fix memory error in testSelectOutput

I found the bug in testSelectOutput with intel.

@matthieu, in the cosmology, you have a method called cosmology_clean that should be called only with cosmology. I guess it would be cleaner if you can call it without cosmology (even if it does nothing). Are you fine if I am implementing this change?

Freeing a null-pointer is not a mistake. So it should be safe to call cosmology_clean().

However, the bug is probably that not all the tables have the pointer set to NULL when initialising the cosmo (with or without cosmo).

added 1 commit

• d5c164e7 - Initialize correctly the scale_factor_interp_table without cosmology

Compare with previous version

Ok, I have a correct fix now

I like this one better. :)

Isn't there a 6th table that is not initialised?

added 1 commit

• 1da789db - fix memory error in testSelectOutput

Compare with previous version

Yes (strange that I did not get an error with this one)

You might just be lucky (or unlucky). It depends on what the value was at this location in memory before. And that may also be why we did not spot it before.

Thanks!

merged

mentioned in commit 1311f784