PowerPC snapshot compilation failures
Glauber de Oliveira Costa
glommer at br.ibm.com
Sun Sep 25 21:03:35 UTC 2005
> > > > - vfork.S fails to compile. If the line #include <error.h> is wipped out
> > > > from libc/sysdeps/linux/powerpc/bits/syscalls.h, everything seems to
> > > > flow okay. As the intel file does not include this file, I'm assuming
> > > > that it should not be there anyway.
>
> what is the failure?
>
> __set_errno is used and INTERNAL_SYSCALL_ERRNO, so maybe errno.h should
> be included instead of error.h
>
It is:
../../../../include/errno.h: Assembler messages:
../../../../include/errno.h:32: Error: Unrecognized opcode:
`__begin_decls'
../../../../include/errno.h:61: Error: Unrecognized opcode:
`__end_decls'
And it happens, as you might know, because the decls macros are not
expanded while compiling the .S file. So, if I run the preprocessor
alone into it, the definitions are there, and are understood as labels.
> > > > ld -shared --warn-common --warn-once -z combreloc -z defs -s -soname=libdl.so.0 \
> > > > -o libdl-0.9.28.so -fini dl_cleanup --whole-archive libdl_pic.a \
> > > > --no-whole-archive ../..//libc/misc/internals/interp.o \
> > > > -L../..//lib -lc ../../lib/ld-uClibc.so.0 /usr/lib/gcc-lib/powerpc-suse-linux/3.3.3/libgcc.a;
> > > > ld: Error: The symbol `__syscall_error' has a R_PPC_REL24 relocation, that means '../..//lib/libc.a(brk.o)' was
> > > > compiled without -fPIC.
>
> This is normally not possible, all the uClibc is compiled w/ -fPIC, unless
> you have changed something in Rules.mak/Makefiles (there is though a
> slightly difference, powerpc uses -fpic, could you make a test, if
> changing in Rules.mak -fpic to -fPIC causes the same, I expect it will).
> Is this a cross-compiler? It does not look like buildroot so better
> describe your build environment/toolchain.
>
Agree. That's what makes it a little strange. I changed nothing in the
Rules, and very early tried changing -fpic to -fPIC (with the same
results).
I'm building it in a IBM OpenPower with a POWER5 processor, with gcc
3.3.3, in a SLES9 box using kernel 2.6.5 (distro's kernel, not my own).
> > >
> > > try a newer snapshot, it should fail at least less times
>
> on the 20. there were changes related to a similar case, I wanted to be
> sure you have catched these too.
BTW, before answering, I downloaded Sep 25 snapshot to be sure the
errors were still there.
> Peter
> --
> Peter S. Mazinger <ps dot m at gmx dot net> ID: 0xA5F059F2
> Key fingerprint = 92A4 31E1 56BC 3D5A 2D08 BB6E C389 975E A5F0 59F2
>
--
=====================================
Glauber de Oliveira Costa
IBM Linux Technology Center - Brazil
glommer at br.ibm.com
=====================================
----- End forwarded message -----
--
=====================================
Glauber de Oliveira Costa
IBM Linux Technology Center - Brazil
glommer at br.ibm.com
=====================================
More information about the uClibc
mailing list