[Bug classpath/31479] New: segfault related to String encoding, JamVM 1.4.5 + Classpath 0.93 on ARM processor

[robert dot dodier at gmail dot com]

I have been running into a persistent error, namely that almost every time I
run a Java program on a Gumstix computer, it runs normally (or dies with a
segfault along the way) and then dies at the end with a segfault. Gumstix is a
small computer which runs a ARM processor.

I am using JamVM 1.4.5 + Classpath 0.93 (most recent versions so far as I
know). Also I am compiling Classpath with Jikes 1.22.5, and compiling my own
programs with Jikes or javac from Java 1.5.0.

The problem seems to be related to string encoding: in some (maybe many cases)
I can prevent the segfault by replacing all System.out.println(s) with
System.out.println(new String(s.getBytes(), "US-ASCII")) i.e. explicitly
specifying an encoding. Specifying -Dfile.encoding=US-ASCII seems to change the
situation somewhat but doesn't fix it.

I can't extract anything useful from strace output. I have run jamvm with the
-verbose option and also recompiled jamvm with --enable-tracedll which is
interesting (shows library access from jamvm) but it doesn't indicate the
source of the problem. valgrind is not available for the ARM processor.

I also tried running the same programs on JamVM + Classpath built on a X86
processor, and there was no segfault.

I know this is vague but it has proven extremely hard to pin down.

Thanks for your help. Robert Dodier


-- 
           Summary: segfault related to String encoding, JamVM 1.4.5 +
                    Classpath 0.93 on ARM processor
           Product: classpath
           Version: 0.93
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: classpath
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: robert dot dodier at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31479