[rvm-core] Getting to the bottom of NPEs during object copying

[Ian R. <rog...@gm...>]

Hi,

we have an intermittent failure where we die during object copying with 
an NPE, for example [1]. I've fixed a few bugs in dump virtual machine, 
so we can now see exactly what the state of all threads is but this 
doesn't seem to make the problem any easier to debug. I believe the 
poisoned tests are trying to get to the root of this problem - is this 
right? Looking at the last x86 sanity results all the poisoned tests are 
failing in the finalizer thread [2][3][4]. RVM-440 captures this problem 
[5]. Is this narrowing in on the finalizer thread being the source of a 
bad reference that's causing what's probably our most common 
intermittent failure now?

Thanks,
Ian

[1] 
http://jikesrvm.anu.edu.au/cattrack/results/rvmx86lnx32.anu.edu.au/perf/3923/production/default/perf-dacapo/lusearch-10/6/Output.txt
[2] 
http://jikesrvm.anu.edu.au/cattrack/results/rvmx86lnx32b.anu.edu.au/sanity/3922/FullAdaptivePoisoned/default/dacapo/lusearch/default/Output.txt
[3] 
http://jikesrvm.anu.edu.au/cattrack/results/rvmx86lnx32b.anu.edu.au/sanity/3922/FullAdaptivePoisoned/default/jni/t3GT3/default/Output.txt
[4] 
http://jikesrvm.anu.edu.au/cattrack/results/rvmx86lnx32b.anu.edu.au/sanity/3922/FullAdaptivePoisoned/default/jni/t3GT3/default/Output.txt
[5] http://jira.codehaus.org/browse/RVM-440