Skip to content

docs(ci): baseline the honesty-limits line the withdrawn-number gate caught (Closes #2311) - #2313

Merged
gHashTag merged 1 commit into
masterfrom
fix/withdrawn-live-baseline-2311
Aug 20, 2026
Merged

docs(ci): baseline the honesty-limits line the withdrawn-number gate caught (Closes #2311)#2313
gHashTag merged 1 commit into
masterfrom
fix/withdrawn-live-baseline-2311

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #2311

What is red, and since when

withdrawn-live-gate.yml -> job withdrawn-live -> step No withdrawn number in a live document has failed on every master run since #178 (400850702,
2026-08-20T02:01:08Z). Still red at master head 9068cf49d (run 32354816054):

FAIL: 1 withdrawn number(s) stated in a live document

  docs/NOW.md:1127
      matches /323(\.[0-9]+)?\s*MHz/ -- ring-oscillator toggle rate, not a GF16 path
      see RESEARCH_CLAIMS.md Retraction 2026-08-18

The job's negative control passes in the same run (self-check: planted hit found = True, clean file silent = True), so the gate is working. Last green was #176
(c3eeefc87) and exactly one commit sits between: 400850702 docs(arxiv): the LUT count was the flip-flop count (Closes #2280) (#2281), which added 67 lines to
docs/NOW.md and touched nothing else.

Why a baseline row is the right remedy here

docs/NOW.md:1127 opens a bullet under ## Honesty limits (BINDING) stating that
the number is disputed and unsettled and that "Nothing here should be read as
evidence for either side." (The number is elided from every quotation in this PR and
in the NOW entry, because quoting it would plant a second occurrence.)

That is the category the gate names in its own failure message: "If the line is text
ABOUT the withdrawal, add it to tools/withdrawn_live_baseline.txt."

tools/withdrawn_live_baseline.txt already carries five rows for this same
document and this same pattern, plus rows for two neighbouring frequencies in it.
This is a sixth of the same kind.

Editing the document instead was rejected: docs/NOW.md is append-only, and
rewriting line 1127 to please a gate would rewrite the record of what was written
that day.

The key was recomputed, not transcribed

The gate keys a row on sha1(" ".join(line.split()))[:12] of the line. The row here
came from running that algorithm over docs/NOW.md as fetched from master. The same
pass reproduced all five existing keys for this pattern -- 18f57dc20b45,
700c6d6aca11, ce84017e739b, e01a7fd68109, f81569f4b61c -- character for
character. The computation was checked against five known-good answers before being
used for an unknown one.

--update-baseline was deliberately not used: it rewrites the whole file from the
current scan and would bless anything else that happened to be hitting. This PR adds
exactly one line.

Simulated before merging

Replaying the gate's own scan and baseline comparison over the pre- and post-change
trees:

OLD: 11 hits in docs/NOW.md, 1 NOT baselined
   line 1127: docs/NOW.md | 323(\.[0-9]+)?\s*MHz | 93e256f28d7c
NEW: 11 hits in docs/NOW.md, 0 NOT baselined
stale docs/NOW.md rows after the change: none

The OLD figure reproduces CI exactly, which is the evidence that the simulation is
faithful. Hit count is unchanged at 11 -- nothing was hidden, one occurrence was
classified.

What this does not do

  • It does not settle the dispute the line is about. specs/numeric/formats_catalog.t27
    and Unbreak FPGA CI and correct two unbacked hardware claims #2081 are untouched.
  • It widens no pattern in tools/withdrawn.txt, relaxes no scan, adds no exclusion
    prefix, and removes no assertion.
  • Keying is on line content, so if that sentence is ever edited the gate re-opens.

Diff

file change
tools/withdrawn_live_baseline.txt +1 / -0 -- one row
docs/NOW.md +72 / -0 -- new top entry, pure prepend

…caught

`withdrawn-live` has been red on master since run #178 (4008507) on a single
line, `docs/NOW.md:1127`, which states that the number it contains is disputed
and unsettled -- text ABOUT the withdrawal, which is the category the gate's own
remedy names for `tools/withdrawn_live_baseline.txt`.

One row is added. The key is the gate's own `sha1(" ".join(line.split()))[:12]`,
computed over `docs/NOW.md` at master; the same pass reproduced all five existing
keys for this pattern, which is what makes the sixth trustworthy. Keying on line
content means the exemption covers one sentence and expires when it changes.

No pattern is widened, no scan relaxed, no exclusion prefix added. Simulated
against the tree: 11 hits in docs/NOW.md before and after, 1 unbaselined before
and 0 after, with no row left stale.

Closes #2311
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-08-20 10:02:57 UTC

Summary

Status Count
Total Open PRs 21
PRs with Failing Checks 6
PRs with All Checks Green 15
READY 7
FAILING 6
PENDING 0

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=cbbfac87dff3 != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@github-actions

Copy link
Copy Markdown
Contributor

📓 NotebookLM Notebook linked to this PR

This notebook contains session context, decisions, and artifacts for this work.

@gHashTag
gHashTag merged commit 07e8d71 into master Aug 20, 2026
21 of 23 checks passed
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.

withdrawn-live is red: one honesty-limits line in docs/NOW.md needs a baseline row

1 participant