This is the mail archive of the
mailing list for the glibc project.
Build failure of 2.2.3 and 2.2.4pre2 on sparc-unknown-linux
- To: libc-alpha at sources dot redhat dot com
- Subject: Build failure of 2.2.3 and 2.2.4pre2 on sparc-unknown-linux
- From: Nix <nix at esperi dot demon dot co dot uk>
- Date: 05 Aug 2001 13:04:29 +0100
[This is a less intemperate version of a massive rant posted to
glibc-linux yesterday, but that list seems to be down. Good; I
said some rather, er, unwise things in there ;) ]
I'm trying to bootstrap a sparc linux box (UltraSparc II CPU, Sun Netra
T1 105), and glibc is giving me grief:
/mnt/tmp/build-glibc/elf/ld-linux.so.2 --library-path /mnt/tmp/build-glibc:/mnt/tmp/build-glibc/math:/mnt/tmp/build-glibc/elf:/mnt/tmp/build-glibc/dlfcn:/mnt/tmp/build-glibc/nss:/mnt/tmp/build-glibc/nis:/mnt/tmp/build-glibc/rt:/mnt/tmp/build-glibc/resolv:/mnt/tmp/build-glibc/crypt:/mnt/tmp/build-glibc/linuxthreads /mnt/tmp/build-glibc/sunrpc/rpcgen -Y gcc -E -c rpcsvc/bootparam_prot.x -o /mnt/tmp/build-glibc/sunrpc/xbootparam_prot.T
make: *** [/mnt/tmp/build-glibc/sunrpc/xbootparam_prot.stmp] Segmentation fault
make: Leaving directory `/mnt/tmp/glibc-2.2.4/sunrpc'
This is with a configure line of
sparc32 ../glibc-2.2.4/configure --prefix=/usr --with-fp --with-elf --enable-shared --enable-omitfp --enable-profile --disable-bounded --enable-add-ons --enable-kernel=2.2.19
I've tried 2.2.4-from-CVS as of just past midnight GMT 2001-08-05 (same
as 2.2.4pre2 as far as I can tell); I've tried GCC-2.95.3 and
GCC-2.95.4-cvs-head (to get the atexit patch), and I'm using
binutils-220.127.116.11.24; I've tried with `-mcpu=v8 -mtune=ultrasparc' and
None work; all give this error. I've applied the patch mentioned last
year for 2.96 involving adding `-traditional' to the definition of
`preprocess-versions' (even though it should have little effect in
2.95.x, and lo, it didn't); I've checked that the patch mentioned last
month (?) to handle the new relocation types generated by new binutils
is in place, and it is.
What else can it be?
(It's been annoying that people persist in saying on the list `oh yes,
you need the atexit patches installed', and then either don't say what
the damned patches *are* or (yesterday) give a URL link to the middle of
a massive flamewar on the GCC list that I know for certain had *dozens*
of candidate patches proposed on it for both atexit and
What on earth compiler, binutils, glibc, weird patches and whatever else
do I need to get glibc bootstrapped? Other people seem to be managing OK
so obviously I am missing something extremely obvious (although nothing
is extremely obvious when one is bootstrapping and has little but a
`It's all about bossing computers around. Users have to say "please".
Programmers get to say "do what I want NOW or the hard disk gets it".'
-- Richard Heathfield on the nature of programming