Skip to content

Conversation

jcarpent
Copy link
Contributor

@jcarpent jcarpent commented Oct 1, 2025

No description provided.

@jcarpent jcarpent marked this pull request as draft October 1, 2025 06:42
@jcarpent jcarpent marked this pull request as ready for review October 1, 2025 06:51
@jcarpent
Copy link
Contributor Author

jcarpent commented Oct 1, 2025

@Lucas-Haubert @jorisv The issue on the Ubuntu CI with the GeneralizedSelfAdjointEigenSolver is not related to this PR, and seems related either to an Eigen issue or to a test issue. @Lucas-Haubert Could you investigate it?

@jcarpent jcarpent requested review from ManifoldFR and jorisv October 1, 2025 06:53
@jcarpent
Copy link
Contributor Author

jcarpent commented Oct 1, 2025

Fixes #586

@jorisv
Copy link
Contributor

jorisv commented Oct 1, 2025

We can wait the conda-forge eigen 5 release so we will be able to reproduce the issue easily.

@traversaro
Copy link

We can wait the conda-forge eigen 5 release so we will be able to reproduce the issue easily.

This is fun, as I ended up here as I wanted to write in that PR "let's wait for downstream projects to start supporting Eigen 5 before upgrading the conda-forge package to Eigen", so that turned in a chicken and egg problems. Jokes apart, I guess this may take a few days, see conda-forge/eigen-feedstock#47 (comment) .

@nim65s
Copy link
Contributor

nim65s commented Oct 3, 2025

I can switch the Nix CI to eigen5 if that help to break the deadlock

@jorisv
Copy link
Contributor

jorisv commented Oct 3, 2025

@nim65s Thank, we were speaking about that with @ManifoldFR

@ManifoldFR
Copy link
Member

Guilhem to the rescue 🙏🏼

@jorisv
Copy link
Contributor

jorisv commented Oct 3, 2025

@traversaro I my point of view the new package should be released, then downstream project can add compatibility to it.

It's a new major version so if it break thing it's normal. People will provide patches. And if there is some bug on eigen side, they will (hopfully) release new patch release (:crossed_fingers:).
I don't really see the difference with a standard conda-forge migration (except the release has been messy and the ABI management you try to add in the conda-forge eigen package).

On CI side we have lockfile to protect us against that. But I know a lot of other projects doesn't have one, so I'm really speaking for myself.

Copy link
Contributor

@jorisv jorisv left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Wait ability to reproduce in CI

@traversaro
Copy link

traversaro commented Oct 3, 2025

@traversaro I my point of view the new package should be released, then downstream project can add compatibility to it.

Sure, I agree in theory, but eigen is a bit specific case for several reasons, including the fact that it is an header only package, so there is no usual migration process (at least until we will have the eigen-abi package, and even in that case the migration will only happen if you use eigen in the public headers, not if you have any transitive dependency that uses eigen, as it happens for "normal" libraries), that it has not being released for a long time, and is used really extensively in most C++ robotics libraries. Indeed my suggestion is just to wait for a few days, not forever for all downstream packages to be fixed (that will probably take months/years).

@jorisv jorisv merged commit 3bd6b04 into stack-of-tasks:devel Oct 3, 2025
36 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants