The subject for this should have said 2026-10-28, the meeting was last night.
On Thu, Oct 29, 2026 at 3:13 PM James Daugherty <[email protected]> wrote: > We meet every week via a google meeting. The times alternate each > week to encourage attendance from multiple time zones. Each week we > will provide a summary here to facilitate historical preservation & > further discussion. If you would like to join this meeting, you can > do so here: https://meet.google.com/her-tjpt-xmf > > The next meeting is at 10 AM Eastern on November 4th. > > > Action Usage > * https://the-asf.slack.com/archives/CBX4TSBQ8/p1792520326623569 > * https://github.com/apache/datafusion/issues/25088 > * https://github.com/apache/grails-core/pull/16779 > * Airflow intends to share its setup for self-hosted runners. Grails will > wait for that work and then adopt it. > > > Incremental Compilation Fix for grails.factories > * https://github.com/apache/grails-core/pull/16780 > * 8.0 or 8.1? 8.0 means we may be able to turn on cached CI sooner once > upstream Gradle changes are adopted, if they are. > > > https://github.com/apache/grails-core/pull/16781#discussion_r4393103167 > * Merge, or fix this before merging? > * Data binding change; we need more reviewers on this one. > > > grails-micronaut release process failure > * https://lists.apache.org/thread.html/poqnkl67g1ko12576zvphbjs0zrt6odn > * https://github.com/apache/grails-core/discussions/16672 > * James Fredley pushed forward but will not have more time this week. > > > CVEs > * External > - > https://github.com/apache/grails-core/actions/workflows/vulnerability-scan.yml > - 7.x.x and 8.0.x have been updated since the last releases. > > > PR Breakdown for Release Discussion > * 7.0.x (4) > - #16726 Publish nothing from a CI run cancelled before its builds > finish, and correct the CI policy docs > - #16771 7.0 (fix): Hold the list of an indexed key in the nested map > that owns it > - #16772 7.0 (fix): Export the entries of a map written by map > subscripts once, in brackets > - #16784 Fix outer session connection closure and exception handling in > withNewSession > * 8.0.x (25) > - #16516 Fix when render(template:) throws "Could not resolve view" > - #16545 Documentation updated for Java 25 memory requirements > - #16559 Hibernate 7: Honour the join type of nested association blocks > in criteria queries > - #16573 Fix when H7 createAlias joins the association twice and ignores > the join type for the alias > - #16624 Validate the constraints block of Bean Validation annotated > domain classes and expose their constraints > - #16627 Fix private trait field access in TraitPropertyAccessStrategy > - #16719 Keep a parenthesized HQL scalar query from being typed as the > domain class > - #16720 Resolve every column of an enum unique-group member > - #16721 Unwrap every Groovy proxy layer when resolving a GORM API > - #16722 Read spring.messages.basename with Binder so a list is not > dropped > - #16723 Query multi-tenant load and proxy only in discriminator mode > - #16725 Docs: Add the mongodb-developer skill and scoped agent guides > for build-logic and grails-gradle > - #16736 Bind flattened dataSource.properties from the Spring Environment > - #16756 Hibernate 7: Write the rows of lists and maps that share a join > table > - #16763 Hibernate 7: Prevent outer session connection closure during > nested withNewSession lazy loading > - #16773 Hibernate 7: Fix subquery correlation with top-level > association blocks > - #16780 Incremental Compilation Support for grails.factories > - #16781 Fix data binding re-evaluating domain constraints on every bind > - #16786 Give a task a copy of the request once the request completes > * 8.1.x (3) > - #15956 Seed URL mappings index with runtime fallback > - #16000 Add optional artefact index reader for startup scanning > - #16510 Address #15956 review: Package-private URL mappings index reader > * 9.0.x (4) > - #16180 Start Grails as a native image without recorded metadata, > checked by an end-to-end i18n module > - #16583 Validate a unique group with a null value against the rows > whose group value is null > - #16732 Treat a dotted configuration