Re: GNOME 2.2 screenshots

[Havoc Pennington <hp redhat com>]

Jeff Waugh <jdub perkypants org> writes: 
> The autotool versions required for GStreamer present an interesting
> technical problem for us in the short term,

So here is how we would ideally move to new automake:
 - all our current scripts would refer to the command "automake-1.4"
   instead of just "automake"
 - any module that wanted could just change their scripts to 
   refer to "automake-1.6" instead
 - people just install both automakes

I'm not sure if that works with upstream automake tarballs yet -
cc'ing Tom. We would need a release of automake 1.4 with the
"automake-1.4" name in it.

If it does work with upstream tarballs, then there's just no problem.
We go through and change all instances of "automake" to either
"automake-1.4" or "automake-1.6", and stuff just works. Modulo
autoconf also working I guess.

Ideally the automake-1.6 package would not even contain a thing called
simply "automake" since it just encourages bugs of the form "I got the
wrong automake" - and the compile failures from this are really hard
to sort out, in my experience.

If we can't do parallel install, we can't migrate to new automake
module-by-module. That means that as soon as we move to 1.6 for HEAD,
then people can't work on the stable branch anymore, unless they have
two totally different build prefixes, with their own copies of
autotools. I don't even have that much disk space free on my computer
at work...

Without the parallel install we're also going to have to endure a
month of everything being broken all the time, since we have to try to
get everyone to migrate at once.

Personally I'm inclined to back gstreamer down to 1.4 rather than go
through the mess. Maybe we can do new automake for GNOME 2.4. ;-) I
have yet to hear what new automake buys us that warrants the enormous
world of migration pain.

> and the decision to require GStreamer at the libgnome level (as
> suggested by many, and patched by Jorn) presents an interesting
> socio-political one. ;-)

My personal belief is that it's too soon for that.  There's ongoing
discussion of this for the interoperability hothouse at LWE for
example.

Plus I just have a general feeling that we should get more de facto
adoption of the API before we plop it in at that low level.

I'm sure we can do a dlopen() module hack to use gstreamer to play
beep noises and such, right?

Havoc