feat(diagnostics): port the SARIF 2.1.0 projection with canonical parity - #324
Conversation
P-022 step 5b. Ports `ownlang/diag_sarif.py` and the two `ownlang/evidence.py`
builders it uses into `own_diagnostics::sarif`, replayed from Python-authored
fixtures with zero Python at steady state.
Modelled as typed structs rather than a JSON value: the shape is checked at
compile time and the crate gains no runtime JSON dependency, matching how
Diagnostic is already done. serde_json stays a dev-dependency.
Three semantics a naive port gets wrong, each with a fixture case:
* Evidence.file falls back on EMPTINESS, not absence. Python writes
`e.file or filename`, so Some("") takes the fallback exactly as None does.
* resource_kind appends on emptiness too: `if d.resource_kind` is truthiness,
so Some("") adds no suffix where Option-presence would emit " [resource: ]".
* results follow input order while rules are sorted — two orderings in one log.
Sorting results to match the catalogue destroys the tie order that equal-code,
equal-message diagnostics at different lines depend on.
Canonical comparison: parse both, remove NOTHING, normalize only object key
order, leave every array in place, compare. The volatile-field list is empty as
a measurement, not a shortcut — the generator censuses for timestamps, GUIDs and
absolute paths and asserts the builder is deterministic. Arrays being left alone
is what makes the comparison strict: a port with the right set of results in the
wrong order fails.
Six mutations, each caught, measured with --no-fail-fast. Worth recording that
the empty-uri invariant does NOT catch the Evidence.file truthiness bug: under
that mutation the resolved file is "", so the step is DROPPED rather than
emitted with an empty uri, and only the reference comparison sees it.
Three of #256's requirements were unbuildable as written and are struck rather
than faked, each verified against the tree:
* .ownreport.json is a buffer report ({module, buffers[]}) with no diagnostics,
Evidence or tool metadata, and porting it needs ast_nodes + buffers.resolve —
which #256's own guardrail forbids this crate from reaching. A standing house
rule already refuses the shape #256 described.
* the DI004/DI005 control comes from ownir.build_sarif, excluded by the same
issue's "no OwnIR bridge".
* "upload Rust-generated SARIF" needs a binary; the workspace has none, and
#256 forbids CLI migration. Replaced by asserting the structural rules an
ingest enforces over the Rust-produced log.
Recorded in P-022 with the measurements, and the index row and crate metadata
moved in the same change.
Refs #256, #250.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CJF7MBi1ijU5m9cJVWgQsM
|
Warning Review limit reached
Next review available in: 50 minutes You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughThe PR adds typed Rust SARIF 2.1.0 generation to ChangesSARIF projection
Estimated code review effort: 4 (Complex) | ~45 minutes Possibly related issues
Possibly related PRs
Sequence Diagram(s)sequenceDiagram
participant Diagnostics
participant build_sarif
participant SarifLog
participant ReplayTests
Diagnostics->>build_sarif: Diagnostic values, filename, severity
build_sarif->>SarifLog: Typed SARIF results and metadata
SarifLog-->>ReplayTests: Generated SARIF document
ReplayTests->>ReplayTests: Compare canonical output and validate invariants
Suggested reviewers: 🚥 Pre-merge checks | ✅ 4✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
@coderabbitai review The automatic pass on Scope note, so the pass is spent on the port rather than on re-opening settled ground: this PR deliberately covers the Most useful places to look: the truthiness fallbacks in Generated by Claude Code |
|
I will keep the review within the stated ✅ Action performedReview finished.
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (4)
rust/crates/own-diagnostics/src/sarif.rs (1)
302-313: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winReuse
Diagnostic::kind_suffixinstead of re-deriving the suffix.
