[jekoorn]

I think in the end it boils down to regenerating the lockfile.
Edit: and the workflows.

Do we want to do something similar as in https://github.com/NNPDF/eko/pull/528/changes
with installing python in the bench, instead of loading an image?

[felixhekhorn]

Do we want to do something similar as in https://github.com/NNPDF/eko/pull/528/changes
with installing python in the bench, instead of loading an image?

in principle yes (to avoid complicated image generation), but as you can see we are getting in trouble with LHAPDF - however, maybe we can get around that thanks to NeoPDF? @Radonirinaunimi ? i.e. we would need to rewrite the benchmarks to use NeoPDF, which should be easy?

(Locally LHAPDF works for you, right?)

[Radonirinaunimi]

in principle yes (to avoid complicated image generation), but as you can see we are getting in trouble with LHAPDF - however, maybe we can get around that thanks to NeoPDF? @Radonirinaunimi ? i.e. we would need to rewrite the benchmarks to use NeoPDF, which should be easy?

Yes 😃 I'd actually advocate to replace all the containers and custom-build of LHAPDF in the CI with NeoPDF 🙈

Making those changes should be as easy as:

from neopdf.pdf import PDF as lhapdf

and the rest should be exactly the same.

[jekoorn]

(Locally LHAPDF works for you, right?)

Locally LHAPDF is giving me issues, too. Unsure yet what exactly is going wrong. It also has trouble finding the pineko.toml (I added it)
Let's see what happens when I use neoPDF