Re: [Qemu-devel] [PATCH] [RESEND2] Qemu unmaintained?

[Avi Kivity <avi@redhat.com>]

On 09/10/2009 10:31 PM, Anthony Liguori wrote:
> Avi Kivity wrote:
>> For qemu.git I'd agree since it's undergoing a lot of churn.  
>> Unfortunately it also feeds qemu-kvm.git which I try very hard to 
>> keep regression-free (and finding and fixing regressions during a 
>> merge is quite horrible), so I'd really appreciate it if qemu.git 
>> quality didn't deteriorate.
>
> Or more accurately, you'd prefer if there were no bugs in qemu so that 
> you only had to deal with the bugs that were introduced in qemu-kvm.

Yes.  Anyone who pulls qemu.git and develops or uses it would agree.

>
> That would be nice for you of course :-)  But it's unrealistic to 
> compare the two.
> $ git log --since='1 Month ago' --no-merges qemu-kvm/master 
> --committer='Avi Kivity' --pretty=format:'%an' | wc -l
> 5
> $ git log --since='1 Month ago' --no-merges qemu-kvm/master 
> --committer='Marcelo Tosatti' --pretty=format:'%an' | wc -l
> 5
>
> $ git log --since='1 Month ago' --no-merges origin/master 
> --committer='Anthony Liguori' --pretty=format:'%an' | wc -l
> 251
>
> So there are going to be at least 25x more regressions introduced from 
> upstream qemu than what are introduced in qemu-kvm.

Well, if we define a regression as something that blocks me from pushing 
(kvm-autotest), then both numbers are zero.  Of course qemu.git is not 
run purely for my enjoyment, but a large number of patches are targeted 
at qemu-kvm.git.  Further, the tests that I run (installing and booting 
some popular OSes) are really the basic minimum functionality, nothing 
fancy there.

>
>> (and we're quite far from catching every regression btw).
>>
>> Anthony, how long are your test cycles?
>
> Depends on the number of regressions.  I can usually get through 
> testing in 4-5 hours when everything works.  Everything usually 
> doesn't work though.

We definitely need to improve this and the 80% reject rate.  Can you 
start being a lot noisier about rejects?  that will at least give some 
visibility into the problem.

-- 
error compiling committee.c: too many arguments to function