Subject: Re: how to kill any O* when compiling
To: Julio M. Merino Vidal <>
From: Charles Swiger <>
List: tech-pkg
Date: 05/04/2005 15:49:24
On May 4, 2005, at 3:36 PM, Julio M. Merino Vidal wrote:
> BTW, I also had problems with perl built with -O2 under Linux/ppc 
> (well,
> in fact it didn't built because miniperl gave problems at runtime, thus
> breaking the build).  Is GCC so buggy under this arch?

Early versions of gcc-3 and GNU binutils had some significant problems 
on 64-bit platforms, for example had a heck of a time trying to get it 
to work sensibly even on the SPARC platform. [1]  However, 
gcc-3.3.{2-4} or later ought to be in pretty good shape, and I've heard 
that Apple has done a lot of work with gcc-4 for the 10.4/Tiger release 
of MacOS X which just came out.

I'm not convinced these issues are entirely GCC's fault, either, since 
there is a lot of code out there which makes assumptions which do not 
apply to 64-bit platforms, and there has been a considerable shakedown 
period getting these cleaned up for the PPC and even the AMD 
Opteron/EM64T targets.  Earlier this week, Kris Kennaway (he's 
<>), mentioned something about the AMD64 ports build 
successfully building over 10,000 of the ~12,600 ports for the first 
time, but even, so about 20% of the software out there still isn't 
64-bit clean. [2]


[1]: The EGCS team started working on a new optimizer/instruction 
scheduler called Haifa, which was starting with the SPARCv8 (v9?) 
architecture and the x86 platform as the primary targets.  The EGCS 
branch merged back into GCC as the start of the gcc-3 releases.

[2]: I have a better handle on the status of FreeBSD ports via and the pointyhat/bento cluster than I have information 
about the pkgsrc builds for NetBSD.  If someone wants to point me 
towards comparable info for pkgsrc, that would be good too.  :-)