[rescue] Re: Being jobless

Jonathan C. Patschke jp at celestrion.net
Mon Jul 28 12:30:31 CDT 2003

On Mon, 28 Jul 2003, Peter Corlett wrote:

> > I agree whole-heartedly, but I'm really starting to take offense at the
> > recent "the world is an x86 running Linux" point of view. If you submit a
> > bug report about compiling it with a vendor's compiler, the first response
> > is "Why don't you use GCC, instead?". Well, gee, if I wanted a database
> > that ran like ass, I'd use MySQL, thanks.
> So... why didn't you use GCC?

  1) GCC generates pathetically pessimized code on RISC systems.
  2) GCC object code will not properly exchange data with the rest of my
     MIPSpro-generated object code on an IRIX system because a member of
     the GCC team got register-pddking backwards.  They admit this is a
     bug (and have since the 2.8 days), but refuse to fix it.
  3) MIPSpro does a much better job of producing 64-bit optimized
     objects and optimizing for a particular instruction set (MIPS-IV in
     this case) than GCC does.

In short: because I like my high-performing computers to continue to
exhibit high-performance.  If I'm going to drop $$$ on a vendor's
compiler that has actually passed an ANSI C certification suite, I'm
going to want to use that rather than GCC, which has only occasionally
been metioned in the same sentence as "ANSI C" and usually has a
negative in-between.

I chuckle at every README I've seen that says "REQUIRES a strictly ANSI
C-compliant compiler (I recommend GCC)".  It's like saying "REQUIRES an
extremely stable operating environment (I recommend Windows 98)".  More
often than not, code bearing such a notice won't even -compile- with
something like SunONE or MIPSpro due to all sorts of GCCisms.

Jonathan Patschke   )  "We're Texans.  We figure out ways to do these
Elgin, TX          (    things..."                    --Bill Bradford

More information about the rescue mailing list