[yocto-ab] Project Status

akuster akuster at mvista.com
Tue May 14 16:31:35 PDT 2019

On 5/14/19 12:51 PM, Philip Balister wrote:
> Regrading patchwork, do you have an estimate of manpower needs to cover
> patchwork?

Patchwork or PatchTest? Patchwork is functional. Patchtest is offline
and has a greatest impact on the project as it provides a first pass
auto-review and it may do more. Both don't have maintainers.  Richard
and Ross now have to do the first level review and checkout.

- armin
> Philip
> On 05/13/2019 03:14 AM, Richard Purdie wrote:
>> Hi All,
>> I wanted to give everyone an update of where the project is at.
>> Current Status
>> ==============
>> 2.7 was released on schedule and on the most part with the features as
>> planned. In particular I want to highlight the improvements in the
>> automated QA situation. We've removed dependencies on legacy tools like
>> testopia and now have a bare bones resulttool which gives us customised
>> results handling. This is very simple code but extremely powerful to
>> let the project do exactly what we need. An example of what this gives
>> us is this automated report:
>> https://autobuilder.yocto.io/pub/non-release/20190512-10/testresults/testresult-report.txt
>> The top chart gives an index summary of all the tests that are run be
>> them selftest, runtime, sdk or esdk. This allows easy comparison with
>> previous builds. During 2.7 we significantly increased the number of
>> tests which were automated and have started working through automation
>> of more of the manual tests.
>> The next section is ptest which gives pass/fail/skip and timing counts
>> for each recipe. There has been a huge amount of work in having stable
>> numbers here but I believe we now have consistent results and can work
>> on reducing the fail counts. These ptests have already caught a kernel
>> bug which was reported upstream and fixed in the 5.0 stable series.
>> Post 2.7 we have developments in master where Armin added ltp support
>> and this report includes those after ptest results. Again, this gives
>> us a baseline which we can investigate and improve upon.
>> Challenges
>> ==========
>> As many know, Stephen who was acting as our program manager in his own
>> time was injured in an accident and we're having to manage without him.
>> Several people have been helping out with different meetings but it has
>> had an impact on 2.8 planning which I'm driving in his absence.
>> Its become apparent we have some resource gaps in some key areas of the
>> project for 2.8:
>> * Patchtest/patchwork offline and no maintainer (bug 13284)
>> * Reproducible builds automated testing (bug 13323)
>> * Build artefact reuse (bug 10682)
>> * License infrastructure improvements (bug 13321, 13322)
>> * Autobuilder code bugs (bug 13332, 13329)
>> To give a bit more information on these:
>> We have no active maintainers for some key pieces of infrastructure.
>> I'm personally covering the autobuilder, patchwork/patchtest have no
>> active coverage.
>> Two key focus areas of the project should be reproducibility and
>> license handling but we have nobody working on these either.
>> Finally, the build artefact reuse would be a *huge* feature win for the
>> project with significant day to day improvement for most users. The
>> base work on the sstate equivalence server has been done, the remaining
>> piece is to rewrite runqueue so that it can take advantage of artefact
>> equivalence mapping. This would benefit sstate cache reuse and improve
>> multiconfig builds. This one is tricky in that it its highly involved
>> changes at the core of bitbake and there are only a limited number of
>> people with the skills. I would love to do it but I can't if I'm
>> covering everything else that I'm doing.
>> I've also not mentioned layer setup here as that is another problem
>> which needs experience to work on it and nobody with the skills has the
>> bandwidth. I believe the artefact reuse would be the bigger win for the
>> project. I've also not going into the hundreds of smaller day to day
>> things we're doing, just the big potential gaps.
>> The team also decided to try and create a new class of "newcomer" bugs
>> which would be straight forward enough that someone new to the project
>> could work on. These can be seen here:
>> https://wiki.yoctoproject.org/wiki/Bug_Triage#Newcomer_Bugs
>> We've not seen much take up on newcomer bugs yet, I'm hoping we can
>> socialise these more and help bring some new people into the project.
>> I want to give the advisory board an indication of the potential missed
>> opportunities we have and where we could use help. Regardless, I'm
>> going to continue to work on things on a priority basis, we have a
>> strong and stable platform for build off now which will help a lot.
>> I've tried to keep the email shorter than longer but am happy to
>> provide more information on anything above, be it via email or
>> potentially a meeting if there is interest.
>> Cheers,
>> Richard

More information about the yocto-ab mailing list