Branch: refs/heads/gh-readonly-queue/master/pr-7218-2889928e6aac5f2ce4079d7288007bbf3120fab5
Home:
https://github.com/google/syzkaller
Commit: ff940bdf86508c3bfe1232795722ed01d7249dac
https://github.com/google/syzkaller/commit/ff940bdf86508c3bfe1232795722ed01d7249dac
Author: Aleksandr Nogikh <
nog...@google.com>
Date: 2026-05-08 (Fri, 08 May 2026)
Changed paths:
M dashboard/app/ai.go
M dashboard/app/ai_report.go
M dashboard/app/ai_report_test.go
Log Message:
-----------
dashboard/app: prevent JobReporting creation for non-patching AI jobs
When a non-patching job (e.g. assessment-security) was manually marked as
correct via the UI, handleAIJobPageCorrectness attempted to upstream it.
Because checkJobUpstreamable did not explicitly forbid non-patching
workflows, a bogus JobReporting entity was created, which subsequently
broke the lore-relay poller.
This updates checkJobUpstreamable to explicitly reject any job type other
than WorkflowPatching and WorkflowPatchIteration. handleAIJobPageCorrectness
now handles this gracefully, recording the manual approval without
spawning a JobReporting stage.
To unsubscribe from these emails, change your notification settings at
https://github.com/google/syzkaller/settings/notifications