Fix namespace budget inflation and degraded reporting for v0.41.28 #480
No reviewers
Labels
No labels
code-review
correctness
dos
performance
security
severity/high
severity/low
severity/medium
tech-debt
Kind/Breaking
Kind/Bug
Kind/Documentation
Kind/Enhancement
Kind/Feature
Kind/Security
Kind/Testing
Priority
Critical
Priority
High
Priority
Low
Priority
Medium
Reviewed
Confirmed
Reviewed
Duplicate
Reviewed
Invalid
Reviewed
Won't Fix
Status
Abandoned
Status
Blocked
Status
Need More Info
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
h-dv/code-index!480
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/namespace-budget-diagnostics"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Mixed-language indexes can skip C# namespace resolution because its lookup estimate multiplies namespace calls by imports from unrelated languages. Count the import multiplier from the same admitted language scope skeleton that TypeLookup loads, while retaining global linear load costs, conservative generation counting and every existing work-budget limit. Namespace binding, grant and ambiguity rules stay unchanged.
Doctor now recognizes both legacy
_degradedkeys and dotted.degradedstage keys. A degraded resolver is reported as a warning instead of beside an OK health label.PHP's measured warm plugin overhead was 1.117x and 1.153x native in performance runs 2050 and 2051, exceeding the former 1.10x budget. Apply the operator-approved increase to 1.20x for PHP only. C# remains 1.15x and other languages 1.10x. Cold budgets, semantic parity, mixed-load ceilings and CPU admission rules are unchanged. This adjusts the accepted budget; it does not claim a runtime performance improvement. All previous failed receipts are retained.
The initial CI attempt failed before checkout because the pinned Rust image disappeared from the registry. Verify the rebuilt Rust/Cargo 1.99.1 image and matching Clippy against every declared prerequisite, then adopt its immutable index digest in all 41 workflow/baseline pins.
Both product regressions fail before the fixes. Local validation includes 598 indexer tests (four existing ignores), 19 doctor tests, 24 C# resolver integration tests, nine version-site checks, 44 release-gate tests, targeted Clippy with warnings denied, formatting and diff checks. Budget tests exercise the reported PHP failure, language-specific headroom and rejection above each limit. Independent cold self-indexes using v0.41.27 and the candidate each processed 1,433 files and 28,611 symbols without parse errors or resolver-budget degradation; they do not reproduce the retained-generation degradation, which the focused regression demonstrates separately.
Prepare v0.41.28 in this PR so CI and performance qualify one release source head. Release source
44d618depassed all four required CI workflows: Windows 2054, core 2055, packages 2056 and pilots 2057. Core ran 5,349 main tests and 53 selected follow-up tests; Windows ran 5,123 main tests. All 28 budget tests passed locally. Fresh performance run 2058 passed semantic parity and both budgets for all seven pinned languages over five paired rounds, plus mixed-load ceilings. PHP measured 1.097752x against its approved 1.20x limit. An independent check recomputed every median, ratio and margin from the full saved log; no failures. Dependency pins, guest inputs, ABI and schema are unchanged. Refs #445 and #53; their broader acceptance criteria remain open.