[openssl-project] update on sporadic test failures

[Benjamin Kaduk]

On Wed, Jan 10, 2018 at 02:12:05PM -0600, Benjamin Kaduk wrote:
> On Wed, Jan 10, 2018 at 09:06:33PM +0100, Richard Levitte wrote:
> 
> > Does this patch make a difference?
> > 
> > diff --git a/util/perl/TLSProxy/Proxy.pm b/util/perl/TLSProxy/Proxy.pm
> > index 99b0dedd5b..9330ec2844 100644
> > --- a/util/perl/TLSProxy/Proxy.pm
> > +++ b/util/perl/TLSProxy/Proxy.pm
> > @@ -300,6 +302,7 @@ sub clientstart
> >              && $ctr < 10) {
> >          if (!(@ready = $sel->can_read(1))) {
> >              $ctr++;
> > +            usleep(100000);     # .1 s sleep, totalling 1 s of attempted reads
> >              next;
> >          }
> >          foreach my $hand (@ready) {
> 
> I will give it a shot, but seeing how unreliable my reproducer is,
> it's hard to say that it will give much confidence of anything.

Well, that was quick:

Waiting for client process to close: 28371
# Subtest: No client extension extended master secret test
    1..4
    #   at ../test/recipes/70-test_tlsextms.t line 233.
    ok 2 - ClientHello extension extended master secret check
    ok 3 - ServerHello extension extended master secret check
    ok 4 - Extended master secret full handshake check
    # Looks like you failed 1 test of 4.
not ok 2 - No client extension extended master secret test

I'll note that I do have some other stuff going on (building openafs
in a loop) to generate some extra load on this system and try to
make races more likely.

-Ben