Several of our js tests rely on spawned shells to run operations in parallel with the main test script. If the interface for managing these spawned shells was a bit more flexible we might be able to make test failures more clear and eliminate some bad failure conditions (for example waiting forever).

Some things which would be immediately helpful:

• being able to check if a shell has finished without waiting for it to finish
• being able to wait for a shell to finish, but with a timeout
• explicit interface to audit errors from the returned shell (via shell return code or another mechanism)

I'm sure other features would be useful too - these are just ones that come to mind immediately.

As an aside - another option would be to allow threading in shells. This might make our tests clearer as well as easier to control (since we don't have to serialize input scripts for separate processes) and errors might be easier to detect as well. This functionality is currently implemented for our v8 mode but not for spidermonkey.