fix(ci): treat missing merge_insert base tag as not generated - #8177
Merged
Conversation
bench_regress has failed on every main push since #8052 because _already_generated assumed Tags.get_version returns None for a missing tag, while the API raises ValueError. Shared GCS datasets left without merge_insert_base then crash datagen before overwrite can recover. Use tags.list() for presence checks and correct the get_version docs.
Wire python/ci_benchmarks/datagen into make test so PR Python CI catches readiness-check regressions. Run the same suite before gen_all in bench_regress so pure-Python failures do not wait on shared GCS generation.
Collecting python/ci_benchmarks/datagen in make test put the source tree on sys.path ahead of the installed wheel, so import lance resolved to pure-Python sources without the native extension and macOS CI failed with BlobDescriptor ImportError. Move the readiness-check unit tests into python/tests, load the installed lance package first, then import ci_benchmarks helpers.
Leaving the source tree on sys.path was inherited by multiprocessing spawn children (fragment progress and torch DataLoader workers), which then imported pure-Python lance without the native extension.
ci_benchmarks is not part of the installed wheel, so PR-suite coverage needed sys.path hacks that broke multiprocessing spawn. The readiness check fix itself is sufficient; skip the non-essential test wiring.
BubbleCal
approved these changes
Aug 3, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
Run Regression Benchmarksonmainhas failed on every push since #8052 landed. The failing step is Generate datasets:Example: https://github.com/lance-format/lance/actions/runs/30790358210/job/91612371208
Root cause
_already_generatedtreated a missingmerge_insert_basetag asNonefromTags.get_version. That API raisesValueErrorinstead. A shared GCS dataset that exists without the tag therefore aborts datagen and never reaches the overwrite path that would repair it.The Python docstring for
get_versionincorrectly documentedOptional[int]/None.Fix
tags.list()in datagen and the merge_insert benchmark fixture.Tags.get_versiondocs/signature to match the raising behavior.No extra unit-test wiring for
ci_benchmarksin the main Python suite: that package is not installed with the wheel, and importing it from PR CI needs path hacks that break multiprocessing spawn tests.After this merges, the next
bench_regressrun should overwrite the incomplete GCS datasets and retag them.