Re: libffi 3.0.11 release candidate 2 available for testing

[David Gilbert <david.gilbert@linaro.org>]

On 13 November 2011 04:54, Anthony Green <green@moxielogic.com> wrote:
> I've just uploaded libffi 3.0.11rc2 for testing here:
>  ftp://sourceware.org/pub/libffi/libffi-3.0.11-rc2.tar.gz
>
> I bumped the sonumber for 3.0.11 because it contains two ABI changes...
>
> 1) a debug function was removed from the non-debug builds
>
> 2) this test release includes David Gilbert's changes to support variadic
> functions.

Hi Anthony,
  Thanks for that - but it looks like a small chunk is missing from
one of the test cases:

On armhf I'm getting:

FAIL: libffi.call/cls_double_va.c -O0 -W -Wall output pattern test, is 7.0
res: 4
2.6
res: 4
? should match 7.0
?es: 4
?.0
res: 4


and looking at cls_double_va.c it's missing the chunk:

@@ -48,6 +48,10 @@ int main (void)
        printf("res: %d\n", (int) res);
        // { dg-output "\nres: 4" }

+       /* The call to cls_double_va_fn is static, so have to use a
normal prep_cif */
+       CHECK(ffi_prep_cif(&cif, FFI_DEFAULT_ABI, 2, &ffi_type_sint,
+               arg_types) == FFI_OK);
+
        CHECK(ffi_prep_closure_loc(pcl, &cif, cls_double_va_fn, NULL,
code) == FFI_OK);

        res     = ((int(*)(char*, double))(code))(format, doubleArg);

that was in my patch; the failures go away with that.

(We also get XPASS: libffi.call/cls_longdouble.c which I think is an
old one, just with it not
knowing it will work on modern ARM - that's both arm and armhf)

Dave (away for the next week)