Repository navigation
Correct six numbers an audit found wrong, at their sources - #3254
Merged
Merged
Conversation
Every figure published during this pass was re-measured on master by five independent readers using their own commands, with a third reading on each disagreement: 10 CONFIRMED, 3 stale-but-were-true, 6 WRONG. competitors.rs: the comment claimed `pub fn beta_competitor` is production code below the first test module and used that to justify the anchor. It is at line 696, INSIDE the raw string `const TWO` which opens at 681 and closes at 705 -- fixture text inside `mod tests`. The file has no production item below the cut. A column-0 `^pub fn ` regex counted a string literal as code. The anchor stays, because the count and absence assertions differ in kind, but its stated reason was false. prose.rs and SKILL.md: "10 of 46 files, 130 items" reproduces under no definition -- 13/160 splitting on the string, 11/140 on the bare attribute line, 9/133 additionally rejecting string literals. The 46 is right; gates.rs holding 79 is right. The file count is now given with its three matchers rather than as one number, because a figure that moves with the matcher carries the matcher. mutate.rs: "SIX doc comments before the attribute" was five -- 216, 355, 358, 371, 375, twelve in the file, attribute at 487. History shows it was never six. Also corrected in the issues, not here: "each push triggers 28 check-runs" was the MINIMUM of a sample whose range is 28-43, median 41 (#3236, where it priced an argument it now strengthens); and "zero Admitted anywhere under coq/" is one, in coq/README.md -- zero in `.v` files, the population the gate reads. Four of the six are a population mismatch rather than an arithmetic error: literals counted as code, a matcher left unstated, a minimum reported as a central value, and `coq/` where `coq/**/*.v` was meant. Same class the pass was hunting. Refs #3253 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
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.
Every figure published during the #3229 / #3232 / #3237 pass was re-measured on
master by five independent readers using their own commands, with a third
reading on each disagreement.
10 CONFIRMED · 3 stale-but-were-true · 6 WRONG. Full accounting in #3253.
Fixed here
competitors.rs— the comment justified the anchor by claimingpub fn beta_competitoris production code below the first test module (line567). It is at line 696, inside the raw string
const TWOwhich opens at681 and closes at 705 — fixture text inside
mod tests. The file has noproduction item below the cut. A column-0
^pub fnregex counted a stringliteral as code.
The anchor stays: the count assertion and the absence assertion genuinely differ
in kind — a call site leaving the slice makes the count read 1 and fail loudly,
while a reintroduced subtraction below the cut would simply not be found. Only
the stated reason was false.
prose.rsandSKILL.md— "10 of 46 files have production items aftertheir first
#[cfg(test)], 130 items in all" reproduces under no definition:#[cfg(test)]The 46 is right by three independent instruments, and
gates.rsholding79 is right. The file count is now given with its three matchers instead of
as one number.
mutate.rs— "SIX doc comments before the attribute" was five: lines216, 355, 358, 371, 375; twelve in the file; attribute at 487. History shows it
was never six.
Corrected in the issues, not in code
range is 28–43, median 41. It priced the argument in SKILL.md is the NOW.md problem again: 182 of 293 commits touch one 865 KB file, and every concurrent PR conflicts on it #3236 — which the
correction strengthens, and that asymmetry is exactly why it survived
unchecked.
Admittedanywhere undercoq/" — there is one, incoq/README.md:26. Zero in.vfiles, the population the gate reads and theone I meant.
Why this is a defect report and not a typo fix
Four of the six are a population mismatch, not an arithmetic error: string
literals counted as code, a matcher left unstated, a minimum reported as a
central value,
coq/wherecoq/**/*.vwas meant.That is the same class the pass was hunting — #3229 was a tally whose classes
did not cover its population, #3237 an assertion evaluated over a region that
need not contain its subject. Every one of the six had already been "checked",
by me, with the instrument that produced it.
Adds a
ci-gatessection recording the audit and its three lessons.cargo test -p tri --bin tri: 776 passed, 0 failed.Refs #3253