Bug 5694 – va_arg doesn't work with idouble and ifloat

Created attachment 927 test case showing the problem In 64bit va_arg does not give back the correct pointer for the type idouble and ifloat. I haven't tested more types so far, but I guess cfloat and cdouble will give problems, too see attached test case. In 32bit the outputs are the same, in 64bit not. I only tested this on D1, but I assume it will have the same problems on D2

Created attachment 928 testcase for ifloat, idouble, cdouble, cfloat I extended the testcase to cfloat and cdouble. It works with cdouble. and it gives Internal error: backend/cgcod.c 576 with cfloat, so I couldn't test that part.

First fix (not the internal error): https://github.com/D-Programming-Language/druntime/commit/bd7b90f1c5dc681d05e872814379b709c1894d75 https://github.com/D-Programming-Language/phobos/commit/2830024ec775485cef643eec4d807dc7c1174d9d

https://github.com/D-Programming-Language/dmd/commit/1ccf5b1e25793e5178643f35e5c16bc2194b566e https://github.com/D-Programming-Language/dmd/commit/4b1357012bc39db2d991e97ff88ede1f74ab1ede Fix for internal error.