[ILUG] Optimizing for Pentium Pt.2

[Liam Bedford]

On Mon, 29 Jul 2002 05:02:23 +0100
Paul Kelly claiming to think:

> Matthew French wrote:
> >>Happiness was restored to the world with gcc 2.95 and later.
> > I do not have the time to follow the compiler "wars", but I notice that I
> > must use egcs to build 64 bit SPARC code.
> 
> 2.95 was where the reintegration effort started and work on the other 
> projects fell away. 3.0 was the target for completion of that work.
> 
> > GCC 3 should do it, but because it is so "buggy"[1] it is not worth trying
> > to use unless you really want to track down those compiler errors... :(
> 
> A lot of those problems seem to have been shaken out by Redhat's ballsy 
> gcc 2.96 stunt. I've been trying out RedHat Limbo for a few weeks now, 
> equipped with gcc 3.1. I haven't fallen foul of compiler issue so far, 
> that I'm aware of anyhow. That said I'm glad I don't do much C++ - seems 
> pretty much every version of gcc (2.95, 2.96, 3.0, 3.1, and the 
> forthcoming 3.2) breaks C++ binary compatibility in some way or other.
> 
diary.codemonkey.org.uk [Dave Jones]:
 Got a mail from Neil telling me about some whizzy new warnings in the cvs 
version of gcc. Pulled it and spent a while playing with gcc, and after a 
few false starts got it to spit out a bunch of warnings when compiling the 
kernel. There's a boatload of stuff that still needs cleaning up for 2.5 
(Things like __FUNCTION__ abuse and the like). 2.4 doesn't even compile due 
to broken asm constraints and other such sillies. As 3.x isn't a recommended 
compiler for 2.4, this isn't too important just yet. As more and more archs 
start using 3.x for their standard compiler though, we should clean up some 
of this stuff for 2.5. The new warning Neil tipped me off about 
(-Wunused-macros) turns out far too much crap. From what I looked at, 
it was correct in most cases, and yes there were macros being defined 
but being unused. Cleaning them up however creates a maintenance nightmare 
for anyone with patches in those areas already. (This sort of cleanup 
particularly hurts people like myself who have huge patches touching large 
parts of the tree). The recent gcc->c99 struct initialisers patch for instance 
made large parts of my tree reject. With more to come, the fun isn't over yet either.

that is the CVS version, but it seems that more breakage is to come with the
kernel :(

(Everyone will want to compile their machine on the latest greatest... I think
they eventually persuaded Compaq's compiler to compile the kernel)...

L.
-- 
     dBP   dBBBBb | If you're looking at me to be an accountant
              dBP | Then you will look but you will never see
   dBP    dBBBK'  | If you're looking at me to start having babies
  dBP    dB' db   | Then you can wish because I'm not here to fool around
 dBBBBP dBBBBP'   | Belle & Sebastian (Family Tree)