Bound resolver work and retain active generations on activation refusal (#53) #496
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!496
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/resolver-cost-model-53"
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?
Repeated import decisions and unstable qualifier/reachability join order made resolver work grow faster than its useful input. Materialize own-import decisions once and probe keyed candidate/admission relations from selective drivers. Independently compared nine pinned corpora: all 1,144,443 reference target/rule/influence rows, symbols/ref counts and graph edges match; measured cold SQLite work drops 69.2% on Django and 35.3% on rust-analyzer.
Publish prepared plugin membership and bounded resolution in one transaction. Budget refusal, cancellation and crashes retain the previous complete projection; parsing and validation stay outside publication. This holds the writer through resolution, with WAL readers serving the prior projection. Legacy interrupted membership-only publications remain recoverable.
Expose process-local checkpoint work, CPU/wall time, populations, relation sizes, generation identity and budget verdicts through additive daemon RPC and
code-index://resolver/costs. Commit sparse/dense/per-site work-growth gates, activation-refusal regression and per-stage VM ceilings in the existing scale artifacts. Lower the seven-corpus work baseline after measured automatic-index reductions. Existing runtime, memory and CI limits remain unchanged; no new CI lanes.Validation: indexer unit and targeted generation/crash/isolation/control/bridge suites; four deliberately bad source mutations caught; all seven corpus cost and deletion/injection/rename controls; tier-3 target ratchet and scale ceilings; real daemon RPC; MCP resource/legacy/payload budgets; source-policy checks; clippy; Windows GNU cross-check. Acceptance details:
_prdoc/records/T-53-resolver-cost-model.md.Implements the remaining #53 work. Keep the ticket open until release publication and installed-build verification complete.