4; - const uint32_t *towrite; - size_t towritelen; - hubbub_error error; - - /* Process any outstanding characters from the previous call */ - if (c->write_len > 0) { - uint32_t *pwrite = c->write_buf; - - while (c->write_len > 0) { - error = hubbub_iconv_codec_write_char(c, pwrite[0], - dest, destlen); - if (error != HUBBUB_OK) { - /* Copy outstanding chars down, skipping - * invalid one, if present, so as to avoid - * reprocessing the invalid character */ - if (error == HUBBUB_INVALID) { - for (ucs4 = 1; ucs4 < c->write_len; - ucs4++) { - c->write_buf[ucs4] = - pwrite[ucs4]; - } - } - - return error; - } - - pwrite++; - c->write_len--; - } - } - - /* Now process the characters for this call */ - while (*sourcelen > 0) { - towrite = (const uint32_t *) (const void *) *source; - towritelen = 1; - ucs4 = *towrite; - - /* Run character we're about to output through the - * registered filter, so it can replace it, if it sees - * fit to do so */ - if (c->base.filter != NULL) { - uint32_t *replacement; - - error = c->base.filter(ntohl(ucs4), - &replacement, &towritelen, - c->base.filter_pw); - if (error != HUBBUB_OK) { - /* Don't eat character -- filter failed, - * so nothing gets written or buffered. - * It's up to the client to ensure that - * the filter works in the case where it - * reprocesses this character after the - * fault is fixed up. */ - - return error; - } - - /* Convert filter output to big endian UCS4 */ - for (ucs4 = 0; ucs4 < towritelen; ucs4++) { - replacement[ucs4] = htonl(replacement[ucs4]); - } - - towrite = (const uint32_t *) replacement; - } - - /* Output current character(s) */ - while (towritelen > 0) { - error = hubbub_iconv_codec_write_char(c, towrite[0], - dest, destlen); - - if (error != HUBBUB_OK) { - ucs4 = (error == HUBBUB_INVALID) ? 1 : 0; - - if (towritelen - ucs4 >= WRITE_BUFSIZE) - abort(); - - c->write_len = towritelen - ucs4; - - /* Copy pending chars to save area, for - * processing next call; skipping invalid - * character, if present, so it's not - * reprocessed. */ - for (; ucs4 < towritelen; ucs4++) { - c->write_buf[ucs4] = towrite[ucs4]; - } - - /* Claim character we've just buffered, - * so it's not repreocessed */ - *source += 4; - *sourcelen -= 4; - - return error; - } - - towrite++; - towritelen--; - } - - *source += 4; - *sourcelen -= 4; - } - - return HUBBUB_OK; -} - -/** - * Decode a chunk of data in an iconv-based codec's charset into UCS4 - * - * \param codec The codec to use - * \param source Pointer to pointer to source data - * \param sourcelen Pointer to length (in bytes) of source data - * \param dest Pointer to pointer to output buffer - * \param destlen Pointer to length (in bytes) of output buffer - * \return HUBBUB_OK on success, - * HUBBUB_NOMEM if output buffer is too small, - * HUBBUB_INVALID if a character cannot be represented and the - * codec's error handling mode is set to STRICT, - * <any_other_error> as a result of the failure of the - * client-provided filter function. - * - * On exit, ::source will point immediately _after_ the last input character - * read, if the result is _OK or _NOMEM. Any remaining output for the - * character will be buffered by the codec for writing on the next call. - * This buffered data is post-filtering, so will not be refiltered on the - * next call. - * - * In the case of the result being _INVALID or the filter function failing, - * ::source will point _at_ the last input character read; nothing will be - * written or buffered for the failed character. It is up to the client to - * fix the cause of the failure and retry the decoding process. - * - * Note that, if failure occurs whilst attempting to write any output - * buffered by the last call, then ::source and ::sourcelen will remain - * unchanged (as nothing more has been read). - * - * There is no way to determine the output character which caused a - * failure (as it may be one in a filter-injected replacement sequence). - * It is, however, possible to determine which source character caused it - * (this being the character immediately at or before the location pointed - * to by ::source on exit). - * - * [I.e. the process of filtering results in a potential one-to-many mapping - * between source characters and output characters, and