2010-04-22 Martyn Russell Updated NEWS Updated NEWS tracker-extract: Fixed double free in Vorbis extractor 2010-04-22 Carlos Garnacho TrackerMinerFS: always set current_parent in ensure_mtime_cache() 2010-04-22 Philip Van Hoof Revert "libtracker-db: Fixed a problem with sqlite3_interrupt usage" This reverts commit f47c31e7f2eff26a30e469bafc01f976741c244a. Using the 'fix' I get sporadic endless loops where the stmt never goes out of SQLITE_INTERRUPTED 2010-04-22 Martyn Russell Bumped version 2010-04-22 Amit Jain Functional Tests: Updated to allow writeback completion before extracting Functional Tests: Operations of files in multilevel subfolders added 2010-04-22 Martyn Russell Updated NEWS 2010-04-22 Carlos Garnacho TrackerDBInterfaceSqlite: do sanity check on is_parent() and is_descendant(). TrackerDBInterfaceSqlite: Check more carefully for is_parent(). Be more intelligent about trailing slashes and don't take exact string matches as a positive. Use tracker:uri-is-descendant() in recursive delete operations. TrackerMinerFS: Get rid of slash_uri in queries where we use tracker:uri-is-*() This is already handled in these functions, so no need for that check in the miner. TrackerMinerFS: Improve condition in ensure_iri_cache(). Remove \n from g_debug and g_criticals. TrackerMinerFS: Make mtime/iri caches query per folder through tracker:uri-is-*(). Also, the query to get all children on directory moving is done through tracker:uri-is-descendant(). TrackerSparqlExpression: add tracker:uri-is-parent/descendant() sparql functions. TrackerDBInterfaceSqlite: Add functions to deal with URIs hierarchy. SparqlUriIsParent and SparqlUriIsDescendant have been added, taking 2 URIs and returning true if the second is a direct/indirect child of the former. TrackerMinerFS: Cache URNs for the processed files. This cache is cleared on processing directory change, late additions are also handled, as well as cache invalidation on files being moved/deleted. TrackerMinerFS: Do not query parent URN frequently if it doesn't exist. This is specially visible in the applications miner, since the container directories aren't actually stored. TrackerMinerFS: Perform one mtime query per folder. Now that crawling/processing order is guaranteed to have all items in one folder to be handled together, perform a mtime query per folder, and compare against these results. 2010-04-22 Philip Van Hoof libtracker-db: Fixed a problem with sqlite3_interrupt usage tracker-search-bar: Comparison with unsigned integer (enum) is always true tracker-extract: Compilation warning fix NFO: Add an index on nfo:fileLastModified for miner-fs's crawling performance 2010-04-21 Jorge González Updated Spanish translation 2010-04-21 Aleksander Morgado Fixes GB#616165: Improved reading PS and PS-GZ files * Don't use a temporary file for the uncompressed ps-gz file contents * Re-use the same buffer when iterating each line in the ps extractor, no need to allocate a new one each time. * Stop reading the uncompressed PS file if reached a max limit (same previous limit of 20MBytes is used, just a safe one) Fixes GB#616403 - Improve & fix reading msoffice/powerpoint files Bugfixes: * CharsAtoms now read as CharsAtoms; and BytesAtoms read as BytesAtoms. * UTF-16 string converted to UTF-8 before normalizing * Fix the Invalid Reads as now generated UTF-8 string comes NIL-terminated Improvements: * Stop reading when max bytes reached * Re-use buffer for reading 2010-04-21 Martyn Russell Merge branch 'async-queries-due' 2010-04-21 Aleksander Morgado Fixes GB#616329: Improve and fix reading msoffice/excel files * Strings which are split into two ExcelExtendedStringRecord are now properly read. * String contents are now converted from CP1252 or UTF-16 to UTF-8 * Reading the string is now done in either 1 or 2 GSF reads, not byte per byte. * Limited the max number of bytes to be read from the stream, to the common safe limit of 3*max_words*max_word_size. * Contents are now normalized and word-counted in a buffered way. * Stop reading the contents when max bytes reached. * Stop reading the contents when max number of words reached. 2010-04-21 Martyn Russell libtracker-miner: Don't translate statuses This also adds a tracker-status -s to list common statuses 2010-04-20 Aleksander Morgado Fixes GB#615948 - Improved reading msoffice/xml files * Don't use heap to store the whole output of libgsf, use just a buffer in stack, and read & parse in a buffered way. * Limit to 20MBytes the max of bytes which could be read from the uncompressed XML file. 2010-04-20 Tshepang Lekhonkhobe tracker-sparql: Fixed typo in man page for command line args tracker-control: Added --reindex-mime-type and --start options to man page docs 2010-04-20 Carlos Garnacho Move nfo:isContentEncrypted to be defined in 33-nfo.ontology. tracker-extract-pdf: set nfo:isContentEncrypted for encrypted docs. 2010-04-20 Aleksander Morgado Fixes GB#616158/doc: Improve reading msoffice doc files * Limit the max number of bytes to be read from the stream, to some safe limit like 3*max_words*max_word_size. * Don't load the whole doc in heap: use a buffer to read the contents, convert to UTF-8, perform normalization and word count (chunk by chunk). * Stop reading the contents when max bytes reached. * Stop reading the contents when max number of words reached. 2010-04-20 Mikael Ottela Fixes bug 161964 - Some special audio files are not displayed in the Music.. Try to check whether a mp4 file is actually only audio, not video. 2010-04-19 Carlos Garnacho tracker-extract-html: Improve