r3394 | joeshaw | 2007-02-07 11:28:58 -0500 (Wed, 07 Feb 2007) | 1 line Update for 0.2.16 release ------------------------------------------------------------------------ r3393 | joeshaw | 2007-02-07 11:24:36 -0500 (Wed, 07 Feb 2007) | 1 line Temporarily back out the opera backend for the 0.2.16 release ------------------------------------------------------------------------ r3392 | joeshaw | 2007-02-07 11:16:43 -0500 (Wed, 07 Feb 2007) | 1 line Bump version number on SVG filter ------------------------------------------------------------------------ r3391 | kkubasik | 2007-02-06 23:23:46 -0500 (Tue, 06 Feb 2007) | 3 lines r195@kjk38-laptop: kjk38 | 2007-02-06 23:24:16 -0500 Initial import of the opera history backend ------------------------------------------------------------------------ r3390 | leonardof | 2007-02-06 21:38:58 -0500 (Tue, 06 Feb 2007) | 6 lines 2007-02-07 Leonardo Ferreira Fontenelle * pt_BR.po: Updated Brazilian Portuguese translation by Igor Pires Soares . ------------------------------------------------------------------------ r3389 | dbera | 2007-02-06 19:27:30 -0500 (Tue, 06 Feb 2007) | 2 lines Batik svg files do not declare xlink namespace. Handle that explicitly. ------------------------------------------------------------------------ r3388 | joeshaw | 2007-02-06 18:19:13 -0500 (Tue, 06 Feb 2007) | 1 line Updated SVG filter, from Alex Macdonald ------------------------------------------------------------------------ r3387 | joeshaw | 2007-02-06 18:17:52 -0500 (Tue, 06 Feb 2007) | 2 lines Use indexable.MimeType rather than filter.MimeType, which isn't canonical and shouldn't really be used outside of the filter itself ------------------------------------------------------------------------ r3386 | joeshaw | 2007-02-06 15:31:43 -0500 (Tue, 06 Feb 2007) | 1 line Roll back our mono requirement to 1.1.13.5 ------------------------------------------------------------------------ r3385 | joeshaw | 2007-02-06 11:38:13 -0500 (Tue, 06 Feb 2007) | 5 lines Only set the mime type in the indexable.ContentUri.IsFile block if it hasn't already been set by a backend. Fixes the problem where Thunderbird mails were being stored as text/plain. ------------------------------------------------------------------------ r3384 | dbera | 2007-02-06 11:24:12 -0500 (Tue, 06 Feb 2007) | 2 lines beagle-hit.c: "(null)" does not deserialize to null Uri. Watch for null ParentUri when serializing. ------------------------------------------------------------------------ r3383 | pgeyleg | 2007-02-06 00:26:45 -0500 (Tue, 06 Feb 2007) | 1 line dzongkha translation added ------------------------------------------------------------------------ r3382 | dbera | 2007-02-05 17:10:20 -0500 (Mon, 05 Feb 2007) | 2 lines Remove debug messages from Semweb. ------------------------------------------------------------------------ r3381 | jmas | 2007-02-05 16:55:18 -0500 (Mon, 05 Feb 2007) | 5 lines 2007-02-05 Jordi Mas * ca.po: Updated Catalan translation. ------------------------------------------------------------------------ r3380 | dbera | 2007-02-05 16:06:43 -0500 (Mon, 05 Feb 2007) | 2 lines Remove some debug messages while parsing image files. ------------------------------------------------------------------------ r3379 | joeshaw | 2007-02-05 15:57:03 -0500 (Mon, 05 Feb 2007) | 1 line Don't set IsIndexing if a folder is being recrawled due to an inotify event ------------------------------------------------------------------------ r3378 | joeshaw | 2007-02-05 15:36:43 -0500 (Mon, 05 Feb 2007) | 1 line Remove some unused code I accidentally committed ------------------------------------------------------------------------ r3377 | joeshaw | 2007-02-05 14:58:03 -0500 (Mon, 05 Feb 2007) | 21 lines Rework the way we do reporting of whether or not we're indexing to make it a more active task for backends, rather than passive. This fixes problems trying to deal with this sanely in the file system backend. * Remove some dead code from QueryDriver.cs. * No longer declare IsIndexing_get as virtual, and replace it with a private