-
Notifications
You must be signed in to change notification settings - Fork 16.4k
Sync v2-4-stable with v2-4-test to release 2.4.0 #26371
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
Conversation
When SHELL variable is not set, kubernetes tests will fall back to using 'bash' (cherry picked from commit b7db311)
The latest docker-compose dropped an alias for `docker-compose` and we need to detect it and use "docker compose" instead. (cherry picked from commit 7f47006)
* Limit eager upgrade of protobuf library to < 4.21.0 Until all the Google client libraries get upgraded to >= 2.0.0, we need to limit the protobuf version. (cherry picked from commit c6053f5)
Since we are not always updating the version suffix in code (rc1/rc2/b1/b2), the tag specified via --tag-build prefix should override the one in code rathe than the other way round. So what's left now - we will just print warning if the suffix does not match. (cherry picked from commit 9cf6f6a)
In v2-4-test it turned out that PROD cache build was a bit too limited - the cache has not been prepared if branch was not main (copy&paste victim). This PR fixes this - cache is always build in merge run regardless of the branch we are in. (cherry picked from commit 4178545)
This makes them proper sections and reorders a couple of them as well. (cherry picked from commit 49fb3ff)
Currently constraints generation is not really showing good progress while the packages are being removed/installed. This adds progress that shows that something happens. (cherry picked from commit d6f473b)
(cherry picked from commit 5066844)
This incorrectly accesses the version *attribute* in airflow.version instead of the airflow.version module itself, breaking compatibility. I actually rewrote the entire __getattr__ hook; much of the logic seems to be quite redundant. (cherry picked from commit b7a603c)
The previous way of building the path depended on the OS path but it was really used to build the URL so we should use urllib instead of os.path.join. (cherry picked from commit 1838602)
(cherry picked from commit aada79e)
When we go to schedule dagruns from the dataset dagrun queue, we assume the events will happen before the queue records, which isn't the case unless we explicitly flush them first. This ensures that dagruns are properly related to their upstream dataset events. (cherry picked from commit 954349a)
(cherry picked from commit 3c9c0f9)
The cause of the failure was a PR race -- the PR that changed migrations was opened and passing tests before the ERD sha was added. (cherry picked from commit 12b5445)
jedcunningham
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We also need the AIP-48 catchall in new features.
6bda3ae to
a8ce509
Compare
Co-authored-by: Konstantin Weddige <konstantin.weddige@lutrasecurity.com> (cherry picked from commit 56e7555)
Previously we had the validation on the Dataset model, but we since moved the "dag" facing class to a separate one. This adds validation to the public class, and extends the validation to not allow space-only strings (cherry picked from commit bd181da)
I'm not sure when we started using attrs directly, but we do and we need >=22.1.0 as we use the min_length validator. (cherry picked from commit fdecf12)
We need to clear any autoregistered DAGs that may have been already registered if we encounter any import errors while parsing a given DAG file. This maintains the behavior before we autoregistered DAGs. (cherry picked from commit 01e3fb7)
While it would ideal to transition to a positional arg like was attempted in #25978, this unfortunately does result in a breaking change so we cannot do it now. By instead marking the existing arg as required, we maintain backcompat while also providing a helpful error message to the user if they forget it. This reverts commit ed6ea72. (cherry picked from commit f6c579c)
a8ce509 to
7ba5963
Compare
3348a3b to
bef7857
Compare
ashb
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh. We need to add datasets to the release notes!
It's the first in new features? |
Sure, but it needs more than just a bullet point :D (i.e. a big call out like we have for the other Significant Changes) Working on it |
potiuk
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Reviewed all, all looks familiar. I additionally cherry-picked one change #26392 which was missing (not sure where and how)
Probably because it doesn't have a milestone |
The change has been cherry-picked so we need to update the docs. (cherry picked from commit 41e48aa)
643a947 to
a63d260
Compare
6dd4ff9 to
2c9109b
Compare
Likely 🤦 |
256dede to
0bcdba0
Compare
|
🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 🎉 |
No description provided.