1. 09 Jan, 2022 1 commit
  2. 04 Jan, 2022 1 commit
  3. 03 Jan, 2022 1 commit
    • Jan Tojnar's avatar
      cli: Filter out hidden files from usage · 8c28c24e
      Jan Tojnar authored
      Nix replaces programs with shell scripts wrappers that execute
      the original files (renamed to .${original_program_name}-wrapped).
      We do not want those files to be listed in help page.
      8c28c24e
  4. 24 Dec, 2021 3 commits
  5. 23 Dec, 2021 4 commits
    • Carlos Garnacho's avatar
      libtracker-data: Flush buffer before resolving INSERT/DELETE/WHERE solution · 26b3633e
      Carlos Garnacho authored
      The triple operations left to be flushed might influence the solution given
      by the WHERE clause of the query being handled, ensure these operations
      turn into SQL commands before we get the WHERE clause solution.
      26b3633e
    • Carlos Garnacho's avatar
      libtracker-data: Honor out variable in early return path · e53487d4
      Carlos Garnacho authored
      When handling the insertion of an individual TrackerResource property that
      points to another TrackerResource, we check if the latter resource is
      already inserted/known and return early in update_resource_single(), but
      we forget to set the "id" out variable, thus go with an uninitialized
      value here.
      
      Shuffle the early return, so the resource ID is already known, and correctly
      returned.
      e53487d4
    • Carlos Garnacho's avatar
      libtracker-data: Handle SQLITE_LOCKED_SHAREDCACHE in the update thread · 1176fcb3
      Carlos Garnacho authored
      Handling this while executing update statements turns out necessary in the
      case of in-memory databases, as it seems the (single) writing thread can
      get this SQLITE_LOCKED extended error if there are already existing read
      locks. This can be seen (or not, it's quite racy) running
      "tracker-miner-fs-3 --dry-run" so it uses an in-memory database.
      
      If this condition happens, we simply retry until we get something else
      than SQLITE_LOCKED (i.e. all readers finished). Handling of
      this error condition could get fancier (e.g. via sqlite3_unlock_notify()),
      but this is already enough to avoid the ill effects.
      
      In order to ensure that this only applies to in-memory connections, check
      that we specifically get SQLITE_LOCKED_SHAREDCACHE, since we only use shared
      caches for in-memory databases. Getting SQLITE_LOCKED in any other
      circumstance is still unexpected and is an error situation, most notably
      WAL databases in disk are not affected by this situation.
      1176fcb3
    • Pekka Vuorela's avatar
  6. 12 Dec, 2021 25 commits
  7. 07 Dec, 2021 1 commit
  8. 05 Dec, 2021 3 commits
  9. 21 Nov, 2021 1 commit