[Toaster] Toaster Weekly Meeting Minutes 02/24/16

[Barros Pena, Belen]

Attendees: Mihail, Ed, Elliot, Sujith, Dave, Michael, David, Brian, Belén

Updates:

Ed: still working on 7880. He wants to bring up as open

Elliot: Worked on 8440 (catching early builds), patches waiting for
review.  Worked on 8842 (build performance data, patches waiting for
review. Also a bit on 8443 (improving display of failed builds), which
depends on 8440. Did a bit more on figuring out bitbake and writing a
beginning guides to it, but nothing published just yet

Sujith: submitted patch for review for 8422

Dave: submitted patches for 9101 and 9111. Opened a bug about adding
packages with dependencies (9156). He has a fix for it but not submitted
yet. Has taken a couple of other bugs.

Michael: reviewing patches and working on build cancellation feature from
Sujith. Sent an RFC about it: general approach is right, but it is
currently blocked by sqlite table locking issues. Time back offs do not
seem to fix the problem. Brian suggests we need to find out which process
is causing the lock.

David: catching up on things. 8037 seems to be obsolete by now

Brian: met with Michael Halstead about layer index. We¹ll be getting layer
index bugs. He is trying to make it easy to push to github and run the
Toaster Selenium tests with Travis in order to test branches under review
and catch regressions.

Mihail: doing release candidate QA. On the Selenium front, they are
updating the tests. Hopefully by next week tests will run again properly.
Good because it affects Brian¹s work on Travis. Updated the runner, so
that you can do --run alltests. Changes from Aníbal Limón might supersede
some of the existing test running code.

Belén: opening image customisation bugs, assigning them to Dave, and
submitting some small UI patches.

Triage bugs:

No triage this week: we had a couple of important opens to discuss. We
need to add layer index bugs to our usual triage.

AR's:

Unassigned - Understand and fix/write-up layer sources¹ odd inheritance
structure. 
Michael - Expend documentation on image customization
Stephen - invite Aníbal to the Toaster call [new]
Brian - set 1:1 with Elliot about layer index [new]

Opens:

Brian: layer index.

We are taking the layer index on because Paul Eggleton no longer has time
to maintain it. The Toaster team will need to fix its bugs. Elliot will be
"gatekeeper": will check the layers submitted to make sure they are
layers, and contact maintainers when layers disappear. Elliot and Brian
will sync off line.

We will also be triaging layer index bugs from now on.

The API changes to bitbake will make the management of layer index
unpleasant in the short term. The online version is not seeing trouble
because it's not reparsing recipes currently, but if you do, you get tons
of errors. 

Michael Halstead contacted Elliot, Michael and Ed to give them access to
infrastructure for development of layer index, which can also be used for
other Toaster things like Patchwork.

Ed: 7880. 

What's the reason of having bitbake server running for Toaster? Brian
thinks that if you have 2 projects in Toaster (dog and cow) you will have
3 build dirs: build_cli, build_dog and build_cow. For build_dog and
build_cow you can use toaster ui. But for build_cli you need bitbake
server running. 

Ed suggests user could specify toaster ui when they want toaster to grab
the information for their cli builds. Another way would be to set some
variable or configuration to make toaster ui the default ui instead of
knotty. How is knotty the default ui? Maybe there is already a variable
you can set. We can set the toaster setup script to do that, and it can be
overridden by passing -u knotty when issuing the bitbake command from cli.
Ed agrees this would be reasonable.

Michael thinks a third option would be a wrapper script for command line
builds to talk to toaster, instead of calling bitbake directly. Brian
thinks Ed's suggestion about the variable to set the default UI is pretty
much the same thing Michael is suggesting. Michael thinks with the wrapper
script we would have more control over the environment, to make sure it's
set up correctly for toaster.

So Ed is correct: we don't need a bitbake memory resident server for each
build directory. That will simplify a lot of things: we can run parallel
builds easier, and it would be possible to run project builds from the cli
if you wanted to. He will figure out if there is already a variable
setting the default bitbake ui, and set it to toaster when toaster is
started.