A follow on from #118452, addressing a similar class of flaky test on iOS.
os.sysconf() appears to have flakiness similar to that observed with posix.confstr(). So far, it has been less common than the test_posix failure; as with the test_posix case, I've been unable to reproduce it locally, and the next CI pass almost always passes.
One of the updates in this PR is in os_helper, which already has a failover case if os.sysconf() isn't available or doesn't work. I've done a test deliberately simulating the failure case (effectively using the constant value of MAXFD regardless of the sysconf value), and the tests that use this utility continue to pass on iOS.
A follow on from #118452, addressing a similar class of flaky test on iOS.
os.sysconf()appears to have flakiness similar to that observed withposix.confstr(). So far, it has been less common than thetest_posixfailure; as with thetest_posixcase, I've been unable to reproduce it locally, and the next CI pass almost always passes.One of the updates in this PR is in
os_helper, which already has a failover case ifos.sysconf()isn't available or doesn't work. I've done a test deliberately simulating the failure case (effectively using the constant value ofMAXFDregardless of the sysconf value), and the tests that use this utility continue to pass on iOS.