render.rsalready definesDiagnostic::kind_suffix, with the identical truthiness rule and the identical" [resource: {kind}]"format. Two copies of one user-visible text contract can drift independently. Call the existing method here.♻️ Proposed refactor
- // Truthiness, not presence: an empty `resource_kind` adds no suffix. - let kind = match diagnostic.resource_kind.as_deref() { - Some(k) if !k.is_empty() => format!(" [resource: {k}]"), - _ => String::new(), - }; + // Truthiness, not presence: an empty `resource_kind` adds no suffix. + // `kind_suffix` owns that rule for both the text and the SARIF renderings. + let kind = diagnostic.kind_suffix();🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rust/crates/own-diagnostics/src/sarif.rs` around lines 302 - 313, Update the SARIF result construction near steps_for to call Diagnostic::kind_suffix for the message suffix instead of matching diagnostic.resource_kind and formatting it locally. Remove the duplicated kind logic while preserving the existing empty-resource behavior and message format.tests/test_sarif_parity_fixtures.py (2)
342-348: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueNarrow the
nt-abs-pathexemption to the fixture path.The
(?!src)lookahead exempts every NT drive path whose first segment issrc. The census is meant to catch a leaked absolute path from a developer machine. A leak such asD:\src\...orC:\srcgen\...passes the lookahead and stays invisible. Anchor the exemption to the two paths the fixture deliberately uses instead.♻️ Proposed refactor
patterns = { "iso-timestamp": r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}", "guid": r"[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-", "posix-abs-path": r'"/(?:home|tmp|Users|var)/', - "nt-abs-path": r'"[A-Za-z]:\\\\(?!src)', + # The windows-drive-path case intentionally carries `C:\src\App\...`; + # every other NT drive path is a leaked machine path. + "nt-abs-path": r'"[A-Za-z]:\\\\(?!src\\\\App\\\\)', }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_sarif_parity_fixtures.py` around lines 342 - 348, Update the nt-abs-path pattern in the patterns mapping to exempt only the two deliberate fixture paths, rather than any Windows path beginning with src. Preserve detection for other absolute paths, including D:\src\... and C:\srcgen\..., while keeping the existing path-matching behavior unchanged.
318-330: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueRemove the unused
canonicalhelper.
canonicalis not called from this module, and no Python code imports it. Delete the helper and keep the canonicalization rationale in the module docstring if needed.
[maintain ibility_and_code_quality]🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_sarif_parity_fixtures.py` around lines 318 - 330, Remove the unused canonical function, including its docstring, from the test module. Preserve any necessary canonicalization rationale in the module-level docstring without retaining the unreferenced helper.rust/crates/own-diagnostics/tests/sarif_replay.rs (1)
344-352: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueExtend the
startLinecheck to the evidence projections.The loop checks only
result.locations.relatedLocationsandcodeFlows[].threadFlows[].locationsalso carry regions, and an ingest validates those the same way. Todayemittabledrops any step withline < 1, so the extra check cannot fail — which is exactly the point: it pins that contract at the invariant layer instead of leaving it to the parity replay alone, as the module docs inrust/crates/own-diagnostics/src/sarif.rsline 41 record.💚 Proposed test extension
+ let regions = result + .locations + .iter() + .map(|l| &l.physical_location) + .chain(result.related_locations.iter().map(|l| &l.physical_location)) + .chain(result.code_flows.iter().flat_map(|cf| { + cf.thread_flows.iter().flat_map(|tf| { + tf.locations + .iter() + .map(|tfl| &tfl.location.physical_location) + }) + })); + for physical in regions { + if let Some(region) = &physical.region { assert!( region.start_line >= 1, "case {name:?}: startLine must be 1-based, got {}", region.start_line ); } }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rust/crates/own-diagnostics/tests/sarif_replay.rs` around lines 344 - 352, Extend the SARIF replay invariant check beyond result.locations to also validate regions in relatedLocations and codeFlows[].threadFlows[].locations. Reuse the existing start_line >= 1 assertion and case context for every evidence projection, preserving the same behavior and diagnostic message style.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rust/crates/own-diagnostics/src/diagnostic.rs`:
- Around line 34-36: Update the SARIF reference in the documentation comment
near Diagnostic::render_pretty to link to the public crate::build_sarif entry
point instead of the private crate::sarif module, preserving the surrounding
description.
In `@rust/crates/own-diagnostics/src/sarif.rs`:
- Around line 292-301: Correct the documentation comment above result_for to
describe the actual severity behavior: requesting "warning" can lower an
intrinsic error to warning, while requesting "error" does not raise an intrinsic
warning, which remains warning. Leave the implementation unchanged.
---
Nitpick comments:
In `@rust/crates/own-diagnostics/src/sarif.rs`:
- Around line 302-313: Update the SARIF result construction near steps_for to
call Diagnostic::kind_suffix for the message suffix instead of matching
diagnostic.resource_kind and formatting it locally. Remove the duplicated kind
logic while preserving the existing empty-resource behavior and message format.
In `@rust/crates/own-diagnostics/tests/sarif_replay.rs`:
- Around line 344-352: Extend the SARIF replay invariant check beyond
result.locations to also validate regions in relatedLocations and
codeFlows[].threadFlows[].locations. Reuse the existing start_line >= 1
assertion and case context for every evidence projection, preserving the same
behavior and diagnostic message style.
