Code review request: 7178649: TEST BUG: BadKdc3.java needs improvement to ignore the unlikely but possible timeout

[Weijun Wang]

Please take a review at

    http://cr.openjdk.java.net/~weijun/7178649/webrev.00/

These tests are meant to verify the behavior of KDC fallback with 
different bad policy settings. Suppose one KDC has a problem, the client 
would try it several times each with a timeout, and then go on with the 
next KDC.

As there are a lot of timeout, I have designed the default timeout value 
to be very short to speed up the whole test. Every now and then, even a 
good KDC might not be able to reply in this short period, and an 
unexpected timeout appears, and the test fails.

This code change is meant to make the verification smarter, so that it 
tolerates such kind of "real" timeout. Suppose KDC1 is off and KDC2 is 
on, then with a max-retries of 3, ideally you will see 3 tries on KDC1 
and 1 try on KDC2, but you might also see 2 tries on KDC2. I'm now using 
the regex "1112{1,2}" to denote this possibility.

Please note that in some cases a single "real" timeout will make the 
whole authentication fail if a KDC is only given max_retries as 1. This 
is the case in BadKdc1.java's 4th and 5th rounds of test, where the 
newly started KDC2 has max_retries being 1 (because of its bad history). 
If a timeout happens on it, there is no way to try it again and the 
whole authentication would fail. Thus I use "-" in the regex pattern as 
a failure.

Thanks
Max



-------- Original Message --------
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7178649
7178649: TEST BUG: BadKdc3.java needs improvement to ignore the unlikely 
but possible timeout

   Product: java
   Category: java
   Subcategory: classes_security

=== *Description* 
============================================================
sun/security/krb5/auto/BadKdc3.java

Test needs improvement to ignore the unlikely but possible timeout. 
Currently it might fail due to such timeout.

sun/security/krb5/auto/BadKdc1.java
has similiar issue