Closed mfherbst closed 3 years ago
This pull request introduces 2 alerts when merging 74082e33f3d2f00f530ae2fa2dfcd03ae75458a3 into 55436b4bbf6c8a716c99add3f15adde341acb9ff - view on LGTM.com
new alerts:
What is wrong with the tests?! The error messages look obnoxious 🔥 🚒
I'm on it. pybind11 found a memory issue in the way we use their numpy interface in the HF provider, so now "our" way of doing it is no longer allowed for safety.
Okay, let me know if I can do anything. Is this because of the new pybind11 2.6?
Yes indeed. The changes were introduced in 2.6.
And now we are hitting the 4th of https://docs.h5py.org/en/stable/whatsnew/3.0.html#breaking-changes-deprecations ...
It's always the same when you just want to change a little thing :smile:
Will the changes be backward compatible for pb11 and h5?
For pybind11 it's a bit tricky. The HDF5 stuff I might manage. Is that important?
Yes, pybind11 is probably not as bad.. but since h5 gives runtime errors...
What do you mean by runtime errors?
Ahh never mind... I thought the pybind11 error was thrown at compile time... but it wasn't.. so yeah it does not matter 😄
Hooray ... nah that was a piece of cake, wasn't it ;)
Utterly frustrating...
During the parsing stage in run_adc, adcc now throws an InputError as a specialised "ValueError"