Debug-ability; getting gnome-panel 3.8 out

[Philipp Kaluza]

Dear Co-Maintainers and other Gnome Flashback stakeholders,

let me start with some good news:
I'm writing this from within a running gnome-flashback session.

It took me quite a while, but I've *finally* made enough progress
building gnome-panel master against a current sid (Debian unstable)
environment to feel that this is going somewhere.

Let me be clear: the debug-ability of Gnome Flashback currently stinks.
The sessions might not appear in gdm, appear but don't get started, or
get started and then show the fail screen. The reason is that gdm is a
system-wide component, and we need to integrate that. And on the other
side gnome-session is a bit of a moving target (it or the components it
tries to launch lost some functionality in the 3.8 timeframe, and
presumably more in 3.10) - and we need to integrate with that as well.

Jhbuilding does not give the required system integration (even when
fiddling with lots of symlinks and session files), building on a Debian
system brought its own set of challanges (version-wise; there was a
freeze...), and testing on an Ubuntu system was no easier. (Fedora might
have helped, but what little time I had there I spent fighting^Wgetting
to know systemd.)

But it's important that interested people can get something basic
running again, or soonish we can declare the whole flashback part of the
stack bitcompost.

So, what can we do about this ?

1) Better logging when searching for errors:
To this extend, I just pushed commit 39b6b48, which adds a wrapper
script named (surprise) gnome-session-flashback. We need a wrapper
script anyhow; if a ~/.xsession-errors-flashback exists, we tee
gnome-sessions STDERR to it. (The normal .xsession-errors did for some
reason not log reliably for me.)

2) Get something into distributions - which means, release a tarball.

There _are_ regressions (nautilus is deactivated currently due to
#695088, most probably others…), but at this point it's more important
to release *something*, so we can get feedback from distributions and
others.

>From an upstream POV, I just want to get a 3.8.0 tarball out (and branch
for 3.8), before the final 3.10 stuff starts to hit gnome ftp.

>From a Debian (and I guess Ubuntu) POV, we need to have a story on how
to deal with approaching influx of 3.8 packages.

2b) It's also conceivable to work on compatibility with 3.10 ASAP, and
jump directly to releasing a gnome-panel 3.10 shortly after gnome 3.10
next month. As I have not even started build-testing gnome-panel in a
3.10 environment yet, I would only consider this path if I felt there
was some strong support from somebody else here for such a path.

@Jonathan, @Josselin: I would like to hear your experiences and
especially your needs wrt. releasing the Debian and Ubuntu packages.
(AFAICT there is no PPA yet with gnome-panel 3.7, right ?) Jeremy pushed
some of his Ubuntu patches last, but indicated to me that he no longer
has time to work on this.

@Everybody: any bugs (preferably with patches) you consider blockers for
a mini-release as described above ?

Thank you for your timely feedback. :-)

Cheers,
Philipp

-- 
Philipp Kaluza
Ghostroute IT Consulting