Re: [Tracker] Fix for segfault

[Jamie McCracken <jamiemcc blueyonder co uk>]

Marcus Fritzsch wrote:
> On 10/5/06, Jamie McCracken <jamiemcc blueyonder co uk> wrote:
> > Marcus Fritzsch wrote:
> > > Hello there.
> > >
> > > With the new tracker I always encountered a segfault with some db code.
> > >
> > >    tracker_db_get_field_def
> > >
> > > can return NULL but
> > >
> > >    tracker_metadata_is_date
> > >
> > > in src/trackerd/tracker-db.c:591 dereferences a field in this
> > > structure w/o checking for NULL.
> > >
> > >
> >
> > thanks have fixed in cvs
>
> When thinking about it a second time --- is that behaviour really ok
> that way? It does prevent the segfault, but it is not a 'sane'
> behaviour i think. The exact log context from such an assertion
> failure is the following:

yes it is right beacuse it indicates a potentially serious bug in one of 
the new metadata extractors!

The segfault or log message would only ever occur if one of the 
extractors was using an invalid metadata name - we need to track these 
down!

Any idea what the mime of the image file was?


> ------snap------
>
> The really disturbing 'Invalid UTF-8' comes from a library call? From
> the UTF-8 conversion? I have not looked in too deep with this, I am
> not sure if this will cause some problems later on.
>
> So, the question is, are there any complications from this UTF-8
> thing? Is this known to someone?

all metadata should be converted to utf8 (from users locales) and 
failing that it must be discarded - this is something that needs to be 
added to the new extractors.

Patches welcome for above two :)

--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/