Panic due to double free was Re: Fwd: BUILD of samba_4_0_test:master BROKEN on SerNet-NetBSD5 with gcc AT REVISION ea510ae16cc7103bb001469b644415b2867ff347

[Kamen Mazdrashki]

Hi Mat,

On Sat, Dec 18, 2010 at 11:26, Matthieu Patou <mat at samba.org> wrote:
> Hello Kamen !
>
>
>> Hi,
>>
>> It turns out it is all my fault for this segfault and unstable work 'make
>> test'
>> execution. I think I've fixed it properly now.
>>
> Great (not you break hein ;-)
>>
>> @Jelmer: Actually this was a regression. Sorry for I've misslead you!
>>
>> I appologies to all of you for introducing this temporary instability!
>> I should have tested it even more thoroughly than I did :(
>
> Does it means that samba4.rpc.samr.large-dc.one(vampire_dc)
> andsamba4.rpc.samr.large-dc.two(vampire_dc) are fixed ?
>
I am afraid I can't claim this test is fixed. I haven't checked why we
receive an
originating update on a deleted object
(this is the "Failed to apply records: (null): No such object" message
in the log).
I still suspect this is why we can't allocate RIDs in timely manner.

> If so can I suggest you to have a look at samba4.drs.fsmo.python(vampire_dc)
>
> It's a flaky test that really poison the result of the build on the build
> farm (and so generate false alert that leads to people ignoring alert
> emails). Matthieu.
>
Yes, this is what I wanted to look into next. I hope I will have something in
next few days.

Thanks for you persistence in chasing those type of flaky tests!

-- 
CU,
Kamen