In `@tests/test_sarif_parity_fixtures.py`:
- Around line 342-348: Update the nt-abs-path pattern in the patterns mapping to
exempt only the two deliberate fixture paths, rather than any Windows path
beginning with src. Preserve detection for other absolute paths, including
D:\src\... and C:\srcgen\..., while keeping the existing path-matching behavior
unchanged.
- Around line 318-330: Remove the unused canonical function, including its
docstring, from the test module. Preserve any necessary canonicalization
rationale in the module-level docstring without retaining the unreferenced
helper.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 9f56ca65-ca33-46f3-855e-204d139a53d0
📒 Files selected for processing (9)
docs/proposals/P-022-rust-core-migration.mddocs/proposals/README.mdrust/crates/own-diagnostics/Cargo.tomlrust/crates/own-diagnostics/src/diagnostic.rsrust/crates/own-diagnostics/src/lib.rsrust/crates/own-diagnostics/src/sarif.rsrust/crates/own-diagnostics/tests/sarif_replay.rstests/fixtures/sarif_parity.jsontests/test_sarif_parity_fixtures.py
… kind_suffix Review round on 984de7d. Every finding checked against the tree first. The one that mattered: the `severity` override was documented backwards. The doc claimed it "only ever raises"; measured against the reference it only ever LOWERS. request intrinsic Error intrinsic Warning "error" (or anything) error warning "warning" warning warning Asking for "warning" turns an error into a warning; asking for "error" cannot turn a warning into an error. The same inversion had been copied into the fixture's case rationale, so both are corrected and the fixture regenerated. Behaviour is unchanged — only the description of it was wrong, which for a user-visible severity contract is worth fixing on its own. Also: * `[crate::sarif]` named the private module, so rustdoc emitted private_intra_doc_links and rendered no link. Points at `crate::build_sarif` now; `cargo doc` is clean. * the SARIF result reused a private copy of the resource-kind suffix rule. Calling the existing `Diagnostic::kind_suffix` removes a second owner of one user-visible text contract — and strengthens coverage as a side effect: breaking the emptiness rule now fails the human render as well as the parity replay, measured. * the volatile-field census exempted every NT drive path starting with `src`, so `D:\src\...` and `C:\srcgen\...` would have passed as clean. Narrowed to the one path the fixture deliberately carries; both probes now flag. * dropped an unused `canonical()` helper from the Python generator — the Rust replay owns canonicalization, and the rationale already lives in the docstring. Declined, with the measurement: extending the `startLine >= 1` assertion to relatedLocations and codeFlows. Both `emittable` and `phys` gate on `line >= 1`, and breaking both together already fails the ingest invariants through the primary `locations` — so the extra assertion cannot be made to fail. It would grow the coverage claim without growing coverage. The mutation table in the module docs was itself stale: it was measured before the ingest-invariant test existed, and said "parity replay only" for a row that now has two catchers. Re-measured in full and corrected. One of those re-runs initially reported zero catchers for the severity mutation, which turned out to be a shell-escaping typo producing a compile error rather than a surviving mutant — re-run cleanly, it is caught. Refs #256, #250. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CJF7MBi1ijU5m9cJVWgQsM
Nitpicks: three taken, one declined with the measurementAll four checked against the tree first. Fixed in Reuse It paid off beyond tidiness, which I only found by re-measuring: breaking the emptiness rule now fails two tests instead of one, because the human render shares the method. The row moved from parity replay only to parity replay and the human render. Deleting the duplicate strengthened coverage. Narrow the I anchored the exemption as a negative lookahead on the whole prefix rather than your Remove the unused Extend You noted yourself that it "cannot fail" today. I tried to make it fail, and it can't be made to:
Both gates are shared between the primary and evidence paths, so any mutation that puts a bad That is the specific thing P-022 rule 2 exists to refuse — an assertion that grows the coverage claim without growing coverage. Recorded in the module docs with the measurement, plus the condition that would make it worthwhile: if the evidence projection ever gets its own location builder rather than sharing One correction to my own work, unpromptedThe mutation table in the module docs was stale. I measured it before adding the ingest-invariant test, then never re-ran it — so it claimed parity replay only for the During that re-run the severity mutation initially reported zero catchers, which would have meant a surviving mutant in the exact semantics this round is about. It was a shell-escaping typo producing a compile error; a mutation that fails to apply is indistinguishable from one nothing catches if you only grep for failing tests. Re-run cleanly, it is caught. Gates after all of it: suite exit 0, ruff clean, mypy clean on 30 files, Generated by Claude Code |
Closes the 58 permissive documents and 9 category mismatches the
re-census opened. The fix is architectural: no arrangement of the
previous design could have passed.
