Bug#1012318: diffoscope 214 produced no output and was killed after running into timeout after 150m

[Chris Lamb]

Mattia Rizzolo wrote:

>> ## output (total time: 413.571s)
>> 413.571s      1 call     html
>
> So it's not really the presenter that takes excessively.  But otoh 413
> seconds are more than 6 minutes. [...]

I'm not quite sure what connection you are making between the signal
processing within diffoscope and HTML output generation. When
diffoscope receives a TERM signal, it only attempts to clean up
temporary files, it does not attempt to try and write any HTML output.
See sigterm_handler in diffoscope/main.py, etc.. (I'm not entirely
sure it should attempt to do so either. Signals are weird.)

We could certainly try and print the profiling output if we receive a
TERM though... aaaand I've pushed something to that effect to master.
I've also added some logging warnings so it's much clearer if we are
in that state.

> If somebody looks into the point above I can change jenkins to wait
> 12 minutes (or 18) before KILLing after the TERM signal.

I'm +1 for increasing the hard upper limit on Jenkins and could be
convinced to increase this TERM/KILL gap too, but as I mention above,
I'm not sure we're following the same rationale yet.


Regards,

-- 
      o
    ⬋   ⬊      Chris Lamb
   o     o     reproducible-builds.org 💠
    ⬊   ⬋
      o