Re-enabled wrtc server tests, using SIGKILL to ensure Electron daemon gets killed

[mappum]

The tests close correctly on Linux now, since we ensure Electron closes by sending it a SIGKILL signal. I'm still not sure what's keeping it open, but it might make sense to just use SIGKILL by default in electron-eval. (Would you rather I make that change and just update the dependency here?)

[mappum]

Not sure why this failed, is the ECONNRESET error actually caused by this PR? (Because it passes for me on multiple machines).

[feross]

@mappum I think defaulting to SIGKILL in electron-eval makes sense, if that's what's required to make it work seamlessly on Linux. I can't think of any downsides to doing that.

Will re-run the tests to see if ECONNRESET was a false positive.

[feross]

I think the SIGKILL is causing the ipc to error somehow, since it's an unclean exit. :(

I'll try hacking around this for now by adding:

wrtc.electronDaemon.on('error', function () {})

But maybe this is an electron / electron-spawn bug that we should file? Or maybe you can just add the above code to suppress errors after the user calls wrtc.close()?

[feross]

Sorry, I had to revert this PR. The ECONNRESET error is indeed cause by this PR. The PR works for me, but commenting out the close('SIGKILL') line makes the tests pass.

Something strange is going on here.

[mappum]

Yeah, could definitely be electron. Closing the process seems to be connected to the ECONNRESET, I think this is a similar problem to this issue: mappum/electron-webrtc#15. That one isn't SIGKILLing, but the process is closing.