The previous door ran `version gate -> all semantic gates -> serde for
all shapes`. That is a third check order, matching neither
implementation. BR-D1 validates each section COMPLETELY before the next
begins, interleaving shape and semantics inside it — so a `components`
shape failure outranks a `services` vocabulary failure, and within a
service `lifetime` outranks `name`. Hoisting semantics in front of serde
gets section-local controls right and every ordering control wrong.
So the strict door is now a sequential validator over the raw document:
own-ir/src/strict.rs primitives + one function per section, applied
in the reference's order
own-ir/src/protocol.rs the obligation ACCEPTANCE grammar
Not 47 transcribed `if`s — eight primitives, each encoding one of
Python's access idioms, and section validators that apply them:
objects / list d.get(k, []) as a container
name_slot isinstance(v, str) and v — a value facts join on
optional_string `is not None and not isinstance` — null tolerated
defaulted_string isinstance(d.get(k, "?"), str) — null rejected
defaulted_int int and not bool
string_array list of str
column the 1-based contract (#317), recursive over flow
sites the {type, file, line} record
The two string primitives are the place a single "policy for optional
fields" would be silently wrong: `resource` rejects an explicit null and
`source_provenance` accepts it, because the reference writes one as a
defaulted isinstance and the other as `is not None and ...`.
serde is now the CONSTRUCTOR, not the arbiter. Once the validator
accepts, a serde failure means a rule lives in the model rather than the
validator — its category and its ordering would both be accidental. That
is marked with a sentinel and asserted against by
`no_control_escapes_into_serde`, which is not decoration: five of the 27
mutations below are caught by it ALONE, because the model still enforced
the rule while the validator no longer did.
`OwnIr::validate` is now `to_value` + the same validator. It costs a
round-trip and buys single-copy-of-the-law — the property whose absence
already produced a false "mutation survived" in this PR, when a planted
mutation hit one copy of a duplicated check and the other caught it.
Protocol scope, exactly as agreed: `parse_protocol`, `parse_matcher`,
`parse_events`, `parse_method` — what the door ACCEPTS. Not the
lattice, the walker, matching, or verdicts. `protocols` and
`protocol_functions` stay raw `Value`s with pure validation beside them,
because nothing consumes a typed representation yet. Two of the grammar's
rules ("can never fire", "barrier equals opens") are well-formedness
rather than shape; they are recorded as `shape` with the taxonomy strain
written down rather than a seventh category invented unilaterally.
Final matrix over 191 controls:
agreed accept 29
agreed reject 162
python reject / rust accept 0
python accept / rust reject 0
category mismatch 0
27 mutations, all caught — one per mechanism, plus three that only
change a category while still rejecting, plus two that only change
ORDER. The order mutations matter most: they are the ones the previous
architecture could not have failed.
Also fixed: `is_none_or` needs Rust 1.82 and the workspace MSRV is 1.74;
three `private_intra_doc_links` of the same class #324 hit; and the
replay now reports all three failure rows in one assertion, since
sequential asserts showed only the first and would have made this a
two-round discovery.
Still open and deliberately not closed here: Python integers are
unbounded, so every line/column field accepts values past 64 bits and
Rust is over-strict above i64::MAX across seven field families. The
ledger pins the boundary where both agree. The range above it is a
Python-first defensive limit, not a reason to thread arbitrary-precision
integers through own-ir and the bridge.
Refs #250, #259.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CJF7MBi1ijU5m9cJVWgQsM
2 participants
Что и зачем
P-022 шаг 5b. Порт
ownlang/diag_sarif.pyи двух builder'ов изownlang/evidence.pyвown_diagnostics::sarif, с replay из Python-авторитетных фикстур и нулём Python в steady state.Типизированные структуры, а не свободный JSON: форма проверяется компилятором, и крейт не получает runtime-зависимости на JSON.
serde_jsonостаётся dev-dependency — ровно как уже сделано дляDiagnostic.Тип изменения
Как проверено
python tests/run_tests.pyruff check .иmypycargo fmt+cargo clippy --all-targets+cargo test --workspace --no-fail-fastPython-выход не менялся: порт читает
diag_sarif, не правит его.Связанные issue
Refs #250, #256. Ничего не закрывает — см. «Что вычеркнуто».
