[Javapathfinder-devel] SF.net SVN: javapathfinder: [277] trunk

[<pcm...@us...>]

Revision: 277
          http://javapathfinder.svn.sourceforge.net/javapathfinder/?rev=277&view=rev
Author:   pcmehlitz
Date:     2007-03-09 11:16:05 -0800 (Fri, 09 Mar 2007)

Log Message:
-----------
* Ok, there are daemons out there - by chance discovered that running TestVMDeadlockJPF
from ant and explicitly via bin/javajpf yielded different results! Ant said Ok, javajpf
actually reported an error.

to make matters worse, the error was real - the artificial RUNSTART (justified by the
"max one CG per insn" constraint) in combination with the Thread.start() native peer
caused an assertion in ElementInfo.lock(), because nobody ever called
ThreadInfo.registerLockContender(ei). And to make that even worse, I now think that
Thread.start has to be a breaker (we had a createThreadStartCG in the SchedulerFactory,
but it was optional and returned null). I think the reason why it has to be a right mover
is that the freshly started thread might well do something (like starting another thread
that interacts with the starter) that causes one of the next insns in the starter to
become a right mover. The real trick with this is that we can't check for
ThreadInfo.isFirstStepInsn in RUNSTART, since it always will be the first insn, so the
block()/registerLockContender() is in Thread.start(), the lock(ti) is in RUNSTART :(
Still need a regression test for that


* finally did bite the bullet and created a 'RawTest' base for our raw tests. Still
it has to run outside JPF, but the main() driver was just too redundant. Now I just
have to update all raw tests to use that (so far, only TestVMDeadlock does)

* added simple asym lock and nested monitor lockout test cases to TestVMDeadlock

* fixed a small bug in ExecTracker, which gets into trouble if we skip insn logging

Modified Paths:
--------------
    trunk/env/jvm/gov/nasa/jpf/jvm/JPF_java_lang_Thread.java
    trunk/extensions/numeric/test/gov/nasa/jpf/numeric/TestNumeric.java
    trunk/src/gov/nasa/jpf/jvm/DefaultSchedulerFactory.java
    trunk/src/gov/nasa/jpf/jvm/ElementInfo.java
    trunk/src/gov/nasa/jpf/jvm/StackFrame.java
    trunk/src/gov/nasa/jpf/jvm/bytecode/RUNSTART.java
    trunk/src/gov/nasa/jpf/tools/ExecTracker.java
    trunk/test/gov/nasa/jpf/mc/TestVMDeadlock.java
    trunk/test/gov/nasa/jpf/mc/TestVMDeadlockJPF.java

Added Paths:
-----------
    trunk/test/gov/nasa/jpf/jvm/RawTest.java


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.