Re: gvfs basic testing

[David Zeuthen <david fubar dk>]

On Tue, 2008-10-14 at 17:29 +0530, CM Reddy wrote:
> Hi 
> I have been working in gvfs for some time. But, I could not find much
> information on the NET. However, I downloaded gvfs-0-2.3 source package
> on my desktop (ununtu) and trying to test some of the gvfs commands to
> familiarize with GVFS architecture. 
> 
> However, I have no pointers on how to go about from this point. I just
> tried mounting the localtest, but it failed with the following error.
> 
> laptop:/usr/local/share/gvfs/mounts$ gvfs-mount localtest://
> Error mounting location: volume doesn't implement mount

We probably need to write some man pages at some point, that's on my
TODO list FWIW. Anyway, typically you need to pass the host part of the
URI if you want to start your backend via gvfs-mount, e.g.

 $ gvfs-mount gphoto2://[usb:001,002]/

Another way is to just start the mount daemon from the command line

 $ ./gvfsd-gphoto2 host="[usb:001,005]"

then you also get debug output to stdout/stderr. This is the preferred
way to do it if you're developing a backend.

> Our goal is to mount our own file system (gvfs) and integrate with our
> rest of the code.

You should be aware that GVfs does not yet support out-of-tree backends
(that's one reason there's no development or files or documentation
installed as part of 'make install'). In other words, assume that the
protocol between the gvfs daemon and the mount daemon can _and_ will
change maybe even in minor releases. Backends in the tree are fine, if
we make a protocol change or whatever, we just fix up all the backends.

At some point I _think_ [1] we want to provide a stable ABI, however at
this point I'm pretty sure we're not ready to make such a commitment.

(Of course, keep in mind I'm only speaking in capacity of a GVfs
contributor, Alex is the maintainer, he might be able to shed some more
light on if and when.)

> Provide me some pointers for good documentation in GVFS, which will
> save my time in developing our own back-end program.

There's no documentation of the backend API other than the code - I
believe fixing that (by providing gtk-doc docs, tutorials, overview-ish
documents) is something that needs to happen before we can declare the
ABI stable. Send patches!

That said, the existing backends in the tree should provide a lot of
hints about how to write a GVfs backend and they are licensed under a
pretty liberal license (LGPLv2+) so if your backend uses a similar
license you should be able to get pretty far in short time using
copy-paste.

Hope this helps.

     David

[1] : but of course it's up to Alex if he wants to provide a stable ABI.
Just FWIW, I'd be 100% supportive if Alex don't want to provide one;
maintaining an ABI really sucks.