Release notes for Recoll 1.15.0/1.15.9

Notes:

• 1.15.9 fixes a startup crash in 1.15.8. This occurs only on some architectures, and if you are not experiencing the problem (it happens every time if at all), you do not need to upgrade.
• 1.15.8 fixes quite a few bugs (see the BUGS file) and also breaks my promise to only fix bugs in minor releases because it also has a few small functional improvements (which actually border on the bug fix category):
  • Searches now allow negative directory filtering (all results except from the specified directory). Other attempts at still impossible negative searches (ie: -mime:) now cause explicit errors messages instead of lame results. The inverted directory filtering is accessible from the query language and by checking a checkbox in the advanced search panel.
  • The GUI advanced search panel now allows specifying a field for each entry (ie: author/recipient, etc).
  • Added filters for .war (Konqueror web archive) and .mhtm (other web archive format).
  • Result table:
    • The detail area now has a popup menu similar to the one in the result list (open parent, save to disk etc.).
    • Estimated result counts are displayed in the status line.
• 1.15.6 and 1.15.7 almost exclusively fix build issues (except for the indication of the stemming language in the menu, fixed in 1.15.7). If you are running 1.15.5, you probably don't need to upgrade.
• 1.15.5 fixes more crashes still in 1.15.2. See the BUGS file
• 1.15.2 fixes the GUI startup crash described below, and an issue with very long path elements, which manifested itself mainly while indexing the Beagle queue.
• 1.15.1 is identical to 1.15.0 except for a very small modification to allow building with qt 4.4.

Caveats

There have been reports of the recoll 1.15 GUI crashing when starting up after an upgrade. This was fixed by making a backup of ~/.config/Recoll.org/recoll.conf and deleting it (Qt stores the QSettings for the recoll GUI in this file, things like the search history and mix GUI options).

If the result list is displayed with strange indentations, try to completely clear the result list paragraph format (in query preferences) to reset it to the default. Next time you open query preferences, the new default format will be up, and you can make any desired changes.

The support for Qt 3 has been removed in this version which uses only native Qt 4 interfaces. If your system has no support for Qt 4.4 or newer, you need to stick with Recoll 1.14, which is still supported.

The defaults for file name searches with multiple fragments changed from OR to AND: see the changes list below for more detail.

Installing over an older version: version 1.15/14/13 indexes are mostly compatible with 1.11 indexes, but some features depend on a full index rebuild. These are:

• Filtering results from a given directory subtree.
• Displaying simple filenames in the new result table.

If installing over 1.10 or older, you need a full rebuild. The best way to do this is to just delete the old .recoll/xapiandb directory, especially if the index was created by an older version. (ie: rm -rf ~/.recoll/xapiandb/). On very old indexes, recollindex -z may sometimes end with a backend doesn't implement metadata error, which is wasteful because you then need to delete xapiandb and run the indexing again.

Changes

Recoll 1.15.0 has major new GUI features and fixes a number of small bugs

• The GUI has a new display for the result list inside a spreadsheet-like table, with configurable columns and sort by any column. The table header right-click menu has the reset sort function and column adding/removing.The "classical" look is still there, you can dynamically switch between list and table by clicking the table-like icon in the toolbar.
• The old sort tool is gone. There are now vertical arrows in the tool bar to directly sort by date ascending or descending, which was its only significant use.
• We added duplication indicators to the result list when results are collapsed because they have bit-for-bit identical contents. The indicater is the collapse count in parentheses, displayed just after the relevancy percentage.
• There is now a menu entry to clear search history.
• File name search: it used to be that multiple fragments where OR'ed together to perform the search. They are now AND'ed, which makes more sense in many cases (remembering several fragments of the file name but not the order), but means that a search for *.doc *.odt will always fail. Use ext:doc OR  ext:odt instead.
• Autophrase now works with the query language where it makes sense.
• Support for lyrics in midi karaoke files. Works better with the Python chardet character encoding identification module.
• Support newer Purple/Pidgin logs using an html format.
• Support thunderbird extreme brokiness in handling the mbox format: naked "^From $" separators are now accepted with mhmboxquirks = tbird.
• A change of method for filtering on directory location makes it much more efficient and faster.
• The utf-8 file name is now a stored field by default.
• We now catch all exceptions in Python filters to avoid crash reports from the system on benign filter failures.
• Indexing now creates a lock/pid file inside the configuration directory.
• Fixed parallel build issue on FreeBSD.