On Mon, Oct 20, 2003 at 03:16:56PM +0100, Paul Brook wrote: > What seems to be happening is that the real component of A is overwritten > before the complex component of A*B is calculated. I'd expect that to be an error in the tree->rtl expanders. r~