Repository navigation
gates mutate: X of Y where X can exceed Y — four lines from a fix I shipped two passes ago - #3221
Merged
Merged
Conversation
SKILL 542 fixed the DENOMINATOR of `tri gates mutate`'s equivalence
report -- claims_seen counted every marker textually, including claims no
mutant was built for. The numerator went untouched, four lines below, in
the same println. The units fan-out found it.
claims_seen += in_scope.len(); // distinct claim LINES
for (dir, _, _, survivors) in &scores { // ... once PER DIRECTION
claims_broken.extend(contradicted_claims(..));
}
println!("{} of {} ...", claims_broken.len(), claims_seen);
"X of Y" is an inclusion statement and both halves must count the same
thing. A claimed line that is a mutable site under two operators and dies
under both contributes two rows to a numerator whose denominator counted
it once. Under --all the command can print "2 of 1".
Measured, not hypothesised: of the eight `# mutant-equivalent:` markers
in tools/, exactly ONE is a site in two directions --
gft_backprop_microcode.py:742, `assert _magsub(10240, 9217) == 9984`,
both an assert site and a boundary site.
Fixed by counting each half in its own unit rather than deleting a
number:
1 of 8 equivalence claim(s) CONTRADICTED, in 2 (claim x operator) row(s):
The rows stay per direction, because which operator killed a claim is the
useful half: an assert mutant dying says something different from a
boundary mutant dying.
Tenth pass running that the wiring outlived the function. Three fresh
unit tests on distinct_claims all passed against a call site reverted to
claims_broken.len(); killed by a structural test reading it.
SKILL 580.
Refs #3195
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 01:40:55 UTC
Summary
Seal Status
|
This was referenced Sep 5, 2026
fix(rust): lower Zig's undefined to Default::default() (+2) — NOT auto-merged, needs your call
#3223
Closed
Merged
Contributor
PR DashboardGenerated at: 2026-09-05 02:02:25 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
This was referenced Sep 5, 2026
gHashTag
enabled auto-merge (squash)
September 5, 2026 02:24
Contributor
PR DashboardGenerated at: 2026-09-05 02:37:12 UTC
Summary
Seal Status
|
# Conflicts: # .claude/skills/ci-gates/SKILL.md
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
1 similar comment
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 02:51:54 UTC
Summary
Seal Status
|
# Conflicts: # .claude/skills/ci-gates/SKILL.md
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 03:03:02 UTC
Summary
Seal Status
|
Contributor
PR DashboardGenerated at: 2026-09-05 03:26:41 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
SKILL 542 fixed the denominator of
tri gates mutate's equivalence report —claims_seencountedevery
# mutant-equivalent:marker textually, including claims for which no mutant was ever built.The numerator went untouched, four lines below, in the same
println!. The units fan-out found it."X of Y" is an inclusion statement, so both halves must count the same thing. They do not. A
claimed line that is a mutable site under two operators and dies under both contributes two rows
to a numerator whose denominator counted it once. Under
--allthe command can print2 of 1.Measured, not hypothesised
Of the eight markers in
tools/, exactly one is a site in two directions:gft_backprop_microcode.pyassert _magsub(10240, 9217) == 9984One in eight, on real data, reachable through a flag the command documents.
Fixed by counting each half in its own unit
The rows stay per direction — which operator killed a claim is the useful half: an
assertmutantdying says something different from a
boundarymutant dying. What changes is that the ratio speaksabout claims, and the rows are counted as rows and labelled as such.
Half of a fix is not the fix
Two passes ago I corrected this function's denominator and wrote a skill section about it. The
numerator sat four lines below and I did not look at it. "The fix does not travel" has a shorter
form here: it did not travel four lines.
The mutant reverting the numerator to
claims_broken.len()survived three fresh unit tests ondistinct_claims, and was killed only by a structural test reading the call site — the tenth pass ina row where the wiring outlived the function.
743 crate tests. SKILL 580. Refs #3195