On Mon, 11 Dec 2006 09:10:26 +1100, Peter Jansen wrote: > So yes gcc-3.2.3 generates broken code in some case, there are many > options to make it generate the correct code, I suspect the problem is > somewhere in the gcc-3.2 language processing part, not in the msp430 > backend.
I just compiled gcc-3.2.3 for x86 and it accesses the value twice as well, which backs up the argument that it's a problem with the backend. I've been using the "copy to local variable" trick. This sucks, but never mind. Rob
