This is the mail archive of the libffi-discuss@sourceware.org mailing list for the libffi project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: Win64 patch to libffi

From: Timothy Wall <twalljava at dev dot java dot net>
To: NightStrike <nightstrike at gmail dot com>
Cc: Kai Tietz <ktietz70 at googlemail dot com>, Kai Tietz <Kai dot Tietz at onevision dot com>, libffi-discuss at sourceware dot org
Date: Sat, 27 Dec 2008 12:07:59 -0500
Subject: Re: Win64 patch to libffi
References: <49469148.3050904@gmail.com> <b609cb3b0812201446m2ab5057et260bf2a6dd01b9df@mail.gmail.com> <b609cb3b0812201555l58180cefqbec12e6a3995aa00@mail.gmail.com> <90baa01f0812210033r1855b87bh618fcf71b444daf@mail.gmail.com> <b609cb3b0812210313m3c2f6300xe51cfcbfacf7b50b@mail.gmail.com> <4235B7D3-54C8-4702-990A-39DECF8D8DFF@dev.java.net> <b609cb3b0812220752m6558a40ex76c83602e517afc2@mail.gmail.com> <96343E9C-FFF6-4D6B-9F7A-A8B043C30841@dev.java.net> <b609cb3b0812231110i711547adsc3e774b88abfa117@mail.gmail.com> <F1A94BCD-8C7E-4146-87E1-29B56E5AE09D@dev.java.net> <b609cb3b0812261840i639e5e3fka211d00b3e784b9a@mail.gmail.com>

It looks like the only thing it's checking is that the libffi stack manipulation has not affected exception handling.

Does the usage of C++ interfaces to C headers cause compiler warnings which in turn cause the test to fail?

Usually, things like this are left as is for existing targets to avoid disrupting older compilers. Ideally you'd check for the availability of the C++ interfaces with an autoconf flag or other predefined macro, but I'm generally autoconf-ignorant so I can't help much there.

On Dec 26, 2008, at 9:40 PM, NightStrike wrote:

Tim,
In looking at the libffi.special tests, unwindtest.cc has the same
casting issues that everything else has.  I didn't fix this yet,
however, because I am not entirely sure how to do it to be consistent
with the rest of libffi.  Mainly, this is c++, yet no c++ is being
used.  Even the headers that serve as an interface to c (for instance,
cstdio) aren't being used, in favor of making it strictly C (stdio.h).
I would expect to use c++ (iostream, cout) instead of C (printf).
But at the very least, C++ interfaces to the C headers should be used
(cstdio).
So, before I go about fixing anything here, can you give some insight
into what exactly I'm even looking at?  I would imagine that you have
a lot more knowledge in this area that I do, and I'm probably missing
something extremely simply, so your input is greatly appreciated.
On Fri, Dec 26, 2008 at 2:06 PM, Timothy Wall <twalljava@dev.java.net> wrote:
XFAIL marked on all longdouble tests for x86_64-*-mingw*.

On Dec 23, 2008, at 2:10 PM, NightStrike wrote:
Note: There is an autoconf 2.62/2.63 version mismatch between your
tarball and the files in CVS
All of the long-double tests should be unsupported or xfailed,
whichever you prefer:
Index: cls_align_longdouble_split.c =================================================================== RCS file: /cvs/libffi/libffi/testsuite/libffi.call/ cls_align_longdouble_split.c,v retrieving revision 1.1 diff -u -r1.1 cls_align_longdouble_split.c --- cls_align_longdouble_split.c 5 Mar 2008 00:07:02 -0000 1.1 +++ cls_align_longdouble_split.c 23 Dec 2008 19:08:46 -0000 @@ -4,7 +4,7 @@ PR: none. Originator: <hos@tamanegi.org> 20031203 */

-/* { dg-do run { xfail mips*-*-* arm*-*-* strongarm*-*-* xscale*- *-* } } */ +/* { dg-do run { xfail mips*-*-* arm*-*-* strongarm*-*-* xscale*- *-* x86_64-*-mingw* } } */ /* { dg-options -mlong-double-128 { target powerpc64*-*-* } } */

#include "ffitest.h"

On Mon, Dec 22, 2008 at 12:08 PM, Timothy Wall <twalljava@dev.java.net > wrote:
I'm on vacation until 1/5/09, but I'll try to apply the patches you send during that time, there just may be a one or two day delay.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]