OBS and PPC64 versus PPC64LE ... sometimes build failures

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

OBS and PPC64 versus PPC64LE ... sometimes build failures

Dr. Werner Fink
Hi,

I've some trouble with building mksh on PPC64 and PPC64LE. Very often
it fails on PPC64 fails with:

  FAIL ./check.t:heredoc-tmpfile-5
         Description:
                 Check that heredoc temp files aren't removed too soon or too late.
                 Backgrounded subshell command with here doc
         unexpected stdout - first difference: line 1, char 1 (wanted 'A', got 'L'
         wanted:
                 A
                 hi
                 B
                 Left overs: *
         got:
                 Left overs: *
                 A
                 hi
                 B

sometimes it works perfect. Also I see sometimes a similar defect on PPC64LE at
exactly the same check of the test suite of mksh.

All build systems seems to be based on qemu/kvm, nevertheless there seem to be
some differences but how to detect those (virt-what does report nothing).
Maybe it depends on the real host on which qemu-system-ppc64 or qemu-system-ppc64le
will be executed.  It could be a defect of qemu based on the hosting system.

Do we have some real PPC64 and/or PPC64LE irons in our build farm or all hosts
based on qemu/kvm?

Werner

--
  "Having a smoking section in a restaurant is like having
          a peeing section in a swimming pool." -- Edward Burr

signature.asc (911 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

Adrian Schröter
On Freitag, 24. November 2017, 07:59:50 CET wrote Dr. Werner Fink:

> Hi,
>
> I've some trouble with building mksh on PPC64 and PPC64LE. Very often
> it fails on PPC64 fails with:
>
>   FAIL ./check.t:heredoc-tmpfile-5
>          Description:
>                  Check that heredoc temp files aren't removed too soon or too late.
>                  Backgrounded subshell command with here doc
>          unexpected stdout - first difference: line 1, char 1 (wanted 'A', got 'L'
>          wanted:
>                  A
>                  hi
>                  B
>                  Left overs: *
>          got:
>                  Left overs: *
>                  A
>                  hi
>                  B
>
> sometimes it works perfect. Also I see sometimes a similar defect on PPC64LE at
> exactly the same check of the test suite of mksh.
>
> All build systems seems to be based on qemu/kvm, nevertheless there seem to be
> some differences but how to detect those (virt-what does report nothing).
> Maybe it depends on the real host on which qemu-system-ppc64 or qemu-system-ppc64le
> will be executed.  It could be a defect of qemu based on the hosting system.
>
> Do we have some real PPC64 and/or PPC64LE irons in our build farm or all hosts
> based on qemu/kvm?

all builds are chained in KVM.

You may check the "osc jobhistory" to see if you can see a pattern which host
works and which not.

--

Adrian Schroeter
email: [hidden email]

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
 
Maxfeldstraße 5                        
90409 Nürnberg
Germany


--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

John Paul Adrian Glaubitz
In reply to this post by Dr. Werner Fink
Hi Werner!

On 11/24/2017 07:59 AM,  Dr. Werner Fink  wrote:

> I've some trouble with building mksh on PPC64 and PPC64LE. Very often
> it fails on PPC64 fails with:
> (snip)
> sometimes it works perfect. Also I see sometimes a similar defect on PPC64LE at
> exactly the same check of the test suite of mksh.
>
> All build systems seems to be based on qemu/kvm, nevertheless there seem to be
> some differences but how to detect those (virt-what does report nothing).
> Maybe it depends on the real host on which qemu-system-ppc64 or qemu-system-ppc64le
> will be executed.  It could be a defect of qemu based on the hosting system.

This could actually be true provided that OBS actually builds on QEMU and not
really iron. Looking at the build logs for mksh for ppc64 [1] and ppc64el [2]
in Debian, I don't see this particular testsuite failure in any of the builds
in the history.

And, knowing Thorsten Glaser, the upstream developer, personally, I am very
confident he would have spotted and fixed such a failure immediately. It might
also be a good idea to contact Thorsten direcly as he is always interested
in bug reports and fixing the bugs.

> Do we have some real PPC64 and/or PPC64LE irons in our build farm or all hosts
> based on qemu/kvm?

If you don't happen to get access from someone at SUSE, I will be happy to
provide you with an account on a powerpc (32-bit) or ppc64 (64-bit, big-
endian) porterbox running Debian unstable. I have also an instance running
openSUSE Tumbleweed ppc64.

If you need access to a ppc64el machine, I could help you getting in touch
with people from IBM who I happen to know through Debian.

Cheers,
Adrian

> [1] https://buildd.debian.org/status/logs.php?pkg=mksh&arch=ppc64
> [2] https://buildd.debian.org/status/logs.php?pkg=mksh&arch=ppc64el
--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

John Paul Adrian Glaubitz
On 11/24/2017 11:25 AM, John Paul Adrian Glaubitz wrote:> This could actually be true provided that OBS actually builds on QEMU and not
> really iron. Looking at the build logs for mksh for ppc64 [1] and ppc64el [2]
> in Debian, I don't see this particular testsuite failure in any of the builds
> in the history.

Forgot to mention: Debian always uses real hardware for the build machines. The
build maintainers and release managers have agreed (not sure whether there is
also an official policy) that packages built with QEMU may not be uploaded to
the archive.

Adrian
--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

Andreas Schwab-2
On Nov 24 2017, John Paul Adrian Glaubitz <[hidden email]> wrote:

> Forgot to mention: Debian always uses real hardware for the build machines. The
> build maintainers and release managers have agreed (not sure whether there is
> also an official policy) that packages built with QEMU may not be uploaded to
> the archive.

KVM is also real hardware.

Andreas.

--
Andreas Schwab, SUSE Labs, [hidden email]
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE  1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

John Paul Adrian Glaubitz
On 11/27/2017 10:31 AM, Andreas Schwab wrote:
>> Forgot to mention: Debian always uses real hardware for the build machines. The
>> build maintainers and release managers have agreed (not sure whether there is
>> also an official policy) that packages built with QEMU may not be uploaded to
>> the archive.
>
> KVM is also real hardware.

Yes, I'm aware of that. What I didn't know was whether QEMU or QEMU-KVM is being
used on OBS.

Adrian
--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: OBS and PPC64 versus PPC64LE ... sometimes build failures

Neal Gompa
In reply to this post by John Paul Adrian Glaubitz
On Fri, Nov 24, 2017 at 5:28 AM, John Paul Adrian Glaubitz
<[hidden email]> wrote:

> On 11/24/2017 11:25 AM, John Paul Adrian Glaubitz wrote:> This could
> actually be true provided that OBS actually builds on QEMU and not
>>
>> really iron. Looking at the build logs for mksh for ppc64 [1] and ppc64el
>> [2]
>> in Debian, I don't see this particular testsuite failure in any of the
>> builds
>> in the history.
>
>
> Forgot to mention: Debian always uses real hardware for the build machines.
> The
> build maintainers and release managers have agreed (not sure whether there
> is
> also an official policy) that packages built with QEMU may not be uploaded
> to
> the archive.
>

For what it's worth, this has also been the policy in Fedora too. All
architectures are built native, even if they're VMs on host hardware
of the same architecture. We don't do emulated builds because they
seem to produce weird output at times.



--
真実はいつも一つ!/ Always, there's only one truth!
--
To unsubscribe, e-mail: [hidden email]
To contact the owner, e-mail: [hidden email]