|
1 | 1 | { |
2 | | - "schema": "worklog/v1", |
3 | | - "date": "2026-09-17", |
4 | | - "taskId": "rustjava-adopt-class-format-mutation-audit-p0", |
5 | | - "summary": "Audit the hand-assembled fixtures for single-defectness — is each file broken in exactly one way? Answered by byte equality against the canonical build rather than by loading, which is both stronger and cheaper. Result: 18 defective fixtures checked, all single-defect; 5 carry no defect at all.", |
6 | | - "changes": [ |
7 | | - "test-data/src/audit-fixture-single-defect.py: repairs each generated fixture's named defect and requires the bytes to equal the canonical build" |
8 | | - ], |
9 | | - "verification": [ |
10 | | - "audit: 18 checked / 18 single-defect / 5 no-defect (valid files) / rc=0", |
11 | | - "mutation M1: plant a second defect (NotFactoryDescriptor also gets reference kind 7) -> MULTI-DEFECT, rc=1", |
12 | | - "mutation M2: flip one byte of a committed fixture -> GENERATOR DRIFT, rc=2", |
13 | | - "cargo test --all unchanged (this round adds no Rust)" |
14 | | - ], |
15 | | - "issues": [ |
16 | | - "The audit is a standalone script, not wired into DoD or CI. An audit nobody runs rots; whether this property is worth a permanent check is a separate decision, argued in the proposal below.", |
17 | | - "Three fixture classes are outside its reach and are named in the script header rather than silently skipped: javac output (no defect), legal-but-unimplemented files (no defect), and fixtures byte-patched inside a test (not on disk)." |
18 | | - ], |
19 | | - "adoptedProposals": [ |
20 | | - "2026-09-16-class-format-mutation-audit#p0" |
21 | | - ], |
22 | | - "proposals": [ |
23 | | - { |
24 | | - "title": "Decide whether single-defectness is a standing check or a one-off", |
25 | | - "plainSummary": "The audit passes today. Nobody runs it tomorrow.", |
26 | | - "userBenefit": "A fixture that grows a second defect stops covering what its test claims, and the test keeps passing — which is exactly the silence this round measured away.", |
27 | | - "why": "Every fixture is generated from a parameterised generator, so the check costs one python3 invocation and no JVM. The repository's own doctrine says a rule kept only in prose is the worst state: 'the rule exists and has no effect'.", |
28 | | - "tradeoff": "It would be an eighth DoD command, and check-dod-ci-parity.py then has to carry it in both directions — the parity lock is deliberately strict about that. Against: the property is stable by construction (the generators take the defect as a parameter), so the check may earn nothing for years.", |
29 | | - "effort": "S", |
30 | | - "target": "CLAUDE.md, .github/workflows/rust.yml" |
31 | | - } |
32 | | - ] |
| 2 | + "schema": "worklog/v1", |
| 3 | + "date": "2026-09-17", |
| 4 | + "taskId": "rustjava-adopt-class-format-mutation-audit-p0", |
| 5 | + "summary": "Audit the hand-assembled fixtures for single-defectness — is each file broken in exactly one way? Answered by byte equality against the canonical build rather than by loading, which is both stronger and cheaper. CORRECTED after gate-2 review (round rustjava-adopt-class-format-mutation-audit-p0-fix): the original claim \"18 checked, all single-defect\" was not a measurement. 14 of the 18 rebuilt `repaired` from canonical arguments instead of deriving it from the fixture's own, so a second defect was discarded by both sides and the verdict could not fail. Only the 4 indy near-misses were really measured. The fix derives `repaired` from `given` for all four families; the 18 committed fixtures still pass, but now that is a result rather than a tautology.", |
| 6 | + "changes": [ |
| 7 | + "test-data/src/audit-fixture-single-defect.py: repairs each generated fixture's named defect and requires the bytes to equal the canonical build", |
| 8 | + "test-data/src/audit-fixture-single-defect.py (p0-fix): `repaired` is now `given` with only the named axis overwritten, for all four families, via a single add() helper", |
| 9 | + "test-data/src/audit-fixture-single-defect.py (p0-fix): `given` is bound to the generator's named parameters with inspect.signature(...).bind(*spec), so a parameter added to a generator is carried into `repaired` instead of being dropped by tuple indexing", |
| 10 | + "test-data/src/audit-fixture-single-defect.py (p0-fix): new AUDIT SPEC STALE verdict (rc=2) when a spec sets a parameter no CANON table names — the guard on the repaired/canonical asymmetry", |
