[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gcc 2.95.2 vs 2.91

>>>>> "Brendan" == Brendan Simon <bsimon@ctam.com.au> writes:

Brendan> "Phillip R. Jaenke" wrote:

Phillip> *sigh* this isn't just powerpc. this is x86 too. I can't get ANY kernels
Phillip> to compile with gcc 2.95 or 2.91. Folks, I think at this point it's
Phillip> SERIOUSLY advisable to just abandon gcc 2.9x as primary and change it to
Phillip> optional. There's just too much stuff that will NOT compile with it. I
Phillip> mean, hell, the damn thing is spitting out asm errors for 'mov' on x86!
Phillip> I've had NO problems with or 2.8.x to date.

Brendan> The real question as far as I can see is whether it is a gcc bug or is it gcc
Brendan> just being more thorough in picking up bad (or potentially bad) code.  If it is
Brendan> the latter, then it is well worth modifying the code to satisfy gcc.  There are
Brendan> two good reasons for this.  1) The code should be less prone to errors.  2)
Brendan> There will be further releases of gcc.  If you want to stay with egcs/gcc-2.91
Brendan> for ever then there is no problem but I think most people will want the benefits
Brendan> that future releases of gcc will offer.

I totally agree. Instead of claiming gcc is broken dot, find the bugs
and report them. I've been using gcc-2.95.2 to compile linux kernel 2.2.12 2.2.13
on Intel without any problem. There were two serious bugs in gcc-2.95.1 for PPC. I found
them while working on RTEMS on MCP750 (PowerPC Compaq PCI borad where I also ported linux).
I reported them and they are now fixed in gcc-2.95.2. One was fixed within a day even
if it took me a week to localize the bug. The code generated by gcc-2.95.2 on PPC is much 
more compac and efficient than with egcs 1.1.2...

Its free software guys. I you submit good bug reports, be sure the cygnus folks and
and other GCC gurus will fix them. Alos note that there are *clear* warnings with the asm
macros on the gcc page...

  /  `                   	Eric Valette
 /--   __  o _.          	Canon CRF - Communication Dept
(___, / (_(_(__         	Rue de la touche lambert
				35517 Cesson-Sevigne  Cedex
Tel: +33 (0)2 99 87 68 91	Fax: +33 (0)2 99 84 11 30
E-mail: valette@crf.canon.fr	http://www.crf.canon.fr

Reply to: