RE: architecture issues

[Nicola Pero]

Hi Mark

thanks for that - I had a quick look at the thread :-)

it seems to be a glitch in SOGO, which should just use the GNUSTEP_HOST_CPU 
variable rather than have separate code trying to determine the same variable 
using 'uname'. ;-)


Inside a GNUmakefile, that variable is always set - in both gnustep-make v1
and v2.

Outside a GNUmakefile (eg, in a configure script), it is set in gnustep-make 
v1, 
but not in gnustep-make v2.  In gnustep-make v2 you can still get the value 
of that variable by executing

  gnustep-config --variable=GNUSTEP_HOST_CPU

the gnustep-config script should always be in your PATH if gnustep-make v2 
was installed correctly.

So I'd recommend you change the SOGO code to use that variable. ;-)

Thanks

-----Original Message-----
From: Mark Pavlichuk <pav5088@internode.on.net>
Sent: Thursday, 29 May, 2008 17:25
To: nicola.pero@meta-innovation.com
Subject: architecture issues

I'm forwarding the end message from a thread on the SOGo (Scalable
OpenGroupware) mailing list ( 
http://mail.opengroupware.org/mailman/listinfo/sogo ).
-----------------
 From Hubert Chathi (Debian gnustep maintainer) :
>> Wolfgang Sourdeau (SOGo maintainer) wrote:
>>   
>>> This would probably work. But the fact that the gnustep-make
>>> package > doesn't default to x86_64-pc-linux-gnu on amd64/x86-64
>>> machines is the > origin of the bug. Personally I had to modify my
>>> gnustep system files > to get the correct architecture, which is
>>> unacceptable.

IMHO, if you calculate a value in two different ways, you can't expect 
them to come out the same.  But I don't know anything about SOGo, so I 
don't know if there's a good reason for doing that.

Anyways, there isn't really much I can do about it with the gnustep-make 
package. It is hard-coded at build time by upstream.  For a pre-build 
distribution like Debian, we have to pick some value, so the only 
reasonable thing for us to do is to set it to the base value for that 
architecture. Having it default to x86_64-pc-linux-gnu on an i486 
wouldn't be any better.

If you have any other suggestions for how to resolve this, please let me 
know. Although this may be something that may be best resolved by 
Nicola, as the upstream maintainer.
-----------------

-- 
Mark Pavlichuk
Strategic IT
ph. (07)47242890
m. 0409 124577