https://bugs.kde.org/show_bug.cgi?id=400498
--- Comment #12 from Matt Fagnani <matthew.fagn...@utoronto.ca> --- (In reply to Stefan Brüns from comment #11) > (In reply to Matt Fagnani from comment #10) > > I read in another post that the failed assertion might indicate a corrupted > > baloo index which seems likely as the crashes occur immediately on starting > > baloo_file or balooctl start. > > In this case its a fault completely inside lmdb. It keeps a list of page no > longer used (i.e. containing stale data), the freelist, and reuses these > pages when writing new data. lmdb apparently has problems managing this list > under some circumstances, and while one issue was fixed, there seem to be > another one. > > Can you please verify you are *really* using the fixed liblmdb package, i.e. > it has not been reverted locally on your system or in the fedora repo? > sudo dnf repoquery lmdb-libs --installed shows lmdb-libs-0:0.9.22-4.fc29.i686 which had the patch suggested applied as shown at https://src.fedoraproject.org/rpms/lmdb/c/8008a8dff332a1faea6be2763eef01e7899c32e5?branch=f29 https://koji.fedoraproject.org/koji/buildinfo?buildID=1160101 > > Some change to how lmdb-libs handles the > > exception from the failed assertion around mdb_assert_fail at mdb.c:1542 or > > mdb_page_dirty at mdb.c:2128 so that the abort signal isn't made might > > prevent the crashes at least. If baloo_file were able to handle that > > exception from the failed assertion, then it might be able to fix or > > recreate the index. > > As the signal never reaches baloo, it has no chance of handling it. lmdb > notices something is wrong and commits suicide, unfortunately taking down > everything else with it ... > I guess that changing how mdb.c of lmdb-libs handles the exception would be necessary then maybe by changing the signal made in mdb_assert_fail at mdb.c:1542. I ran ls -la ~/.local/share/baloo/ total 21096 drwxrwxr-x. 2 matt matt 4096 Nov 9 10:04 . drwxr-xr-x. 78 matt matt 4096 Nov 7 21:34 .. -rw-rw-r--. 1 matt matt 21585920 Feb 16 2018 index -rw-r-----. 1 matt matt 8192 Nov 9 10:32 index-lock Since ~/.local/share/baloo/index hadn't been updated since Feb 16, I guessed that it might've been corrupted or something similar. I ran mv ~/.local/share/baloo/index ~/.local/share/baloo/index.old balooctl start The baloo_file and baloo_file_extractor started running without crashing. ls -la ~/.local/share/baloo/ now shows total 23288 drwxrwxr-x. 2 matt matt 4096 Nov 9 10:42 . drwxr-xr-x. 78 matt matt 4096 Nov 7 21:34 .. -rw-r-----. 1 matt matt 2244608 Nov 9 10:42 index -rw-r-----. 1 matt matt 8192 Nov 9 11:13 index-lock -rw-rw-r--. 1 matt matt 21585920 Feb 16 2018 index.old The crashes might've been due to how mdb.c of lmdb-libs was handling the likely corrupted ~/.local/share/baloo/index -- You are receiving this mail because: You are watching all bug changes.
