Need advice finding Java-Plugin regression

Hello,


Since Java-6 (5.0_u3 worked) there's a regression in the Unix mozilla
browser plugin which has the following impacts:
- FireFox waits while loading the javaplugin with 100% cpu
- If an applet-start fails, mozilla/firefox dies (firefox never gets
out of the 100% state).
This is the really critical part, its enough to let an applet throw an
exception in init() and the browser is dead, I reported the bug about
a year before: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6471789

Some profiling with oprofile showed that most time was wasted in kernel-code:
303179   39.6945  no-vmlinux               no-vmlinux               (no symbols)
44400     5.8132  libpthread-2.6.so        libpthread-2.6.so
pthread_cond_broadcast@@GLIBC_2.3.2
42011     5.5004  anon (tgid:892 range:0xb7fe5000-0xb7fe6000)
firefox-bin              (no symbols)

So the conclusion I draw from this is that it seems some waiting-code
which is responsible for waiting until the java-plugin has loaded is
somehow crippled.
I'll also write to the mozilla developers to get some build
instructions how to get a better debugable firefox-binary.

Any idea where to look at first? My intention is not to bring
attention to this bug but to fix it myself, however its quite hard
without knowing where to start searching ;)
Is there some waiting code, and where should I start searching.
Some help would be greatly apprciated.

* I already built my own FireFox-2.0.0.6, however although I said
--diable-optimize and --enable-debug-modules it was compiled with -O2
and of course without debuging symbols. Is there a better way to
create a not-optimized debug-friendly build?

* Since FireFox3-Alpha(6?)7 starting java-applets (6.0_u1) immediatly
crashes firefox on my machine. Is this a known problem (thats was my
thought, thats why I didn't report it).


Thank you in advance, lg Clemens