ci: use geometry latest images rather than fixing the usage of 25R2

[RobPasMue]

As title says.. Otherwise CI/CD will continue to fail as shared here https://github.com/ansys/actions/actions/runs/22067805128/job/63853386883?pr=1158

[moe-ad]

@RobPasMue The reason for testing against a stable container version it is to limit external impacts. Latest container builds might not be as stable and pulling that might start impacting the actions pipelines more frequently.

Can we do 261 instead?

[RobPasMue]

@RobPasMue The reason for testing against a stable container version it is to limit external impacts. Latest container builds might not be as stable and pulling that might start impacting the actions pipelines more frequently.

Can we do 261 instead?

As one of the main maintainers of the PyAnsys Geometry repository, I have to disagree on that one..

The *-core-latest images are verified against the entire test suite before they are tagged as such. By default the repository receives *-core-latest-unstable coming from internal sources. If the entire test suite passes, then and only then, is the image promoted to *-core-latest.

In fact, using a "version specific" image is not ensured to be stable. Take *-core-271 for example.. that tag gets pushed every time a commit is done inside the server team. And I can assure you that you will hit problems if you use that one 😄

On the other hand, if you use an old version (such as the 26R1 image), I can also assure you that the documentation might fail to build, as it is happening currently.. because we are constantly adding new examples and some of them might require features from 27R1.

So, all that being said, I suggest we stick to the *-core-latest image.

[moe-ad]

@RobPasMue The reason for testing against a stable container version it is to limit external impacts. Latest container builds might not be as stable and pulling that might start impacting the actions pipelines more frequently.
Can we do 261 instead?

As one of the main maintainers of the PyAnsys Geometry repository, I have to disagree on that one..

The *-core-latest images are verified against the entire test suite before they are tagged as such. By default the repository receives *-core-latest-unstable coming from internal sources. If the entire test suite passes, then and only then, is the image promoted to *-core-latest.

In fact, using a "version specific" image is not ensured to be stable. Take *-core-271 for example.. that tag gets pushed every time a commit is done inside the server team. And I can assure you that you will hit problems if you use that one 😄

On the other hand, if you use an old version (such as the 26R1 image), I can also assure you that the documentation might fail to build, as it is happening currently.. because we are constantly adding new examples and some of them might require features from 27R1.

So, all that being said, I suggest we stick to the *-core-latest image.

I assumed, thanks for explaining the geometry case... 😄
IIRC, I had to start avoiding using latest for the containers we pull because the daily update for some projects sometimes starts breaking things here (of course it also breaks things on the project's side as well when that happens), and using a tagged version sort of shielded ansys/actions from all of that.

[moe-ad]

LGTM

[RobPasMue]

I assumed, thanks for explaining the geometry case... 😄
IIRC, I had to start avoiding using latest for the containers we pull because the daily update for some projects sometimes starts breaking things here (of course it also breaks things on the project's side as well when that happens), and using a tagged version sort of shielded ansys/actions from all of that.

No problem @moe-ad -- I wasn't expecting you to know the internals and pipelines of PyAnsys Geometry - otherwise you'd be a maintainer of that library 😄 (looking for a job? I've got work for you there if you're interested 😄)

Your assumption is good for most libraries nonetheless. Many won't assume that the latest image is "stable". Most probably the contrary. But in our case, it is ensured to be stable. And if things break there... rest assured that the one place where it will get fixed will be directly in the main branch of the PyAnsys Geometry repo.. so you'll see the fix sooner than anybody else