Glibc 2.3.2 failures and possible fixes?

John, Tausq,

Succeeded in building glibc 2.3.2 with syscall cancellation support, but I'm pretty sure it's wrong.

The current list of failures includes:

iconvdata/bug-iconv3.out

- This is a new one.

math/test-fenv.out
math/test-float.out
math/test-double.out
math/test-ifloat.out

- We've had the math test failures for a while, would you like it if I tried building glibc with gcc head? It's been so long I've forgotten if this was really the delayed fp stuff that I fogot to implement :} Will get around to testing this more near the end of April.

stdlib/tst-strtod.out
stdlib/bug-strtod.out

- These are not so new, but indicate problems.

libio/tst-mmap-eofsync.out
libio/tst-mmap-fflushsync.out

- New kernel with Tausq's mmap flushing changes should make these go away... have to go into the office to install a new kernel.

dlfcn/default.out
dlfcn/tststatic.out

- New, possibly related to syscalls.

posix/annexc.out

- Expected.

linuxthreads/ex8.out
linuxthreads/ecmutex.out
linuxthreads/ex14.out
linuxthreads/ex15.out
linuxthreads/bug-sleep.out
linuxthreads/tst-cancel1.out
linuxthreads/tst-cancel2.out
linuxthreads/tst-cancel3.out
linuxthreads/tst-cancel4.out
linuxthreads/tst-cancel5.out
linuxthreads/tst-cancel6.out
linuxthreads/tst-popen.out
linuxthreads/tst-popen2.out
linuxthreads/tst-cancel-static.out
linuxthreads/tst-cancel-wrappers.out
rt/tst-shm.out
rt/tst-clock.out
rt/tst-aio.out
rt/tst-aio64.out
rt/tst-aio2.out
rt/tst-aio3.out
rt/tst-aio4.out
rt/tst-aio5.out
rt/tst-aio6.out
rt/tst-aio7.out

- I broke a lot of stuff... :}

elf/vismain.out
elf/neededtest.out
elf/neededtest2.out
elf/neededtest3.out
elf/neededtest4.out
elf/circleload1.out

- These have been around for a while, any idea about how to fix them?

Cheers,
Carlos.

I think this is the mismatch in sizeof(long double) problem again.

They've certainly been around since I was hacking on glibc ...

"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

I noticed that a couple of libstdc++ tests that used to fail because of the long double mismatch are fixed with 2.3.1-14 :)

Are you testing with 3.3? Some elf tests in the binutils suite were failing because of the need to canonicalize function pointers. There were also some tests that failed because we absolutely need '-fPIC' for tests with shared libraries. I'd be interested to know what's causing these.

Dave

J. David Anglin                                  [email protected]
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Really? I didn't even put in patches for -14? :) This is all strictly from upstream glibc.

I can rebuild it using 3.3 and see if this goes away. Right now I'm building with 3.2.3. I sure hope getting rid of these is as easy as building with 3.3!

On a second note, I've completely revamped the sysdep-cancel support since I must have been smoking crack when I wrote it... so I'm going to compile another round that should come out much better (hopefully removing all the linuxthread failures).

Thanks for the feedback!

c.

Hi Carlos,

I am less luky.

I grab src of glibc-2.3.2 from ftp.gnu and aply your patches and try to rebuild on an unstable debian (update this 20030324).

unfortunately I got the following err:

gcc-3.2 dl-reloc.c -c -std=gnu99 -O2 -Wall -Winline -Wstrict-prototypes -Wwrite-strings
-fstrict-aliasing -g -pipe -I../include -I.
-I/Develop/tools/glibc/glibc-2.3.2/hppa-linux/obj/elf -I.. -I../libio -I/Develop/tools/glibc/glibc-2.3.2/hppa-linux/obj -I../sysdeps/hpp
a/elf -I../linuxthreads/sysdeps/unix/sysv/linux/hppa -I../linuxthreads/sysdeps/unix/sysv/linux -I../linuxthreads/sysdeps/pthread -I../sys
deps/pthread -I../linuxthreads/sysdeps/unix/sysv -I../linuxthreads/sysdeps/unix -I../linuxthreads/sysdeps/hppa -I../sysdeps/unix/sysv/lin
ux/hppa -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps
/unix/sysv -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/hppa/hppa1.1
-I../sysdeps/wordsize-32 -I../sysdeps/ieee754/flt-32 -I../sysde
ps/ieee754/dbl-64 -I../sysdeps/ieee754/ldbl-128 -I../sysdeps/hppa/fpu -I../sysdeps/hppa -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I.
./sysdeps/generic -nostdinc -isystem /usr/lib/gcc-lib/hppa-linux/3.2.3/include
-isystem /usr/src/linux-2.4.20-pa32-64/include -D_LIBC_RE
ENTRANT -include ../include/libc-symbols.h -o /Develop/tools/glibc/glibc-2.3.2/hppa-linux/obj/elf/dl-reloc.o
In file included from dynamic-link.h:49,

                 from dl-reloc.c:207:

../sysdeps/hppa/dl-machine.h: In function `_dl_relocate_object':
../sysdeps/hppa/dl-machine.h:646: conflicting types for elf_machine_rela_relative' dynamic-link.h:37: previous declaration of elf_machine_rela_relative' make[3]: *** [/Develop/tools/glibc/glibc-2.3.2/hppa-linux/obj/elf/dl-reloc.o] Error 1
make[3]: Leaving directory `/Develop/tools/glibc/glibc-2.3.2/glibc-2.3.2/elf' make[2]: *** [elf/subdir_lib] Error 2
make[2]: Leaving directory `/Develop/tools/glibc/glibc-2.3.2/glibc-2.3.2' make[1]: *** [install] Error 2
make[1]: Leaving directory `/Develop/tools/glibc/glibc-2.3.2/hppa-linux/obj'
make: *** [/Develop/tools/glibc/glibc-2.3.2/hppa-linux/installed-binaries] Error 2

I also find:

into dynamic-link.h
auto void __attribute__((always_inline)) elf_machine_rela_relative (ElfW(Addr) l_addr, const ElfW(Rela) *reloc,

                           ElfW(Addr) *const reloc_addr);

into sysdeps/hppa/dl-machine.h:
auto void __attribute__((always_inline)) elf_machine_rela_relative (struct link_map *map, Elf32_Addr l_addr,
const Elf32_Rela *reloc, Elf32_Addr *const reloc_addr)

Thanks for advise,
Joel

---------------------------------
Vous surfez avec une ligne classique ? Economisez jusqu'à 25% avec Tiscali Complete ! Offre spéciale : première année d'abonnement offerte. ... Plus d'info sur http://complete.tiscali.be

Missing patch. My bad.
I'll update in a few hours with a new sysdep-cancel patch.

Thanks for testing Joel!

c.

Don't worry.

I will waiting for :)

You know how much am i interested. I know my knowledge is limited but I try to do my best.

Always wellcome and thanks for your attention,
Joel

---------------------------------
Vous surfez avec une ligne classique ? Economisez jusqu'à 25% avec Tiscali Complete ! Offre spéciale : première année d'abonnement offerte. ... Plus d'info sur http://complete.tiscali.be

Patches updated:
http://www.baldric.uwo.ca/~carlos/glibc-2.3.2-patches.tar.gz

Still no luck on getting sysdep-cancel to work. I'll need another pair of eyes to look at the assembly in sysdep-cancel.h

c.