Dr Andrew John Hughes wrote:
On 11:00 Wed 19 Jan , Joe Darcy wrote:
Dr Andrew John Hughes wrote:
snip...
[snip]
Thanks Joe. Pavel pointed me to the discussion on the bug on the HotSpot lists.
It was on hotspot-runtime-dev which is why I missed it:
http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2010-November/001477.html
As you can see from the discussion, the test is at fault. It only
works on 32-bit platforms and needs to be more generic to handle all
situations. Pavel's fix in IcedTea works for GNU/Linux platforms, but
AIUI will still run into difficulties on Solaris x86_64 where the
compiler produces 32-bit binaries by default.
Just a quick comment on the test, the original test was clearly faulty
was not running properly on all supported platforms. In particular,
when I have to write a shell test like this, if it doesn't make sense to
the run the test on, say, Windows, I'll have the test check for that
platform (Windows or Cygwin) and vacuously pass in that case.
-Joe