key past a value that is not a map > as absent > - #16760 Let Spring Boot register the character encoding filter > * Decision: Call for votes for 7.0.x, 7.1.x, 7.2.x, 8.0.1, 8.1.0, and > 9.0.0-M1 24 hours after merging the 7.x PRs. GitHub Actions and the > IntelliJ plugin need votes too. > > > CI Policy Change discussion of changes from #16673 ( > https://github.com/apache/grails-core/pull/16673) to #16726 ( > https://github.com/apache/grails-core/pull/16726) > * [skip tests] on pull requests > - What's new: Before #16673 ( > https://github.com/apache/grails-core/pull/16673), the marker only worked > on pushes. The old check read github.event.head_commit.message, which a > pull request run doesn't have, so a PR's tests always ran. After #16673 ( > https://github.com/apache/grails-core/pull/16673), the policy reads the > PR's head commit, so a contributor can skip the tests of their own PR. That > includes PRs into 8.0.x and above, where tests otherwise always run. > - Example: > - A fork PR into 8.0.x has one commit titled: [skip tests] Fix X. > - PR run: Every test job is skipped. The build jobs show green, and > the test jobs show as skipped. The "Tests did not run" notice appears only > on the run's summary page. The reviewer can't override it because the > marker wins over the run ci label. > - Merge: With a merge commit ("Merge pull request #N …"), the push > runs the tests; that's James's point. A squash merge reuses the commit's > title for a single-commit PR, so the commit is [skip tests] Fix X (#N), and > the push skips the tests and publishes the snapshot anyway. A rebase merge > keeps the commit unchanged, so the push skips them too. > - Result: That change never gets its own test run. The next push to > 8.0.x tests it mixed in with later commits. > - The push half isn't new. Before #16673 ( > https://github.com/apache/grails-core/pull/16673), [skip tests] anywhere > in the message skipped a push's tests. Squash messages include every > commit's message, so the old check was even easier to trigger. #16673 ( > https://github.com/apache/grails-core/pull/16673) made the push side > stricter: The marker must now start the first line. James's merge-commit > argument is about this half. What's new is the PR half. > - Options: > - Leave it. Reviewers watch for the marker. > - Members only (Scott's idea). On a PR, the marker counts only when > pull_request.author_association is OWNER, MEMBER, or COLLABORATOR. > Committers keep the shortcut, and outside contributors can't skip their own > tests. One caveat: A committer whose organization membership is private may > show up as CONTRIBUTOR. That fails safe because their tests just run. > - Ignore the marker on PRs, as before #16673 ( > https://github.com/apache/grails-core/pull/16673). > * run ci on drafts > - It has nothing to do with publishing: PR runs never publish. The issue > is that run ci does two jobs: > - The draft gate in every workflow: A draft runs nothing unless it > carries run ci. > - A test request in ci-policy.yml: On any PR, run ci runs the tests > even where the branch rule would skip them. > - PR into 8.0.x and above: > - Draft: Nothing. > - Draft + run ci: Full tests. > - Ready: Full tests. > - Ready + run ci: Full tests. > - PR into 7.0.x/7.1.x/7.2.x or a side branch: > - Draft: Nothing. > - Draft + run ci: Full tests. > - Ready: Build only. > - Ready + run ci: Full tests. > - For a draft below the default branch, there's no way to get the > build-only run. To check that a draft backport into 7.0.x compiles, you > either run the full matrix (hours of runner time) or mark it ready. Before > #16673, every draft ran the full matrix, so nothing was lost; it's a gap in > the new design, not a regression. > - The fix, if wanted: A separate run tests label for the test request, > so run ci only opens the draft gate. A draft into 7.0.x with run ci would > then get the build-only run. The cost is one more label, and run ci stops > meaning "run the tests" on ready PRs into 7.x, which is how AGENTS.md and > RELEASE.md describe testing a backport now. A draft with only run tests > would also need run ci, unless every workflow's gate is changed too. >
