deps: update libuv to 1.2.0

[bnoordhuis]

R=@piscisaureus or @saghul?

[piscisaureus]

Probably unrelated but:

=== release test-process-active-wraps ===
Path: parallel/test-process-active-wraps
assert.js:100
  throw new assert.AssertionError({
        ^
AssertionError: 1 == 0
    at Immediate._onImmediate (D:\iojs\test\parallel\test-process-active-wraps.js:69:16)
    at processImmediate [as _immediateCallback] (timers.js:342:17)

[piscisaureus]

lgtm

[saghul]

LGTM. (and tests pass here on OSX, FWIW)

[bnoordhuis]

@piscisaureus Curiously enough, the same test is failing on the CI. I've been going over the changes in libuv but I don't see an obvious culprit.

[bnoordhuis]

@piscisaureus I let the CI do some ad hoc bisecting and it looks like the regression was introduced in commit 94e1475, the joyent/v0.12 merge. Maybe you can run a git bisect locally to track down the offending commit?

[cjihrig]

I'll look into this too. I recently changed that test in b636ba8. It works fine on OS X, but fails on Windows.

[bnoordhuis]

@cjihrig It's probably unrelated but there was a merge conflict around line 840 in lib/net.js when I did the merge. It's possible I screwed up conflict resolution although it doesn't look like it (and like you say, the test passes on other platforms.)

[cjihrig]

The problem seems to be the removal of this logic.

@piscisaureus any clue why that would leave an extra handle around on Windows? It was removed to try to eliminate hard coded IPv4 addresses.