On Thu, 16 Sep 1999, Herbert Xu wrote:
herber>John Lapeyre <[EMAIL PROTECTED]> wrote:
herber>>
herber>> Hmm. Well my two potato systems are slightly different. One just
herber>> compiled 2.0.36 with the patch. But the other one failed with the
herber>> message
herber>> fixed or forbidden register 2 (cx) was spilled for class CREG,
herber>
herber>This means that you're not using gcc272.
I used the default compiler, which is "2.95". The patch is supposed
to allow old kernels to be compiled with new compilers. So this is what
I wanted. I also compiled it on a slightly older potato system with
a version "egcs-2.91.66", and this has produced a kernel, which, so far,
seems stable. (btw, both systems have binutils 2.9.1.0.25-2)
There is a note here about the problem with 2.95:
http://egcs.cygnus.com/faq.html/#asmclobber
However, it just gives an example, but does not enumerate all of the
violations in the linux source.
This is what I get when I try to compile the patched 2.0.36 source with
gcc272. This is similar to what I get when I try to compile un-patched
source with gcc272:
homey 38 > make 'CC=gcc272' zImage gcc272 -Wall -Wstrict-prototypes -O2
-fomit-frame-pointer -fno-strength-reduce -pipe -m486 -malign-loops=2
-malign-jumps=2 -malign-functions=2 -DCPU=586 -c -o init/main.o
init/main.c herber>-- init/main.c:23: linux/head.h: No such file or
directoryherber>Debian GNU/Linux 2.1 is out! ( http://www.debian.org/ ) In
file included from /usr/include/linux/sched.h:14,herber>Email: Herbert Xu
~{PmV>HI~} <[EMAIL PROTECTED]>
from init/main.c:20:herber>Home Page:
http://gondor.apana.org.au/~herbert/
/usr/include/linux/timex.h:159: field `time' has incomplete type
The bottom line, from my point of view, is that make 'CC=gcc272'
will not compile any kernel on potato, and that 2.95 will not
compile even a patched (see http://www.suse.de/~florian/kernel+egcs.html)
2.0.36 kernel.
There is still the possibility that my kernel crash is a hardware
failure that is triggered under 2.2.x, but not under 2.0.36. Still,
from a pratical point of view, I (or a user) should not be forced to
upgrade the kernel because I want to build one more module, which is
almost what happened. (I just happened to have a system with egcs-2.91.66
available.)
John Lapeyre <[EMAIL PROTECTED]>
Tucson,AZ http://www.physics.arizona.edu/~lapeyre
