[opensuse-factory] openblas build failure
Hello, openblas package [1] has strange build problems for x86_64 architecture without any other errors:
collect2: error: ld returned 1 exit status
Unfortunately I can't reproduce this situation outside of openSUSE:Factory repository. Are there any ideas? [1] https://build.opensuse.org/package/show/openSUSE:Factory/openblas -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday 2016-07-26 14:53, Dmitry Roshchin wrote:
collect2: error: ld returned 1 exit status
Unfortunately I can't reproduce this situation outside of openSUSE:Factory repository. Are there any ideas?
Works in a local build, so it is probably a out of memory thing (there are four lds running concurrently, and it gets passed an .a archive which is likely very huge.) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 2016-07-26 at 16:00 +0200, Jan Engelhardt wrote:
On Tuesday 2016-07-26 14:53, Dmitry Roshchin wrote:
collect2: error: ld returned 1 exit status
Unfortunately I can't reproduce this situation outside of openSUSE:Factory repository. Are there any ideas?
Works in a local build, so it is probably a out of memory thing (there are four lds running concurrently, and it gets passed an .a archive which is likely very huge.)
The .a files in my local build are all ~63MiB. The generated test
executables are ~25MiB. Can this really break ld on an OBS build host?
Martin
-- Dr. Martin Wilck
On Tuesday, 26 July 2016 16:00:08 MSK Jan Engelhardt wrote:
Works in a local build, so it is probably a out of memory thing (there are four lds running concurrently, and it gets passed an .a archive which is likely very huge.)
I don't expect memory consumption problems, maximum used memory in science project ~1200mb. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday, 26 July 2016 21:52:04 MSK Dmitry Roshchin wrote:
I don't expect memory consumption problems, maximum used memory in science project ~1200mb.
And looks like failure is host-related, only on lambXX 2016-06-15 15:09:26 openblas meta change succeeded 1h 14m 6s cloud122:1 2016-07-01 08:51:58 openblas new build failed 27m 33s lamb08:5 2016-07-03 11:21:45 openblas new build failed 27m 49s lamb14:1 2016-07-05 09:36:45 openblas new build succeeded 1h 12m 23s cloud116:2 2016-07-07 14:07:14 openblas new build failed 30m 52s lamb10:2 2016-07-09 11:35:58 openblas new build unchanged 2h 46m 40s build73:4 2016-07-10 23:59:05 openblas new build unchanged 4h 1m 13s cloud113:3 2016-07-12 22:34:47 openblas new build failed 25m 53s lamb03:5 2016-07-14 08:32:10 openblas new build failed 28m 27s lamb16:5 2016-07-23 23:54:44 openblas new build succeeded 1h 6m 17s cloud112:3 2016-07-24 20:55:18 openblas new build succeeded 2h 29m 34s build27:2 2016-07-26 11:27:56 openblas new build failed 23m 58s lamb04:6 2016-07-26 14:51:11 openblas new build failed 30m 0s lamb01:6 -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 2016-07-26 at 22:38 +0300, Dmitry Roshchin wrote:
On Tuesday, 26 July 2016 21:52:04 MSK Dmitry Roshchin wrote:
I don't expect memory consumption problems, maximum used memory in science project ~1200mb.
And looks like failure is host-related, only on lambXX
I made a successful compilation in a KVM guest with 4G RAM. The OBS
build KVM that failed hat 8GB. So it's probably not an OOM problem.
Martin
-- Dr. Martin Wilck
On Tue, 2016-07-26 at 22:33 +0200, Martin Wilck wrote:
On Tue, 2016-07-26 at 22:38 +0300, Dmitry Roshchin wrote:
On Tuesday, 26 July 2016 21:52:04 MSK Dmitry Roshchin wrote:
I don't expect memory consumption problems, maximum used memory in science project ~1200mb.
And looks like failure is host-related, only on lambXX
I made a successful compilation in a KVM guest with 4G RAM. The OBS build KVM that failed hat 8GB. So it's probably not an OOM problem.
Another local attempt succeeded over night with as little as 1GB RAM and 1GB swap. Martin
Martin
-- Dr. Martin Wilck
, Tel. +49 (0)911 74053 2107 SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
-- Dr. Martin Wilck
On Tue, 26 Jul 2016 21:38:07 +0200, Dmitry Roshchin wrote:
On Tuesday, 26 July 2016 21:52:04 MSK Dmitry Roshchin wrote:
I don't expect memory consumption problems, maximum used memory in science project ~1200mb.
And looks like failure is host-related, only on lambXX
lamb* are the new AMD machines, IIRC. Does the package have any CPU-specific code? It's still strange that ld fails silently, though. It might be a platform problem. We had mysterious build errors on these hosts in the past, and it was addressed by the new firmware. Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Wednesday, 27 July 2016 09:28:19 MSK Takashi Iwai wrote:
lamb* are the new AMD machines, IIRC. Does the package have any CPU-specific code?
This package contains big amount of cpu-specific code, as this is optimized mathematics library.
It's still strange that ld fails silently, though. It might be a platform problem. We had mysterious build errors on these hosts in the past, and it was addressed by the new firmware.
But I don't see any problems with lambXX hosts for science/openblas builds. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Wed, 2016-07-27 at 10:41 +0300, Dmitry Roshchin wrote:
But I don't see any problems with lambXX hosts for science/openblas builds.
Mysterious - what's the difference between science/openblas and
factory/openblas?
-- Dr. Martin Wilck
participants (4)
-
Dmitry Roshchin
-
Jan Engelhardt
-
Martin Wilck
-
Takashi Iwai