SIGINT (sent with CTRL+c) doesn't always stop the test.

[jettisonjoe]

Usually it does, but not always. I haven't identified the specific conditions that cause SIGINT handling to break yet.

[fahhem]

This is usually a threading problem, they don't always like to die on their own. You can either set them to daemon mode or explicitly pass some "die now" data into them

[jettisonjoe]

Yeah, I suspect the threads that aren't dying are in the Rocket instance in the HttpHandler. Once we flatten multi-cell support in the framework, there will be fewer places to look for the cause of this behavior.

[jettisonjoe]

More info on this. It actually still happens with the bare-bones frontend API server we now have, so it wasn't specific to Rocket. SIGINT handling consistently fails if SIGINT happens while execution is inside one of the HTTP handlers.

[wallacbe]

Provide a DUT ID in order to start the test.

^CReceived SIGINT. Stopping everything.

Exception happened during processing of request from ('127.0.0.1', 57906)
Traceback (most recent call last):
File "/usr/lib/python2.7/SocketServer.py", line 295, in _handle_request_noblock
self.process_request(request, client_address)
File "/usr/lib/python2.7/SocketServer.py", line 321, in process_request
self.finish_request(request, client_address)
File "/usr/lib/python2.7/SocketServer.py", line 334, in finish_request
self.RequestHandlerClass(request, client_address, self)
File "/usr/lib/python2.7/SocketServer.py", line 647, in init
self.setup()
File "/usr/lib/python2.7/SocketServer.py", line 699, in setup
self.rfile = self.connection.makefile('rb', self.rbufsize)
File "/usr/lib/python2.7/socket.py", line 217, in makefile
return _fileobject(self._sock, mode, bufsize)
ThreadTerminationError

[wallacbe]

Observing this also with the bazel build environment

[wallacbe]

Found that SIGQUIT ( Ctrl+) will kill the process.

[fahhem]

Does this only happen when the frontend is being accessed or can it be any time?

[wallacbe]

My test cases about were without the front end running.

Also SIGQUIT may not be best method since tear down methods would be skipped.

[wallacbe]

Additional observation from head:

If I start the tester like this then do CTRL+C it hangs
test.Execute(test_start=triggers.PromptForTestStart())

If I disable http and execute CTRL+C stops execution immediately:
test.Configure(http_port=None)
test.Execute(test_start=triggers.PromptForTestStart())

Does this point to http server being the issue?

[jettisonjoe]

Yes, I believe the issue is in the http server. @grybmadsci and I have some changes planned for the HTTP API that will give us a good opportunity to investigate and resolve this.

[jettisonjoe]

Is anyone still seeing issues with CTRL+c in the current PyPI version?

[jettisonjoe]

I believe this is fixed now.