You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Scheduled automation workers never start when the pane shell is PowerShell #507
On Windows with "shell": "pwsh" or "powershell", a scheduled automation never starts its agent. The worker pane receives:
"C:\...\luvus.exe" __automation-worker t1 a1 r1
PowerShell reads a quoted string followed by more tokens as an expression and stops:
ParserError: Unexpected token '__automation-worker' in expression or statement.
The run and its ORCH task then stay running, and nothing reports the failure. With the default overlap: skip, every later occurrence is skipped. An --every 60 automation recorded r1 as running (stuck), then r2 and r3 as skipped with previous run is still active.
Measured
0.14.3 (52def96), Windows 11, PowerShell 7.6, isolated LUVUS_HOME. It fails before any agent starts, so it affects every agent. The same automation runs with "shell": "cmd".
Cause
automation_agent_launch_line (src/app/board.rs) always emits "<exe>" __automation-worker … on Windows. Manual tasks already handle this: manual_agent_runner_line uses & "$env:LUVUS_BIN_PATH" __task-worker for pwsh/powershell and "%LUVUS_BIN_PATH%" __task-worker for cmd.
Proposal
Build the automation launch line per shell in the same way, using the call operator & for PowerShell. Add tests for pwsh, powershell, cmd, and the POSIX form.
I'd like to fix this and will open a PR that closes this issue.
Activity