Lukas Krenz

I think that Ravil's general approach is good for our purposes for the following reasons: - This is only supposed to be a temporary solution. While I do agree that...

METIS_RANDOM_SEED is a constant already (defined in the same file), so the metis call itself should be deterministic. There is also the compile option -DMINI_SEISSOL=OFF which sets the weights of...

Sure, but it is quite easy to reorder elements in post-processing, especially for small simulations. You can steal the logic from here: https://github.com/SeisSol/SeisSol/blob/master/postprocessing/science/compute_diff_seissol_data.py In any case, I believe that -DMINI_SEISSOL=OFF...

It is not documented for a reason - I am not sure that this actually solves the mesh reordering issue and because it incurs a severe performance penalty on some...

Maybe you could try and check whether you can ping something (other than archive.ubuntu.com) in your container? It sounds like it might be related to access problems/firewall settings. I can...

There's also CMAKE_CXX_FLAGS_DEBUG. Apparently Cmake doesn't set -O0 (but I thought it was default anyway on most compilers?) Apparently, gcc recommends -0g (see [here](https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html))

According to the man pages, -O0 is default for gcc and -02 is default for intel. You can find the latter under the documentation for the argument -ftrapuv: > which...

Without actually reviewing the PR, I can see the following issues: - You've commited many pyc files - You actually didn't add new submodules, but rather the entire content of...

Which recent PR are you referring to? Haven't touched the HDF5 writer, ever (I think). Afaik the problem is known, I've always used posix on NG.

No, that PR should have no effect on this. I had the issue already before that PR (for at least 1 year!). @uphoffc You had the issue as well, right?