[gentoo-dev] gcc 3.1 inconsistencies :

[gentoo-dev] gcc 3.1 inconsistencies :

[Spider]

[-- Attachment #1: Type: text/plain, Size: 1315 bytes --]

Hello, I did a new gcc3.1 bootstrap when I was forced offline, and
theese are my notes:


bootstrap went nice first time, but I managed to abort it and mess up a
few times, generally my fault to set the wrong flags (1.3 gcc 3.1 image)

Texinfo needs to be fixed and unmasked for the gcc3 profile (it doesn't
break any old system so it should work nicely)

have gcc 2.95.3 hide it , gcc 3.1 depend on it... works nice :)

after a time of recompiling system with new flags I noted some ncurses
stuff started to break... couldn't compile gpm, zsh, cscope and other
things, was odded by this until I checked ncurses (clean new install )
with gcc 3.1's headers... 

xfree has some problems with a source not found for the latest masked
verison


Wonderfully enough, this was solved by recompiling -gcc- ..... 
fixincludes script did something there I suspect.

unmask gnome2 and emerge gnome works. (not a single breakage!! )
gdm works perfect once Xfree was started. 

Generally the whole progress felt really smooth. I'm still recovering
and hope my connection will last through this.

soo, right now I'm running march=athlon-tbird -O3 -pipe for most things
:)


//Spider

--
begin  .signature
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] gcc 3.1 inconsistencies :

[Dave Lee]

Spider wrote:
> soo, right now I'm running march=athlon-tbird -O3 -pipe for most things

In case anybody is curious like I was:

I didn't know the difference between the various athlon arch settings,
athlon-tbird, athlon-xp, etc.  I found that there are two athlon 
variations, those with the sse instruction set and those w/o sse.

no sse: athlon, athlon-tbird
sse: athlon-4, athlon-xp, athlon-mp

`grep sse /proc/cpuinfo` to see if your cpu has sse, though I don't know
how reliable cpuinfo is.  I have a duron 1GHz, and it has sse, but earlier
durons, <1GHz, don't have sse.

also, when you do either of -msse -m3dnow gcc automagically does -mmmmx
for you.

from what I see, athlon and athlon-tbird are identical arch settings, as
are athlon-4, athlon-xp, and athlon-mp.

so can anyone verify this?  of course its possible the different names are
there for finer grained code gen in the future.

Dave

Re: [gentoo-dev] gcc 3.1 inconsistencies :

[fthieme]

> after a time of recompiling system with new flags I noted some ncurses
> stuff started to break... couldn't compile gpm, zsh, cscope and other
> things, was odded by this until I checked ncurses (clean new install )
> with gcc 3.1's headers... 

This is because of the headers generated by fixincludes in
/usr/lib/gcc-lib/i686-linux/3.1/ I removed ncurses.h and all the other files starting with the
same fixincludes message. Now all the ncurses stuff works fine.

> xfree has some problems with a source not found for the latest masked
> verison

No problem for me, too. But there is at least one bug on bugzilla according
to the fixincludes headers.

Bye...Frank