In `@cmd/agent/workspace/up.go`:
- Around line 60-93: The init failure can leave credential files behind because
cleanupCredentials is only deferred after a successful init; change the
call/handling so credentialsDir is always captured and cleaned up even on error:
call initWorkspace(...) and immediately defer
cmd.cleanupCredentials(credentialsDir) after the call (before checking err) or
modify initWorkspace to return the credentialsDir even when it returns an error,
then ensure UpCmd.Run defers cleanupCredentials(credentialsDir) right after
initWorkspace returns so both the success path and the cmd.handleInitError(err,
workspaceInfo, logger) path always remove credentials; update any tests or
callers expecting empty credsDir on error accordingly.
- Around line 414-437: The shouldConfigureDockerDaemon method currently returns
err != nil && !local which skips daemon config when Local parsed successfully as
false; change the logic so it returns err == nil && !local (i.e., configure when
parsing succeeded and Local is false). Update
workspaceInitializer.shouldConfigureDockerDaemon to call
workspaceInfo.Agent.Local.Bool(), return true only when err == nil && !local,
and optionally log or handle the parse error from
workspaceInfo.Agent.Local.Bool() instead of treating it as a successful false.
- Around line 121-132: handleInitError can be called with a nil logger when
initializeTunnel fails, causing DeleteWorkspaceFolder to panic when it calls
log.Errorf; before calling clientimplementation.DeleteWorkspaceFolder in
handleInitError, guard against a nil logger (logger == nil) and substitute a
no-op or default logger instance (e.g., a NOP logger or a logger created via the
same log package used elsewhere) so DeleteWorkspaceFolder and its use of logger
(via log.Errorf) won’t panic; reference handleInitError, initializeTunnel,
DeleteWorkspaceFolder, DeleteWorkspaceFolderParams and workspaceInfo when
locating the change.