Repository navigation
Copy base-out recursively in the Modal base hand-off blocks - #1677
Conversation
Since #1635 a base writes its gate-shadow files in `base-out/gate_battery_shadow/`, and a Modal receipt lists every file under the state tree. Step 7 of the Modal stage runbook and the 18 October rerun record copied `base-out/*` without `-R`, so `cp` exited 1 on that directory and, under `set -euo pipefail`, ended the block before `verify-receipt`. Both blocks now copy with `-R`, and the rerun record keeps `-c` for the APFS clone. A new hand-off test gives the Modal run that directory, lists its files in the receipt, and runs the runbook's block as the other hand-off tests do. axiom: n/a: documentation and a test; no rule, parameter or dataset change Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
No hand-off has run on a base built since #1635, so the old copy has stopped none yet. The earlier commit's message reports a local run of both copies on a small tree. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Main gained #1650, #1653, #1559 and #1635 after the old pin, c7a5f96. 198 of the log's 200 code links now cite main at bcb5cba, and two dated observations keep the commits their runs used. The log drops the `uv sync` entry. Since #1650 the driver syncs with `--python "$UV_SYNC_PYTHON"` and unsets `UV_FROZEN`, so neither cause reaches `uv sync in $WT`. The code removed that stop; no run passed it. New entries: - 2 (`predicted`): the 2023-2025 pool needs `census_cps_2025.h5` under `STORAGE` and `asecpub26csv.zip` under `EDU` (#1559). - 6 (`from the code`): a rerun whose worktree synced with another interpreter stops at the sync marker (#1650). - 9 (`predicted`): the Modal hand-off's copy stops on the base's `gate_battery_shadow/` directory (#1635). #1677, still open, makes that copy recursive. Moved or rewritten: - The emitter entry moves up to 5. The driver now checks the installed emitter in the commit check, before it creates a worktree (#1650). - The gate-shadow entry, now 8, states what main does, since #1635 merged. - Entry 1 quotes the message #1653 added for a file the driver cannot hash. - Entry 4 quotes the feed and crosswalk blockers as #1650 reworded them. - Entry 7 adds the 2023-2025 Modal template and the reference a hand-off needs. - Entries 3, 10 and 11 keep their stops at new line numbers. - "The run" loses its paragraphs on #1650 and #1653; the entries carry them. axiom: n/a: documentation only Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
|
Merged at |
…o the engine refresh main brings #1681 (rulings 1 and 3 on #1672: a build has no tier and certification chooses it; certification may run at a descendant of the build's commit that changes only tests, docs, changelog fragments and the standing evidence owners register) and #1677 (the Modal base hand-off copies the gate shadow directory). No textual conflicts. Checked where the two sides meet in tools/build_us_fiscal_refresh_release.py: the release still writes the base's household-location attributes after the written-H5 stored-input check and before the build record binds the H5's sha256, and certification still reads them back from the written H5. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
The independent review of this change came back after the merge with one defect and three smaller points. The defect sits in the 18 October record's hand-off block. Its guard, #1695 holds the smaller points: a failed digest that step 7 does not stop on, symlinks in Two corrections to the description here. The dated block had already run a strict |
Summary
Since #1635 merged, a US base writes its gate-shadow files in
<out>/gate_battery_shadow/(build_us_puf_support_base.py#L1003-L1014).
A Modal base's
--outsits atstate/base-out(modal_us_stage_plan.py#L1026-L1027),
and the runner lists every regular file under the state tree in the receipt
(#L2963-L2984).
On main, two hand-off blocks copy
base-out/*without-R:(us-modal-stage-runbook.md#L1391);
(us-route-a-rerun-20261018.md#L471).
cpexits 1 on a directory it gets without-R. Both blocks run underset -euo pipefail, so the copy would endeach block before
verify-receipt --prefix base-out --strict, and the base would not reach Route A. No hand-offhas run on a base built since the merge.
The runbook's block names the reference file for commit
4b57d15a2(us-modal-stage-runbook.md#L1387-L1389). For a base from another
commit,
compare-lineagerefuses the receipt's commit(#L4213-L4217) until the operator names that commit's reference,
and the copy stops the block next.
What changes
cp -pR. The rerun record copies withcp -pRc, which keeps the APFS clone.test_the_hand_off_copies_the_base_s_gate_shadow_directorygives the fixture's Modal run agate_battery_shadow/directory, lists its files in the receipt, and runs the runbook's block as the otherhand-off tests do. It asserts exit 0, both files under
$RUN/base-out/gate_battery_shadow/, andbase-sup/ACCEPTED.Checks
base-outof two files and one directory three ways.cp -pexited 1 withis a directory (not copied), andverify_receipt(strict=True, prefix="base-out")then reported both nestedfiles missing.
cp -pRandcp -pRcexited 0,verify_receiptreported no problem, and the nested fileskept their modification times.
compare-lineagechecks that the receipt holds each output the local reference lists(#L4256-L4264), so the receipt's extra entries pass it.
ruff formatandruff checkpass on the test file. The new test has not run locally: CI'sengine-freejobs run it.
axiom: n/a: documentation and a test; no rule, parameter or dataset change
🤖 Generated with Claude Code