http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51741
--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-01-03 15:25:32 UTC --- The bug reporting guidelines at http://gcc.gnu.org/bugs/ ask for the following which you have not provided: * the exact version of gcc * the exact command line * the complete preprocessed source, not some code snippets (what is pos?) (In reply to comment #0) > (&((knot*)((((*pos).branches)[i]).where))) > > (compiler sas lvaluve needed for unary & operator) The compiler is correct. The cast to (knot*) creates an rvalue, so you cannot take its address. > ((knot**)(&((((*pos).branches)[i]).where))) Here you take the address of an lvalue, then cast it to (knot**), that's legal. > will. > both seem like legal C to me. No, the first one is not. The bug is in your code, not GCC. > (((knot*)((((*pos).branches)[0]).where)).father)=pos; > > won't compile . > compiler thinks I try to adress .father for something thats not a structure or > a union. even though I castet it to a knot* before trying to adress .father. Maybe you mean ->father not .father. Just because your code doesn't compile, that doesn't mean it's a GCC bug.
