Repository navigation
Expose bounded key reuse through the protected residency owner - #90
Merged
Merged
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
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 |
MrScripty
marked this pull request as ready for review
October 21, 2026 05:46
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The protected residency facade could not select the existing bounded workload-key cache: its constructor always chose fair serial residency with reuse off. Add
EmbeddedSerialCpuResidency::new_with_bounded_key_reuse(config)and delegatescheduler_configuration()to the existing immutable value query. Existingnew(config), config struct literals, private owners, run/resume feature gates and shutdown behavior remain intact.Exact source head
c45d103fedc08f567e2892e8d71b1eda49cfe881, tree540618b75e7ebbde9a5713c8ac14f05e77936f80, based on PR89's normal mergedc64b895180260ceef8619bedd1ac6d8b756a3ec. Independent pre-code and exact staged-tree source reviews approved publication, and a refreshed independent review approved normal merge. All 22 exact-head check runs and all four PR workflows passed; CodeRabbit status supplies no independent source-review credit. PR89's actual mergedc64b895independently passed all 20 post-merge check runs and four workflows.Each actual member still takes its own fair opportunity, preparation, reservation, final Start fence, native execution/drain and matched cleanup. This selects pure planning reuse; it enables no grouped native forward, new admission/resource policy or complete-memory guarantee. The query exposes no owner handle and remains configuration after acknowledged shutdown.
The canonical CPU plan now links a finite five-row completion checklist, explicitly selecting ordinary-member planning reuse as the supported CPU envelope. It distinguishes required software/consumer functionality, the still-required prospectively fixed actual CPU baseline-versus-key-reuse comparison, and later representative model/hardware qualification. Complete allocator/drop/alias/scratch instrumentation is required for a future complete-memory claim, not for cost-free C2a. No further allocator slice is selected; broader M5/M6 acceptance remains open.
Local validation passed: default facade module 8 cases; full embedded-runtime library 607 passed with 2 existing ignored controlled cost probes; inventory-feature facade module 6 cases; all-feature/all-target own-package Clippy with warnings denied; all-feature own-package Rustdoc with warnings denied; formatting, 33 resolved doc links, traceability (7 paths/0 mapped impacts), and declared pre-push tests (261/one existing ignored + 168 passed). The inventory suite's existing native lifecycle case uses its original constructor; it does not qualify native key reuse. Initial feature-selection/Option-access and restricted-home Pumas setup failures are retained in the report. Corrected test runs use pinned Pumas's supported process-local registry override in writable scratch; no HOME, permission or security setting changed. PR89's baseline inference strict lint/Rustdoc limitations and denied historical log evidence remain preserved and unretried.
The completion checklist owns exact remaining gates and stop conditions. Native qualification producer/protocol work is still software work; no external input blocks it. All earlier branches, PR84 lineage and unrelated work remain retained-protected. No dependency, manifest/lockfile, CI, IPC/binding or default-policy changes.
Exact-head hosted workflows: Quality Gates, CPU Composition Contract, Headless Workflow Contract, and Runtime Separation Check. Fresh Rust-focused CI at integration commit
7fb1c05d5d542438a715a7b58ffb8d8a743069db(same approved tree) passed both new facade cases; the hosted embedded-runtime summary also confirms 607 passed, zero failures and the same two existing ignored cost probes. This source is merged; no overall M5/M6, native key-reuse performance or complete-memory acceptance is claimed.Normal guarded integration completed as
ecfed47da7c51d03b968bed739e2abc06eb4a6de, parentsdc64b895180260ceef8619bedd1ac6d8b756a3ecandc45d103fedc08f567e2892e8d71b1eda49cfe881, identical approved tree540618b75e7ebbde9a5713c8ac14f05e77936f80. All 20 actual-main check runs and all four workflows passed: Quality Gates, Headless Workflow Contract, Runtime Separation, Push on main. Integration receipt SHA2569d7e20d34d5a1a55811134dd35058f3795f0f966a0fdf49b0c05b95b84a38daf; eight prior clean worktrees and all twelve prior remote branch tips are retained. Denied historical log reads were not retried. The separate prospective native comparison is a distinct acceptance gate; this integration does not qualify it.