At 4:58 PM +1030 1/9/04, Jeremy Begg wrote:
>I have an Itanium rx2600 running VMS 8.1, so I thought I'd try and build
>this release of Perl there.  To summarise ...
>
> - I ran @CONFIGURE and answered all the questions.  I said "Y" to the
>   questions on 64bit support, but the configure procedure crashed with
>   and RMS-F-EOF error shortly after displaying the message
>   "Checking whether your NVs can preserve your UVs...".

Interesting test, one I haven't tried.  Obviously there's some work
to do in this area.  Does this still work on Alpha?  I haven't tried
it there recently but it used to work fine.

> - So I re-ran CONFIGURE.COM, this timing declining it's requests for
>   64bit support.  It ran to completion, and invited me to run MMS to
>   build Perl.
>
> - I ran MMS which chugged away for a while and built Perl without
>   any complaints.

Ditto in my work on the testdrive system.

> - I ran MMS TEST which (eventually) crashed the system :-(
>
>After the system rebooted I logged in again and ran MMS TEST again, with
>exactly the same result ... system crashed at the same point in the MMS TEST
>output.

Ditto, again, in my work on the testdrive system.  I should have
mentioned something about this, but I didn't realize folks would
already be chomping at the bit to run Perl on their I64 systems.
This was an OS bug that has been fixed.  I reported the problem and
handed reproducer details off to the system manager of the testdrive
system, who passed it along to engineering, who fixed it within a few
days.

The subject line of one of the replies I got was "A: 75-101-661 PERL
test suite causes system crash" which I assume is the internal bug
number and description.  The crasher was an inconsistent state
brought on by attempting to restore floating point registers at AST
delivery time without  a correct list of which registers needed
restoring.  In other words, just the sort of thing you would expect
them to still be working out at this stage of development on an
architecture that has completely different (and a lot more) registers
than Alpha.

The testdrive system has new images installed and no longer crashes
when I run the Perl test suite.  I don't know whether there will be
any publicly available fix before v8.2.  I now see the following with
RC1 on the testdrive system:

t/io/openpid.........................FAILED at test 11
t/op/magic...........................FAILED at test 3
ext/POSIX/t/posix....................FAILED at test 9
ext/POSIX/t/sigaction................FAILED at test 7
ext/Time/HiRes/t/HiRes...............FAILED at test 17
lib/Memoize/t/expmod_t...............FAILED at test 16
lib/Net/t/hostname...................FAILED at test 3
lib/vmsish...........................FAILED at test 25
Failed 8 test scripts out of 760, 98.95% okay.

Several of these are subprocesses that exit with SS$_BADSTACK, so
there are still some rough edges in the infrastructure, but it's
getting there.
-- 
________________________________________
Craig A. Berry
mailto:[EMAIL PROTECTED]

"... getting out of a sonnet is much more
 difficult than getting in."
                 Brad Leithauser

Reply via email to