[dpdk-dev] [PATCH v1] maintainers: update for AMD xgbe and ccp crypto

Aaron Conole aconole at redhat.com
Tue Apr 28 15:31:12 CEST 2020


Thomas Monjalon <thomas at monjalon.net> writes:

> 27/04/2020 14:58, Aaron Conole:
>> Ferruh Yigit <ferruh.yigit at intel.com> writes:
>> > [1] Two issues reported
>> > a) ninja: build stopped: Error writing to build log: Disk quota
>> > exceeded.
>> 
>> This occurs on some of the ARM64 builds.  Travis is aware of the issue,
>> but don't seem urgently fixing it. :-/  I think we have a series
>> outstanding that should be applied to drop those builds for now.
>
> I saw some "Disk quota exceeded" on x86 static builds.
> So I changed my mind, I am not sure dropping only Arm from Travis
> is a solution, even temporary.

Okay.

>
>> > b) No output has been received in the last 10m0s, this potentially indicates a
>> > stalled build or something wrong with the build itself.
>> 
>> This is a future enhancement I need to work on.  When a build has jobs
>> that are stalled / errored, we need to restart them (at least once).  Or
>> otherwise flag it.  It especially happens when internal travis
>> infrastructure fails.
>
> Yes Travis infrastructure fails. Always.
> Travis is free, but it is not a reason to provide a randomly failing tool.

It has been quite useful - it shows us that we still have some
incomplete unit tests and also some racy unit tests.  I agree, it
shouldn't be failing the way it is.

> I suggest switching to other platforms like OBS or our community lab,
> preferring the most reliable one.

OBS and community lab isn't as easy for individual developers to
integrate with.  One thing that's been useful for me is to push to my
github repo and see the results of all the build configurations.  It
doesn't require much effort from my side to ensure that code works.  I
don't always do this (but that's my bad habit), but when I do it proves
very useful.

Maybe the new github-ci (I think it's called 'actions') would be good to
investigate.  Also, Cirrus-CI lets us test on windows and freebsd.  And
there's no reason we can't use all of them.



More information about the dev mailing list