[gst-devel] Overwriting files in sinks

[Stefan Kost]

Hi,

Sebastian Dröge wrote:
> Am Samstag, den 29.09.2007, 13:38 +0300 schrieb Stefan Kost:
>> shouldn't this be done as a GInterface? Then an application can reliably detect
>> that it supports it.
> 
> How would this interface look like? Can interfaces contain signals?
>
Yes, they can.
GstFilesystemSinkIface could be a interface name - dunno what else could fit in
there.

> 
> Am Samstag, den 29.09.2007, 19:38 +0100 schrieb Tim Müller:
>> This issue is also being discussed in
>>
>>   http://bugzilla.gnome.org/show_bug.cgi?id=334290
>>
>> just in case you haven't seen that yet.
> 
> Thanks, didn't know about this one yet :)
> 
> IMHO this should get it's own interface though as GstURIHandler
> is for sources and sinks but this would only be useful for sinks.
> 
>>> For 0.11 this should be changed in gnomevfssink as this way the app has
>>> to know about GnomeVFS unfortunately.
>> In 0.11 gnomevfssink will hopefully have been replaced by giosink.
> 
> Right, would be surprised if gnomevfs was still used at that time :)
> 
>>> Does anybody have better ideas? Otherwise I'll go ahead and implement
>>> this in filesink and giosink.
>> It would IMHO be good if a solution was found that didn't require the
>> application to check the actual type of the element in order to know
>> what the second argument is.
> 
> What about a new signal then that all elements implement with gchar*
> as second parameter? gnomevfssink would only have two signals for the same
> then.
> 
>> Also, emitting a signal is a bit evil, even if it's usually not going to
>> be emitted from a streaming thread in this case, since applications will
>> probably show a confirmation dialog from the signal callback and run a
>> main loop via gtk_dialog_run() or similar, which means there's at least
>> the potential for all kind of reentrancy issues that an application
>> developer is unlikely to expect.  Not sure what a better solution would
>> be though.
> 
> Right, but other options do we have? A GstMessage would be nice but we
> need to get feedback from the application somehow...

A method in the iface for the reply would be nasty :(

Maybe the application should check itself before if the file exists or not?

Stefan