RE: [Lksctp-developers] Looking for non-x86 platform tester(s).

[Fan, A. <ard...@in...>]

Hi Jon

I'm just wondering whether the existed kerneltest can cover any small =
corners of rfc2960. And some test is only a test tool like sctp_darn. =
Users must base the test result upon their own SCTP knoledge and =
judgement. No PASS or FAIL result is just there. Do you know whether the =
other implementation, such as from Randy Stewart, has any such kind of =
conformance test suites?
As for the Arm/Xscale, since it is an embedded platform, ususally with a =
small kernel and simple OS. We need a build machine with a cross compile =
tool chain to build kernel and library, then download the binary to the =
target platform. Your understanding of it is right. Just hope my =
description makes more sence. :)
But you may misunderstand the issue I want to bring forward. Usually =
user space library, application use autoconf as the building tool. =
Autoconf will check a lot of system environment before creating the =
Makefile or some header file. For example: AC_CHECK_SIZEOF(int) or check =
endian, the check result is based on running some built in test program. =
So cross compiled test program cannot run on the build machine, then =
check result cannot be reached. I have tried the lksctp's sctp-tools for =
cross compile, no such problems found up to now.
And the endian issue may occur in some data structure declaration with =
member's size smaller than one byte. In this case, big/little endian =
should have different data structure definition. I just thought of =
something here FYI. :)

Regards/Ardelle


-----Original Message-----
From: Jon Grimm [mailto:jg...@us...]
Sent: 2003?3?27? 23:36
To: Fan, Ardelle
Cc: lks...@li...
Subject: Re: [Lksctp-developers] Looking for non-x86 platform tester(s).


"Fan, Ardelle" wrote:
>=20
> Hi Jon
>=20
> From your request, I tought of conformance test suite.
> I mean if we want to validate whether lksctp is fully enabled on any =
platform, including x86, a conformance test suite for rfc2960 seems to =
be needed.
> Is there any existed one or similar one?
> What should be used now? The frame test and the kernel test?
>=20

The kerneltest are pretty authoritative.  The frametest let us get at
places where its difficult to otherwise.   However, the frametest is
very platoform dependant, I'm guessing there would be a bit of work to
get the frametest running on anything non-x86.  The kernel tests
theoretically work if we've done our job well.. ;-)  The kernel tests
are probably just fine as a conformance suite, but am always happy to
have new tests added!


> As for the non-x86 platform, I could do it on Arm/Xscale. But since =
the Arm/Xscale is usually an embedded platform, cross compiling is more =
offten used than native build on the target platform. So configuration =
may meet many issue, for example you cannot try to run a test program =
cross compiled... But cross compiling kernel is more simple.

ARM/XSCALE would be great.  I don't think I understand what you mean
here though.  Cross-compilation needs to occur for any application
wanting to run on the target platform, right?   Ah, maybe I understand
--- out testsuite 'driver' is our Makefile. Execution won't occur on the
build machine.  The answer would be to break out execution into its own
kerntest.sh script.  The crosscompiled testsuite and 'driver' script
then get put to the target for execution.  Am I understanding the issue
correctly?

> And Arm can support both little endian and big endian. Does lksctp =
addresses endian strictly? especially the user library and header file =
definition.
>=20

We've tried hard to do endianess right.  It helps that x86 isn't network
order, so it forces us to get things right, just to communicate with
anyone else.  Its a pretty safe bet that you'll find something in
compilation and/or in execution even.   I've got a to-do from my boss
this year to validate our execution on PPC (32 or 64), but want to see
it across as many platforms as possible to weed out various issues.

Thanks,
Jon=20


> Regards/Ardelle
>=20
> -----Original Message-----
> From: Jon Grimm [mailto:jg...@us...]
> Sent: 2003?3?25? 22:58
> To: lks...@li...
> Subject: [Lksctp-developers] Looking for non-x86 platform tester(s).
>=20
> Hi there,
>         Thought I'd put out a request to see if there are any =
ambitious
> individuals willing to test/port lksctp on some non-x86 platform(s).
> You'll definately want to be comfortable with building linux-2.5 =
kernels
> on that platform before such an undertaking and otherwise be willing =
to
> get your hands dirty.
>=20
>         If interested either post back or send a note if you have =
questions.
>=20
> Best Regards,
> Jon Grimm
>=20
> -------------------------------------------------------
> This SF.net email is sponsored by:
> The Definitive IT and Networking Event. Be There!
> NetWorld+Interop Las Vegas 2003 -- Register today!
> http://ads.sourceforge.net/cgi-bin/redirect.pl?keyn0001en
> _______________________________________________
> Lksctp-developers mailing list
> Lks...@li...
> https://lists.sourceforge.net/lists/listinfo/lksctp-developers