solfec-mpi runtime error

[bencannell]

Hi @tkoziara

I think there may be a bug in solfec changeset db5b5ef.

solfec-mpi exits with a "Segmentation fault" as soon as the analysis begins.

I do not get an error running exactly the same analysis script using changeset fd106b2.

Best regards,

Ben

[tkoziara]

This is related to #72: I must have had reasons not to just do what I did in the last commit:) I have now reverted it and reopened #72:) If you can provide the crushing input deck that may help.

[bencannell]

I can reproduce the error by running the file "81array.py" in your "examples" folder using solfec-mpi with -np 8.

I've attached the output from each of the simulations for your info (note that I manually terminated the "81array-fd106b2" run since it seemed to be running OK).

This runs ok:
81array-fd106b2.log

This does not run ok:
81array-db5b5ef.log

[tkoziara]

Great, thanks:)

[bencannell]

Hi @tkoziara,

Did this and #72 get fixed?

Thanks,

Ben

[tkoziara]

Hi Ben. My understanding is that after reverting to the original code - solfec-mpi would run the 81array and other examples - I made a quick test now and it seemed ok. Providing a fix to #72 goes a little deeper into nuances of Solfec implementation and I will need some dedicated time to make it work without disrupting stability. Is this a priority for you? Such constraints are not necessarily physical but they can have meaning in modelling. Tomek

[bencannell]

Hi,

Ok, we can work around #72 so it's not a priority to change the source. I was just checking in case you had already done this in a different version of Solfec (which worked with the resolution to #74).

Thanks,

Ben

[tkoziara]

Ben, my understanding is, that I can close this issue, as it was caused by and ad hoc resolution of #72, which I subsequently reverted. Could you confirm that it is ok to close the current issue?