Commit 655b07e
committed
Pin Rust coverage threshold to CI reality (96), drop its auto-ratchet
The Rust coverage check now passes the python3 parser (good), but it exposed a
phantom threshold: tarpaulin's line attribution is platform-divergent — macOS
reports 67/67=100% while Linux CI reports 66/68=97.1% for the SAME code (lines
like 'match name {' and struct-literal fields are attributed differently by the
host LLVM; the code is already exercised by existing tests). The 99% threshold
had been auto-ratcheted from the inflated macOS number, so CI could never meet it.
- coverage-thresholds.json: Rust threshold 99 -> 96 (floor(CI 97.1%)-1, matching
the convention used by every other project's threshold).
- Makefile _coverage_check: make the Rust block CHECK-ONLY (no auto-ratchet
write). Auto-ratcheting platform-divergent coverage captures the highest
(macOS) measurement and pins an unmeetable CI threshold. F#/TS ratchets are
unchanged (their coverage is platform-stable).
Verified: make _coverage_check -> all OK exit 0, and coverage-thresholds.json is
NOT mutated by a macOS run (no re-inflation). CI's 97.1% >= 96 passes.1 parent 4fa5229 commit 655b07e
2 files changed
Lines changed: 9 additions & 9 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
271 | 271 | | |
272 | 272 | | |
273 | 273 | | |
| 274 | + | |
| 275 | + | |
| 276 | + | |
| 277 | + | |
| 278 | + | |
| 279 | + | |
274 | 280 | | |
275 | 281 | | |
276 | 282 | | |
| |||
281 | 287 | | |
282 | 288 | | |
283 | 289 | | |
284 | | - | |
285 | | - | |
286 | | - | |
287 | | - | |
288 | | - | |
289 | | - | |
290 | | - | |
291 | | - | |
| 290 | + | |
292 | 291 | | |
293 | 292 | | |
294 | 293 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
19 | 19 | | |
20 | 20 | | |
21 | 21 | | |
22 | | - | |
| 22 | + | |
| 23 | + | |
23 | 24 | | |
24 | 25 | | |
25 | 26 | | |
0 commit comments