Re: [SCM] master: Add a test testMemoryBufferCapacity() to TestByteBuffer.java

[Mark Wielaard <mark at klomp dot org>]

Hi Pearly,

On Wed, 2007-11-28 at 16:24 +0800, Zhao Shujing wrote:
> It's been updated and the test is passing now.

Congrats, it does PASS on x86 and x86_64. And more congrats, since you
seem to have caught a bug on 32bit-on-64bit systems! If you run the test
on a x86_64 system with ./TestRunner -arch 32
frysk.proc.live.TestByteBuffer.testMemoryBufferCapacity you get:

Caused by this piece of code in frysk-sys/frysk/sys/cni/Ptrace.cxx:

jlong
frysk::sys::Ptrace$AddressSpace::length ()
{
  return -1UL;
}

Which is correct for 32 bit systems when frysk is compiled on a 32 bit
system and correct for 64 bit systems when compiled on 64 bit systems.
But not for an 32 bit address space on a 64 bit system.

Unfortunately I don't immediately know the correct way to fix this at
this point. So I filed it as
http://sourceware.org/bugzilla/show_bug.cgi?id=5432
If someone knows how to fix this properly you now have a testcase!

Cheers,

Mark