Skip to content

docs(mirror): record #4746 Item A batch 3 (#4751) and the ratchet-slack decision point - #4752

Merged
phasetr merged 1 commit into
mainfrom
docs/mirror-4746-batch3
Jul 28, 2026
Merged

docs(mirror): record #4746 Item A batch 3 (#4751) and the ratchet-slack decision point#4752
phasetr merged 1 commit into
mainfrom
docs/mirror-4746-batch3

Conversation

@phasetr

@phasetr phasetr commented Jul 28, 2026

Copy link
Copy Markdown
Owner

Summary

Records #4746 Item A batch 3 (PR #4751, main 051aa857) in the local issue mirror.
.self-local/issues/4746.md only; no Lean source, no docs/, no tex/, no
scripts/ file is touched, so no library gate can move.

What the entry records

  • The ten deleted modules (724 lines / 39 declarations), the ten dropped import lines,
    and the Moved: doc-block corrections in eight surviving siblings.
  • Condition (ii): the eight survival-count retractions in docs/index.md and
    tex/proof-guide.tex, plus the eight stale survival-location module names the
    follow-up commit corrected, and the three that were deliberately left alone.
  • Condition (i): --lean was not reported in the PR body, so the tier-1 audit ran it
    against a green pre-deletion build — no consumer seen by Lean was missed on a
    safe-to-delete verdict
    .
  • Gates re-run on the merged main: lake build 5,015 jobs / 0 warning / 0 error
    (5,025 before, difference exactly ten), citation_audit.py PASS with 960 gating
    findings and ratchet 0/0, dead_candidate_scan.py --self-test 100 OK,
    test_audit_gate.py 166 OK, lake exe GKSTest pass, #print axioms std-3 only.
  • The one PR-body number that was wrong (grep -rn "sorry" IsingModel/ claimed 4 prose
    hits, actual 2) and its correction in the squash message.

Two findings carried into the mirror

The recorded Item A denominator no longer reconciles. Subtracting batches 1-3 from
the recorded 64 leaves 39, but a full-tree sweep on the merged main reports 205
fully-dead modules excluding Lemma_17_5_2 (231 on the pre-merge main), and
recomputing the same quantity from the tier-2 report's own stored verdicts gives 178
against a recorded 75. The rule behind the recorded figure is not recoverable from the
stored data, so the denominator should be re-derived and written down before batch 4 is
scoped.

F1 is now due rather than deferrable. Measured slack after batch 3 is 36 / 36 / 46
against CHECKED_FILE_FLOOR / LIB_FILE_FLOOR / the new V4_FILE_FLOOR; one more
ten-module batch takes it to 26 / 26 / 36 and two more exhaust the named floors.

Also carries a cascade warning: 30 declarations became reference-0 as a consequence of
batch 3, six of them abstract GJ §18 statements in
IsingModel/ClusterExpansion/MayerCore/Terms.lean, which a later batch must not sweep.

Part of the mirror upkeep for #4746.

🤖 Generated with Claude Code

…ck decision point

Appends the batch-3 progress entry to the local #4746 mirror: the ten deleted
modules, the docs/tex retraction and the survival-location corrections, the
independently re-run gates on the merged main, and the two corrections the tier-1
audit made (the `--lean` condition that the PR body omitted, and the wrong prose
`sorry` count that was fixed in the squash message).

Also records two findings the batch surfaced. First, the recorded Item A
denominator no longer reconciles: subtracting the three batches from the recorded
64 leaves 39, while a full-tree sweep on the merged main reports 205 fully-dead
modules excluding `Lemma_17_5_2`, and the rule behind the recorded figure is not
recoverable from the stored tier-2 data. Second, F1 is now due rather than
deferrable, since the measured slack is 36/36/46 and one more ten-module batch
takes it to 26/26/36.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@phasetr
phasetr merged commit bfc4e32 into main Jul 28, 2026
1 check passed
@phasetr
phasetr deleted the docs/mirror-4746-batch3 branch July 28, 2026 06:39
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