Fix nil pointer dereference on pipe close

[m-kostrzewa]

Fixes #44

We don't nil the file pointer until after the connectPipe() function returns.

[dgageot]

Hi @m-kostrzewa, do you think it's possible to write a unit test that demonstrate the issue?

[m-kostrzewa]

Hi @dgageot

I don't see a way to do this without introducing mocks, and I don't see them being used it current test suite. I could theoretically write a test that spawns like 1000 goroutines and detect the race that way, but it would be flaky and not really a unit test anymore. Not to mention it wouldn't really demonstrate the issue.

I wonder why this Sleep exists: https://github.com/Microsoft/go-winio/blob/master/pipe_test.go#L172. Maybe it was due to flaky test?

[m-kostrzewa]

@dgageot I asked some of my friends and they're not sure how to demonstrate this issue with unit tests either... it's a race condition, after all.

For what it's worth, we merged this patch to our repos 1.5 months ago and the issue never appeared again, and it would happen every other test run before that.

[jstarks]

Thanks for the contribution, sorry for the delay. We decided to take PR #63, which should address the same issue.