Чеклист
feat:,fix:,docs:…)Три семантики, на которых наивный порт ломается
Каждая закреплена своим case в
tests/fixtures/sarif_parity.json.Evidence.filee.file or filename— truthiness:Some("")берёт fallback так же, какNoneartifactLocation.uri, и весь лог непроцессируем для code scanningresource_kindif d.resource_kind— тоже truthiness: пустая строка не даёт суффикс" [resource: ]"в пользовательском тексте сообщенияresultsв порядке входа,rulesотсортированы — два разных порядка в одном логеresultsпод каталог и уничтожить tie-порядокКанонное сравнение
#256 требует перечислить каждое удалённое поле и запрещает wildcard «ignore metadata». Перечень пуст, и это измерение, а не срезанный угол:
Поэтому нормализуется ровно одна ось — порядок ключей объекта, единственное, где два корректных сериализатора имеют право разойтись. Массивы не трогаются:
results,relatedLocationsиthreadFlows[].locations— упорядоченные контракты, и сортировка стёрла бы ровно то, ради чего шаг существует. Именно это делает сравнение строгим: порт с правильным множеством результатов в неправильном порядке здесь падает.Отдельный тест мутационно защищает сам канонизатор: если бы он сортировал массивы, контракты порядка молча перестали бы проверяться.
Mutation-проверка
--no-fail-fast, чтобы увидеть все слои-ловцы, а не первый упавший таргет.Evidence.fileкак Option-presenceresource_kindкак Option-presenceresultsотсортированы по кодуregionдля file-level строкиrulesусечены (висячийruleId)Первая строка стоит отдельного слова, потому что очевидная интуиция неверна: тест «нет пустых
artifactLocation.uri» не ловит truthiness-баг. Под мутацией file резолвится в"", и шаг отбрасывается фильтром, а не эмитится с пустым uri — симптом в молча пропавшем evidence-шаге. Порт, который отгрузил бы uri-инвариант и посчитал ловушку покрытой, ошибался бы насчёт собственного покрытия.Контроли из #256
Покрыты: пустой результат-сет, диагностика без Evidence, multi-file
relatedLocations, multi-stepcodeFlows, Windows-путь и Unix-путь, не-ASCII путь, равные code+message в разных строках, стабильный порядок таблицы правил, warning/error mapping,Owenкак tool identity.Про «note»: у
Severityв ядре есть толькоERRORиWARNING— уровеньnoteэтим путём не производится. Это измерение, а не пропуск.Что вычеркнуто, а не подделано
Три требования #256 оказались невыполнимыми как написано. Каждое проверено прогоном против дерева.
.ownreport.json. Issue описывает «schema/version, tool and run metadata, diagnostics and ordered Evidence». В дереве это{module, buffers[]}— отчёт о буферах. Диагностики дают четыре булевыхchecksна буфер; ни Evidence, ни tool-метаданных, ни schema version, ни одного пути. Плюсreport.pyимпортируетast_nodesиbuffers.resolve— то есть честный порт требует AST и резолвер политик, чего guardrail самого же #256 крейту не разрешает.И на эту форму уже стоит явный запрет: «НЕ перегружать
.ownreport.json» вAGENTS.execution-surfaces.md, плюс acceptance-строка «build_reportuntouched» вdocs/tasks/evidence-coverage.md. То есть половину про отчёт не просто не сделали — её просили в форме, от которой проект уже отказался.Поэтому она вычеркнута из шага 5b, а не отложена: cutover не требует Rust-овского buffer-отчёта, а изобрести диагностико-несущий отчёт под текст issue значило бы протащить новое поведение под видом порта.
DI004/DI005. Эти коды даётownir.build_sarifнадownir.Finding— путь OwnIR, исключённый guardrail'ом «no OwnIR bridge» того же issue. Уезжает с мостом (#259).Загрузка Rust-SARIF в Code Scanning. В Rust-воркспейсе нет ни одного binary target — писать лог физически нечем, а #256 запрещает CLI-миграцию. Добавить одноразовый бинарник ради галочки значило бы пересечь ту самую границу. Вместо этого проверяются структурные правила, которые реально энфорсит ingest: каждый
ruleIdразрешается в каталоге драйвера,levelиз SARIF-энума,startLine≥ 1, uri непустой и без обратных слэшей — над Rust-произведённым логом. Вместе с байтовой идентичностью Python-логу, который CI уже успешно загружает, acceptance следует из идентичности, а не из непроверенного утверждения.Всё это записано в P-022 с измерениями, и строка индекса плюс метаданные крейта (
description, докстрингEvidence, обещавшие «SARIF — это позже») переставлены тем же изменением.Generated by Claude Code
Summary by CodeRabbit
New Features
Bug Fixes
Documentation