Skip to content

fix(gates): close the last six boundaries (#2492) - #2507

Merged
gHashTag merged 3 commits into
masterfrom
fix/six-boundaries
Aug 23, 2026
Merged

gHashTag merged 3 commits into
masterfrom
fix/six-boundaries

Conversation

@gHashTag

Copy link
Copy Markdown
Collaborator

All twelve boundary survivors from #2503 are gone. The suite stands at 111 mutants across 13 gates — 36 silent, 21 loud, 33 invert, 21 boundary — with two survivors, both proven equivalences that carry their proof in the row.

The six closed here, and what they had in common

Every one needed an input no case had a reason to build, and in each the missing input was the degenerate one:

boundary the input nobody planted
v[0] > 0 in the prose census a file with zero cases
total > 0 and data == 0 for a new file a new file with zero cases — announced as prose-only under >=, on a tree with nothing wrong
total == 0 and old[0] > 0 a file already empty when the ledger was written. An emptying is a transition; under >= a file that never made one is announced EMPTIED on every run forever
len(parts) > 1 parsing the ledger a line with no separator at all — under >= the index raises and the gate dies on a healthy tree
len(fixed) > 5 exactly five repaired seals

Nothing exotic. A control author plants the fault they are testing for, which is by construction a non-degenerate example: a file with the wrong contents, not a file with no contents. The boundary operator's whole yield is the empty case, the equal case, and the one-off-the-edge case — the inputs that are nobody's example.

Two boundaries, one planted tree

The prose count appears in --update-baseline's header and again in the verify path's census line. A single fixture carrying a zero-case file at record time pins both, because they are the same question asked on either side of the ledger. That case needed its own function for the same reason the refusal case does: what differs is when the fault arrives, and every table case plants after the ledger is written.

Cosmetic closed rather than declared, twice

Both "(+N more)" guards change a message and nothing else. Both got a case anyway — this campaign has twice found a written-down limitation to be invented, and a declared exception costs every future reader more than a case costs to write once.

Refs #2492

@github-actions

Copy link
Copy Markdown
Contributor

📓 NotebookLM Notebook linked to this PR

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

@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-08-23 06:47:23 UTC

Summary

Status Count
Total Open PRs 6
PRs with Failing Checks 2
PRs with All Checks Green 4
READY 3
FAILING 2
PENDING 0

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=e905ba8eeb53 != 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).

@gHashTag
gHashTag merged commit d5ff83e into master Aug 23, 2026
24 of 26 checks passed
@gHashTag
gHashTag deleted the fix/six-boundaries branch August 23, 2026 06:48
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