Repository navigation
typescript: sync @useblacksmith/sandbox - #8
Conversation
6b6685f to
b6e5600
Compare
|
b6e5600 to
4ed1778
Compare
GitOrigin-RevId: ebf48032142a7dd390a1994fcba685c20e64bf27
4ed1778 to
39f6f42
Compare
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 39f6f42. Configure here.
| await this.raceAbort( | ||
| this.waitUntilReady(cutoff - Date.now(), waitSignal, cutoff), | ||
| waitSignal, | ||
| ); |
There was a problem hiding this comment.
Health probe consumes resume redial
Medium Severity
redial() runs immediately before waitUntilReady, so the health probe consumes redialNext. The replayed agent call can then reuse a pooled connection to the host the resume left. After a transport-triggered wait, withRecheck has already spent its transport replay and surfaces that failure even though the sandbox is running.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 39f6f42. Configure here.
| return fail( | ||
| "timeout", | ||
| new SandboxTransportError("GET", `${this.client.baseUrl}/health`, 1, unreachable), | ||
| ); |
There was a problem hiding this comment.
Ready wait discarded at cutoff
Medium Severity
After waitUntilReady succeeds, waitThroughPauseResume still fails the wait when Date.now() has reached cutoff. A health probe that finishes inside the 250ms slack is treated as unreachable, so the user call errors even though the agent already answered.
Reviewed by Cursor Bugbot for commit 39f6f42. Configure here.




Sync @useblacksmith/sandbox