Repository navigation
test(sessions): I cover tied-timestamp persistence gaps - #7492
Draft
jabrailkhalil wants to merge 1 commit into
Draft
jabrailkhalil wants to merge 1 commit into
jabrailkhalil wants to merge 1 commit into
Conversation
I capture fresh-reader ordering, recent-event limits and same-timestamp stale writers with strict xfails while schema alignment remains open. Related: google#7286.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Link to Issue or Description of Change
Related: #7286.
I prepared this tests-only draft following the suggested regression-first approach.
I reproduce two persistence gaps:
(timestamp, id)ordering changes a tied-timestamp append history and its recent tail; usingevent.timestampas the revision marker lets a stale writer append after another same-timestamp commit.I added file-backed tests with independent service instances, 40 events in forward/reverse ID order, recent limits
None/0/1/5/100, and same/increasing-timestamp writer controls. I mark only the five known failures with strict, assertion-onlyxfail. I retain the existing ID-order tests and leave production code, schemas and migrations unchanged while the ordering/schema decision remains open.Testing Plan
I ran the full tox matrix and repository hooks on Linux:
I used parallel pytest within tox and the same
lancedbinstall exclusion as upstream CI. The validation workflow is separate from this PR.I also ran:
7 passed, 5 xfailedon every matrix version; with--runxfail, exactly the five intended assertions fail.586 passed, 6 xfailed.pre-commit run --all-files: passed; mypy comparison: 828 errors on both base and head, no new errors.uv build, a clean wheel installation/session round-trip, and the six model-free Workflow replay scenarios in both development and wheel environments: three successful controls and three expected replay divergences.I reviewed the diff. I tested SQLite-backed storage; I have not tested PostgreSQL/MySQL/Spanner or an actual concurrent scheduling race. I am recording current failures, not claiming a production fix.