Re: [ESPResSo-devel] el2d.tcl test fails

[Axel Arnold]

Hi!

As the ELC with dielectric contrasts is in any case still buggy, I will anyways take a closer look at the code. However,  I first have to read through ~200 pages of theses and papers...

Axel
 
On 19.09.14 09:05, Olaf Lenz wrote:

I have been bug hunting for a while, but I cannot catch it. It seems to be a Heisenbug that disappears as soon as you start looking for it. I have compiled with -g and -O0, and it seems to occur much more rarely. When the debugger actually finds a problem, it seems to be hidden somewhere in the MPI-stuff.

Olaf

2014-09-18 16:20 GMT+02:00 Christoph Junghans <address@hidden>:

2014-09-18 6:29 GMT-06:00 David Schwörer <address@hidden>:

> Hi,
>
> in the current master, el2d test fails sometimes:
>
> for i in {1..100}; do ./runtest.sh -p 4 el2d.tcl|grep "el2d.tcl -"; done
>                            el2d.tcl - ok
>                            el2d.tcl - ERROR (code=5)
>                            el2d.tcl - ok
>
> error codes are 5 or 139 ...
>
> the probability of failure seems to depend on the number of processes:
> 1 process: 100 % success
> 2 processes: 92 % success
> 4 processes: 25 % success
> 8 processes: 65 % success
> 16 processes: 66 % success
>
> This test also fails with the default configuration.

Same here:
<https://travis-ci.org/espressomd/espresso/jobs/35509461#L1389>

Christoph
>
> Sincerly
> David
>



--
Christoph Junghans
Web: http://www.compphys.de

--

Dr. rer. nat. Olaf Lenz

Institut für Computerphysik, Allmandring 3, D-70569 Stuttgart

Phone: +49-711-685-63607

-- 
JP Dr. Axel Arnold
ICP, Universität Stuttgart
Allmandring 3
70569 Stuttgart, Germany
Email: address@hidden
Tel: +49 711 685 67609