Repository navigation
Upstream v0.1.6: reviewer findings in upstream code the overlay does not touch #162
Description
The review of the v0.1.6 rebuild (#152, head 0ddf6f4628) found ten defects in upstream T3 Code v0.1.6 code. The overlay does not touch any of these files: git diff 2359643cff <head> -- <file> is empty for each of them on e8098266fc. The rebuild therefore carries them as upstream ships them.
The rebuild has not reproduced any of them: each one is unverified here and quoted as the reviewer reported it. Stage 9 of docs/operations/lazurio-fork-release.md decides which ones go upstream as reports or fixes. Anything that should be fixed in the fork before then would be a new overlay row and needs its own decision.
| # | Upstream file (v0.1.6) | Reported defect | Since |
|---|---|---|---|
| 1 | apps/server/src/peer/RemoteDelegation.ts:559 |
A remote Stop refused for its mode is logged and recorded as stoppedThere: true, so the local task says the remote agent stopped while it keeps working (RemoteDelegation.test.ts:1372 asserts that state). |
new after v0.1.3 |
| 2 | apps/server/src/peer/PeerLinkRequests.ts:165 |
An unexpired local record answers linked after the other end revoked the link, so a request by Environment ID never shows a new pairing card. |
new after v0.1.3 |
| 3 | apps/server/src/orchestration-v2/ProviderSessionManager.ts:2362, ThreadManagementService.ts:1233 |
Force stop compares entry objects; after a sibling turn replaces the entry, release can report released while the process stays live. |
Force stop is new in v0.1.6 (pingdotgg#27252) |
| 4 | apps/server/src/device/DeviceService.ts:935 |
Shutdown removes sessions before powering off without the lifecycle lock, so an overlapping open can leave a session on a powered-off device. |
file exists on v0.1.3 |
| 5 | packages/source-control-gitlab/src/server/GitLabPullRequestCli.ts:1331 |
The bridges request omits input.host, so with several GitLab hosts child checks can come from another host or be dropped. |
file exists on v0.1.3 |
| 6 | apps/server/src/vcs/GitVcsDriverCore.ts:3691 |
Branch creation always passes --no-track and then tracks every remote start point, ignoring branch.autoSetupMerge. |
file exists on v0.1.3 |
| 7 | apps/server/src/vcs/GitVcsDriverCore.ts:2574 |
The merge ref and the remote are written separately; a failed second write can leave the branch tracking the wrong upstream. | file exists on v0.1.3 |
| 8 | scripts/mobile-behavior.ts:123, scripts/mobile-behavior/grade.ts:209 |
A run that exits nonzero after some rows still grades as complete; missing cases are not counted. | new after v0.1.3 |
| 9 | apps/server/src/mcp/toolkits/device/handlers.ts:91 |
The recovery text points to Settings > Integrations > Devices; the switch is under Settings > Devices (the ACC-01 eval expects the same wrong route). | file exists on v0.1.3 |
| 10 | apps/desktop/src/app/DesktopStartupRecovery.ts:220 |
Copy details puts literal \n in the clipboard instead of line breaks (non-blocking in the review). |
new after v0.1.3 |
"Since" comes from git cat-file -e v0.1.3:<file>. It says only whether the file exists on v0.1.3, not whether the reported defect does.
Next step: in stage 9, reproduce each one on the then-current upstream release before reporting it upstream. Close an item here when it is reported, fixed upstream, or found not to reproduce.
Activity