Ross Barnowski

Thanks @dschult for the explanation. I think the docstring accurately reflects the state of affairs here, though of course proposed wording suggestions are always welcome! Since there is nothing clearly...

This has come up again (see #5133) so maybe an additional note in the docstring is warranted - wording suggestions welcome!

> Oh, I'm sorry @rossbar for merging this prematurely. No worries at all, the merge definitely wasn't premature on your end, the comment was extremely delayed on my end :upside_down_face:...

There is nothing specific to Enums in the standard, so I suppose they follow the class docstring recommendations. Note that some of the validation checks are extra nit-picky and are...

Thanks for reporting - this indeed does look like a real issue related to the authentication layer. The model downloading component implements a simple cache, but the model *extraction* component...

Thanks for getting the discussion going @Schefflera-Arboricola ! Another potential challenge, at least for NetworkX, is that some input validations might be computationally expensive. For example, even relatively straightforward things...

> It is ultimately up to you but I am working on Linux today and, if I can get it to the same state as macos and Windows, it would...

Your best bet to build wheels locally is to follow the [recipe in CI](https://github.com/pygraphviz/pygraphviz/blob/6cdd1eb676a67de697bf23d6948eabb525052eee/.github/workflows/test.yml#L151). Fundamentally, the process boils down to: 1. Install graphviz on your system 2. Figure out how...

> I recommend using [cibuildwheel](https://github.com/pypa/cibuildwheel) for cross building and publishing wheels. Its quite easy to set up a github workflow for this: https://cibuildwheel.pypa.io/en/1.x/setup/#github-actions Oh we're trying :) - see e.g....

> It looks like _graphviz can't be imported which I think of as not compiling to get _graphviz.so Yeah the CI failures on windows are a bit misleading... it's actually...