As David pointed out, my libffi patch
http://gcc.gnu.org/ml/gcc-patches/2003-10/msg00412.html
broke bootstrap for mipsisa32el-linux (or in fact for all mips*-*-{irix,
linux}* platforms):
http://gcc.gnu.org/ml/gcc-patches/2003-10/msg00566.html
As described in
http://gcc.gnu.org/ml/gcc-patches/2003-10/msg00599.html
the mips port should consistently define and use _ABIO32 for _MIPS_SIM
instead of relying on an external definition of _MIPS_SIM_ABI32, which is
only present in <sgidefs.h>. The following patch does this all over the
tree.
I'll better fire off another full bootstrap on mips-sgi-irix6.5; perhaps
someone else could do the same on some MIPS/Linux system to make extra sure
this doesn't break anything else?
Ok for mainline if it passes?
Rainer
-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University
Wed Oct 8 13:57:17 2003 Rainer Orth <ro@TechFak.Uni-Bielefeld.DE>
boehm-gc:
* mips_sgi_mach_dep.s: Use _ABIO32 instead of external
_MIPS_SIM_ABI32.
gcc:
* config/mips/iris6.h (TARGET_OS_CPP_BUILTINS): Define _ABIO32.
Use it in _MIPS_SIM definition.
* config/mips/linux.h (TARGET_OS_CPP_BUILTINS): Likewise.
* config/mips/mips.h (CRT_CALL_STATIC_FUNCTION): Likewise.
libffi:
* src/mips/ffi.c: Use _ABIN32, _ABIO32 instead of external
_MIPS_SIM_NABI32, _MIPS_SIM_ABI32.
libstdc++-v3:
* config/cpu/mips/atomicity.h (__atomic_add): Use _ABIO32 instead
of external _MIPS_SIM_ABI32.