Re: 64 bit ABI issue on Solaris

[Stefan Schaeckeler <schaecsn%gmx.net@localhost>]

Hello,
 
> On Wed, Feb 16, 2011 at 08:16:18AM -0800, Stefan Schaeckeler wrote:
> > Hi,
> > 
> > > On Mon, Feb 14, 2011 at 06:40:54AM +0100, Stefan Palm wrote:
> > > > A far as I remember the problem was that pkgsrc reported 
> > > > MACHINE_GNU_PLATFORM as "sparc-sun-solaris2" (instead of 
> > > > "sparc64-sun-solaris2") to gmp. My "workaround" was to use a fake uname 
> > > > ("#!/bin/sh; echo `/usr/bin/uname $@ | /usr/xpg4/bin/sed 
> > > > 's/sparc/sparc64/g'`") during bootstrap to get the 
> > > > MACHINE_ARCH/MACHINE_GNU_PLATFORM fixed.
> > > 
> > > 
> > > Out of interest, what does running the script found at
> > > 
> > >   
> > > http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD
> > > 
> > > print?
> > 
> > sparc-sun-solaris2.9
> 
> I think I should have asked for uname -a too!

SunOS unknown 5.9 Generic_118558-34 sun4u sparc SUNW,Ultra-5_10

Similiar to Stefan's uname wrapper, I added this to bootstrap/bootstrap, and 
now most packages compile:

        if [ $abi"x" = 64"x" ]; then
             machine_arch=uname -p | sed -e 's/sparc/sparc64/g'  
        fi

So far, only gmp and gcc44 still have ABI issues.

 Stefan 


-- 

Scotty: Captain, we din' can reference it!
Kirk:   Analysis, Mr. Spock?
Spock:  Captain, it doesn't appear in the symbol table.
Kirk:   Then it's of external origin?
Spock:  Affirmative.
Kirk:   Mr. Sulu, go to pass two.
Sulu:   Aye aye, sir, going to pass two.