identification of - * individual output characters is impossible.] - * - * If STRICT error handling is configured and an illegal sequence is split - * over two calls, then _INVALID will be returned from the second call, - * but ::source will point mid-way through the invalid sequence (i.e. it - * will be unmodified over the second call). In addition, the internal - * incomplete-sequence buffer will be emptied, such that subsequent calls - * will progress, rather than re-evaluating the same invalid sequence. - * - * ::sourcelen will be reduced appropriately on exit. - * - * ::dest will point immediately _after_ the last character written. - * - * ::destlen will be reduced appropriately on exit. - * - * Call this with a source length of 0 to flush the output buffer. - */ -hubbub_error hubbub_iconv_codec_decode(hubbub_charsetcodec *codec, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen) -{ - hubbub_iconv_codec *c = (hubbub_iconv_codec *) codec; - hubbub_error error; - - if (c->read_len > 0) { - /* Output left over from last decode - * Attempt to finish this here */ - uint32_t *pread = c->read_buf; - - while (c->read_len > 0 && *destlen >= c->read_len * 4) { - *((uint32_t *) (void *) *dest) = pread[0]; - - *dest += 4; - *destlen -= 4; - - pread++; - c->read_len--; - } - - if (*destlen < c->read_len * 4) { - /* Run out of output buffer */ - size_t i; - - /* Shuffle remaining output down */ - for (i = 0; i < c->read_len; i++) { - c->read_buf[i] = pread[i]; - } - - return HUBBUB_NOMEM; - } - } - - if (c->inval_len > 0) { - /* The last decode ended in an incomplete sequence. - * Fill up inval_buf with data from the start of the - * new chunk and process it. */ - uint8_t *in = c->inval_buf; - size_t ol = c->inval_len; - size_t l = min(INVAL_BUFSIZE - ol - 1, *sourcelen); - size_t orig_l = l; - - memcpy(c->inval_buf + ol, *source, l); - - l += c->inval_len; - - error = hubbub_iconv_codec_read_char(c, - (const uint8_t **) &in, &l, dest, destlen); - if (error != HUBBUB_OK && error != HUBBUB_NOMEM) { - return error; - } - - - /* And now, fix everything up so the normal processing - * does the right thing. */ - *source += max((signed) (orig_l - l), 0); - *sourcelen -= max((signed) (orig_l - l), 0); - - /* Failed to resolve an incomplete character and - * ran out of buffer space. No recovery strategy - * possible, so explode everywhere. */ - if ((orig_l + ol) - l == 0) - abort(); - - /* Handle memry exhaustion case from above */ - if (error != HUBBUB_OK) - return error; - } - - while (*sourcelen > 0) { - error = hubbub_iconv_codec_read_char(c, - source, sourcelen, dest, destlen); - if (error != HUBBUB_OK) { - return error; - } - } - - return HUBBUB_OK; -} - -/** - * Clear an iconv-based codec's encoding state - * - * \param codec The codec to reset - * \return HUBBUB_OK on success, appropriate error otherwise - */ -hubbub_error hubbub_iconv_codec_reset(hubbub_charsetcodec *codec) -{ - hubbub_iconv_codec *c = (hubbub_iconv_codec *) codec; - - iconv(c->read_cd, NULL, NULL, NULL, NULL); - iconv(c->write_cd, NULL, NULL, NULL, NULL); - - c->inval_buf[0] = '\0'; - c->inval_len = 0; - - c->read_buf[0] = 0; - c->read_len = 0; - - c->write_buf[0] = 0; - c->write_len = 0; - - return HUBBUB_OK; -} - -/** - * Feed a UCS4 character through the registered filter and output the result - * - * \param c Codec to use - * \param ucs4 UCS4 character (big endian) - * \param dest Pointer to pointer to output buffer - * \param destlen Pointer to output buffer length - * \return HUBBUB_OK on success, - * HUBBUB_NOMEM if output buffer is too small, - * <any_other_error> as a result of the failure of the - * client-provided filter function. - */ -hubbub_error hubbub_iconv_codec_filter_decoded_char(hubbub_iconv_codec *c, - uint32_t ucs4, uint8_t **dest, size_t *destlen) -{ - if (c->base.filter != NULL) { - uint32_t *rep; - size_t replen; - hubbub_error error; - - error = c->base.filter(ntohl(ucs4), &rep, &replen, - c->base.filter_pw); - if (error != HUBBUB_OK) { - return error; - } - - while (replen > 0 && *destlen >= replen * 4) { - *((uint32_t *) (void *) *dest) = htonl(*rep); - - *dest += 4; - *destlen -= 4; - - rep++; - replen--; - } - - if (*destlen < replen * 4) { - /* Run out of output buffer */ - size_t i; - - /* Buffer