Ilhan Polat

Hi @rebeccafair > Scipy >= 1.2.0 from PyPI, and only with certain numbers/sizes of matrices. What are those sizes? Which dtype are the matrices (complex or complex128)? and are they...

Segfaults typically hard to pin down indeed however it is also hard for us to help you if we cannot narrow the problem scope. One thing you can try is...

I don't know if I understood correctly but eigh (and the underlying LAPACK routines) expects no complex numbers on the diagonals (otherwise they are not hermitian) but `dmats` is having...

Could you please use a seed and fix the array for every run that leads to a segfault? Then we can bisect and find which array slice is causing the...

Now that the array is fixed, And the error appears at the later part of the slice, could you please take the first and the last part of the array...

Or there is an offending data in the last part of the array. If certain parts are coming out consistently OK then get rid of the parts that returned OK....

I don't have enough time/patience to troubleshoot it anymore so better ignore it for now.

Cool. Didn't know about `dll-diagnostics`, looks really useful. > unless you're using the dev.py script, and the --win-cp-openblas flag, that does that job for you. Another interesting flag that I...

I used the flag and for the build-install I think it does something correct though I couldn't find the place it copied the DLL and it said OpenBLAS copied. But...

Yes that's true but even then I don't see it in my build-install directory after dev.py finishes off with OpenBLAS copied and Installation successful. That is to say, it says...