>=2.5.66 compiling errors on Alpha

>=2.5.66 compiling errors on Alpha

[Christian]

Hi,

i don't know if it a known issue but perhaps this helps to find out.
i was able to compile 2.5.65 with gcc version 2.95.4 on an Alpha 
EV45/Avtanti and it is running quite stable.

but compiling a 2.5.66 or 2.5.67 fails with gcc (GCC) 3.2.3 20030309 :

(before patching i always do "make clean". patches applied with no 
errors. same config used.)

-----
lila:/usr/src/linux-2.5.x# export LANG=C
lila:/usr/src/linux-2.5.x# make vmlinux
make -f scripts/Makefile.build obj=scripts
make -f scripts/Makefile.build obj=arch/alpha/kernel 
arch/alpha/kernel/asm-offsets.s
make[1]: `arch/alpha/kernel/asm-offsets.s' is up to date.
   Starting the build. KBUILD_BUILTIN=1 KBUILD_MODULES=
make -f scripts/Makefile.build obj=init
   CHK     include/linux/compile.h
make -f scripts/Makefile.build obj=usr
make -f scripts/Makefile.build obj=arch/alpha/kernel
make -f scripts/Makefile.build obj=arch/alpha/mm
make -f scripts/Makefile.build obj=arch/alpha/math-emu
make -f scripts/Makefile.build obj=kernel
   gcc -Wp,-MD,kernel/.sys.o.d -D__KERNEL__ -Iinclude -Wall 
-Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common 
-pipe -mno-fp-regs -ffixed-8 -msmall-data -mcpu=ev4 -Wa,-mev6 
-fomit-frame-pointer -nostdinc -iwithprefix include 
-DKBUILD_BASENAME=sys -DKBUILD_MODNAME=sys -c -o kernel/sys.o kernel/sys.c
kernel/sys.c:226: conflicting types for `sys_sendmsg'
include/linux/socket.h:245: previous declaration of `sys_sendmsg'
kernel/sys.c:227: conflicting types for `sys_recvmsg'
include/linux/socket.h:246: previous declaration of `sys_recvmsg'
make[1]: *** [kernel/sys.o] Error 1
make: *** [kernel] Error 2
lila:/usr/src/linux-2.5.x#
----

do you need further infos to debug this?
compiling a kernel on this machine is very slow, i think i can't take 
the "BUG-HUNTING" approach.

Thank you,
Christian.

>=2.5.66 compiling errors on Alpha

[Christian]

[...]

the same thing happens when compiling 2.5.67 with gcc 2.95.4.
i did "make mrproper", same .config again and the first "make vmlinux" 
failed again. the output of a second "make vmlinux" is shorter so i can 
post it here again:

------------
lila:/usr/src/linux-2.5.x# make vmlinux
make -f scripts/Makefile.build obj=scripts
   gcc -Wp,-MD,scripts/.modpost.o.d -Wall -Wstrict-prototypes -O2 
-fomit-frame-pointer   -c -o scripts/modpost.o scripts/modpost.c
   gcc  -o scripts/modpost scripts/modpost.o scripts/file2alias.o
make -f scripts/Makefile.build obj=arch/alpha/kernel 
arch/alpha/kernel/asm-offsets.s
make[1]: `arch/alpha/kernel/asm-offsets.s' is up to date.
   Starting the build. KBUILD_BUILTIN=1 KBUILD_MODULES=
make -f scripts/Makefile.build obj=init
   CHK     include/linux/compile.h
make -f scripts/Makefile.build obj=usr
make -f scripts/Makefile.build obj=arch/alpha/kernel
make -f scripts/Makefile.build obj=arch/alpha/mm
make -f scripts/Makefile.build obj=arch/alpha/math-emu
make -f scripts/Makefile.build obj=kernel
   gcc -Wp,-MD,kernel/.sys.o.d -D__KERNEL__ -Iinclude -Wall 
-Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common 
-pipe -mno-fp-regs -ffixed-8 -mcpu=ev4 -Wa,-mev6 -fomit-frame-pointer 
-nostdinc -iwithprefix include    -DKBUILD_BASENAME=sys 
-DKBUILD_MODNAME=sys -c -o kernel/sys.o kernel/sys.c
kernel/sys.c:226: conflicting types for `sys_sendmsg'
include/linux/socket.h:245: previous declaration of `sys_sendmsg'
kernel/sys.c:227: conflicting types for `sys_recvmsg'
include/linux/socket.h:246: previous declaration of `sys_recvmsg'
make[1]: *** [kernel/sys.o] Error 1
make: *** [kernel] Error 2
lila:/usr/src/linux-2.5.x# gcc --version
2.95.4
lila:/usr/src/linux-2.5.x#
------------


Thank you,
Christian.

