PATCH: ffi_call argument/return types

Tom Tromey tromey@redhat.com
Thu Aug 16 08:33:00 GMT 2001


>>>>> "Jeff" == Jeff Sturm <jsturm@one-point.com> writes:

Jeff> My only misgiving about Mauve vs. the libgcj testsuite is the
Jeff> former isn't used as often.  At least I don't run Mauve every
Jeff> time I bootstrap a build, and I certainly doubt random GCC
Jeff> hackers do.

I agree that is a problem.

Jeff>         ffi_call(&cif, FFI_FN(return_sc), &rint, values);

Jeff> Note the return type is specified as ffi_type_schar, but an (int
Jeff> *) is provided to store the return value regardless (long long
Jeff> for a 64-bit ABI).  The way libffi is currently implemented it
Jeff> _cannot_ be smaller than this.

I see what you mean.  I looked at README but it doesn't really
describe exactly how the arguments to ffi_call should be laid out.
That is, should we be using the real types or the promoted types?  FYI
it does mention that for Irix 6.2 the return value pointer should
point to a `long long'.

Anthony, help?

Jeff, how does what you are proposing affect the "raw" or "closure"
APIs?

Tom



More information about the Java-patches mailing list