Re: [Tracker] Review request, bug fixed by a downstream integrator (Jolla)

[Carlos Garnacho <carlosg gnome org>]

Hey Philip,

Sorry I missed this...

On Tue, Jul 8, 2014 at 9:48 AM, Philip Van Hoof <philip codeminded be> wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi guys,

As you know I'm tailing the Tracker package for the Jolla device and I
noticed that there a bugfix was made by Richard Braakman for a change
that happened recently:

https://github.com/nemomobile-packages/tracker/pull/29

Commit here:
https://github.com/amtep/tracker/commit/419d680619b5e0d4f3ae308a087dd313e8ce252e

I see how spurious ::file-deleted might happen, the TrackerFileSystem caches data for directories, so if a reindex happens on a directory that happened to preserve data, tracker_file_system_traverse() would recurse deep in the old tree while the crawling phase only updated the most direct children at that time.

I however think a better fix would be to add a max_depth parameter to tracker_file_system_traverse(), and pass the same depth that's given on tracker_crawler_start() (and implicitly performed in queries). AFAICS the approach in the patch would still fail for files deeper than MAX_DEPTH, as those won't be yet in the pending_files array. TrackerFileSystem is private, so API changes are just fine there.

 

Cheers,

  Carlos