[gnome-flashback] The case for other WMs (was: Re: [gnome-panel] Pass --session-name from desktop files.)

[Philipp Kaluza <floss ghostroute eu>]

Hi Alberts, hi All,

Am 08.12.2013 12:59, schrieb Alberts Muktupāvels:
> >     Who tested this ?
> >
> >     In short, I don't think this should ever have passed the bugzilla
> >     review
> >     - there was a reason Jeremy Bicha and myself both independently
> >     introduced a wrapper script, and that reason is that it is needed
> >     - not
> >     all display managers can pass on parameters in this way. (If memory
> >     serves, GDM3 is one that cannot, and is most certainly a target
> >     DM for us.)
> >
> >     So in short, NAK, I'll revert this
> I tested, but tested only with lightdm. My fault. :(

So it turns out you did not get the patch reviewed _at all_.

This could have easily been caught by following the 4-eyes-principle.

And this is not just a problem for that one patch, it is the case for
multiple of your own patches you pushed yesterday and today.
One of them also broke libpanel-applet API and ABI, without dicussing an
ABI break beforehand here, and also without bumping the version number.

Let me be clear: *This is not okay.*

In this subproject we follow the best practices of the overall Gnome
project, that a lot of sweat and experience went into. (Except
time-based releases, and I hope to get there again eventually.)

What do you hope to achieve by just pushing it ? Force someone to review
it immediately, and after-the-fact ? You can ask/re-ask nicely for a
review on this list, but nobody (me included) magically has more time
for Gnome Flashback just because you create a big mess in master. Force
me to test it all before pushing out 3.9.91 ? All you accomplished is
that I will feel much less confident about the state of master, and will
either need to revert everything comitted by you at once, or postpone
the 3.9.91 release until I have A-LOT-OF-SPARE-TIME™.

If you keep abusing your commit privileges like this, they can and will
be revoked. :-(

> Ok, but that means we need wrapper script for each session? One more
> thing, TryExec is used to determinate if this session should be shown
> or not. Am I correct? If so than should not we change TryExec line for
> at mutter and compiz session to TryExec=mutter and TryExec=compiz. We
> don't want show these sessions if these window managers are not available.
The tryexec thing is a good idea, though you might want to list absolute
paths.

So basically, nobody except Ubuntu cares about compiz at this point.

And we have only very limited support capabilities, so I am not
interested in supporting multiple configurations upstream.

Is anybody here interested in supporting other WMs except
metacity/mutter ? If so, please speak up now, and make the case for
supporting other WMs.


Otherwise, we should make it perfectly clear to downstreams that it is
not OK to ship a "Gnome Flashback (Compiz)" - the Gnome Flashback name
is specifically reserved for a replacement for the old fallback mode.
Let them name it "Gnome + Compiz" or whatever.

About mutter:
I know the mutter codebase is interesting, in that it has already done
the hard work of porting to GTK+ 3. But current mutter cannot run
without OpenGL, which goes somewhat against our initial mission
statement. And I'm not aware if anybody actually regularly runs mutter
without gnome-shell.
Again: if anybody has any experience to share here and is willing to
support it, please make the case for it. Otherwise I'm inclined to veto
this as well.

Cheers
  Philipp

-- 
Philipp Kaluza
Ghostroute IT Consulting