Skip to content

test: reproduce normal-wheel runtime dependency gap - #299

Closed
Pigbibi wants to merge 1 commit into
mainfrom
test/phase1-normal-wheel-baseline
Closed

Pigbibi wants to merge 1 commit into
mainfrom
test/phase1-normal-wheel-baseline

Conversation

@Pigbibi

@Pigbibi Pigbibi commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Disposition

Closed without merging after recording the clean-install baseline. The separate
adoption proposal is draft #300.
That proposal uses a strict successful-import gate and contains no diagnostic
acceptance mode. Neither draft was merged by this task.

Baseline evidence

This diagnostic head, 7025582e5ba99781197379415abbc9a8001a78a0, changes only
verification tooling over main ec0b330081213f56c3ff98164ec00ba26c10dfc9.
Package metadata, runtime source, dependency versions and the lockfile are
unchanged.

Check Observed result
Normal wheel, Python 3.11 Normal pip install and pip check succeed. Public entrypoint import fails with missing Pandas; NumPy and Pandas distributions/modules are absent.
Normal wheel, Python 3.12 The same independently installed result.

Both wheel builds have SHA-256
f990c3b458beb8e1f8ea07e9ad2acae58da85ba3efb956d9ad0f4618f976afc9.
Both reports verify installed module/distribution origins under the fresh
virtual environment and QPK commit
c7646a7168b3dafa763ef7751a182d23e8de7790.

Each job built the normal wheel, created a new virtual environment, and installed
only the wheel and declared normal dependencies through pip. No research extra,
manual numeric installation, editable package, source-path overlay or skipped
missing import supplied this measurement. The probe used python -I outside
the checkout, recorded zero guarded socket/DNS attempts and evaluated no
strategy.

The JSON explicitly retains runtime_import_succeeded: false,
missing_dependency: pandas and expected_gap_reproduced: true. A green
baseline job means that this exact gap was observed. It does not report the
runtime import as successful. Unexpected errors, contamination, source or pin
mismatches, and observed network attempts fail the diagnostic check.

Verification and limits

All five diagnostic checks completed successfully, including the existing unit,
actionlint and documentation jobs. Ten offline verifier tests and an independent
review checked the negative-result classifier and provenance requirements.

The earlier executor attempt failed while fetching unchanged QPK because its
configured proxy was unreachable. That attempt remains an environment failure;
the CI installations above provide the reproduced dependency gap.

ADR 0002 in #300 records the evidence and proposed correction. Adoption remains
subject to the owner's approval. This closed diagnostic PR requires no merge,
deployment, production configuration change or dependency adoption.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant