Ask the agent to run git status in legacy terminal in an empty repository - he will say that he received an empty output, although you will see a one-line output of the command
Sorry for the delay here. Indeed, looking into the logs I can see the missing final line of your scripts.
I’m still having trouble reproducing this on my end. Could you explain why you’re using the Legacy Terminal Tool? Do you have different behavior locally when you switch it off?
The new terminal launches in read-only mode, which prevents the user from entering any commands into the same terminal.
Based on the previous point, if the Agent accidentally or intentionally runs a command that requires user input, the command cannot be executed. Also, I cannot send a KeyboardInterrupt signal.
The new terminal does not monitor running processes, leaving behind zombie processes.
The new terminal ignores color coding, which reduces the convenience and speed of reading the terminal output for the user.
I filed bug reports about all these problems as soon as the new terminal was released in Early Access, and the only solution offered by the team was a button to switch back to the legacy terminal.
The only thing I don’t like about the legacy terminal is the inability to disable the auto-kill process function, which can kill the terminal during long compilations or manual testing.