[OE-core] [PATCHv3 00/30] oeqa core and oe-selftest threaded enablement
Aníbal Limón
anibal.limon at linux.intel.com
Fri Jul 21 16:15:10 UTC 2017
On 07/21/2017 03:11 AM, Richard Purdie wrote:
> On Mon, 2017-07-17 at 16:45 -0500, Aníbal Limón wrote:
>> This v3 makes changes on tinfoil busy status patch and adds an
>> environment variable to oe-seltest script to specify number of
>> threads.
>> There are patches for bitbake [3] and meta-yocto-bsp [4] that was
>> send
>> to their specific ML.
>>
>> The first 6 patches previous enable support of threads in oeqa core
>> are independent and can be merged first.
>>
>> The v2 address comments did by Patrick about add better explanation
>> on the commit messages and some typos.
>>
>> Other interesting information to add is that with this series we will
>> able to execute oe-selftest in around half [1] of the original time
>> [2], (9406.782s vs 17303.014s).
>
> I tested this, admittedly on top of the server rework changes and see
> some odd failures on the autobuilder. Things aren't looking very
> deterministic and I'm a little worried about merging them.
>
> (e.g. https://autobuilder.yocto.io/builders/nightly-oe-selftest/builds/
> 397)
>
> The second issue I'd like to ask questions about is whether the tinfoil
> changes and server rework we've just done remove the need for some of
> this patchset?
Yes if the tinfoil API can work within a thread we can remove the
functionality to run into the main thread, but the end_thread
functionality needs to stay because the devtool test content will need
to be modified to don't touch meta/ folder.
>
> I ask since now I'm hoping the "main thread" issues might be resolved
> and we truly can have independent tinfoil connections. There were
> issues there but I believe Paul has resolved those.
>
> For that reason I'm going to defer this patchset from M2 (I did take
> some of the initial patches) and ask that we review the patchset in
> light of the server rework and tinfoil changes that just merged.
>
> I do know that for example, "oeqa/selftest/case: Support bitbake memres
> mode in per build director" is no obsolete as we removed the memres
> init script and the -t option to bitbake.
Ok, i understand, after merge the patches from Paul and you, i will
rebase and test again searching for remove has much code as i can.
Cheers,
Anibal
>
> Cheers,
>
> Richard
>
More information about the Openembedded-core
mailing list