Killing plots.py does not kill inference subprocesses

[Yan Wong]

If an inference subprocess stalls, and the main plots.py routine is killed, the inference subprocess carries on running. When the main process is killed, it should probably kill all child processes too. Is there an easy way to do this, @jeromekelleher ?

[Jerome Kelleher]

Using ctrl-D instead of ctrl-C is the easy way to do this @hyanwong. I'm not sure why, but it kills all child processes as well as the main process.

[Yan Wong]

Ah, thanks!

[Yan Wong]

How do I debug a subprocess? I have a seemingly hanging tsinfer process (see my email)

[Jerome Kelleher]

The trick with subprocesses is to get things working properly with a -p=1 first (which doesn't invoke any subprocesses) and then try it with many processes. Tracing errors is tricky with subprocesses.

I wasn't able to reproduce your issue with hanging tsinfer.

[Yan Wong]

Yes, thanks. I left it overnight and it was fine in the morning. Although I did think I had run it with p as the default, i.e. p=1, and it still fired off a subprocess. From my reading, the time_cmd() function allways invokes a subprocess (even if p=1), so that it can get time and memory usage back. Yan

…

On 24 Jan 2017, at 08:23, Jerome Kelleher ***@***.***> wrote: The trick with subprocesses is to get things working properly with a -p=1 first (which doesn't invoke any subprocesses) and then try it with many processes. Tracing errors is tricky with subprocesses. I wasn't able to reproduce your issue with hanging tsinfer. � You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[Jerome Kelleher]

Ah yes, that's true. I thought you were talking about worker processes.

[Yan Wong]

Reopening, as Cntl-D doesn't terminate the process on my machine.

[Jerome Kelleher]

I've just pushed a change that might help. It's a bit messy, but if you hit CTRL-C now it should clean up all descendant processes.

[Yan Wong]

OK, seems to work. closing