-
Notifications
You must be signed in to change notification settings - Fork 36
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
iris
currently pinned to iris>=3.1.0,!=3.2.0.post0,!=3.2.0
- rethink pin for future iris
bugfix release
#1509
Comments
that looks dodgy and a half, I can't believe it wasn't picked up by any of iris' tests - I commented on the issue, it'd be useful to look at dask and what version produces that behaviour |
Since no other package than I think we should pin iris quickly so that we are able to perform another round of tests for v2.5. |
One additional comment: for our tests with |
I just had a discussion with @remi-kazeroni about this: our approach now is to wait for a response of the iris developers and postpone our release if necessary by 1-2 weeks so that we can include a fixed version of iris. It would be really a shame not to include the many new features that iris 3.2 offers. In addition, we thought that it might be good to decouple our release from the iris releases. Having a new iris version in the middle of our testing phase really isn't optimal. |
sounds reasonable to me, do we have any pressing IS-ENES3 matters?
their release schedule is slightly more hectic than ours, and that'd mean we'll have to dance by their music, I am not sure this is something I'd like, we need to talk about it |
One more comment: I guess one of the reasons this doesn't break our entire preprocessing chain is that we use lots of |
We didn't mean to speed up or slow down our release schedule, but rather be aware of iris' schedule. |
Yep, that's it! 👍 |
note that it's half term here in the UK, and the MO as a gov institution is observing the week off (well, not all of them) so it may be that the issue gets delayed |
The difficulty in trying to plan our releases around the iris releases is that from my experience they are quite unpredictable and can be delayed by months. I would propose to deal with collisions on a case by case basis rather than counting on them sticking to a particular schedule. The alternative to pinning iris to <3.2 could be to pin to iris !=3.2, like that we will still get iris 3.2.1 one once it's released. But I agree that it is a very good idea to wait for the iris developers to respond to the issue before making any decisions about our own release. |
We had more discussion about this and thought that |
It seems the consensus here is to give the Iris folks a few more days to react, particularly in light of the holidays. Shall we wait till Tuesday or so before we make a decision? |
But what kind of action by the iris developers would make us not pin iris to We will definitely delay the release, but pinning this rather sooner gives us more time for the tests. |
they will tag a bugfix release of the rc they have, it is possible it will be |
How about |
I suggest we pin extremely strictly eg |
Pinning to |
for short term yes, for long term we should have a pin for |
Cheers guys, will open the PR now! Note that we need both pins |
huh? |
Can you elaborate on that? 😄 |
hahaha, sorry dude, had to dash out - yeah, why do we need to pin on the more general |
If I don't do that the version |
let's keep this open since it's just a temporary pin, we want to revist this once all matters settle on the iris side |
iris<3.2
iris
currently pinned to iris>=3.1.0,!=3.2.0.post0,!=3.2.0
- rething pin for future iris
bugfix release
iris
currently pinned to iris>=3.1.0,!=3.2.0.post0,!=3.2.0
- rething pin for future iris
bugfix releaseiris
currently pinned to iris>=3.1.0,!=3.2.0.post0,!=3.2.0
- rethink pin for future iris
bugfix release
The risk of pinning with My suggestion to move forward would be: pin
I don't think so. The project is currently under review and what matters is everything that was delivered until December 2021. |
Good points, @remi-kazeroni. In some sense, we want the core to be slightly ahead of the tool to facilitate the idea that new tool developments can benefit from new core features. The support for UGRID and irregular grids in Iris 3.2 would be a welcome addition. Perhaps we could consider a 2.5.1 release? Or a work-around (since all we need to do is realize the data before saving)? |
good points both from @remi-kazeroni and @zklaus - I too support a bugfix release
Would that not be good enough for a |
I like this idea @remi-kazeroni ! As far as I can tell we are not using any of the new iris features in the core at the moment, so pinning it to <3.2 for the release and immediately remove this pin afterwards for new developments sounds fine to me. The alternative approach would be to leave the pins as they are for the current release and hope that the bugfix release of iris doesn't break anything for us. If it does we would need a |
We (the release managers) had another discussion about this. For us, pinning Our main arguments for this are:
I will open a PR with these changes now (see #1525). |
Fully support this because ⬆️ Cheers for all the work, guys! |
During the latest round of tests I found a pretty severe bug in
iris
: SciTools/iris#4599If a cube with a lazy aux coord is saved as netCDF file, this gives an error. The error does not appear in
iris=3.1.0
. I suggest pinningiris<3.2
for this release.@ESMValGroup/esmvaltool-coreteam
The text was updated successfully, but these errors were encountered: