Re: [Banshee-List] GIT Build has stopped playing:

["Andrés G. Aragoneses" <knocte gmail com>]

Andrés G. Aragoneses escribió:
> Alex Bennee escribió:
>> 2009/11/14 Bertrand Lorentz <bertrand lorentz gmail com>:
>>> On Sat, 2009-11-14 at 14:29 +0000, Alex Bennee wrote:
>>>> I keep getting this:
>>>>
>>>>
>>>> Any ideas?
>>> What are you tring to do ? Play a Last.fm station I guess ?
>>> In that case, are the 5 tracks displayed in the track list ?
>> Yes. And I'm a paid subscriber.
> 
> Ok let's stick with the library problem and not worry about LastFM until
> we fix that one.
> 
> 
>>> Are you able to play files from your library ?
>> Uhh, no, it looks like all playback is broken. Although when trying to
>> play a library file I get an "Object is Null" dialog box.
> 
> Mmm, then I guess you're also being affected by this bug:
> 
> https://bugzilla.gnome.org/show_bug.cgi?id=598223
> 
> Can you confirm?
> 
> Also it would be nice if you could try this workaround:
> 
> 1) Backup your database (copy it to another location).
> 2) With Banshee closed, delete the database so as next time you start it
> it doesn't find the data.
> 3) Import your files again.
> 
> If now playback works, we would really appreciate that you send us (as
> Bertrand also requests on the bug) the problematic database so we can
> determine the cause of the problem.
> 
> That being said, I would like to chat with the rest of the developers
> about this. Clearly what is happening here is that there is some code
> path in some migration step that was not tested as much as we wanted to.
> I'm now wondering if the fast-failing methodology would have helped here
> and would have let us detect the bug much sooner, just by removing the
> some try-catch blocks. Sure, this would make the first user that has a
> problem receive a very ugly error, but that would make him report a very
> clear problem instead of side-effects of a bigger problem which is what
> we are getting right now.
> 
> In particular, in the file BansheeDbFormatMigrator.cs I think we should
> remove every try-catch that has an empty catch block. And, in respect to
> this concrete bug, I believe the problem may come from the method
> Migrate_31, which converts relative URIs to absolute, and has a
> try-catch which just sends the error to the console. How about removing
> that try-catch as well, so the user would not be able to run Banshee and
> this way he would report the problem much faster?

Maybe I spoke too soon. Not sure if the Hyena.Log.Exception () method is
actually sending the error to a dialog box instead of just the console.
However, regardless of this question, the exception should be re-thrown
so as the migration operation can rollback right?

	Andrés

--