[Toaster] [PATCH 00/15] Build cancellation

[Brian Avery]

Hi!


On Mon, Mar 21, 2016 at 3:34 AM, Ed Bartosh <ed.bartosh at linux.intel.com> wrote:
> On Sun, Mar 20, 2016 at 10:12:22AM -0700, Brian Avery wrote:
>> a couple of other thoughts after looking at it a little more today:
>>
>> 1) pid of bitbake started in proj builds dir for cleanup? do we want
>> to save these?
> not until we use them.
>
>> bitbake.lock with lsof might mean we don't need these. however, having
>> them might negate the lsof requirement...)
> I'll look what I can do. I was planning to reuse what I did in my old
> ed/toaster/bec branch: to generate shell script for cloning all repos
> and running the build. It would be much more easy to deal with
> pid files in the shell script than in current python code.
> However, I think it's better not to do right now as we already have
> quite a bit of changes and very close to release.
>
k. was more of a me musing about it than anything else...
>> 2)should poky oe-init be a #!/bin/bash if dash fails?
> No, we shouldn't as it works with zsh.
>
k. btw, how did it fail? I just ran oe-init and bitbake
core-image-minimal in an ubuntu:14.04 container and it seemed fine.
/bin/sh was liked to dash...

>> 3) toasterui event loop should probably check for liveness of server
>> connection…. or do we get an error on waitEvent if the server  has
>> died (segv’d, kill -9 by user, etc)?
> We run sever only during the build. We probably need to set build state to
> 'Failed' if server is killed or crashed.
yep. though it can easily be a follow on bug.
>
> Regards,
> Ed

My thoughts:
since Ed has fixed the ../bitbake/bin issue and if
https://lists.yoctoproject.org/pipermail/toaster/2016-March/004162.html
fixes the CI not working I'm good with upstreaming and filing
additional bugs for the corner cases since this fixes alot.

TY!
-brian
-an intel employee