…eel (#95)

The post-release job verified the publication against two different PyPI
endpoints and treated them as one. 'Verify PyPI identity and digests' reads
https://pypi.org/pypi/<name>/<version>/json; 'Install and exercise the
published wheel' reads https://pypi.org/simple/ through uv. They propagate
independently, so the first can pass on a version the second has not listed.

Measured on 0.1.7: publish completed 11:59:22Z, the install failed 24s later
with 'no version of ml4t-backtest==0.1.7', and a re-run of the same job at
12:03:22Z passed unchanged. Nothing was wrong with the release; the check was
faster than the index.

Retry the install for up to 12 attempts over 120s, the same budget the
deployed-documentation recheck already uses below it, and fail with an
explicit message naming the simple index if it never appears. A genuinely
absent artifact still fails, two minutes later.


Claude-Session: https://claude.ai/code/session_01UPDj1nioTuD4xATMVyhZxB

Co-authored-by: Claude Opus 5 (1M context) <[email protected]>

… failing silently (#97)

* Reconcile time zones before the as-of join, so trade enrichment stops failing silently

enrich_trades_with_signals joins trade times against signal timestamps, and the two sides
reach it from different places: trade times come out of the engine carrying the bar clock
without its zone, while a prediction panel is read from parquet with its zone intact.
join_asof refuses a naive left key against an aware right key, and BacktestProfile wraps
this call in a try that degrades to an empty frame with a UserWarning - so the reader does
not get a traceback, they get a tear sheet quietly missing its ML trade-alignment charts.

Found in case_studies/nasdaq100_microstructure/20_strategy_analysis.ipynb, whose committed
output carries 'Prediction enrichment failed: datatypes of join keys don't match -
entry_time: datetime[us] on left does not match timestamp: datetime[us, UTC] on right'.

A naive column here is the same wall clock with the zone dropped, so it is localized and
never shifted. Measured on that case study: reading the naive trade times as UTC matches 703
of 905 trades to an exact (symbol, instant) row in the prediction panel, and reading them as
America/New_York matches 199; the residual 202 are trades entered between two prediction
instants, which is what a backward as-of join is for. Shifting instead of localizing would
still return rows, so getting this backwards would be silent rather than loud. Where both
sides are aware and disagree the signals are converted, because two aware columns already
name instants.

The negative control neutralizes the helper with a monkeypatch and asserts the join still
raises, so the other six tests are evidence that this path is load-bearing rather than
assertions polars would satisfy on its own.

* Re-derive the retained correctness evidence under the current engine source

The timezone-alignment change moves the engine tree digest, and provenance.py stamps
that digest into every retained record, so all 68 read stale against the working tree.
Correctness is the kind of evidence that is genuinely re-derived rather than certified:
it is cheap and tolerant of a contended machine.

66 pass, 2 unsupported, unchanged from the previous matrix. Compared leaf by leaf against
the committed copy: 231,740 values, 201 differ, and every one of the 201 is a digest, a
commit or a timestamp. No correctness value moved.

Every record is stamped dirty:false, so this was produced from a clean tree.

* Rebuild the real-strategy evidence under the current engine, and certify the timings against it

The timezone-alignment change moves the engine tree digest, and the three retained evidence
files answer to it in three different ways. Correctness was re-derived in 281370c. This is
the other two.

REAL_STRATEGY_RESULTS.json has a hard equality on the digest and no certification path, so
it is rebuilt from runs: all 17 required pairs regenerated and compared, 17/17 pass, and the
promoted file carries commit 281370c with dirty false. The first rebuild today was refused
by its own gate - it ran while CORRECTNESS_RESULTS.json was still uncommitted, and
provenance.py stamps dirty from git status taken at build time - so the driver that produced
this one refuses on a dirty tree rather than discovering it eighteen minutes later.

REAL_STRATEGY_PERFORMANCE.json is certified rather than re-timed. The timed region is the
single line result = engine.run() at validation/real_strategy_runner.py:386, the declared
timing_policy excludes result extraction and reporting, and enrich_trades_with_signals - the
only caller of the new _align_join_time_zones - has no caller anywhere in src/ and runs after
that call returns. So the added code cannot execute inside the measured interval. That
mechanism is the stored reason; correctness parity alone would not be sufficient and
certify_source's own docstring says so.

generate_parity_claims.py --check passes against the regenerated claims.

Bumps the github-actions group with 1 update: [astral-sh/setup-uv](https://github.com/astral-sh/setup-uv).


Updates `astral-sh/setup-uv` from 10.0.1 to 10.1.0
- [Release notes](https://github.com/astral-sh/setup-uv/releases)
- [Commits](astral-sh/setup-uv@20cfd1b...bec219d)

---
updated-dependencies:
- dependency-name: astral-sh/setup-uv
  dependency-version: 10.1.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: github-actions
...

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Stefan Jansen <[email protected]>

Closes #100