> On Sat, Mar 29, 2008 at 09:58:51PM -0400, Dennis Clarke wrote:
>>
>> So with all that being said in 2008, here I go again with this hardware :
>>
>> [EMAIL PROTECTED]:~$ cat /proc/cpuinfo
>> processor       : 0
>> cpu             : G2_LE
>> clock           : 396.000000MHz
>
>  Ooh, I don't think I could afford the  necessary amount of coffee
> to build on that!  If patience is a virtue, you will be exceedingly
> virtuous.

Have a search in this msg for "yes really" :
http://gcc.gnu.org/ml/gcc-testresults/2007-08/msg00318.html

So, been there, done that, get me some wings and a harp :-)

>> revision        : 1.4 (pvr 8082 2014)
>> bogomips        : 65.53
>> timebase        : 33000000
>> platform        : Efika
>> machine         : EFIKA5K2 CHRP PowerPC System
>> revision        : 2B3
>> vendor          : bplan
>>
>
>  If you haven't already done so, please check if the debian kernel,

My understanding is that everything I need is in linux-2.6.24.3 and that was
why I stopped building a whole new tool chain on my own and then thought
"geez, let's just do this all from scratch" which brought me back here to
LFS.

> and their version of yaboot, need any non-mainline patches to boot on
> Efika.

I'm not using yaboot nor grub. I did a port of GRUB2 to ppc a while back for
the PowerPC project at OpenSolaris and that worked pretty well. The presence
of good firmware means that I don't need a boot loader like yaboot. I just
place the boot command in the firmware "boot" environment variable and then
the machine boots fine.

>  I can remember various threads on the powerpc-dev list which
> suggested certain things needed to be done differently on it -
> maybe it's all now been merged upstream.

yep.  That is the rumour but I have yet to see proof.

>> http://gcc.gnu.org/ml/gcc-testresults/2008-03/msg01254.html
>>
>  The number of failures isn't extreme, but it is slightly high.

No surprise. My builds of autoconf failed a few test later because of faulty
fortran. I was in touch with the autoconf people and we are looking into a
few things.

> Your report doesn't show which tests failed.  Could be all manner of
> things, including old tools or even missing locales.

Well, I was doing a two pass process :

[ pass zero ]
doesn't count for much. Install Debian and see what
I get from that. It has a pile of basic things in the
build_essentials debian package. Like gcc 4.1.2 20061115
and a starting point.

[ pass one ]
build my own toolchain with whatever I have on hand.

[EMAIL PROTECTED]:~/build$ ls
first_pass  second_pass

[EMAIL PROTECTED]:~/build$ ls first_pass/
autoconf-2.61    flex-2.5.35              mpfr-2.3.1
automake-1.10.1  gcc                      mpfr-2.3.1-enable_shared
binutils         gmp-4.2.2                readline-5.2
bison-2.3        gmp-4.2.2-enable_shared  sed-4.1.5
coreutils-6.10   gzip-1.3.12              smake-1.2
dejagnu          libiconv-1.12            star-1.5
env.sh           m4-1.4.10                tar-1.19
expect-5.43      make-3.81                tcl8.4.16

all of that stuff builds and most of it passes all of their testsuites
neatly. I have never seen GCC pass its own testsuite so therefore I am not
surprised when I see failures in its tests.  The best I can hope for is that
C and C++ compiles as expected.

[pass two]
use the tools from pass one to compile everything again and install them
into /usr/local this time.  Again I simply hope for the best when it comes
testsuites.

After all of that is done ... I think its time to try to build a new Linux
kernel.

> I haven't used 4.2.3 so far, and my last ppc build was using
> recent clfs (with e.g. upstream patches in binutils).  I only
> enable c,c++.  For gcc-4.2.2 I had the following failures:
>
> FAIL: gcc.dg/tree-ssa/loop-19.c scan-tree-dump-times MEM.(base: &|symbol:
> )a, 2
> FAIL: gcc.dg/tree-ssa/loop-19.c scan-tree-dump-times MEM.(base: &|symbol:
> )c, 2
> FAIL: g++.dg/other/unused1.C scan-assembler
> (string|ascii?)z?\t"class2("|\\\\000)

Well, like I was saying. I have never seen GCC pass its own testsuites and
so I'm not surprised when it doesn't. In fact, if it ever does pass all of
its testsuite for C and C++ then I think I'll probably faint.

>  Enjoy the build.

It will be something that I do on that one little embedded thingy and I
expect it will happen over time. Lots of time.

>  The jh branch covers ppc (except for bootloaders)
> if you don't want to use clfs.

What do you mean by "jh" ?

Dennis

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page

Reply via email to