Debug Buildslave Maintenance

[Colin Finck]

Hello all,

The Debug Buildslave will be down for maintenance throughout the next 
week. We hope to have it fully working again by next weekend.

In the process of this maintenance, the OS will be reinstalled to match 
the one we already use on our other servers. Besides, more of our 
administrators will get access to the machine, so that further problems 
can be debugged more easily.

- Colin

[Timo Kreuzer]

Hi,

What about fixing release buildbot first?
And is there a chance, maybe in the future, to switch buildbots more 
easily, to have a backup solution running?
One week without debug builds is a serious thing.

Timo

Am 19.08.2011 21:43, schrieb Colin Finck:
> Hello all,
>
> The Debug Buildslave will be down for maintenance throughout the next 
> week. We hope to have it fully working again by next weekend.
>
> In the process of this maintenance, the OS will be reinstalled to 
> match the one we already use on our other servers. Besides, more of 
> our administrators will get access to the machine, so that further 
> problems can be debugged more easily.
> Colin
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

[Pierre Schweitzer]

Inlined.

Timo Kreuzer [e-mail] wrote on Sat, August 20th, 2011, 10:25 AM:
> Hi,

Hi
> What about fixing release buildbot first?

AFAIK, this is not doable yet.
> And is there a chance, maybe in the future, to switch buildbots more 
> easily, to have a backup solution running?
> One week without debug builds is a serious thing.

Given the new way debug build slave will be administrated, having one week without debug builds will be impossible: simple configuration, monitored host & services, several administrators.
> Timo

Regards,
Pierre

What about Testslave? Lack of KVM is currently the worst thing.

Best regards

On Sat, 20 Aug 2011 13:32 +0200, "Pierre Schweitzer" [e-mail] wrote:
> Inlined.
>
> Timo Kreuzer [e-mail] wrote on Sat, August 20th, 2011, 10:25
> AM:
> > Hi,
>
> Hi
> > What about fixing release buildbot first?
>
> AFAIK, this is not doable yet.
> > And is there a chance, maybe in the future, to switch buildbots more 
> > easily, to have a backup solution running?
> > One week without debug builds is a serious thing.
>
> Given the new way debug build slave will be administrated, having one
> week without debug builds will be impossible: simple configuration,
> monitored host & services, several administrators.
> > Timo
>
> Regards,
> Pierre
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev
-- 
With best regards
Caemyr

[Colin Finck]

[e-mail] wrote:
> What about Testslave?

It's still the same machine as Debug slave, so it's the thing we 
actually try to fix.

- Colin

[Colin Finck]

Timo Kreuzer [e-mail] wrote:
> What about fixing release buildbot first?

It's still a private machine owned and administered solely by Christoph. 
As long as I can't even reach him by phone, we can only wait.
> And is there a chance, maybe in the future, to switch buildbots more
> easily, to have a backup solution running?

First of all, we hope that we can get the same reliability as our other 
servers after reinstalling the server OS, giving more ReactOS admins 
access to the machine and adding monitoring.
The current OS was meant to be reinstalled for quite a long time, but up 
to now, nobody with physical access to the machine had any time to do it.

If such problems continue to exist afterwards, we can try to set up a 
fallback system, but this would require an equally configured and 
powerful Linux machine first.
> One week without debug builds is a serious thing.

I hope you're aware that builds are still properly uploaded, it's only 
the testing step which fails.

- Colin

[Pierre Schweitzer]

Hi,

finally.
Colin and I are pleased to announce you that ReactOS Linux KVM tests are back online and working. You can find the first tests results (on r53383) sent tonight by the testbot on testman: http://www.reactos.org/testman/.

Regards,
Pierre.

