Re: libffi merge

[Andrew Haley <aph at redhat dot com>]

Timothy Wall wrote:
> On Jun 4, 2009, at 1:03 PM, Andrew Haley wrote:

>> OK, I backed out that patch.  I'm awaiting some explanation of what
>> on Earth is going on.  I'm guessing that Timothy Wall changed the
>> version upstream but not the version in gcc, and then you changed
>> the version in gcc but not the version upstream.  So, we have a two
>> sets of changes, and I can't tell which is supposed to go where.
> 
> I applied the STDCALL closure patch to libffi, but not to GCC.  I
> believe Dave applied the STDCALL closure patch to libffi, expanded
> somewhat to include the extra EH info.

OK, so the version in libffi ought to be correct, then.

But I'm looking at src/x86/win32.S in libffi, and I see:

        .globl _ffi_closure_STDCALL
_ffi_closure_STDCALL:
        pushl   %ebp
        movl    %esp, %ebp
        subl    $40, %esp
        leal    -24(%ebp), %edx
        movl    %edx, -12(%ebp) /* resp */
        leal    12(%ebp), %edx  /* account for stub return address on stack */
        movl    %edx, 4(%esp)   /* args */
        leal    -12(%ebp), %edx
        movl    %edx, (%esp)    /* &resp */
        call    _ffi_closure_SYSV_inner
        movl    -12(%ebp), %ecx
        jmp     .Lcls_return_result
.ffi_closure_STDCALL_end:

which is the code that Dave is complaining about.

Andrew.