Re: unaligned access: a request

[Angus Leeming <a.leeming@ic.ac.uk>]

Please disregard this test as it stands. I've discovered
that the program main.cpp

	int main()
	{
		return 1;
	}

will produce the unaligned traps under LinuxAlpha! Please
excuse my previous message. I'm not used to Linux, usually
using Digital Unix. (There are other problems associated
with xtl with Digital Unix, but I'll come to them in due
course).

If however, you edit alltests.h and uncomment all the
commented out data fields, then the three programs raw, xdr
and giop will produce no new unaligned traps. All well and
good. The last two programs, xdr and giop, will FAIL when
run on a LinuxAlpha system however.

aleem@acinus:angus$ xdr
Testing xdr: Failed
{1 ÿ  ý 4d -5 6d -9 10ld fffffffffffffff5 1cbe991a14 13.140000 15.160000}
{1 ÿ  ý 4d -5 6d 4294967287 10ld 70e9ff1f 70e9ff1f 13.140000 0.000000}
aleem@acinus:angus$ giop
Testing giop: Failed
{1 ÿ  ý 4d -5 6d -9 10ld fffffffffffffff5 1cbe991a14 13.140000 15.160000}
{1 ÿ  ý 4d -5 6d 47244640247 10ld fffffffffffffff5 1cbe991a14 13.140000 15.160000}                                                                             

Angus

> José, (or indeed anyone else reading this list)
> 
> I have a request to make.
> 
> Attached is a stripped down version of the code in
> xtl-1.3/examples/tests
> 
> You'll see that the only test I do is on class numbers and
> the only data member of numbers is a bool.
> 
> I've compiled this code on a LinuxAlpha machine without
> problems.
> All three tests, raw, xdr and giop pass.
> All three tests produce the same error messages in
> /var/log/kern.log:
> 
> Apr  4 15:52:27 acinus kernel: raw(3909): unaligned trap at 000002000000e954: 000002000027632c 29 1
> Apr  4 15:52:27 acinus kernel: raw(3909): unaligned trap at 000002000000e95c: 000002000027632c 2d 1
> Apr  4 15:52:27 acinus kernel: raw(3909): unaligned trap at 000002000000e954: 0000020000278eac 29 1
> Apr  4 15:52:27 acinus kernel: raw(3909): unaligned trap at 000002000000e95c: 0000020000278eac 2d 1
> Apr  4 15:52:43 acinus kernel: xdr(3912): unaligned trap at 000002000000e954: 000002000027632c 29 1
> Apr  4 15:52:43 acinus kernel: xdr(3912): unaligned trap at 000002000000e95c: 000002000027632c 2d 1
> Apr  4 15:52:43 acinus kernel: xdr(3912): unaligned trap at 000002000000e954: 0000020000278eac 29 1
> Apr  4 15:52:43 acinus kernel: xdr(3912): unaligned trap at 000002000000e95c: 0000020000278eac 2d 1
> Apr  4 15:52:49 acinus kernel: giop(3913): unaligned trap at 000002000000e954: 000002000027632c 29 1
> Apr  4 15:52:49 acinus kernel: giop(3913): unaligned trap at 000002000000e95c: 000002000027632c 2d 1
> Apr  4 15:52:49 acinus kernel: giop(3913): unaligned trap at 000002000000e954: 0000020000278eac 29 1
> Apr  4 15:52:49 acinus kernel: giop(3913): unaligned trap at 000002000000e95c: 0000020000278eac 2d 1 
> 
> You'll note that the addresses are identical in all three
> cases.
> 
> My request is this:
> Could you run the test on an x86 machine and then look to
> see if there are any error messages in /var/log/kern.log.
> I'm interested to see whether this is an Alpha-specific
> problem.
> 
> Regards,
> Angus
>