libffi 3.2 and disruptive changes

libffi 3.2 and disruptive changes

[Anthony Green]

I'd like to move ahead with Richard's proposed changes for GO, complex
types, and the numerous other clean ups he's proposed.

These changes will break msvc builds as well as iOS/OSX.

As long as we're breaking things, I'd like to use this opportunity to
change the header files so we have a shared set of .h files in
/usr/include.

Once things settle down from these changes, I'm hoping somebody can help
with the msvc and iOS/OSX ports.

In preparation for all of these changes, I've made a quick 3.2 release.
There were a few nice fixes that were sitting around in the tree, and
I'm not sure how long this next round of hacking will take.

The 3.2 release should be in pretty good shape *except* for the win32/64
msvc builds which are known to be broken.  See
https://github.com/atgreen/libffi/issues/138 for details.

Thanks,

AG

Re: libffi 3.2 and disruptive changes

[James Greenhalgh]

On Tue, Nov 11, 2014 at 02:53:35PM +0000, Anthony Green wrote:
> 
> I'd like to move ahead with Richard's proposed changes for GO, complex
> types, and the numerous other clean ups he's proposed.
> 
> These changes will break msvc builds as well as iOS/OSX.
> 
> As long as we're breaking things, I'd like to use this opportunity to
> change the header files so we have a shared set of .h files in
> /usr/include.
> 
> Once things settle down from these changes, I'm hoping somebody can help
> with the msvc and iOS/OSX ports.
> 
> In preparation for all of these changes, I've made a quick 3.2 release.
> There were a few nice fixes that were sitting around in the tree, and
> I'm not sure how long this next round of hacking will take.
> 
> The 3.2 release should be in pretty good shape *except* for the win32/64
> msvc builds which are known to be broken.  See
> https://github.com/atgreen/libffi/issues/138 for details.

AArch64 linux builds are still broken (Alan Lawrence posted a patch at
  https://sourceware.org/ml/libffi-discuss/2014/msg00091.html )

Could we please have that patch in for the 3.2 release?

Thanks,
James

Re: libffi 3.2 and disruptive changes

[Alan Lawrence]

In the meantime I believe non-iOS AArch64 builds are still broken, too?

Fix posted independently by myself and Richard Henderson:

https://sourceware.org/ml/libffi-discuss/2014/msg00134.html
https://sourceware.org/ml/libffi-discuss/2014/msg00091.html

Cheers, Alan

Anthony Green wrote:
> I'd like to move ahead with Richard's proposed changes for GO, complex
> types, and the numerous other clean ups he's proposed.
> 
> These changes will break msvc builds as well as iOS/OSX.
> 
> As long as we're breaking things, I'd like to use this opportunity to
> change the header files so we have a shared set of .h files in
> /usr/include.
> 
> Once things settle down from these changes, I'm hoping somebody can help
> with the msvc and iOS/OSX ports.
> 
> In preparation for all of these changes, I've made a quick 3.2 release.
> There were a few nice fixes that were sitting around in the tree, and
> I'm not sure how long this next round of hacking will take.
> 
> The 3.2 release should be in pretty good shape *except* for the win32/64
> msvc builds which are known to be broken.  See
> https://github.com/atgreen/libffi/issues/138 for details.
> 
> Thanks,
> 
> AG
> 

Re: libffi 3.2 and disruptive changes

[Dominik Vogt]

On Tue, Nov 11, 2014 at 09:53:35AM -0500, Anthony Green wrote:
> I'd like to move ahead with Richard's proposed changes for GO, complex
> types, and the numerous other clean ups he's proposed.

Richard proposed fixes for complex types?  Can you point me to the
messages with these patches please?  I must have overlooked them.

> msvc builds which are known to be broken.  See
> https://github.com/atgreen/libffi/issues/138 for details.

I've added a lengthy comment to that message report refering to
complex support and __STDC_NO_COMPLEX__.  That issue needs to be
resolved.

Ciao

Dominik ^_^  ^_^

-- 

Dominik Vogt
IBM Germany

Re: libffi 3.2 and disruptive changes

[Richard Henderson]

On 11/12/2014 08:16 AM, Dominik Vogt wrote:
> On Tue, Nov 11, 2014 at 09:53:35AM -0500, Anthony Green wrote:
>> I'd like to move ahead with Richard's proposed changes for GO, complex
>> types, and the numerous other clean ups he's proposed.
> 
> Richard proposed fixes for complex types?  Can you point me to the
> messages with these patches please?  I must have overlooked them.

They're mixed in with the Go closures patch sets, since we will
of course want libgo to make use of this new complex support.

Which will fix, e.g.

>     case GO_COMPLEX64:
> #ifdef __alpha__
>       runtime_throw("the libffi library does not support Complex64 type with "
>                     "reflect.Call or runtime.SetFinalizer");



r~

Re: libffi 3.2 and disruptive changes

[Anthony Green]

James Greenhalgh <james.greenhalgh@arm.com> writes:

> AArch64 linux builds are still broken (Alan Lawrence posted a patch at
>   https://sourceware.org/ml/libffi-discuss/2014/msg00091.html )
>
> Could we please have that patch in for the 3.2 release?

I've just tagged and released 3.2.1 with this fix. 

Thanks,

AG