Four files describe crates/fmw-noise/src/cliffs/connections.rs as a "445-line module". It is 578 lines (wc -l, measured on main at 56a56f5).
Where
| file |
line |
scripts/verify-rust.sh |
305 |
docs/rust-wasm-port.md |
1108 |
docs/rust-wasm-port-history.md |
2505 |
docs/ci-sharding-measurements.md |
35 |
They all carry the same sentence, which exists to justify keeping the most expensive test in the crate - the cliff connection fixture test, 33s normally and 93s under poison. The argument is that without it, cliffs::connections would be a large module with unit tests and no measurement against anything.
What is right, and what is not
The argument is unaffected and the rest of the claim checks out. Verified rather than assumed:
mod fixtures is gated #[cfg(test)] (crates/fmw-noise/src/lib.rs:43-44), and crates/fmw-noise/src/fixtures.rs:4050 is the module's only consumer.fmw-wasm never references it - grep -rn 'connections::' crates/fmw-wasm/src/ returns nothing.
So "on no render path, only consumer is the test harness" is correct. Only the line count is wrong, and it is low by 24%.
Why it is worth correcting anyway
The number is the load-bearing half of the sentence: it is what makes "unit tests and no measurement" sound like it matters. A reader checking the claim finds a different file than the one described, which is the same failure mode notes-must-say-how-they-were-measured exists to prevent - and this repo has been bitten repeatedly by figures that went stale in place (the verify total was wrong three times, engine.wasm's byte count twice, the testTimeout counts twice).
Not urgent
Nothing depends on it. It is prose in four files, one of which is a pure-archaeology document (rust-wasm-port-history.md) where the 445 may well have been true when written - worth checking git log on that line before overwriting it, rather than assuming it was always wrong.
Found while landing #389; deliberately left out of that PR to keep its diff scoped to the rustdoc gate.
Four files describe
crates/fmw-noise/src/cliffs/connections.rsas a "445-line module". It is 578 lines (wc -l, measured onmainat 56a56f5).Where
scripts/verify-rust.shdocs/rust-wasm-port.mddocs/rust-wasm-port-history.mddocs/ci-sharding-measurements.mdThey all carry the same sentence, which exists to justify keeping the most expensive test in the crate - the cliff connection fixture test, 33s normally and 93s under poison. The argument is that without it,
cliffs::connectionswould be a large module with unit tests and no measurement against anything.What is right, and what is not
The argument is unaffected and the rest of the claim checks out. Verified rather than assumed:
mod fixturesis gated#[cfg(test)](crates/fmw-noise/src/lib.rs:43-44), andcrates/fmw-noise/src/fixtures.rs:4050is the module's only consumer.fmw-wasmnever references it -grep -rn 'connections::' crates/fmw-wasm/src/returns nothing.So "on no render path, only consumer is the test harness" is correct. Only the line count is wrong, and it is low by 24%.
Why it is worth correcting anyway
The number is the load-bearing half of the sentence: it is what makes "unit tests and no measurement" sound like it matters. A reader checking the claim finds a different file than the one described, which is the same failure mode
notes-must-say-how-they-were-measuredexists to prevent - and this repo has been bitten repeatedly by figures that went stale in place (theverifytotal was wrong three times,engine.wasm's byte count twice, thetestTimeoutcounts twice).Not urgent
Nothing depends on it. It is prose in four files, one of which is a pure-archaeology document (
rust-wasm-port-history.md) where the 445 may well have been true when written - worth checkinggit logon that line before overwriting it, rather than assuming it was always wrong.Found while landing #389; deliberately left out of that PR to keep its diff scoped to the rustdoc gate.