[Buildroot] [PATCH v5 00/11] autobuild-run improvements

[Thomas De Schampheleire]

Hi Thomas,

On Sun, Mar 15, 2015 at 2:34 PM, Thomas Petazzoni
<thomas.petazzoni at free-electrons.com> wrote:
> Dear Thomas De Schampheleire,
>
> On Fri, 12 Dec 2014 21:04:45 +0100, Thomas De Schampheleire wrote:
>
>> Thomas De Schampheleire (11):
>>   scripts: add python module docopt
>>   autobuild-run: use docopt for argument parsing
>>   autobuild-run: add option --make-opts for custom Buildroot options
>>   autobuild-run: use **kwargs to avoid explicit parameter passthroughs
>>   autobuild-run: check-requirements does not need to know the login
>>     details
>>   autobuild-run: set LC_ALL=C to not use locale settings of host machine
>>   autobuild-run: improve the logic to generate build-end.log
>>   autobuild-run: save config.log files for failed package
>>   autobuild-run: extend TODO list
>>   autobuild-run: kill all children on SIGTERM
>>   autobuild-run: catch KeyboardInterrupt in the same way as SIGTERM
>
> I've applied your patch series. A few comments, though:
>
>  1/ I also applied your two follow-up patches
>     http://patchwork.ozlabs.org/patch/444865/ and
>     http://patchwork.ozlabs.org/patch/444866/. As requested in the
>     commit log, the second patch was squashed into the patch adding the
>     support for killing subprocesses.
>
>  2/ I had to rework patch http://patchwork.ozlabs.org/patch/444866/
>     because subprocess.check_output() doesn't exist in Python 2.6. So
>     I've used a different construct.
>
>  3/ Killing all subprocesses apparently still doesn't work completely
>     correctly, I got the backtrace of the attached file once when
>     interrupting the autobuild-run script with Ctrl+C, on a system that
>     uses Python 2.6. And the child processes (timeout processes) were
>     still running.

To understand better, some questions:
- you saw this situation once, with Python 2.6, but have been able to
successfully stop autobuild run with Ctrl-C with Python 2.6 on other
occasions?

- did you happen to use Python 2.7 with these patches on a certain
machine? Did you see issues?

In my testing I have been using Python 2.7.
It could be that there is something different in Python 2.6, not sure.
I would need to look into this deeper, unless André who seems to be
much more knowledgeable about Python has an idea?

Thanks,
Thomas