[2.20] [6/6] Do not terminate default test runs on test failure

[Joseph S. Myers]

On Mon, 10 Feb 2014, Brooks Moses wrote:

> True, and I strongly agree with the sentiment!  However, often the
> failures get sorted out in trunk (or in the Linux kernel), and we are
> working on an older release branch, and even if fixes get backported
> the right answer is often to xfail locally while discussing upstream.
> 
> For what it's worth, here are some discussion threads related to tests
> we've disabled locally:
> http://cygwin.com/ml/libc-alpha/2010-09/msg00030.html
> http://cygwin.com/ml/libc-alpha/2012-02/msg00466.html
> https://sourceware.org/ml/libc-alpha/2012-10/msg00021.html
> 
> Beyond those, we also disabled nptl/tst-robust8 because of flakiness
> (I don't know details there), and rt/tst-cpuclock1 and
> rt/tst-cputimer{1,3} because of virtual-machine issues in the test
> environment, and I think that's all.

Well, if any of these aren't resolved in current glibc, the usual 
principle applies of keeping pinging and working on driving things to 
consensus until the issue is resolved.  And if a fix depends on a new 
kernel feature it may be appropriate to make the tests laxer when run on 
older kernels - as long as there are appropriate __ASSUME_* conditionals 
on the laxity so that it's clear when it can be removed because the older 
kernels are no longer supported.

-- 
Joseph S. Myers
joseph@codesourcery.com