Repository navigation
fix(test): isolate upgrade installer fault fixtures - #834
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
@codex review |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Codex Review: Didn't find any major issues. What shall we delve into next? Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
🛡️ Codex Security Review · Automatically triggeredSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
|
Ready for lead: 34b7780 HACK-2239: private extraction parents remove shared-temp snapshots; consumer-driven response faults replace timed server errors and released-port assumptions. All assertions and timeouts retained, with five additional assertions. Recovery fixture compilation stays inside its owned temp directory.
Based on next 5965b23. No manual macOS dispatch; Mac/Docker execution remains deferred by the light-PR policy. Direct fault fixtures use the existing request seam, while compiled CLI recovery retains real HTTP success-path coverage. No real TCP-reset qualification claimed. Scratch cleaned and durable validation evidence retained. No merge; lead owns the batch. |
|
Lead note: verified on the Mac batch tree at this head ( 🤖 Generated with Claude Code |
The installer fixture failed on Linux shard 7 of run 43578179726 (PR #830 at dc55ab0) although its installer code was unchanged. Structured attempt-1 annotations identify:
stable update refuses a release with wrong checksum before publicationand the correspondingwrong versioncase: line 153 compares the whole OS temporary directory. The snapshots gain or lose another process'shack-update-*directory.stable download restarts a truncated transfer and retries 503 within one budgetand its prerelease counterpart: line 173 raises the intentional interruption asynchronously inside the HTTP server. The successful rerun still annotated both fixture errors.Extraction tests now pass a private temporary parent; the implementation still creates a fresh owned directory and preserves the OS-temp default. Retry fixtures use the existing
ReleaseFetchseam and fail on the consumer's second stream pull, after reading back the written five-byte prefix. Assertions retain exact attempts, complete final bytes and no leftovers. Offline refusal is injected deterministically with exactly three attempts, removing the released-port assumption. Recovery fixture builds run from their own temporary directory, keeping Bun's transient build files outside the checkout.All 17 tests remain, now with 146 assertions (previously 141). Direct fault cases exercise the transfer/installer logic without sockets; the compiled CLI hard-kill/recovery cases retain real HTTP success-path coverage. This does not claim a real TCP-reset qualification. No assertion or timeout was relaxed.
Validation: 100 consecutive full-file passes in separate Bun 1.4.4 processes (17 tests / 146 assertions each, zero failures/skips/retries), while 610 other 9-test batches passed, over 523.8 seconds. Concurrent background batches run
release-download-installer,self-updateandrelease-boundary; deliberate global-temp directory churn challenges shared-path assumptions. Root typecheck, check/privacy and changed-file lint passed. Persistent Claude review CLEAN (rounds 52–53). The first concurrent attempt exposed checkout-write guard failures from staging after launch and Bun's transient build path; that attempt was retained as failed evidence. The final sequence started after both fixes and contains no retries or reset. Hosted Linux shard and head Codex review pending. No manual macOS dispatch.HACK-2239. Separate from #832. Lead owns merge.
Open in [code]smith · Need help on this PR? Tag
@codesmith-botwith what you need.