[parisc-linux] Problems with glibc

[parisc-linux] Problems with glibc

[Andreas Thienemann]

Hi,

while trying to compile the current glibc snapshot from the puffin-cvs, make
segfaults:

make[2]: Entering directory `/mnt/lfs/usr/src/glibc/sunrpc'
/mnt/lfs/usr/src/glibc-build/elf/ld.so.1 --library-path
/mnt/lfs/usr/src/glibc-build:/mnt/lfs/usr/src/glibc-build/math:/mnt/lfs/usr/src/glibc-build/elf:/mnt/lfs/usr/src/glibc-build/dlfcn:/mnt/lfs/usr/src/glibc-build/nss:/mnt/lfs/usr/src/glibc-build/nis:/mnt/lfs/usr/src/glibc-build/rt:/mnt/lfs/usr/src/glibc-build/resolv:/mnt/lfs/usr/src/glibc-build/crypt:/mnt/lfs/usr/src/glibc-build/linuxthreads
/mnt/lfs/usr/src/glibc-build/sunrpc/rpcgen -Y `x=\`gcc
-print-file-name=cpp\`; test x$x = xcpp && x=\`which cpp\`; echo $x | sed
's,/cpp$,,'` -c rpcsvc/bootparam_prot.x -o
/mnt/lfs/usr/src/glibc-build/sunrpc/xbootparam_prot.T
make[2]: *** [/mnt/lfs/usr/src/glibc-build/sunrpc/xbootparam_prot.stmp]
Segmentation fault
make[2]: Leaving directory `/mnt/lfs/usr/src/glibc/sunrpc'
make[1]: *** [sunrpc/others] Error 2
make[1]: Leaving directory `/mnt/lfs/usr/src/glibc'
make: *** [all] Error 2

Has anyone ever experienced a similar problem? This segfault is
reproducable and it's the first time that happened. gcc did compile
correct, the same with the kernel itself.
Therefore I'd guess that glibc is at fault here.
This happened while building the glibc on the hppa box (with the
baseplus-20010404 nfsroot).
Building the cross-compiler toolchain which does include building glibc
was successful though.

Any ideas?

bye,
 andreas

Re: [parisc-linux] Problems with glibc

[Jes Sorensen]

>>>>> "Andreas" == Andreas Thienemann <andreas@thienemann.net> writes:

Andreas> Has anyone ever experienced a similar problem? This segfault
Andreas> is reproducable and it's the first time that happened. gcc
Andreas> did compile correct, the same with the kernel itself.
Andreas> Therefore I'd guess that glibc is at fault here.  This
Andreas> happened while building the glibc on the hppa box (with the
Andreas> baseplus-20010404 nfsroot).  Building the cross-compiler
Andreas> toolchain which does include building glibc was successful
Andreas> though.

You don't see it when cross compiling because this is only run when
building natively. The sunrpc stuff is the first place where the newly
generated dynamic linker is exercised, your cross compiled binaries
are unlike to work either. The can be either a glibc (most likely
ld.so) or due to a change in syscall or compiler ABI behavior.

Jes

Re: [parisc-linux] Problems with glibc

[Matthew Wilcox]

On Fri, Apr 20, 2001 at 07:10:43PM +0200, Andreas Thienemann wrote:
> while trying to compile the current glibc snapshot from the puffin-cvs, make
> segfaults:
> Has anyone ever experienced a similar problem? This segfault is
> reproducable and it's the first time that happened. gcc did compile
> correct, the same with the kernel itself.

Yes, I was getting it too.  After my recent kernel checkin to fix the
stat64 struct, I was able to build glibc successfully.  My theory is that
glibc was trying to call the *stat64 syscalls and the kernel was happily
writing all over user memory.  I would be grateful if you updated your
(running) kernel to head of CVS and tried building glibc again.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Problems with glibc

[Andreas Thienemann]

Hi Matthew,

On Sun, 22 Apr 2001, Matthew Wilcox wrote:

> Yes, I was getting it too.  After my recent kernel checkin to fix the
> stat64 struct, I was able to build glibc successfully.  My theory is that
> glibc was trying to call the *stat64 syscalls and the kernel was happily
> writing all over user memory.

> I would be grateful if you updated your (running) kernel to head of
> CVS and tried building glibc again.
Jepp. That did it. glibc was just build successful.
But you wouldn't believe how long it took on that system. Just 101MHz..

But thanks for the advice.
Now back to my quest on installing linux from  scratch. ;)

bye,
 andreas