Re: [rvm-research] ZipException when running Dacapo 2006-10-MR2 eclipse on Ubuntu 13.04

Hi Erik,

On 08/09/2013 02:26 PM, Erik Brangs wrote:
> Hi,
>
> On 07.08.2013 12:04, Erik Brangs wrote:
>> I recently upgraded to Xubuntu 13.04 which seems to have introduced an
>> error in running eclipse from Da Capo 2006-10-MR2. When I try to run
>> DaCapo eclipse, it immediately fails with a ZipException (stack traces
>> are attached). Sometimes eclipse works fine. Other benchmarks seem to
>> work. An exception to this is luindex which also shows a ZipException
>> (with a different stack trace) from time to time.
>>
>> The problem is reproducible in a freshly installed and upgraded Xubuntu
>> 13.04 in VirtualBox on a another computer. It does not occur in a
>> freshly installed and upgraded Xubuntu 12.10. I've tried the following
>> to narrow down the problem:
>> -change version of JDK used for build (Java 6 vs. 7)
>> -use GNU Classpath from git
>> -use Jikes RVM 3.1.3 and Jikes RVM 3.1.2
>>
>> None of that seems to have made any difference in behaviour.
> I've tried to debug this issue further but have not met with success.
> The only thing I could find out was that it might be a heap corruption
> issue. The contents of the byte array window[] in the StreamManipulator
> class from GNU Classpath seem to get overwritten. This leads to errors
> when decoding the eclipse.zip file (which is contained in the jar) and
> thus to to the ZipException.
>
> I'd appreciate any tips for debugging.
>
> I have considered trying Valgrind but I have never used it before.
> Moreover, the newest release supports only up to glibc 2.16 (Ubuntu
> 13.04 uses glibc 2.17) so I would have to build from source. Could
> Valgrind help in debugging this problem?

Some random thoughts:

Have you tried MarkSweep to see if objects getting moved (or 
generational GC) is the problem? I know that Jikes handles I/O by using 
duplication to handle a buffer that might get moved, but maybe a 
previously true assumption is now getting violated.

The option -X:gc:eagerMmapSpaces=true can be useful to keep C and Jikes 
memory spaces from conflicting. On the other hand, if that were 
happening, it seems like the resulting errors are usually 
random/arbitrary crashes.

I wonder if running with gdb (i.e., by running "rvm -gdb") would yield 
anything interesting?

Cheers,
Mike

>
>
> Kind regards,
>
> Erik Brangs
>
> ------------------------------------------------------------------------------
> Get 100% visibility into Java/.NET code with AppDynamics Lite!
> It's a free troubleshooting tool designed for production.
> Get down to code-level detail for bottlenecks, with <2% overhead.
> Download for free and get started troubleshooting in minutes.
> http://pubads.g.doubleclick.net/gampad/clk?id=48897031&iu=/4140/ostg.clktrk
> _______________________________________________
> Jikesrvm-researchers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers


------------------------------------------------------------------------------
Get 100% visibility into Java/.NET code with AppDynamics Lite!
It's a free troubleshooting tool designed for production.
Get down to code-level detail for bottlenecks, with <2% overhead. 
Download for free and get started troubleshooting in minutes. 
http://pubads.g.doubleclick.net/gampad/clk?id=48897031&iu=/4140/ostg.clktrk