| 11 | + "test-data/src/audit-fixture-single-defect.py (p0-fix): docstring gains a fourth coverage limit (the canonical arguments and the named axis are human annotations) and a note that the MULTI-DEFECT byte count is a loudness figure, not an edit distance" |
| 12 | + ], |
| 13 | + "verification": [ |
| 14 | + "audit: 18 checked / 18 single-defect / 5 no-defect (valid files) / rc=0 — SUPERSEDED: gate 2 showed 14 of those 18 verdicts could not fail, so this line reports 4 measurements and 14 vacuous passes. Re-run after the p0-fix round gives the same numbers, and they now mean what they say.", |
| 15 | + "mutation M1: plant a second defect (NotFactoryDescriptor also gets reference kind 7) -> MULTI-DEFECT, rc=1", |
| 16 | + "mutation M2: flip one byte of a committed fixture -> GENERATOR DRIFT, rc=2", |
| 17 | + "cargo test --all unchanged (this round adds no Rust)", |
| 18 | + "p0-fix: baseline unchanged on the real tree — 18 checked / 18 single-defect / 5 no-defect / rc=0", |
| 19 | + "p0-fix mutation M-A (ldc, LdcDynamicOldMajor also gets a duplicate BootstrapMethods): before rc=0, after rc=1 MULTI-DEFECT 13 bytes", |
| 20 | + "p0-fix mutation M-B (cp, generator grows a `major` parameter and UnreferencedTag13 gets major 45): before rc=0, after rc=2 AUDIT SPEC STALE — the audit refuses to judge a parameter it does not know the canonical of", |
| 21 | + "p0-fix mutation M-B2 (M-B plus telling the audit that major 52 is canonical): rc=1 MULTI-DEFECT 1 byte — the cp family is judgeable once the CANON table is current", |
| 22 | + "p0-fix mutation M-C (indy near-miss, NotInvokeStaticFactory also gets the wrong method name): rc=1 before and after — the family that already worked still works", |
| 23 | + "p0-fix mutation M-D (indy LINKED, MakeConcatWrongDescriptor also gets a static argument): before rc=0, after rc=1 MULTI-DEFECT 4 bytes", |
| 24 | + "p0-fix: each mutation ran in a scratch copy with the fixture regenerated, so built()==committed still holds and the run is not a GENERATOR DRIFT" |
| 25 | + ], |
| 26 | + "issues": [ |
| 27 | + "CORRECTED: the original 18/18 was 4/18 measured. See the p0-fix entries above.", |
| 28 | + "The audit is a standalone script, not wired into DoD or CI. An audit nobody runs rots; whether this property is worth a permanent check is a separate decision, argued in the proposal below.", |
| 29 | + "Three fixture classes are outside its reach and are named in the script header rather than silently skipped: javac output (no defect), legal-but-unimplemented files (no defect), and fixtures byte-patched inside a test (not on disk).", |
| 30 | + "p0-fix: the verdict now depends on the CANON tables in the script being right. A wrong canonical produces a confident false MULTI-DEFECT where the old form was silently green — louder, but a new way to be wrong.", |
| 31 | + "p0-fix: growing a generator now blocks the audit with AUDIT SPEC STALE until its CANON table is updated. Deliberate — the alternative is guessing — but it is a standing cost on anyone adding a generator parameter.", |
| 32 | + "p0-fix: the committed count did not go down (18 still pass). What went down is the set of inputs that could pass: structurally-unfailable cases 14 -> 0." |
| 33 | + ], |
| 34 | + "adoptedProposals": [ |
| 35 | + "2026-09-16-class-format-mutation-audit#p0" |
| 36 | + ], |
| 37 | + "proposals": [ |
| 38 | + { |
| 39 | + "title": "Decide whether single-defectness is a standing check or a one-off", |
| 40 | + "plainSummary": "The audit passes today. Nobody runs it tomorrow.", |
| 41 | + "userBenefit": "A fixture that grows a second defect stops covering what its test claims, and the test keeps passing — which is exactly the silence this round measured away.", |
| 42 | + "why": "Every fixture is generated from a parameterised generator, so the check costs one python3 invocation and no JVM. The repository's own doctrine says a rule kept only in prose is the worst state: 'the rule exists and has no effect'.", |
| 43 | + "tradeoff": "It would be an eighth DoD command, and check-dod-ci-parity.py then has to carry it in both directions — the parity lock is deliberately strict about that. Against: the property is stable by construction (the generators take the defect as a parameter), so the check may earn nothing for years.", |
| 44 | + "effort": "S", |
| 45 | + "target": "CLAUDE.md, .github/workflows/rust.yml" |
| 46 | + } |
| 47 | + ] |
33 | 48 | } |
0 commit comments