Latest IcedTea7 forest sync

[Mark Wielaard]

On Wed, 2012-11-07 at 15:07 -0500, Andrew Hughes wrote:
> I'd like to know that we haven't broken SystemTap support in the process,
> so could someone please verify that everything is still working as it should
> with IcedTea7 HEAD? :-)

Thanks for the integration and yes they do work fine.
Jon wrote jstaptest.pl which is ran with make check (check-tapset
target, which tests both the tapsets and jstack support, both of which
rely on the SDT markers being there). make check-tapset shows one odd
failure, but except for that odditiy everything seems fine:

$ make check-tapset
/home/mark/src/icedtea7/test/tapset/jstaptest.pl \
  -B /home/mark/build/icedtea7-obj/openjdk.build -A amd64 \
  -S /home/mark/src/icedtea7/test/tapset \
  -a test/check-stap.log -p
Compiling tests.
Check whether we have enough privs to run systemtap script...
OK
Testing if systemtap can match probes.
..................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
Testing if detected probes work as expected.  This may take a while...
............................................................................................................................................................................................................................................
Probe hotspot.jni.FatalError failed.
......................................................................................................................................................................................................................................................................
Working probes:     497
Broken probes:     1
Some tests did not work as expected.  See file test/check-stap.log for details.

The hotspot.jni.FatalError probe triggers right before a JNI using
program tells the JVM to abort(). Inspecting the failing testcase by
hand shows it does trigger, but instead of printing the message, it
prints FatalErrormsg=<unknown>. Which is odd, because when you print
$arg2 it correctly shows the address of the msg (0x40a182 in my case)
and running the same under gdb shows that is correct:
#3  0x00007ffff7648aaa in jni_FatalError (env=<optimized out>, 
    msg=0x40a182 "Intentional Crash: Ignore.")

at /usr/local/build/icedtea7-obj/openjdk/hotspot/src/share/vm/prims/jni.cpp:870
870	  os::abort(); // Dump core and abort
(gdb) print msg
$1 = 0x40a182 "Intentional Crash: Ignore."

So, this is something in the stap runtime (maybe it really cannot read
the user string at that point for some odd reason). Anyway, it isn't an
issue with the patch. Everything else works nicely and this is a
somewhat odd SDT marker probe anyway because the very next thing that
will happen is the abort() call.

I'll see if I can trick Jon into setting up his buildbot fedora testers
to have full systemtap privileges so that these tests are automagically
ran by his buildslaves.

Cheers,

Mark