Closed mbargull closed 7 months ago
(FYI: I restarted failed CI jobs twice. Some Linux-aarch64
builds stalled for some reason. https://github.com/conda-forge/miniforge/actions/runs/6935425583/attempts/1 )
Restarted all the CI jobs. Will merge once it passes.
Restarted all the CI jobs. Will merge once it passes.
Shall we rather wait a couple of days and discuss on the next core if we want to go with conda=23.10
or rather wait for a 23.11
which will (hopefully) include fixes for better conda-build
<->conda-libmamba-solver
integration?
(Even though conda-build
is not part of Miniforge, it's still used for build setups and conda>=23.10
's default solver change can thus affect those.)
I don't really see the point in waiting a couple of days.
Generally speaking, I wanted to stop users from being "force to downgrade any package" in the case that boa needed to be installed.
This is a standard usecase for conda's base installer.
However, today in most feedstocks:
conda 23.10.0 py310hff52083_1 conda-forge
mamba 1.5.3 py310h51d5547_2 conda-forge
The latest build I triggered was in https://github.com/conda-forge/boa-feedstock/pull/76, thus I feel like we should update this.
I have an other fear, if we do this these updates too quickly, we don't give time to catch all the conda bugs.
Thus I typically like to "wait" a little bit.
I don't follow.
The latest build I triggered was in conda-forge/boa-feedstock#76, thus I feel like we should update this.
Why "thus"?
I have an other fear, if we do this these updates too quickly, we don't give time to catch all the conda bugs.
Thus I typically like to "wait" a little bit.
We do have the case that 23.10
has (impactful) bugs (surfaced via the solver: libmamba
default change) -- which means we are in a "catch conda bugs" phase right now warranting to "wait a little bit"...
It is rather strange to ignore raised concerns.
@hmaarrfk, shall we wait until Wednesday to discuss the changes and keep the release as pre-release for now?
Sorry about this. sigh. ok. I think I need to spend more time following the condo/boa/libmamba bugs.
let's leave it as a draft until things are resolved....
just I don't doubt we will have more bugs in 23.11
Why "thus"?
more like "it didn't give me any problems" and this upgrade "avoids a regular cycle of updates for every build".
however. I was unaware of the impact of the latest bugs.
however. I was unaware of the impact of the latest bugs.
Ah, I thought you were aware (through core chat, my comment on your original PR, etc.).
To give context: These bugs aren't even new, but just more likely to encounter since conda=23.10
changed its default to solver=libmamba
.
As such, it is more of a "make things less likely/immediately to break for users" thing than "ship things broken" as of now.
I don't know whether we get a conda=23.11.0
/conda=23.10.1
quickly, but I hope to get answers to that on Wednesday.
23.11 should be tagged this week, AFAIK.
Did we add this to the agenda tomorrow?
Edit: Turns out I added it 2 weeks ago and now it is first up 🙂
Conclusion from the meeting today is that we wait until there is a new Conda release (tentatively planned this week)
Edit: This is also being added to the next meeting in 2 weeks for follow-up (unless it has already been handled before then)
Looks like Conda 23.11.0 just came out 🎉
xref: https://github.com/conda-forge/conda-feedstock/pull/228
Checklist
0
(if the version changed)conda-smithy
(Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)closes gh-496 closes gh-511
Continuation of gh-496 using
setup-miniconda
for macOS and updating toconda-libmamba-solver=23.11.1
.