CLAUDE.md claimed ~31% of lines. Measured on main today: 29.8% (629/2113 lines, 408/1424 branches).
It went down, and that is the interesting part. The review filed R30 as PLAUSIBLE with "direction certain, magnitude unmeasured" — the assumption being that 122 → 276 tests must have raised it. It did not:
The fixes added Reattachment, JobSnapshots, JobTags, InputQuery, StagingNames, StagingSweep, NativeLoadFailure and an Application class. The denominator outran the numerator.
That is not an argument against the tests — the new code is largely why the defects are fixed. It is an argument against quoting a coverage percentage from memory, which is exactly how a stale figure survived for weeks. The line now carries the measurement, its date, and an instruction to re-measure rather than assume.
This is why R30 was deliberately held to the end of the run: measuring it before the new tests existed would only have made it stale again, and guessing the direction would have got it backwards.
Closes **R30 (#39)** — the last above-cut finding.
`CLAUDE.md` claimed **~31% of lines**. Measured on `main` today: **29.8% (629/2113 lines, 408/1424 branches)**.
**It went down, and that is the interesting part.** The review filed R30 as PLAUSIBLE with "direction certain, magnitude unmeasured" — the assumption being that 122 → 276 tests must have raised it. It did not:
| | 903b43c | now |
|---|---|---|
| JVM test files | 11 | 43 |
| main source | 4,114 lines | 5,715 lines |
| line coverage | ~31% (claimed) | **29.8% (measured)** |
The fixes added `Reattachment`, `JobSnapshots`, `JobTags`, `InputQuery`, `StagingNames`, `StagingSweep`, `NativeLoadFailure` and an `Application` class. **The denominator outran the numerator.**
That is not an argument against the tests — the new code is largely *why* the defects are fixed. It is an argument against quoting a coverage percentage from memory, which is exactly how a stale figure survived for weeks. The line now carries the measurement, its date, and an instruction to re-measure rather than assume.
This is why R30 was deliberately held to the end of the run: measuring it before the new tests existed would only have made it stale again, and **guessing the direction would have got it backwards**.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Closes R30 (#39) — the last above-cut finding.
CLAUDE.mdclaimed ~31% of lines. Measured onmaintoday: 29.8% (629/2113 lines, 408/1424 branches).It went down, and that is the interesting part. The review filed R30 as PLAUSIBLE with "direction certain, magnitude unmeasured" — the assumption being that 122 → 276 tests must have raised it. It did not:
903b43cThe fixes added
Reattachment,JobSnapshots,JobTags,InputQuery,StagingNames,StagingSweep,NativeLoadFailureand anApplicationclass. The denominator outran the numerator.That is not an argument against the tests — the new code is largely why the defects are fixed. It is an argument against quoting a coverage percentage from memory, which is exactly how a stale figure survived for weeks. The line now carries the measurement, its date, and an instruction to re-measure rather than assume.
This is why R30 was deliberately held to the end of the run: measuring it before the new tests existed would only have made it stale again, and guessing the direction would have got it backwards.
🤖 Generated with Claude Code