Re: How to deal with failing 'make test-bseline'

[Thomas Morley]

Am So., 9. Feb. 2020 um 19:48 Uhr schrieb David Kastrup <address@hidden>:
>
> Thomas Morley <address@hidden> writes:
>
> > I now tried
> >  2014  make test-clean
> >  2015  make test
> > with no useful infos, like before
> >
> > but
> >  2020  make test-clean
> >  2021  CPU_COUNT=9 make -j9 test
> >  2022  grep sourcefilename `grep -L systems.texi
> > out/lybook-testdb/*/*log|sed s/log/ly/g`
> > finally pointed to input/regression/display-lily-tests.ly
> >
> > I always thought a multi-core run is more difficult to debug ...
> > Of course 9 cores is pretty pointless here. I've two behaving like 4...
>
> Assuming you have enough memory, conventional wisdom is to use 5 then.
> I have 4 pretending to be 8, so that's what the 9 is about.  The next
> generation Thinkpad would have allowed getting a processor doing that
> without exceeding thermal design power.  But then I don't have a
> discrete GPU and don't do a lot of video processing.  I still have the
> original 2-core somewhere in a drawer.

I've rerun with
CPU_COUNT=3 make -j3 test
with success, i.e. the errored file is grepped.

I usally use 3 not 5, because of limited memory.


Cheers,
  Harm