aarch64: buildbot failure since fe05e1cb

aarch64: buildbot failure since fe05e1cb

[Siddhesh Poyarekar]

Hi Adhemerval,

It looks like glibc make check has been failing in the aarch64 buildbot
slave[1] since fe05e1cb that you committed on 20th October.  It looks
like a failure in tst-spawn2.  Can you please take a look?

Thanks,
Siddhesh

[1]
http://glibc-buildbot.reserved-bit.com/builders/glibc-aarch64-linux/builds/760

Re: aarch64: buildbot failure since fe05e1cb

[Adhemerval Zanella]

On 15/11/2017 12:34, Siddhesh Poyarekar wrote:
> Hi Adhemerval,
> 
> It looks like glibc make check has been failing in the aarch64 buildbot
> slave[1] since fe05e1cb that you committed on 20th October.  It looks
> like a failure in tst-spawn2.  Can you please take a look?
> 
> Thanks,
> Siddhesh
> 
> [1]
> http://glibc-buildbot.reserved-bit.com/builders/glibc-aarch64-linux/builds/760
> 

It was reported failures in all architectures in fact and I changed back
to old semantic of not using WNOHANG on aa95a2414e4. Is it still failing?

Re: aarch64: buildbot failure since fe05e1cb

[Siddhesh Poyarekar]

On Wednesday 15 November 2017 08:18 PM, Adhemerval Zanella wrote:
>> [1]
>> http://glibc-buildbot.reserved-bit.com/builders/glibc-aarch64-linux/builds/760
>>
> 
> It was reported failures in all architectures in fact and I changed back
> to old semantic of not using WNOHANG on aa95a2414e4. Is it still failing?

Yes, all checks since this one have been failing.

Siddhesh

Re: aarch64: buildbot failure since fe05e1cb

[Szabolcs Nagy]

On 15/11/17 14:54, Siddhesh Poyarekar wrote:
> On Wednesday 15 November 2017 08:18 PM, Adhemerval Zanella wrote:
>>> [1]
>>> http://glibc-buildbot.reserved-bit.com/builders/glibc-aarch64-linux/builds/760
>>>
>>
>> It was reported failures in all architectures in fact and I changed back
>> to old semantic of not using WNOHANG on aa95a2414e4. Is it still failing?
> 
> Yes, all checks since this one have been failing.
> 

does not fail for me after the WNOHANG fix.

Re: aarch64: buildbot failure since fe05e1cb

[Adhemerval Zanella]

On 15/11/2017 12:54, Siddhesh Poyarekar wrote:
> On Wednesday 15 November 2017 08:18 PM, Adhemerval Zanella wrote:
>>> [1]
>>> http://glibc-buildbot.reserved-bit.com/builders/glibc-aarch64-linux/builds/760
>>>
>>
>> It was reported failures in all architectures in fact and I changed back
>> to old semantic of not using WNOHANG on aa95a2414e4. Is it still failing?
> 
> Yes, all checks since this one have been failing.
> 
> Siddhesh
> 

It is really unexpected because the semantic of spawni should be equal 
to the one before fe05e1cb. I am trying to reproduce it on my aarch64 
environment (4.4.0-62-generic), but without much success.  Also I am 
not seeing this issue on powerpc64le (which Tulio raised an issue with
fe05e1cb), neither on sparc64 (which I also observed intermittent 
failures). I am also checking on powerpc and arm.

In fact I did see an issue on aarch64, but it is unrelated to the
spawni itself. I am running test-spawn2 in a continuous loop on a
container system and after 26556 I got a 'Cannot fork' message
(most likely due a pid limit from container).

Re: aarch64: buildbot failure since fe05e1cb

[Siddhesh Poyarekar]

On Wednesday 15 November 2017 09:08 PM, Adhemerval Zanella wrote:
> It is really unexpected because the semantic of spawni should be equal 
> to the one before fe05e1cb. I am trying to reproduce it on my aarch64 
> environment (4.4.0-62-generic), but without much success.  Also I am 
> not seeing this issue on powerpc64le (which Tulio raised an issue with
> fe05e1cb), neither on sparc64 (which I also observed intermittent 
> failures). I am also checking on powerpc and arm.
> 
> In fact I did see an issue on aarch64, but it is unrelated to the
> spawni itself. I am running test-spawn2 in a continuous loop on a
> container system and after 26556 I got a 'Cannot fork' message
> (most likely due a pid limit from container).

I just broke the buildbot while doing an upgrade, so I guess I'll have
to look into this over the weekend.  I'll let the list know if I find
anything.

Thanks,
Siddhesh