instance variable. It's now the backends' responsibility to set IsIndexing as the status changes, rather than waiting for the getter to be called and reporting the current state of things. * Update the file system backend to the new scheme. Should fix problems where the "currently indexing" info box stays on in beagle-search even after the initial crawl has finished. * Also update Evolution mail, EDS, Gaim, and Konversation backends to the new scheme. * Improve the status message when indexing files inside a directory. ------------------------------------------------------------------------ r3376 | dbera | 2007-02-05 13:12:15 -0500 (Mon, 05 Feb 2007) | 2 lines * libbeagle/beagle/beagle-hit.c: Add ParentUri to snippet_request->hit. This is needed by some backends that check for parent uri while generating snippets. ------------------------------------------------------------------------ r3375 | dbera | 2007-02-05 12:11:20 -0500 (Mon, 05 Feb 2007) | 4 lines * Konversation backend: Add server and channel name as properties. Somehow missed this before. * LuceneIndexingDriver: Back out a change from yesterday. Its incorrect. ------------------------------------------------------------------------ r3374 | aihana | 2007-02-05 07:30:59 -0500 (Mon, 05 Feb 2007) | 4 lines 2007-02-05 Takeshi AIHANA * ja.po: Updated Japanese translation. ------------------------------------------------------------------------ r3373 | dbera | 2007-02-04 17:23:12 -0500 (Sun, 04 Feb 2007) | 4 lines Correctly handle filter versions while indexing. * FileSystemQueryable.cs, FileAttributesStore.cs: If filter information is not present for a file, resubmit it for indexing. Ideally, we would like to resubmit whenever a new filter version is present, but that is becoming too costly. * FileAttributes.cs: Valid filter can have version 0+. Fix the version checking. ------------------------------------------------------------------------ r3372 | dbera | 2007-02-04 12:21:51 -0500 (Sun, 04 Feb 2007) | 2 lines Konversation backend: Better protection against old log files. Also, fix a bug where one log file would be indexed twice. ------------------------------------------------------------------------ r3369 | dbera | 2007-02-03 18:20:51 -0500 (Sat, 03 Feb 2007) | 2 lines Filter.cs: Explicitly warn Filters if they try to register a type in constructor. ------------------------------------------------------------------------ r3368 | dnylande | 2007-02-02 18:45:45 -0500 (Fri, 02 Feb 2007) | 1 line sv.po: Swedish translation updated ------------------------------------------------------------------------ r3367 | dbera | 2007-02-02 18:06:38 -0500 (Fri, 02 Feb 2007) | 2 lines KMailQueryable.cs: Fix checking whether a directory is a valid kmail directory. Fixes #401767 and #391647. ------------------------------------------------------------------------ r3366 | dbera | 2007-02-02 16:10:32 -0500 (Fri, 02 Feb 2007) | 2 lines Konversation backend: return true for IsIndexing when performing initial indexing. ------------------------------------------------------------------------ r3365 | dbera | 2007-02-02 15:13:17 -0500 (Fri, 02 Feb 2007) | 2 lines Really old konversation logfiles have bad timestamp strings. Ignore those lines completely, preventing crash if you have these old files. ------------------------------------------------------------------------ r3364 | joeshaw | 2007-02-02 13:54:11 -0500 (Fri, 02 Feb 2007) | 6 lines Mono doesn't handle JITted code running in a child process after a fork() very well, and can often hang. Change our process limiting code to execute in C after the fork, rather than in C# in the runtime. It's a little less flexible in theory, but has higher reliability. ------------------------------------------------------------------------ r3363 | icq | 2007-02-02 09:12:23 -0500 (Fri, 02 Feb 2007) | 2 lines Updated Galician Translation ------------------------------------------------------------------------ r3362 | joeshaw | 2007-02-01 16:07:35 -0500 (Thu, 01 Feb 2007) | 1 line Also run pdftotext -q (for quiet) again ------------------------------------------------------------------------ r3361 | joeshaw | 2007-02-01 16:05:29 -0500 (Thu, 01 Feb 2007) | 2 lines Bleh, r3360 didn't work and just inverted the deadlock. Just ignore and don't redirect stderr for now ------------------------------------------------------------------------ r3360 | joeshaw | 2007-02-01 15:50:24 -0500 (Thu, 01 Feb 2007) | 7 lines pdftotext completely spams stderr on some documents, which causes a deadlock between beagled-helper and pdftotext: pdftotext is blocking on additional buffer space to write out more warnings while beagled-helper is blocking on data from stdout. Fix this by processing data from stderr before processing stdout in DoPull(), rather than deferring it all until DoClose(). Fixes Jan's part of bnc #402065 ------------------------------------------------------------------------ r3357 | joeshaw | 2007-01-31 18:44:13 -0500 (Wed, 31 Jan 2007) | 2 lines Don't include a pre-configured beagle-crawl-system in the tarball. Fixes bgo #401504 ------------------------------------------------------------------------ r3356 | joeshaw | 2007-01-31 16:07:31 -0500 (Wed, 31 Jan 2007) | 1 line Change a couple of chatty Warns to Debugs in the KMail backend ------------------------------------------------------------------------ r3355 | joeshaw | 2007-01-31 15:08:29 -0500 (Wed, 31 Jan 2007) | 4 lines Add --no-snapshot to beagled to keep it from automatically snapshotting when --heap-shot is provided. ------------------------------------------------------------------------ r3354 | joeshaw | 2007-01-31 14:56:27 -0500 (Wed, 31 Jan 2007) | 17 lines Put the looping bug out to pasture once and for all. In FSQ.DoneCrawlingOneDirectory(), in the event that we get a null file attribute back -- which should no longer happen -- complain loudly but still mark the directory as clean. This will prevent the daemon from looping on that one directory. Fix our sqlite schema for our fallback FileAttributesStore. We were declaring our columns as STRING -- which isn't valid -- instead of TEXT. This meant that fields were defaulting to being integer-like. Filenames like "0009" became "9" in the database. Bump up the schema version and fix that. Also make our file_path index unique, so we don't have to deal with potential multiple entries in the database. Add debugging output to the FileAttributesStore. ------------------------------------------------------------------------ r3353 | joeshaw | 2007-01-30 17:32:00 -0500 (Tue, 30 Jan 2007) | 2 lines Use a better GenericName in the beagle-settings .desktop file, lose the redundant comment ------------------------------------------------------------------------ r3352 | joeshaw | 2007-01-30 16:34:22 -0500 (Tue, 30 Jan 2007) | 10 lines Only do the position check hack in the archive filter for gzip archives; looks like the position is pretty much worthless for bzip2 archives and it triggers this code path. Fix a bug where we were leaking a tempfile when decompressing archives of more than 30 files. Improve debug output when storing streams in Indexable.cs. ------------------------------------------------------------------------ r3351 | joeshaw | 2007-01-30 16:07:16 -0500 (Tue, 30 Jan 2007) | 6 lines Workaround archive files that cause infinite loops. When reading from an archive stream, if we are extracting the maximum amount of data and our position in the stream isn't moving, break out of the loop and flag the file as bad. Don't create child indexables for such entries. ------------------------------------------------------------------------ r3350 | joeshaw | 2007-01-30 15:37:11 -0500 (Tue, 30 Jan 2007) | 1 line Add more info to the ExceptionHandlingThread name ------------------------------------------------------------------------ r3349 | joeshaw | 2007-01-30 15:11:13 -0500 (Tue, 30 Jan 2007) | 4 lines Always dispose of the SqliteCommand, even if we get an exception. Log the SQL that caused an exception inside of DoNonQuery() ------------------------------------------------------------------------ r3348 | joeshaw | 2007-01-30 15:01:05 -0500 (Tue, 30 Jan 2007) | 20 lines GetLastWriteTimeUtc() no longer throws an IOException if the file isn't there in 2.0. It returns a silly date from 1601. Fix various places (mostly in the file system backend) where we were depending on that behavior. Slightly reorder the flow in FilterFactory.FilterIndexable so that we avoid unnecessary work, and also check in here if the file has disappeared out from underneath us. Fix a DateTime UTC conversion problem that was still hanging around in FileAttributesStore. Fix our ParseExact() in StringFu. We want to AssumeUniversal AND AdjustToUniversal so that dates we get back are right. Mono is broken with these options right now -- see http://bugzilla.ximian.com/show_bug.cgi?id=80677 -- but things work well enough now with the broken behavior, and the AdjustToUniversal is necessary for when it's fixed. ------------------------------------------------------------------------ r3347 | dbera | 2007-01-30 12:14:11 -0500 (Tue, 30 Jan 2007) | 3 lines * F-Spot code: More protection against bad jpegs. * XdgMime.cs: Catch other valid exceptions when detecting mimetype. ------------------------------------------------------------------------ r3346 | joeshaw | 2007-01-29 15:52:35 -0500 (Mon, 29 Jan 2007) | 3 lines Always return false instead of true from KonversationQueryable.IsIndexing, because otherwise the 'your data is being indexed' info box is always present. ------------------------------------------------------------------------ r3345 | joeshaw | 2007-01-29 14:17:01 -0500 (Mon, 29 Jan 2007) | 1 line Workaround a mono bug (bxc #80320) when dealing with dates around MinValue ------------------------------------------------------------------------ r3344 | dbera | 2007-01-28 22:14:18 -0500 (Sun, 28 Jan 2007) | 2 lines Fix crash when parsing extremely small files with no head or body. ------------------------------------------------------------------------ r3343 | dbera | 2007-01-28 11:30:59 -0500 (Sun, 28 Jan 2007) | 2 lines Ignore Bim entries with marker PHUT. F-Spot Bim parser does not recognise them. ------------------------------------------------------------------------ r3342 | dbera | 2007-01-28 11:13:07 -0500 (Sun, 28 Jan 2007) | 2 lines Handle javascript comments properly, namely the comments can contain quote characters which should be ignored and not assumed part of a quote. Error in original HtmlAgilityPack. There is still scope of error in script and style elements. They will be reported as "System.Exception: Load() was not paused previously" and need to be fixed on demand. ------------------------------------------------------------------------ r3341 | dbera | 2007-01-27 13:39:27 -0500 (Sat, 27 Jan 2007) | 2 lines Add the changes. ------------------------------------------------------------------------ r3340 | dbera | 2007-01-27 13:35:34 -0500 (Sat, 27 Jan 2007) | 2 lines Check iptc entry for null before trying to parse it. Close memorystream after use. ------------------------------------------------------------------------ r3339 | dbera | 2007-01-26 19:57:06 -0500 (Fri, 26 Jan 2007) | 2 lines Fix a bug in extraction of JFIF comments from Jpeg files. ------------------------------------------------------------------------ r3338 | joeshaw | 2007-01-26 16:16:26 -0500 (Fri, 26 Jan 2007) | 4 lines Update some assembly requirements in configure.in, and remove some unnecessary ones from beagled/Makefile.am ------------------------------------------------------------------------ r3336 | joeshaw | 2007-01-25 18:40:53 -0500 (Thu, 25 Jan 2007) | 1 line 0.2.15.1, brown paper bag release ------------------------------------------------------------------------ r3335 | joeshaw | 2007-01-25 18:35:51 -0500 (Thu, 25 Jan 2007) | 2 lines We need to bundle the thunderbird files in any case, regardless of whether or not it's enabled. ------------------------------------------------------------------------ r3332 | joeshaw | 2007-01-25 17:36:12 -0500 (Thu, 25 Jan 2007) | 1 line Update changelog for 0.2.15 ------------------------------------------------------------------------