[rfc][icedtea-web] rewritten xfork test

[Pavel Tisnovsky]

Jiri Vanek wrote:
> Inspired by Omair's effort to reduce failing tests, I have rewritten
> deadlock tests, which were failing quite randomly in half of the cases.
> It was caused by delayed removing of zombies after process have time-outed.
> 
> By this patch, ignorelist which I have proposed two days ago (and will
> do again soon as separate patch) to ignore no-testable classes (eg 
> interfaces), and fixes from [RFC][icedtea-web] fixes to reproducer
> engine, we will be on three  failing reproducers ("spaces" ) and  with
> occasional cache test fill.
> 
> All those five -  two sometimes, three regularly - failing reproducers
> are the same issue we are talking  in [icedtea-web] RFC: disable tests
> that are known to fail thread. - they are representing issue., but...:)
> 
> 
> 
> 2012-03-22  Jiri Vanek <jvanek at redhat.com>
> 
>     Rewritten DeadLockTestTest to stop failing in more then 1/2 of cases
>     All assassinated processes were hanging as zombies, killed forcibly by
>     kill -9 now.
>     *
> /tests/jnlp_tests/simple/deadlocktest/testcases/DeadLockTestTest.java:
>     (countJavaInstances) now return pids of found javas.
>     (killDiff) new method killing zombie javas forcibly.
> 

Hi Jiri,

could you please fix following typos in the test tool?

"This process is hanging more th_e_n"

"...but isnt!"
          ^    (on more places)

"proces inside background deadlock cant be null. Was."
                                     ^           It ^

"lunched JVMs must be exactly 2"
  ^  (this one is funny :-)


Thank you in advance,
Pavel