This is the mail archive of the crossgcc@sourceware.org mailing list for the crossgcc project.

See the CrossGCC FAQ for lots more information.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Glibc make tests failing on sh3


Hi,
I have successfully build gcc-3.4.5 and gcc-4.1.0 both with glibc-2.3.6
for sh3. I already built a userland with them, so they seem to work.

Unfortunatelly 'make tests' for glibc is failing in the math folder when
compiling bug-nextafter.c , because FE_OVERFLOW isn't defined. (this is
the first, not the only one failing)
It seems to be floating-point related, because for other architectures
this macro is defined in <<ARCH>>/<<SPECIFIC-ARCH>>/fpu/bits/fenv.h
However floating point has been disabled with --without-fp and my sh3
board doesn't have a fpu.

I managed to compile the tests with a _very_ dirty hack. I replaced line
102 in Makeconfig 
export full_config_sysdirs := $(addprefix $(..),$(config-sysdirs))
With the following:
export full_config_sysdirs := $(addprefix $(..),sysdeps/sh/sh4/fpu
$(config-sysdirs))

So gcc will use the includes for sh4, which has a fpu.

Usually the following files are included from bug-nextafter.c with
'#include <fenv.h>'

include/fenv.h
	math/fenv.h
		include/features.h
		sysdeps/generic/bits/fenv.h

With my hack, sysdeps/sh/sh4/fpu/bits/fenv.h will be included instead of
sysdeps/generic/bits/fenv.h.

So I'm using some constants for another cpu. That's not acceptable for
reliable tests.

Any Ideas?
Has anybody tried the tests on a board without fpu?

Thanks in advance

Markus

--
For unsubscribe information see http://sourceware.org/lists.html#faq


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]