>=2.5.66 compiling errors on Alpha

[Christian]

[...]

the same thing happens when compiling 2.5.67 with gcc 2.95.4.
i did "make mrproper", same .config again and the first "make vmlinux" 
failed again. the output of a second "make vmlinux" is shorter so i can 
post it here again:

------------
lila:/usr/src/linux-2.5.x# make vmlinux
make -f scripts/Makefile.build obj=scripts
   gcc -Wp,-MD,scripts/.modpost.o.d -Wall -Wstrict-prototypes -O2 
-fomit-frame-pointer   -c -o scripts/modpost.o scripts/modpost.c
   gcc  -o scripts/modpost scripts/modpost.o scripts/file2alias.o
make -f scripts/Makefile.build obj=arch/alpha/kernel 
arch/alpha/kernel/asm-offsets.s
make[1]: `arch/alpha/kernel/asm-offsets.s' is up to date.
   Starting the build. KBUILD_BUILTIN=1 KBUILD_MODULES=
make -f scripts/Makefile.build obj=init
   CHK     include/linux/compile.h
make -f scripts/Makefile.build obj=usr
make -f scripts/Makefile.build obj=arch/alpha/kernel
make -f scripts/Makefile.build obj=arch/alpha/mm
make -f scripts/Makefile.build obj=arch/alpha/math-emu
make -f scripts/Makefile.build obj=kernel
   gcc -Wp,-MD,kernel/.sys.o.d -D__KERNEL__ -Iinclude -Wall 
-Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common 
-pipe -mno-fp-regs -ffixed-8 -mcpu=ev4 -Wa,-mev6 -fomit-frame-pointer 
-nostdinc -iwithprefix include    -DKBUILD_BASENAME=sys 
-DKBUILD_MODNAME=sys -c -o kernel/sys.o kernel/sys.c
kernel/sys.c:226: conflicting types for `sys_sendmsg'
include/linux/socket.h:245: previous declaration of `sys_sendmsg'
kernel/sys.c:227: conflicting types for `sys_recvmsg'
include/linux/socket.h:246: previous declaration of `sys_recvmsg'
make[1]: *** [kernel/sys.o] Error 1
make: *** [kernel] Error 2
lila:/usr/src/linux-2.5.x# gcc --version
2.95.4
lila:/usr/src/linux-2.5.x#
------------


Thank you,
Christian.

Re: >=2.5.66 compiling errors on Alpha

[David S. Miller]

On Sat, 2003-04-12 at 06:15, Christian wrote:
> do you need further infos to debug this?
> compiling a kernel on this machine is very slow, i think i can't take 
> the "BUG-HUNTING" approach.

The Alpha folks need to fix their definition of cond_syscall().
Copying over the asm-i386/*.h definition will probably "just work".

-- 
David S. Miller <davem@redhat.com>

Re: >=2.5.66 compiling errors on Alpha

[Måns Rullgård]

"David S. Miller" <davem@redhat.com> writes:

> > do you need further infos to debug this?
> > compiling a kernel on this machine is very slow, i think i can't take 
> > the "BUG-HUNTING" approach.
> 
> The Alpha folks need to fix their definition of cond_syscall().
> Copying over the asm-i386/*.h definition will probably "just work".

That won't work.  There have been two (equivalent) patches posted to
this list recently.  Will someone pick them up?

-- 
Måns Rullgård
mru@users.sf.net

Re: >=2.5.66 compiling errors on Alpha

[Marc Zyngier]

>>>>> "Måns" == Måns Rullgård <mru@users.sourceforge.net> writes:

Måns> That won't work.  There have been two (equivalent) patches
Måns> posted to this list recently.  Will someone pick them up?

Looks like Ivan Kokshaysky is opposed to this approach (see the
discussion that happended on axp-list last week about the same patch I
posted to both axp-list and lkml).

http://www.redhat.com/mailing-lists/axp-list/msg20606.html

Regards,

        M.
-- 
Places change, faces change. Life is so very strange.

Re: >=2.5.66 compiling errors on Alpha

[Måns Rullgård]

Marc Zyngier <mzyngier@freesurf.fr> writes:

> Måns> That won't work.  There have been two (equivalent) patches
> Måns> posted to this list recently.  Will someone pick them up?
> 
> Looks like Ivan Kokshaysky is opposed to this approach (see the
> discussion that happended on axp-list last week about the same patch I
> posted to both axp-list and lkml).

It's strange that this is bad on Alpha, but ok for all other machines.
Whatever the reason, it would be nice to have a fix in 2.5.68.

-- 
Måns Rullgård
mru@users.sf.net