[oe] release 2010-12: tar error

[Martyn Welch]

On 21/12/10 22:58, Graham Gower wrote:
> On 21 December 2010 22:20, Martyn Welch <martyn.welch at ge.com> wrote:
>> On 06/12/10 20:47, Denys Dmytriyenko wrote:
>>> On Fri, Dec 03, 2010 at 11:44:14AM -0800, Khem Raj wrote:
>>>> On Fri, Dec 3, 2010 at 7:36 AM, Frans Meulenbroeks
>>>> <fransmeulenbroeks at gmail.com> wrote:
>>>>> Not sure if this is reproducible or a race but when building console-image
>>>>> for minimal for calamari I got:
>>>>>
>>>>> ERROR: Function staging_packager failed
>>>>> NOTE: Task failed: ('function staging_packager failed',
>>>>> '/home/hudson/jobs/FM_TEST/workspace/tmp/work/calamari-oe-linux-gnuspe/linux-2.6.30-r6/temp/log.staging_packager.27026')
>>>>> ERROR: Task 601
>>>>> (/home/hudson/jobs/FM_TEST/workspace/openembedded/recipes/linux/linux_2.6.30.bb,
>>>>> do_package_stage) failed with 256
>>>>> NOTE: oestats: task failed, see
>>>>> http://tinderbox.openembedded.net/packages/1176040/
>>>>> ERROR: Logfile of failure stored in:
>>>>> /home/hudson/jobs/FM_TEST/workspace/tmp/work/calamari-oe-linux-gnuspe/linux-2.6.30-r6/temp/log.staging_packager.27026
>>>>> Log data follows:
>>>>> | tar: .: file changed as we read it
>>>>>
>>>>> NOTE: package linux-2.6.30-r6: task do_package_stage: Failed
>>>>>
>>>>> Didn't have time to dig into this. Sry.
>>>>
>>>> yes this happens time to time. Its some sort of race condition. I get
>>>> it sometimes in different packages but never went in deep to fix it.
>>>
>>> I can confirm that this is a long standing race condition, presumably in
>>> packaged staging - there were several attempts at fixing it w/o much success.
>>>
>>
>> I recently repeatedly hit this. I'm in the process of setting up some
>> automated testing which made this particularly annoying. I had
>> BB_NUMBER_THREADS and PARALLEL_MAKE setup in my local.conf. I think this is
>> arises when BB_NUMBER_THREADS is set. I've removed it and taken the hit in
>> build time instead (preferable in my situation). This seems to have stop it
>> being an issue, but I need to do quite a few more runs to make sure.
>>
>> Martyn
>>
> 
> If you have a good way to reproduce any race, please provide instructions.
> 
> I have PARALLEL_MAKE = "-j 12" and BB_NUMBER_THREADS = "12" in my
> local.conf and have never seen this race.
> 
> -Graham
> 

I really wish I could say I had. I believe I have had PARALLEL_MAKE = "-j 2"
and BB_NUMBER_THREADS = "2" in my configuration for quite a while now and
didn't see this problem until I started to script some of our testing.

Martyn

-- 
Martyn Welch (Principal Software Engineer) | Registered in England and
GE Intelligent Platforms                   | Wales (3828642) at 100
T +44(0)127322748                          | Barbirolli Square, Manchester,
E martyn.welch at ge.com                      | M2 3AB  VAT:GB 927559189