This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

All contributors have signed the DCO ✍️ ✅
_{Posted by the DCO Assistant Lite bot.}

I have read the DCO document and I hereby sign the DCO.

recheck

/ok to test 6e7e986

/ok to test 9abd176

gator-agent

PR Review Status

Validation: project-valid. The PR author is in @NVIDIA/openshell-maintainers, the PR is not draft, DCO is passing, /ok to test has already been posted for the current head SHA, and duplicate searches did not find overlapping open or closed issues/PRs.
Head SHA: 9abd176bcdd777b60fe2c7bfef5456917affbc10

Review findings:

• Warning: crates/openshell-core/src/config.rs appears to still miss one case called out in the PR description. The new Podman detection probes Podman-specific socket candidates, but Docker detection can still treat Docker-compatible socket paths such as /var/run/docker.sock, $HOME/.docker/run/docker.sock, $XDG_RUNTIME_DIR/docker.sock, or DOCKER_HOST=unix://... as Docker based on socket existence. If /var/run/docker.sock is a symlink to a Podman machine socket, Podman probing can return false, Docker probing can return true, and auto-detect can still select the wrong driver.
• Suggested fix: include Docker-compatible socket candidates in the Podman Libpod probe before falling through to Docker detection, and add a unit test for a Libpod server reachable through a docker.sock-named path or symlink.

E2E decision: this affects Podman/Docker driver auto-detection. After the review finding is addressed, test:e2e should be applied before moving to pipeline watch.

Next state: gator:in-review.

It's true that the Docker driver can fail if Docker is selected and /var/run/docker.sock points at a Podman machine socket. That was also true before this change: the Docker driver uses Bollard's local defaults, which resolve to DOCKER_HOST or /var/run/docker.sock, and this PR does not change Docker endpoint resolution.

This change is scoped to Podman autodetection. Podman autodetection probes the same Podman-specific socket paths that the Podman driver will use by default, plus OPENSHELL_PODMAN_SOCKET when explicitly set. It intentionally does not treat Docker-compatible socket paths like /var/run/docker.sock as Podman discovery inputs.

I think we should fix the Docker autodetect and socket selection in a separate change; Bollard's defaults aren't sufficient. And in that change we should do the same as this change does for Podman, keeping the auto-detection's candidate sockets and the client's actual socket selection aligned.

gator-agent

PR Review Status

Validation: project-valid. The PR remains scoped to Podman autodetection and the author clarified that Docker-compatible socket selection should be handled in a separate Docker autodetect/socket-selection change.
Head SHA: 9abd176bcdd777b60fe2c7bfef5456917affbc10

Review findings:

• No blocking findings remain. The earlier Docker-compatible socket concern is not treated as a blocker for this Podman-scoped PR after the author's clarification.

E2E decision: this affects Podman/Docker driver auto-detection, so test:e2e is being applied before pipeline watch.

Next state: gator:watch-pipeline.

Label test:e2e applied for 9abd176. Open the existing run and click Re-run all jobs to execute with the label set. The run will execute the standard E2E suite after building the required gateway and supervisor images once. The matching required CI gate status on this PR will flip green automatically once the run finishes.