Specifically:
Apple LLVM version 7.0.0 (clang-700.0.72)
Target: x86_64-apple-darwin14.5.0
Thread model: posix

I'm assuming this is a Mac issue since none of the buildbots are having a similar issue.

Brett, was this on OS X 10.10.x or a 10.11 pre-release?  test_eintr does not fail fail for me on either using the most recent Xcode 7.

It is 10.10, but apparently a reboot fixed it (software, am I right? =/ ).

This is still an intermittent problem for me. Maybe it's related to issue #25122 ?

Hi Brett, if you are motivated, you can try the C program that I wrote to reproduce the bug on FreeBSD. It's attached to #25122. If the C program hangs too, it would mean that Darwin has also a bug (same bug??) in its code to handle signals in the code handling FIFO devices.

Or you can just add a skip mentioning this issue.

On my FreeBSD VM, it took a few minutes and multiples instances of the C program to reproduce the bug. On physical machines, other reported that a seconds and a single instance is enough. Well, see the FreeBSD bug report for more information ;-)

I can't get your testing program to compile, Victor, so if I run into it again I will add a skip to the test (harder to detect that it's the same test since test_eintr runs another test suite as if it's a single test).

The Yosemite ICC 3.5 buildbot seems to be intermittently hanging test_os_open() and test_open(). Currently it shows five builds. Build 396 hung at test_os_open(); build 400 hung at test_open(), and the other three passed.

http://buildbot.python.org/all/builders/x86-64%20Yosemite%20ICC%20Debug%203.5

Slave information
Buildbot-Slave 0.8.12
OSX 10.10.3 (Yosemite) Mac Mini VMWare VM 2x2.86 GHz 8GB

I'm still getting the occasional hang for this so I'm going to skip the test under OS X.

> I'm still getting the occasional hang for this so I'm going to skip the test under OS X.

Go ahead. I'm not interested to analyze bugs of closed source kernels.

New changeset 7fc1296d1acf by Brett Cannon in branch 'default':
Issue #25234: Skip test_eintr.test_os_open under OS X.
https://hg.python.org/cpython/rev/7fc1296d1acf

Brett, do you think you should also skip test_open()? The same buildbot recently failed again in that method.

Sure, that test can be skipped as well if it is hanging.

New changeset 9d65a195246b by Brett Cannon in branch 'default':
Issue #25234: Skip test_eintr.test_open() under OS X to avoid hanging
https://hg.python.org/cpython/rev/9d65a195246b

New changeset ddce15b21c21 by Victor Stinner in branch '3.5':
Backport fixes on test_eintr
https://hg.python.org/cpython/rev/ddce15b21c21