OpenSSL issues with jill installs in Python, possibly due to v1.12?

See for example diffeqpy test failures:

https://github.com/SciML/diffeqpy/actions/runs/18408384170/job/52453820503?pr=162#step:6:1097

You're on quite an old juliapkg there.

The error comes from ensuring that the version of OpenSSL is compatible between Julia and Python, but no version is compatible for the versions of Julia and Python used.

If you upgrade, it will ensure you use a version of Julia whose OpenSSL is compatible with Python.

I am forcing the latest version in SciML/diffeqpy#162 and it still happens. Is there some other requirement I have to bump?

Just looked again and see this in the logs:

juliapkg==0.1.17

That's not the latest.

I don't directly depend on it, so that's an issue with this package?

https://github.com/SciML/diffeqpy/blob/master/setup.py#L31

I don't get your point. If you require this feature then depend on it directly. I think your test environment is just out of date.

juliacall==0.9.26

is also not the latest and it is a direct dependency.

Maybe CI caching means you're getting old versions of things?

The test environment is the most up to date it can be? I don't restrict the versions at all, that's not in the repo:

https://github.com/search?q=repo%3ASciML%2Fdiffeqpy+0.9.26&type=code

The only versions I set are right here: https://github.com/SciML/diffeqpy/blob/b0ce7ce2a43866100ba7d2c447a4b9a1a20fe25b/setup.py#L31

and that's lower bounds.

Maybe there is some incompatibility between dependencies of juliacall and jill, so you're being forced back to an old version.

Maybe locally try installing them both in a venv pinned to the latest version and see what happens.

If I were you I'd just remove jill and use the install-julia GitHub action instead.

Also you should put upper bounds on your compats :)

Fix CI to use latest package versions

If I were you I'd just remove jill and use the install-julia GitHub action instead.

We were trying to test this system because it's the 99% user's way of doing it.

So then it seems like jill is just bugged / old dependencies and we need to change tests to use setup-julia and tell users that manual installation of Julia is required again?

So then it seems like jill is just bugged / old dependencies

Have you actually determined this by trying?

You don't need your users to download Julia at all, juliapkg will do it for them. However the preferred system is that users install juliaup themselves and juliapkg will use that.

Have you actually determined this by trying?

Yes, that's exactly what I am showing with the CI results. It used to but it's not working anymore. If I force the installation of Julia it runs, but if I use the automated Python-side Julia installs that we suggest users do from the package, that runs into all of these OpenSSL jill versioning issues.

What version of Julia is jill installing, vs what version is setup-julia installing?

Also it is worth noting that the version of OpenSSL_jll is locked on Julia 1.12.0 because it got turned into a stdlib for...some reason...

I just tried locally, and there doesn't seem to be a compat issue between jill, juliacall and juliapkg - uv at least is happy to install them all at the latest version:

p> uv add jill==0.12.0 juliacall==0.9.28 juliapkg==0.1.22
Using CPython 3.13.6
Creating virtual environment at: .venv
Resolved 24 packages in 480ms
Prepared 17 packages in 1.31s
Installed 23 packages in 201ms
 + anyio==4.11.0
 + attrs==25.4.0
 + certifi==2025.10.5
 + click==8.3.0
 + colorama==0.4.6
 + filelock==3.20.0
 + h11==0.16.0
 + httpcore==1.0.9
 + httpx==0.27.2
 + idna==3.11
 + jill==0.12.0
 + jsonschema==4.25.1
 + jsonschema-specifications==2025.9.1
 + juliacall==0.9.28
 + juliapkg==0.1.22
 + python-gnupg==0.5.5
 + referencing==0.37.0
 + rpds-py==0.27.1
 + semantic-version==2.10.0
 + semver==3.0.4
 + sniffio==1.3.1
 + tomli==2.3.0
 + tomlkit==0.13.3

Aha! I just noticed your failing test is on Python 3.8. JuliaCall/JuliaPkg do not support Python 3.8 because it is deprecated, so you're falling back to an old version that did support it. Python 3.9 is also about to be deprecated, so I'd recommend bumping your tests to 3.10.

Gotcha, tests are now running with that change.

Navigation Menu

Repository navigation

OpenSSL issues with jill installs in Python, possibly due to v1.12? #689

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

OpenSSL issues with jill installs in Python, possibly due to v1.12? #689

Description

Activity

cjdoris commented on Oct 10, 2025

ChrisRackauckas commented on Oct 10, 2025

cjdoris commented on Oct 10, 2025

ChrisRackauckas commented on Oct 10, 2025

cjdoris commented on Oct 10, 2025

ChrisRackauckas commented on Oct 13, 2025

cjdoris commented on Oct 13, 2025

cjdoris commented on Oct 13, 2025

cjdoris commented on Oct 13, 2025

ChrisRackauckas commented on Oct 13, 2025

ChrisRackauckas commented on Oct 13, 2025

cjdoris commented on Oct 13, 2025

ChrisRackauckas commented on Oct 13, 2025

asinghvi17 commented on Oct 15, 2025

cjdoris commented on Oct 16, 2025

cjdoris commented on Oct 16, 2025

ChrisRackauckas commented on Oct 16, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions