Restore the release H5 write that #437 dropped - #444
Merged
Conversation
…tifact) Fixes populace#443. #437 collapsed dataset_path = artifact_root / DATASET_FILENAME release_engine.write_dataset(export_frame, dataset_path, period=PERIOD) into the bare assignment while inserting the batched pre-export raise. Since then no run has written its export H5: Build M attempts 13/14 smoke-scored attempt 10's stale July-15 artifact (keogh +$11,809, byte-identical to pre-swap runs, despite 13 verified carriers at $148.97M in the frozen selection), and a green run would have sha-pinned the stale file into the release manifest. The smoke gate firing on stale content is the only reason nothing shipped. Restore the write at the same point: after the batched raise (a gate-failed run must not produce the H5) and before the smoke (which must score the just-written file). CI passed on #437 because no test runs main() past the batched gates. Until the green-path main() harness exists (populace#443 remedy), an AST-level test pins the ordering contract - exactly one write_dataset in main(), strictly between the single batched terminal raise and the reform-coverage smoke - and fails when the call is absent. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #443
What happened
#437 (my batching PR) accidentally deleted the export H5 write while inserting the batched pre-export raise:
Since #437 merged, no release run has written its export H5. Build M attempts 13/14 smoke-scored attempt 10's stale July-15 artifact — keogh scored +$11,809, byte-identical to the pre-swap runs, despite the frozen selection carrying 13 verified keogh carriers at $148.97M — and a green run would have sha-pinned the stale file into the release manifest. The smoke gate firing on stale content is the only reason nothing shipped: the gate did its job end to end.
What
release_engine.write_dataset(export_frame, dataset_path, period=PERIOD)at the same point in the pipeline: after the batched raise (a gate-failed run must never produce the H5 — the Batch pre-export terminal gates into one report per run #437 contract) and before the reform smoke (which must score the just-written file). A comment pins why the line sits exactly there.test_release_h5_write_sits_between_batched_raise_and_smoke: AST-level ordering contract onmain()— exactly onewrite_datasetcall, strictly between the single batched terminal raise andus_reform_coverage_smoke_gate. Fails on the pre-fix code (verified). This is the stopgap until the green-pathmain()harness (#437 dropped the release H5 write; smoke gate scored a stale artifact (attempts 13/14) #443 remedy) exists — CI passed on Batch pre-export terminal gates into one report per run #437 precisely because no test runsmain()past the batched gates.Full builder suite passes locally (111 tests).
🤖 Generated with Claude Code