remaining output */ - c->read_len = replen; - - for (i = 0; i < replen; i++) { - c->read_buf[i] = htonl(rep[i]); - } - - return HUBBUB_NOMEM; - } - - } else { - if (*destlen < 4) { - /* Run out of output buffer */ - - c->read_len = 1; - c->read_buf[0] = ucs4; - - return HUBBUB_NOMEM; - } - - *((uint32_t *) (void *) *dest) = ucs4; - *dest += 4; - *destlen -= 4; - } - - return HUBBUB_OK; -} - -/** - * Detect if a codec's charset is Unicode capable - * - * \param c Codec to consider - * \return true if a Unicode variant, false otherwise - */ -bool hubbub_iconv_codec_is_unicode(hubbub_iconv_codec *c) -{ - static uint16_t ucs4; - static uint16_t ucs2; - static uint16_t utf8; - static uint16_t utf16; - static uint16_t utf16be; - static uint16_t utf16le; - static uint16_t utf32; - static uint16_t utf32be; - static uint16_t utf32le; - - if (ucs4 == 0) { - ucs4 = hubbub_mibenum_from_name("UCS-4", SLEN("UCS-4")); - ucs2 = hubbub_mibenum_from_name("UCS-2", SLEN("UCS-2")); - utf8 = hubbub_mibenum_from_name("UTF-8", SLEN("UTF-8")); - utf16 = hubbub_mibenum_from_name("UTF-16", SLEN("UTF-16")); - utf16be = hubbub_mibenum_from_name("UTF-16BE", - SLEN("UTF-16BE")); - utf16le = hubbub_mibenum_from_name("UTF-16LE", - SLEN("UTF-16LE")); - utf32 = hubbub_mibenum_from_name("UTF-32", SLEN("UTF-32")); - utf32be = hubbub_mibenum_from_name("UTF-32BE", - SLEN("UTF-32BE")); - utf32le = hubbub_mibenum_from_name("UTF-32LE", - SLEN("UTF-32LE")); - } - - return (c->base.mibenum == ucs4 || - c->base.mibenum == ucs2 || - c->base.mibenum == utf8 || - c->base.mibenum == utf16 || - c->base.mibenum == utf16be || - c->base.mibenum == utf16le || - c->base.mibenum == utf32 || - c->base.mibenum == utf32be || - c->base.mibenum == utf32le); -} - -/** - * Read a character from the codec's native charset to UCS4 (big endian) - * - * \param c The codec - * \param source Pointer to pointer to source buffer (updated on exit) - * \param sourcelen Pointer to length of source buffer (updated on exit) - * \param dest Pointer to pointer to output buffer (updated on exit) - * \param destlen Pointer to length of output buffer (updated on exit) - * \return HUBBUB_OK on success, - * HUBBUB_NOMEM if output buffer is too small, - * HUBBUB_INVALID if a character cannot be represented and the - * codec's error handling mode is set to STRICT, - * <any_other_error> as a result of the failure of the - * client-provided filter function. - * - * On exit, ::source will point immediately _after_ the last input character - * read, if the result is _OK or _NOMEM. Any remaining output for the - * character will be buffered by the codec for writing on the next call. - * This buffered data is post-filtering, so will not be refiltered on the - * next call. - * - * In the case of the result being _INVALID or the filter function failing, - * ::source will point _at_ the last input character read; nothing will be - * written or buffered for the failed character. It is up to the client to - * fix the cause of the failure and retry the decoding process. - * - * ::sourcelen will be reduced appropriately on exit. - * - * ::dest will point immediately _after_ the last character written. - * - * ::destlen will be reduced appropriately on exit. - */ -hubbub_error hubbub_iconv_codec_read_char(hubbub_iconv_codec *c, - const uint8_t **source, size_t *sourcelen, - uint8_t **dest, size_t *destlen) -{ - size_t iconv_ret; - const uint8_t *origsrc = *source; - size_t origsrclen = *sourcelen; - uint32_t ucs4; - uint8_t *pucs4 = (uint8_t *) &ucs4; - size_t sucs4 = 4; - hubbub_error error; - - /* Use iconv to convert a single character - * Side effect: Updates *source to point at next input - * character and *sourcelen to reflect reduced input length - */ - iconv_ret = iconv(c->read_cd, (char **) source, sourcelen, - (char **) (void *) &pucs4, &sucs4); - - if (iconv_ret != (size_t) -1 || - (*source != origsrc && sucs4 == 0)) { - /* Read a character */ - error = hubbub_iconv_codec_filter_decoded_char(c, - ucs4, dest, destlen); - if (error != HUBBUB_OK && error != HUBBUB_NOMEM) { - /* filter function failed; restore source pointers */ - *source = origsrc; - *sourcelen = origsrclen; - } - - /* Clear inval buffer */ - c->inval_buf[0] = '\0'; - c->inval_len = 0; - - return error; - } else if (errno == E2BIG) { - /* Should never happen */ - abort(); - } else if (errno == EINVAL) { - /* Incomplete input sequence */ - if (*sourcelen > INVAL_BUFSIZE) - abort(); - - memmove(c->inval_buf, (const char *) *source, *sourcelen); - c->inval_buf[*sourcelen] = '\0'; - c->inval_len = *sourcelen; - - *source += *sourcelen; - *sourcelen = 0; - - return HUBBUB_OK; - } else if (errno == EILSEQ) { - /* Illegal input sequence */ - bool found = false; - const uint8_t *oldsrc; - size_t oldsrclen; - - /* Clear inval buffer */ - c->inval_buf[0] = '\0'; - c->inval_len = 0; - - /* Strict errormode; simply flag invalid character */ - if (c->base.errormode == HUBBUB_CHARSETCODEC_ERROR_STRICT) { - /* restore source pointers */ - *source = origsrc; - *sourcelen = origsrclen; - - return HUBBUB_INVALID; - } - - /* Ok, this becomes problematic. The iconv API here - * is particularly unhelpful; *source will point at - * the _start_ of the illegal sequence. This means - * that we must find the end of the sequence */ - - /* Search for the start of the next valid input - * sequence (or the end of the input stream) */ - while (*sourcelen > 1) { - pucs4 = (uint8_t *) &ucs4; - sucs4 = 4; - - (*source)++; - (*sourcelen)--; - - oldsrc = *source; - oldsrclen = *sourcelen; - - iconv_ret = iconv(c->read_cd, - (char **) source, sourcelen, - (char **) (void *) &pucs4, &sucs4); - if (iconv_ret != (size_t) -1 || errno != EILSEQ) { - found = true; - break; - } - } - - if (found) { - /* Found start of next valid sequence */ - *source = oldsrc; - *sourcelen = oldsrclen; - } else { - /* Not found - skip last byte in buffer */ - (*source)++; - (*sourcelen)--; - - if (*sourcelen != 0) - abort(); - } - - /* output U+FFFD and continue processing. */ - error = hubbub_iconv_codec_filter_decoded_char(c, - htonl(0xFFFD), dest, destlen); - if (error != HUBBUB_OK && error != HUBBUB_NOMEM) { - /* filter function failed; restore source pointers */ - *source = origsrc; - *sourcelen = origsrclen; - } - - return error; - } - - return HUBBUB_OK; -} - -/** - * Write a UCS4 character in a codec's native charset - * - * \param c The codec - * \param ucs4 The UCS4 character to write (big endian) - * \param dest Pointer to pointer to output buffer (updated on exit) - * \param destlen Pointer to length of output buffer (updated on exit) - * \return HUBBUB_OK on success, - * HUBBUB_NOMEM if output buffer is too small, - * HUBBUB_INVALID if character cannot be represented and the - * codec's error handling mode is set to STRICT. - */ -hubbub_error hubbub_iconv_codec_write_char(hubbub_iconv_codec *c, - uint32_t ucs4, uint8_t **dest, size_t *destlen) -{ - size_t iconv_ret; - uint8_t *pucs4 = (uint8_t *) &ucs4; - size_t sucs4 = 4; - uint8_t *origdest = *dest; - - iconv_ret = iconv(c->write_cd, (char **) (void *) &pucs4, - &sucs4, (char **) dest, destlen); - - if (iconv_ret == (size_t) -1 && errno == E2BIG) { - /* Output buffer is too small */ - return HUBBUB_NOMEM; - } else if (iconv_ret == (size_t) -1 && errno == EILSEQ) { - /* Illegal multibyte sequence */ - /* This should never happen */ - abort(); - } else if (iconv_ret == (size_t) -1 && errno == EINVAL) { - /* Incomplete input character */ - /* This should never happen */ - abort(); - } else if (*dest == origdest) { - /* Nothing was output */ - switch (c->base.errormode) { - case HUBBUB_CHARSETCODEC_ERROR_STRICT: - return HUBBUB_INVALID; - - case HUBBUB_CHARSETCODEC_ERROR_TRANSLIT: - /** \todo transliteration */ - case HUBBUB_CHARSETCODEC_ERROR_LOOSE: - { - pucs4 = (uint8_t *) &ucs4; - sucs4 = 4; - - ucs4 = hubbub_iconv_codec_is_unicode(c) - ? htonl(0xFFFD) : htonl(0x3F); - - iconv_ret = iconv(c->write_cd, - (char **) (void *) &pucs4, &sucs4, - (char **) dest, destlen); - - if (iconv_ret == (size_t) -1 && errno == E2BIG) { - return HUBBUB_NOMEM; - } else if (iconv_ret == (size_t) -1 && - errno == EILSEQ) { - /* Illegal multibyte sequence */ - /* This should never happen */ - abort(); - } else if (iconv_ret == (size_t) -1 && - errno == EINVAL) { - /* Incomplete input character */ - /* This should never happen */ - abort(); - } - } - break; - } - } - - return HUBBUB_OK; -} - -const hubbub_charsethandler hubbub_iconv_codec_handler = { - hubbub_iconv_codec_handles_charset, - hubbub_iconv_codec_create -}; |