Demonstration of signal handling race

[metcalf]

This is a demonstration test for a signal handling race. I'll follow up with more details in an issue once I get a Travis build failure to link.

[tarcieri]

This doesn't look like a "race" so much as a spurious wakeup. Anything that can interrupt a system call will cause the selector to wake up. nio4r intentionally does not gloss over these wakeups, but assumes you're using the selector in a loop which can handle them and retry selecting in the event they are spuriously awoken.

Perhaps this behavior can be better documented, but at least if I'm understanding your example case this is a WONTFIX/"works as intended" for me.

The real answer to providing better behavior is to abstract nio4r in a higher level framework which implements a reactor/proactor loop.

[metcalf]

@tarcieri: I added more details in #134. Prefer to continue the discussion over there?

[tarcieri]

Cool, let's continue discussion on #134