cv2(gpu-death): compose guest GPU-death fix onto the deployed capstats+rearm lineage (deploy-ready superset) #42
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "cv2/gpu-death-recycle"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Deploy-ready superset for the CV2 WebRTC ≥90% gate
The deployed cb-chromium image
cr7727-f5dcb6087fa9= this base branchcapstats-plus-rearm(getCaptureStats health-gate + capture-rearm). The last measured residual (~15-20%) is the guest run11 zero-production GPU death — root-caused live: full WebRTC handshake completes (offer+answer+ICE delivered, ICE connected) yet the guest sends 0 RTP.The fix (guest self-detect GPU death →
kSessionUnhealthyenvelope → recycle; physics consumer #157 already landed) lives oncv2/gpu-death-recycle. It was merged tomain(PR #41), but main lacks getCaptureStats — building from main would regress the poisoned-golden health-gate the isolator + recycler depend on.This PR composes the fix onto the correct lineage:
capstats-plus-rearm+ the gpu-death commit = getCaptureStats + capture-rearm + gpu-death (verified all three coexist; merge-base is #37 so only the gpu-death delta lands). Merging here yields the exact image to build →cr7727-<sha>→ cb-rootfs bake → pin → roll → re-measure ≥90%.🤖 Generated with Claude Code