Repository navigation
fix: resolve seven regressions from the merged PR audit - #159
Merged
Merged
Conversation
Protect worktree files and shared checkpoint refs, preserve safe branch metadata, coordinate inference route persistence, recover summary retries, and support legacy OpenCode Morph credentials. Add regression coverage for all seven findings.
|
This was referenced Nov 18, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 audit of the last 40 merged PRs found seven regressions still present on
main: worktree recovery could delete newly arrived files, a seeded worktree development home could prune the original home's checkpoints, live branch updates bypassed the fork's safety checks, failed inference configuration writes could leave mixed state, conversation-summary recovery could stall in two cases, and Morph evaluations could miss valid OpenCode credentials.This PR combines all confirmed corrections, as requested. Every finding was reproduced on unchanged production source before its fix. Regression cases extend existing owning tests; the Morph helper gets an isolated configuration test.
Corrections
WorktreeBranches.syncafter the existing stale-event check, through the existing provided service layer.waitingMessages = 0, so it was permanently excluded from recovery.providers.morph.settingsbut rejected the supported v1provider.morph.optionsformat.Regression evidence
notes.mdcreated during project lookup. The fixed table also covers a newly appearing.gitand a vanished folder.7ef0e6d4fails the primary-route write regression; the original fallback and catalog blocks separately fail published-then-error cases. The final 11 proxy/route/HTTP tests pass, including later settings-write failure, readers before and after settings commit, health during publication, and cancellation. Independent review checked the host → route → settings lock order and that network work stays outside the route snapshot lock.Audit scope
Pinned base:
7ef0e6d44b886604bc04b5048fcf55356399e671. The scope is the latest 40 merged PRs authored byScriptedAlchemy, selected by merge time from the full closed-PR inventory:#158, #157, #156, #155, #154, #153, #152, #151, #150, #149, #146, #148, #147, #144, #145, #142, #143, #141, #139, #140, #137, #138, #135, #136, #134, #127, #133, #131, #129, #125, #132, #130, #128, #126, #124, #122, #123, #121, #115, #117.
Reviews covered inference and provider execution, model evaluations, memory/LCM and migrations, group chat and subagent state, mobile composer behavior, process/platform changes, dependency/CI changes, and upstream integration. The two large upstream syncs received targeted integration review across these areas; this is not a claim of exhaustive manual inspection of every imported upstream line. The already merged #157/#158 corrections remain intact and are not counted again. The withdrawn #142 empty-config suspicion was not reopened.
Validation
vp run check:changed origin/main --concurrency 1. This includes the changed service tests, runtime-layer integration, and Agent Context Budget. These cases overlap the audit totals above and are not added to them.seccomp_unotify. The same package command,effect-tsgo tsrs --noEmit, was then run directly in apps/server, evals, and scripts; all three exited 0. The direct checks caught one Effect diagnostic in restoration; corrected it toEffect.undefinedand reran the affected checks. No CI configuration or validation gate was weakened.git diff --check. GitHub CI will validate the published PR head.The audit also passed 176 group/model-evaluation tests, 30 mobile unit tests, and 42 client group-state tests. Native Windows/macOS/iOS/Android behavior and private/live-model behavior were reviewed from source where relevant, but were not independently exercised on those runtimes. Eleven unrelated baseline cases are blocked by the environment: one GitHubRest Unix-socket fixture cannot bind (
EPERM), and ten desktop WSL configuration cases cannot enumerate network interfaces (the same error reproduces in standalone Node). These are not counted as passing. No model-facing prompts changed.Inference rollback remains best effort if storage continues failing: the original error and previous in-memory pool are retained, and a sanitized warning reports a failed rollback. The worktree cleanup fix closes the reproduced lookup-to-cleanup gap; it does not claim an atomic filesystem transaction.
Model: GPT-6 Astra Pro. Harness: ChatGPT with GitHub connector and isolated local Git worktrees.