On Fri, Sep 28, 2007 at 12:17:56PM +0200, Martin Michlmayr wrote: > > From: David Schleef <[EMAIL PROTECTED]> > > > > even if there isn't. The binary will only fail at run time, i.e. when > > > you actually execute it. This check therefore needs to be converted > > > to a runtime check. > > > > This is correct. It's a test whether or not binutils can compile > > floating point code on purpose. This is because binutils targetted > > for some embedded targets doesn't allow fp code. > > Are you sure about that? I believe I've built my toolchain correctly > for powerpc soft-float, but binutils doesn't complain at all about > this code. As a result, liboil builds for FPU but then later on GCC > fails to compile some source file in liboil because of assembler > constraints that cannot be met.
Ah, then you've built a non-fp gcc, but an fp binutils. I hadn't thought of that possibility. Anyway, the test is wrong, since it doesn't check that gcc can handle fp register constraints, which is just as important as being able to assemble fp instructions. dave... -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
