Closed paleolimbot closed 2 months ago
Hmm, I don't directly see an option for the command line to allow overwriting existing files (I know that for scientific python index at https://anaconda.org/scientific-python-nightly-wheels/repo, many packages use an identical dev version, but that's of course using a different tool to upload).
So I assume the only option to get it working with fury is to ensure we get unique version names ..
One problem is that we don't have tags on he main branch, so git describe
based approaches don't work here
On the Apache Arrow repo after bumping versions we push the dev
tag:
https://github.com/apache/arrow/blob/main/dev/release/post-11-bump-versions.sh#L124-L129
This is all me (for managing my tags improperly!). I think ADBC rebases on the release commit, which I never quite got to. The heuristic I use for the changelog is to find the version bumping commit and use that:
ADBC doesn't actually rebase on the release commit, we just have a script that manually bumps a version string in a bunch of places
Git describe seems to work for ADBC:
...so I assume I'm just putting the tag in the wrong place? The "rebase" comment is my misinterpretation of this chunk:
That git describe command is not working for me locally for ADBC .. (not sure if I am doing something wrong)
But as @raulcd mentioned, we can probably do the same as Arrow and push a dev tag to main
I'm game for either wiring up the logic in changelog.py
or by pushing a dev tag (but one of you might have to point me to where this happens in the Arrow repo).
edit: 🤦 I see that Raul already linked it
Git describe seems to work for ADBC:
...so I assume I'm just putting the tag in the wrong place? The "rebase" comment is my misinterpretation of this chunk:
Hmm, that's a weird way of writing git pull --ff-only apache main
I believe, given we don't make any other commits beforehand and local main shouldn't have diverged from upstream main.
After #427 we now upload wheels! This job seems to now be failing:
https://github.com/apache/arrow-nanoarrow/actions/runs/8768760925/job/24063892961#step:5:78
The issue seems to be that the version is identical:
cc @jorisvandenbossche @raulcd