Diffractor.jl icon indicating copy to clipboard operation
Diffractor.jl copied to clipboard

bump compat for AbstractDifferentiation to 0.6

Open github-actions[bot] opened this issue 2 years ago • 9 comments

This pull request changes the compat entry for the AbstractDifferentiation package from 0.5 to 0.5, 0.6. This keeps the compat entries for earlier versions.

Note: I have not tested your package with this new compat entry. It is your responsibility to make sure that your package tests pass before you merge this pull request.

github-actions[bot] avatar Oct 13 '23 01:10 github-actions[bot]

Yes, I tried this the other day. This fixes a bunch of stuff that was erroring due to getting rid of closures, Which means we don't need mutation support for basic functioning. (we still want it, but we don't need it to just support the AD.jl interface)

oxinabox avatar Oct 13 '23 12:10 oxinabox

Ok I removed the things that were marked as broken that were fixed when i last tried 0.6-DEV but now new things are failing and I think it will need a little more investigation on what is broken and what extra change has been included in 0.6 since i last looked

oxinabox avatar Oct 13 '23 13:10 oxinabox

@oxinabox happy to help debug the 0.6 changes, this issue recently popped up on discourse

gdalle avatar Jan 16 '24 07:01 gdalle

Since I put this on the backburner I have completely forgotten what the more subtle broken things were.

Feel free to take it over @gdalle

oxinabox avatar Jan 16 '24 09:01 oxinabox

I don't know anything about diffractor though, I was mostly talking about problems arising from AbstractDiff. Not sure I can help debug internals over here

gdalle avatar Jan 16 '24 11:01 gdalle

also it doesn't help that the Diffractor tests seem to be failing by default?

gdalle avatar Jan 16 '24 11:01 gdalle

I think for julia 1.10 they should be working.

oxinabox avatar Jan 16 '24 11:01 oxinabox

I think for julia 1.10 they should be working.

That's only mildly reassuring :rofl:

gdalle avatar Jan 16 '24 12:01 gdalle

I played with this, and it's quite weird. There's a bunch of things where if I run the tests in the REPL they work fine, but if I do ] test it blows up, including tests that go from passing to failing, and also tests that make the testset fail even though they're locked behind a @test_broken.

MasonProtter avatar Jan 31 '24 11:01 MasonProtter