Major blunder on my part: I had Eclipse using a classpath with build/classes instead of build/exposed.

Is it safe to keep build/classes in the classpath at all?
The wiki seems a little outdated with respect to Eclipse setup ( but I don't have a convenient configuration to offer yet.

- Ken

On Wed, Sep 8, 2010 at 6:47 AM, Ken Fox <> wrote:
Thanks--it's tough starting on a large project because I'm never sure if I have things setup properly. Nice to know you don't have hazing rituals. :)

The latest svn updates (7113) didn't fix my problem, but I've narrowed it down some and will dig into it. I found a small JUnit test, org.python.core.
WrappedFloatTest, that reproduces the null defaultSystemState exception in PythonInterpreter.initialize.

I can get it to make a bit more progress (fails testAdd, but passes testDiv and testMod) with this hack. This seems to indicate a static initializer bug.

--- src/org/python/core/    (revision 7113)
+++ src/org/python/core/    (working copy)
@@ -34,7 +34,12 @@
             ts = new ThreadState(t, newSystemState);
-            newSystemState.registerThreadState(threadLocal, ts);
+            if (newSystemState == null) {
+                Py.writeWarning("init", "Py.defaultSystemState is null");
+            }
+            else {
+                newSystemState.registerThreadState(threadLocal, ts);
+            }
             return ts;

On Tue, Sep 7, 2010 at 5:57 PM, Jim Baker <> wrote:

It looks like you are seeing the same issue recently raised on #jython; see No it's not part of some Jython hazing ;) , because it applies to all of us. (And in this case, it's because I broke the build, so my apologies!) It's just a problem when we have too much unreliable dev infrastructure to immediately catch failures like this one.

Fortunately, we are catching them reasonably quickly, either manually thanks to people like you or when an instance comes back on line.

Anyone following this list or IRC will recognize that we have had problems with our Hudson instance, buildbots, IRC log bots, and issue tracking system in the last 6 months. Some of these might be best resolved by moving to an appropriate forge, but I'm not aware of a forge that supports continuous integration or the log bot, but actually these are both very important to us.

Maybe the ideal solution is to have an organization donate some reliable, administered hardware resources that don't randomly break or have intermittent network connectivity. If we can get that sort of support, that would be much appreciated.

- Jim

On Tue, Sep 7, 2010 at 2:06 PM, Ken Fox <> wrote:
I'm trying to understand Jython (svn trunk) in Eclipse Helios under Java 6 (Mac OS X) so I can eventually help fix bugs. Running org.python.util.jython.main fails due to Py.defaultSysemState being null when used in getThreadState. I may have a bad project setup, but I followed the wiki--there's a "Bootstrap types weren't encountered in bootstrapping" warning to the console.

This seems like a circular dependency bug because the PySystemState constructor calls __dict__.invoke which seems to rely on Py.defaultSystemState which is only set after the constructor finishes.

Unless this is part of Jython developer initiation/hazing, I'd really appreciate a little nudge in the right direction.


- Ken

This Dev2Dev email is sponsored by:

Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
Jython-dev mailing list