Re: [gentoo-user] Bug reporting questions

[Leo]

Dave Nebinger wrote:
I'm getting an error with sys-libs/glibc-2.3.4.20050125-r1:
[snipped]
You should not do either of these things.  First you should fix your CFLAGS
to remove the -mcpu flag and replace with the appropriate -march value.
 

OK what do you suggest as an appropriate -mach value? 

Some time ago I started getting this warning.
Secondly, since you're using -mcpu=686 yet gcc appears to be i386 based, I
would guess that your system setup from the ground up is poorly set up.
 

I set up the system from Gentoo ver 2004.2  At the time I chose what the 
documentation recommended as reasonable defaults.  I am not familiar 
with what these flags really do for gcc.  I understand that certain 
optimizations for 686 exist that make incompatible binaries for earlier 
processors ie i586.

My make.conf has
CHOST="i386-pc-linux-gnu"
I don't know if gcc depends on any specific part of this value.  AFAIK 
it's simply a name (handle).  Maybe so you can have several versions of 
gcc around and chose which one you want to use?

If you're still in the early stages of the system build, I'd suggest
starting over with valid CFLAGS; trust me, having a solid foundation ensures
that you'll end up with a stable gentoo box when you're done.
 

Yeah I though that I might have to tweak my CFLAGS...  I was never sure 
if it was optimized or not.

Right now I use
CFLAGS="-O2 -mcpu=i686 -fomit-frame-pointer"
Clearly -mcpu=i686 is depreciated.  What -march setting is good?
How does -fomit-frame-pointer effect the compilations?
I don't put -O[3-6] because I read that gcc doesn't always define such 
optimizations it falls back to -O2.  Also, I don' t know if any of the 
higher optimizations are experimental or not. 

Anyway what do you suggest?
I have a vanilla P4 Willemite...
If the system has been operational for awhile, then I think you should count
yourself lucky, not emerge the glibc update, and keep your misconfigured box
running as it is.
 

I've emerged glibc about 4 times already.  I remember a similar error 
occured a while ago but was fixed by a subsequent release.

TNX for your attention
Leo
--
gentoo-user@gentoo.org mailing list