sound problems

[Kevin Kofler]

Suren Karapetyan wrote:
> Everything was fine, BUT there was no sound.
> He plays with it a few days and switches to something which has sound.

If he is knowledgeable about Fedora, he doesn't do that, because he knows
Fedora gets version updates and thus a fix should be forthcoming fairly
quickly.

> Jane installs F10 and is quite satisfied until the day security update
> breaks sound (the same update fixes sound for Bob, but he'll not know it
> until F11,  cause he is sure F10 is broken).
> She has good education and is computer-literate, so she goes to
> bugzilla, reports her problems.
> Waits for a few days and gets nothing.
> Then she calls Bob and asks what he did to fix his problem.

No, she just uses the old working kernel as long as no new working one comes
out. Kernels can be installed in parallel (and you can set how many kernels
should be kept at a time in yum.conf, I recommend just setting
installonly=0, i.e. no limit, and cleaning up by hand once you're sure you
don't need the old kernels anymore). For example, I skipped all the 2.6.26
kernels on my laptop because iwl4965 wireless was broken, the 2.6.27
updates fixed that, I just stayed on 2.6.25 until those came out.

Regressions are definitely an annoyance, but they're not the unsolvable
problem you make them out to be. They're better than unfixed bugs, because
with regressions, at least you have the option to revert to the old
version, with unfixed bugs, there's no working version to revert to.

> And now to Your point.
> I agree: downgrading is wrong.
> But upgrading was even wrongER.
> Just look at https://admin.fedoraproject.org/updates/F10/FEDORA-2008-11593
> The new kernel spent ~30 hours in updates-testing.
> While it's quite normal for a security fix it's unacceptable for such a
> huge change as ALSA upgrade.

I agree that the mixing of this upgrade with the security fixes was
unfortunate from a testing standpoint and that unfortunately this is not an
isolated incident with kernel updates (it's pretty much unavoidable though:
there are lots of changes to the kernel, some of which are security fixes,
some which aren't, so the security update will always also include
non-security stuff), but reverting the ALSA upgrade is still not the proper
way to fix this, fixing the regressions is.

Now if you think ALSA shouldn't get upgraded at all during a stable release,
you're using the wrong distribution.

> And a question to You.
> If I become a fedora maintainer and (in some way manage to) upgrade F10
> to kernel-2.6.29 are we going to revert my update or try to fix all the
> regressions which will take ~2 months.

Nobody is going to upgrade F10 to 2.6.29 now. The kernel has locked ACLs to
prevent you from doing that. ;-) Those people on the ACL know very well
that 2.6.29 is not ready for production use at all yet. But if you mean
2.6.28:

> P.S. 2.6.28 fixes suspend-wakeup issues for my notebook, but I'm not
> asking anyone to upgrade F10 to it.

F10 _will_ be upgraded to 2.6.28, and in fact this is likely to happen very
soon. (And 2.6.29 will be pushed too, when it's ready, which is of course
not now.) Fedora does push out new versions of some packages as updates,
and the kernel is one of those packages. If that's not OK for you, then
Fedora is not for you.

And once this upgrade is out, you can be fairly sure it will *not* be
reverted. Reverting it would be highly impractical for several reasons, in
particular it would require introducing an Epoch in the kernel package and
it would reintroduce all the bugs fixed by the update (which would also be
a nightmare from a bug triaging standpoint). The "try to fix all the
regressions" method is how *all* kernel updates in the past have been
handled.

        Kevin Kofler