On 7/20/06, Ken Moffat <[EMAIL PROTECTED]> wrote:

 The problem shouldn't really apply to LFS-6.1.1, so if you haven't
accidentally missed some of the packages or otherwise deviated from
the book I'll go with Dan's guess that it's the dreaded
via-and-gcc-disagree-about-i686 problem (no 'cmov' instruction, I
believe - marked as optional in the technical docs, but gcc assumes
it exists) - you might need to look at the crosscompiling-x86.txt
hint and build for i486 (ISTR Alan Cox saying that 486 code was
faster than 586 on via cpus).  If that is indeed the problem, you
can perhaps see a segfault if you run

/bin/echo "hello world!"

(you need to specify the path because echo is a built-in in bash)

 - if you *don't* see an error from /bin/echo, or any other evidence
   that i686 binaries are failing, then building for i486 is probably
   not going to help.

It's still going to help to see the errors. Might be that this has
nothing to do with cmov. After writing, I came across this extremely
informative document that says that this is not a problem on newer VIA
processors.

http://radagast.bglug.ca/epia/epia_howto/index.html
http://radagast.bglug.ca/epia/epia_howto/x1099.html#pentium_compatible

Also, while looking through the gcc bugs, it seems that gcc knows that
the c3 does not have cmov and disables it. However, I think this only
happens if march/mtune is unset or set to c3. Could certainly be wrong
there, though.

--
Dan
--
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