[renatoathaydes]

What

devpod status --all shows the status of all workspaces of the current context in one command:

                      NAME                       | PROVIDER   | STATUS
  -----------------------------------------------+------------+---------
  curity-ai-plugins-devpod-git-identity          | aks        | Stopped
  k8s-ai-plugins                                 | kubernetes | Running

• Workspaces are ordered most recently used first, like devpod list.
• Statuses are retrieved concurrently (up to 8 at a time), each within --timeout. A timeout is reported as "timed out after 30s, you can wait longer via --timeout". --container-status still applies. The status of a machine that several workspaces share is only retrieved once.
• The table, or the hint when there are no workspaces, goes to stdout (like devpod list), so it can be piped; the reasons for Error rows are warnings on stderr. --output json prints a list of the same WorkspaceStatus objects devpod status --output json prints for one workspace.
• A workspace whose status can't be retrieved, including one whose workspace.json can't be loaded, is still listed:
  • In the plain table it shows as Error with a warning, and after the others are shown the command exits with an error.
  • In JSON it gets a new optional error field, and the command succeeds, because the errors are part of the output and callers can still use the other statuses.
• --all together with a workspace argument is an error. With no workspaces, plain output prints a hint and JSON prints [].

Why

Checking which dev containers are running used to need one devpod status per workspace.

Notes for reviewers

• pkg/workspace.ListClients returns a client for each workspace of the context, most recently used first, or the error that prevented it. It loads the providers once, not once per workspace, so warnings about a broken provider aren't repeated for every workspace. ListLocalWorkspaces and loadExistingWorkspace share their directory listing and machine loading with it, and FindProvider shares its lookup.
• workspace.SortByLastUsed is now the one place for the most-recently-used ordering, used by devpod list too.
• client.StatusOptions gains MachineStatuses *MachineStatusCache (json:"-"). The workspace client uses it, when set, to share a machine's status between workspaces; every other caller leaves it nil and behaves as before.
• client.WorkspaceStatus gains Error string \json:"error,omitempty"`, which only --all` sets, so existing JSON output is unchanged.

Testing

• Ran against real Kubernetes/AKS workspaces. Also tried a 1ms timeout, and, with a temporary DEVPOD_HOME, a workspace with a missing provider, a corrupt workspace.json, no workspaces, an invalid argument and an invalid output format, all in both output formats.
• The up-docker e2e spec "should start a new workspace with existing image" now also checks that the workspace is Running in status --all; it passes locally.
• New unit tests cover collecting and printing the statuses (order, errors, per-workspace timeouts, the concurrency limit, cancellation), ListClients, SortByLastUsed and the machine status cache, and pass with -race.
• golangci-lint is clean on the changed packages, and the unit tests pass.

🤖 Generated with Claude Code