vallis / lisasolve Goto Github PK

What steps will reproduce the problem?
1. I have been trying out for code for an unrelated project, and I noised a 
some quite annoying behavior.
2. When using your pymultinest module, and I run a problem where the global log 
evidence ends up being negative, which theoretically should be ok, it just 
means that the evidence term is really small.
3. The problem seems to run to completion.

What is the expected output? What do you see instead?
I get that the stats and post-separate files are malformed. The stats file only 
have three lines, and the post-seperate file are empty.

What version of the product are you using? On what operating system?
I have tried this on both my MacBook labtop, and my Ubuntu box, I used gfortran 
to compile, with no MPI (the -DMPI flag not set)

The problem does not seem to occur when using the original compiled interface 
to MultiNest, without any python involved.

Has anyone seen similar behaviour, and can anyone confirm this behaviour?

Best regards, Kennet