fix(coverage): restore specialized domain ownership and floors

This commit is contained in:
2026-10-14 10:35:41 +01:00
parent bad942fb25
commit ff73f8ad39
5 changed files with 846 additions and 13 deletions
@@ -58,8 +58,33 @@ or symlink aliases while producing plausible percentages.
**Trade-offs:**
New crates require explicit ownership. Reports with missing domains, duplicate
sources or zero instrumented lines are rejected before threshold comparison.
The current source map assigns 232 files to 15 exclusive owners without
changing the original eight domain baselines.
The initial ownership extension assigned 232 files to 15 exclusive owners
without changing the original eight domain baselines. Source counts grow with
the repository; the current evidence records the measured layout.
### Exclusive ownership must also be the intended ownership
**What:**
Module entry files and their sibling directories belong to the same specialized
domain. Memcard descendants belong to system/fs; asset-driver descendants
belong to asset/bank. Both residual owners exclude the entry file and the whole
module tree. Regression tests call the production auditor's ownership function
for current and future paths, and audit independent reports containing absent
and zero-line descendants.
**Why:**
A prefix ending in `.rs` covers the entry file but not the directory created
when a module is split. A residual owner can absorb every new descendant and
still pass a generic exactly-one-owner audit. The resulting percentages look
valid but measure the wrong responsibility. Expected-owner assertions detect
this failure before instrumentation or percentage comparison.
**Trade-offs:**
Correcting ownership moves the denominator and may reduce a specialized
domain's margin. Preserve reviewed floors and investigate fresh failures with
behavioral evidence. Historical measurements remain useful only when their
original source map and fingerprints remain intact; they should not be
silently rewritten to match the repaired map.
## Patterns and Algorithms
@@ -80,6 +105,8 @@ changing the original eight domain baselines.
- Allowing an absent or zero-line report to pass a zero baseline.
- Assigning the whole VM to the system integration domain, or counting one
source in both a specialized owner and its enclosing residual crate owner.
- Treating exactly one owner as proof that a module retained its intended
owner after its implementation moved into sibling source files.
- Treating local host measurements as evidence for unexecuted platforms.
## Published Extension and Evidence
@@ -97,6 +124,17 @@ Mathematics and its portable MPFR-generated fixture now exercise vm/core;
bytecode has its own gate. The standalone MPFR qualification remains separate.
This evidence does not claim Linux or Intel Mac coverage execution.
The module-tree ownership correction derives from accepted DEC-0059. Its
[current measured evidence](../../../docs/vm-arch/COVERAGE_DOMAINS_CURRENT.json)
preserves the earlier evidence records and restores system/core to 87.71%
without changing the other 14 floors. The fresh local map has 257 source files
and 15 exclusive domains. All domain gates pass: system/fs measures 78.73%,
system/core 93.69%, asset/bank 90.52% and drivers/core 89.17%.
Ten policy tests cover both generic validity and intended specialized owners.
The existing repository target passes 1,021 tests with 11 ignored; strict
workspace Clippy still reports existing VM diagnostics. Coverage qualification
must not be presented as completed strict workspace qualification.
## Takeaways
- Runtime-edge coverage is safer when CI enforces domain ownership instead of trusting a workspace aggregate.