ReactOS Development List [e-mail] wrote on Sat, August 20th, 2011, 4:24 PM:
> Timo Kreuzer [e-mail] wrote:
> > What about fixing release buildbot first?
>
> It's still a private machine owned and administered solely by Christoph. 
> As long as I can't even reach him by phone, we can only wait.
> > And is there a chance, maybe in the future, to switch buildbots more
> > easily, to have a backup solution running?
>
> First of all, we hope that we can get the same reliability as our other 
> servers after reinstalling the server OS, giving more ReactOS admins 
> access to the machine and adding monitoring.
> The current OS was meant to be reinstalled for quite a long time, but up 
> to now, nobody with physical access to the machine had any time to do it.
>
> If such problems continue to exist afterwards, we can try to set up a 
> fallback system, but this would require an equally configured and 
> powerful Linux machine first.
> > One week without debug builds is a serious thing.
>
> I hope you're aware that builds are still properly uploaded, it's only 
> the testing step which fails.
> Colin
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

@Eric

http://reactos.org/testman/compare.php?ids=7429,7579
http://reactos.org/testman/detail.php?id=2639952

Please find that the crash is not CMake related, and CMake build is not inherently broken.

On Thu, 25 Aug 2011 23:03 +0200, "Pierre Schweitzer" [e-mail] wrote:
> Hi,
>
> finally.
> Colin and I are pleased to announce you that ReactOS Linux KVM tests are
> back online and working. You can find the first tests results (on r53383)
> sent tonight by the testbot on testman: http://www.reactos.org/testman/.
>
> Regards,
> Pierre.
>
> ReactOS Development List [e-mail] wrote on Sat, August 20th,
> 2011, 4:24 PM:
> > Timo Kreuzer [e-mail] wrote:
> > > What about fixing release buildbot first?
> >
> > It's still a private machine owned and administered solely by Christoph. 
> > As long as I can't even reach him by phone, we can only wait.
> > > And is there a chance, maybe in the future, to switch buildbots more
> > > easily, to have a backup solution running?
> >
> > First of all, we hope that we can get the same reliability as our other 
> > servers after reinstalling the server OS, giving more ReactOS admins 
> > access to the machine and adding monitoring.
> > The current OS was meant to be reinstalled for quite a long time, but up 
> > to now, nobody with physical access to the machine had any time to do it.
> >
> > If such problems continue to exist afterwards, we can try to set up a 
> > fallback system, but this would require an equally configured and 
> > powerful Linux machine first.
> > > One week without debug builds is a serious thing.
> >
> > I hope you're aware that builds are still properly uploaded, it's only 
> > the testing step which fails.
> > Colin
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://www.reactos.org/mailman/listinfo/ros-dev
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

With best regards
Caemyr

[Eric Kohl]

Hi Olaf,

I am already investigating the services test issues. One of my first 
findings is that the current widl seems to mess up value ranges in some 
cases. This is the cause of the following errors:
  err:(dll/win32/rpcrt4/ndr_marshall.c:6496) value exceeded bounds: 918, 
low: 0, high: 514

<rant>
It is pretty annoying that jgardou updated rpcrt4, widl and other 
components without proper testing. At least he should have posted a bug 
list BEFORE he comitted the new stuff.
</rant>

Regards
Eric
> @Eric
>
> http://reactos.org/testman/compare.php?ids=7429,7579
> http://reactos.org/testman/detail.php?id=2639952
>
> Please find that the crash is not CMake related, and CMake build is not inherently broken.

[Aleksey Bragin]

On Aug 26, 2011, at 3:01 AM, Eric Kohl wrote:
> <rant>
> It is pretty annoying that jgardou updated rpcrt4, widl and other  
> components without proper testing. At least he should have posted a  
> bug list BEFORE he comitted the new stuff.
> </rant>

I also ranted about that. In fact that's why I stopped syncing rpcrt4  
some time ago - because new version always gave problems. So I wanted  
to solve problems first and only then commit.

But, OK, as Olaf said - if something needs to be done, let's simply  
do that and fix everything which broke ;).

WBR,
Aleksey.