On Wed, Apr 15, 2009 at 12:59 AM, Jean-Claude Beaudoin jean.claude.beaudoin@gmail.com wrote:

One of the main new features of ECL 9.4 is its support for C type "long long" on Intel x86.

Here is a patch against CFFI 0.10.4 that gives access to this long awaited foreign type.

I tried your patch on x86-64 and it fails a lot of the "long long" tests. Did you run the test suite?

16 out of 192 total tests failed: FUNCALL.LONG-LONG, FUNCALL.VARARGS.DOUBLE, DEFCFUN.LONG-LONG, DEFCFUN.NOOP, DEFCFUN.VARARGS.FLOAT, DEFCFUN.VARARGS.DOUBLE, FOREIGN-GLOBALS.REF.LONG-LONG, FOREIGN-GLOBALS.SET.LONG-LONG, DEREF.LONG-LONG, DEREF.UNSIGNED-LONG-LONG, DEREF.NONCONST.LONG-LONG, DEREF.NONCONST.UNSIGNED-LONG-LONG, STRING.ENCODING.UTF-16.BASIC, STRING.ENCODINGS.ALL.BASIC, MISC-TYPES.STRING+PTR, MISC-TYPES.WRAPPER.

Had to disable the callback tests altogether. (In any case, ECL says it doesn't long-long for callback return types.) I've attached the patch I used.

-- Luís Oliveira http://student.dei.uc.pt/~lmoliv/