Serious regression in staging

Hi,

After the merge of the latest GNU Classpath SableVM/staging seems to
have problems with GUI applications (or at least all command-line apps
I tested worked, but none that included some GUI).

My guess is that the problems have been introduced by removal of
SecurityManager and advent of VMStackWalker, which is now used to
retrieve the current class loader.

We used to have 3 (!) distinct implementations of what is now
gnu.classpath.VMStackWalker.getClassContext() method, but apparently
they were not used to such extent (not so many thing relied on them)
and the old code is not good enough anymore (see
http://sablevm.org/bugs/89 for one of the problems). 

There is also something weird happening, that results in having an
object instance, with instance->vtable->type == null, which gives
segfaults on monitor_enter().  I suspect it also might have much to
do with VMStackWalker, which is now heavily used in j.l.Class.

I am trying to debug this issue, but I feel quite sick since a few
days, and it goes rather slowly ATM.

Etienne wanted to revert staging to before-merge state right away, but
I am not a big fan of reverting things on the first (even serious)
problem.  At the moment staging is considered broken, and if we don't
find a fix until sunday evening, I'll revert it to the previous state.
The fact is, however, that we have to find the fix anyway, though the
sooner, the better.

Help is very much welcomed.

				Grzegorz B. Prokopski

-- 
Grzegorz B. Prokopski           <[email protected]>
SableVM - Free, LGPL'ed Java VM  http://sablevm.org
Why SableVM ?!?                  http://sablevm.org/wiki/Features
Debian GNU/Linux - the Free OS   http://www.debian.org