Re: F-spot network sync problems

["John Carr" <john carr unrouted co uk>]

On Tue, Aug 5, 2008 at 8:05 PM, Olof Sjobergh <olofsj gmail com> wrote:
> Hi again,
>
> Thanks for your reply. I've looked a bit more at it today.
>
>> John is busy refactoring our file handling code to make use of GIO.
>> You should be able to patch the tempfile set the correct mtime, but
>> make sure you do so against trunk or its all in vain. (Its worth
>> checking with John if he has more GIO merges pending, im sure he will
>> reply shortly... after waking up). There should be code showing how
>> this works in File.py already.
>>
>> Renaming the temp file is a little trickier, but if you are confident
>> you won't encounter name collisons you should be able to do this too.
>> Again, im certain an example is in there already.
>
> The methods force_new_mtime and force_new_filename defers the change
> for  tempfiles, but then the F-spot importer imports the file before these
> deferred changes are applied. As there might be name collisions I
> understand this. But could it be possible to create subdirectories in
> the tempdir to avoid collisions and be able to rename the files?

You could create tempdirs to avoid this. I'm also wondering if you can
just adjust the interface so that we can send these details over dbus
when passing the file to f-spot.

> Another thing i noticed is that the network sync first copies the file
> to a local tempfile named something like "netsyncVDPrxj", and
> then this file is copied again to a tempfile "conduitpIXAGs". Is there
> a reason for this? Seems a bit unnecessary to me, but I haven't looked
> much at the network sync module.

I think this is just an accidental result of using layers of abstraction.

>> For the third question, you have no choice but to cook up an f-spot
>> patch. The dbus addin maps useful f-spot database model functionality
>> to dbus for us to use to sync f-spot, and is available here:
>>
>> http://svn.gnome.org/svn/f-spot/trunk/extensions/DBusService/
>>
>> I'm sure you'll be able to figure out what is needed, and that Thomas
>> will help you.
>
> I'll try to have a look at this later on.
>
> I was also thinking about adding support for F-spot versions of photos.
> It would be nice to have them synced as well. What happens with versions
> now? Do they show up as different photos, or is only the latest version
> synced?
>
> Best regards,
>
> Olof Sjöbergh

Oh wow, if you do this I will definitely owe you beer. My dream is
that Conduit can sync f-spot together seamlessly with a full database
history... so this is pretty much a requirement. As far as I know, you
will definitely need to enhance the f-spot dbus interface. At the
moment I think we just think about the latest photo.

John