Re: ARC & GNOME [Was: How we make decisions...]

[Havoc Pennington <hp redhat com>]

On Tue, 2004-11-30 at 15:14 +0000, Ghee Teo wrote:
> >   
>     I see leaving it to QA have 2 problems:
>     - more complete QA is generally done after the GNOME release has
> been adopted by the various distros.
>     - lot of these breakages are only exposed when users upgrades from
> one release to the next.
>       For example the gnome-panel patch which Mark has nicely fixed.
>          http://mail.gnome.org/archives/desktop-devel-list/2004-
> November/msg00318.html
>         That bug was really been raised after 2 releases later.

Right, by "QA" I mean "QA during the devel process," not "QA after
release" - what I'm saying is that if one of us companies really decides
that these problems are important, we will allocate a few people to be
testing the devel tree in tinderbox style (nightly test suites, etc.)
and also run manual QA on the devel tree.

>     If somewhere alone here we can come up with a checklist where the
> developers  can go through before
> they release a new module or version such as:
> 
> - Have I changed my library version?
> - Have I created/removed any binary?
> - Have I created/removed any configuration file?
> - Have I added/changed any gconf key?
>     ... so on
>     I am sure others can come up with a longer relevant yet simple
> list.

This would be helpful, but with hundreds of contributors some of them
will still screw this up. If we really want to solve the problem then we
need ongoing tinderbox-style verification of the devel tree, so when an
interface breaks a bug is filed in a matter of days.

Havoc