https://pantsbuild.org/ logo
Join Slack
Powered by
I notice that the `publish-tag-to-commit-mapping` ...
# development
h
I notice that the 
publish-tag-to-commit-mapping
CI job failed for 2.15.0rc2, yet the mapping exists (and installing that version with scie-pants succeeds).
Copy code
$ aws s3 cp <s3://binaries.pantsbuild.org/tags/pantsbuild.pants/release_2.15.0rc2> -
4e0a3fd845ca9c71792ad3d032542924a6cdcfd2
I wonder if we're running this twice, and the failure is on an attempt to sync the same file again when it exists
It's consistently failing for every release, yet the mapping exists for all of them: https://github.com/pantsbuild/pants/actions/workflows/release.yaml
e
When folks announce, I run that.
h
Ah, so you've been manually backfilling
e
This is a dead man walking after all. Next work is pants scie.
There was talk in other channels of pants native client work, but it does not seem to be appreciated that pants scie is needed 1st to support that.
h
Why is this a dead man walking? Do we not need this in perpetuity?
e
Releases for scies will be (currently) 4 PEXes on GH releases (+ lift manifest and - later + native client) with 0 use of S3. The PEX will have all wheels.
So it's dead at that cutover point and the last mapping will have been done.
h
To clarify my understanding from looking at the scie-pants code, it currently installs Pants from pypi with find_links from S3? So using the GH release pexes is future work?
e
Yes, that is what make pants scies. Today you just have scie pants which uses the old install method of pip + venv.
So this is ~pure Pants side work.
There will be an update to scie pants to handle the two forms of Pants release.
I did the bare minimum to get scie-pants working in other words. Pants is still not a scie and we still abuse PyPI.
Once Pants community signs off on scie pants, the pants scie work can begin. You started pushing that along a few days ago; so maybe time to start planning / detailing that work.
h
Got it, thanks for the context. Is there a ticket for this work?
e
No.
This is part of why I was beside myself re native client. It seemed clear no one was thinking at any level of detail here
h
I certainly didn't have this context.
e
This is not personal context though. It's dictated by public facts
h
Public facts that I was not fully aware of, alas
In any case, to rewind back to the start of this thread, that job is failing because we're not setting the region. I'll send a fix shortly.
Not sure why this works in the wheel-build jobs though
Open in Slack
PreviousNext
Powered by