Maik Riechert

I agree, yes, it makes more sense to use `_delete`.

I just noticed that this only happens when the array in the patch has a bigger size than in the input. If it's equal or smaller everything works.

With patch you can also add new keys, and for those it uses the default indentation.

Looks good! What happens if I include `_indent` in a version of f90nml that doesn't support this feature yet? Is it silently ignored?

After adding `libexecinfo-dev` I'm seeing the same as in https://discuss.pytorch.org/t/compiling-master-from-source-on-alpine-fails-with-undefined-reference-to-backtrace/64676. I tried `LDFLAGS="-lexecinfo" python setup.py install` in the hope that it trickles down to CMake and uses that as extra...

I found a work-around. You can define the environment variable `CFLAGS="-D__EMSCRIPTEN__"` which causes PyTorch to not assume backtrace/execinfo is available. It's a hack though and ideally the PyTorch folks should...

For SCITT (see [draft spec](https://datatracker.ietf.org/doc/html/draft-birkholz-scitt-architecture-00.html)), the proposal is to use DID both as mechanism to establish issuer identifiers and to do key discovery (via DID resolution), while not being constrained...

Huh, DCO is complaining again. Looking at it more closely, this is because existing commits in `main` have DCO issues. Apart from re-writing main, I think the only sane solution...

No demo project but I successfully applied it to one of my projects: https://github.com/Reading-eScience-Centre/leaflet-coverage Key takeaways for me: - When you test server-side only with Mocha (or something else) (and...

Doesn't change anything for me, in fact produces more errors but the stacktraces are the same.