Jack,
As I said in IRC, I agree with the concept of these changes --
that fix will need to be applied many places. My suggestion is to
sed/perl-ify the "It's-not-a-Puma-configure" patch to make it easier to
apply to other packages without affecting patch-MD5s (which you've already
done in some places).
Thanks,
Fang
> David.
> I've updated gcc49, gmp5 and libmpfr4 to all build with the
> configure patch applied to insure that they are not linked as if Puma was
> in use.
> The new revision bump of gcc49 now produces prestine fortran test suite
> results for the first time under 10.10.
> Jack
> ps I've worked through my package set and updated those needing the patch
> as well as r-base30 and r-base31.
> Moving on to a clean bootstrap to see how much of that and the base package
> set are linking as Puma.
>
> On Sun, Oct 26, 2014 at 12:08 PM, Jack Howarth <howarth.at.f...@gmail.com>
> wrote:
>
>> David,
>> It appears that the hot fixes to mpfr seem to break the gcc49 build.
>> Attached is the previous
>> copy of libmpfr4-3.1.2-2 which only adds the yosemite symbol lookup patch.
>> Lets use that version
>> for now.
>> Jack
>>
>>
>> On Sun, Oct 26, 2014 at 11:34 AM, Jack Howarth <howarth.at.f...@gmail.com>
>> wrote:
>>
>>> David,
>>> We need to push out some hot fixes to repair breakage on Yosemite.
>>> MacPorts is using patches to configure to address the inappropriate linkage
>>> of shared libraries with...
>>>
>>> -Wl,-flat_namespace -Wl,-undefined -Wl,suppress
>>>
>>>
>>> instead of
>>>
>>>
>>> -Wl,-undefined -Wl,dynamic_lookup
>>>
>>>
>>> when MACOSX_DEPLOYMENT_TARGET is set to 10.10 as fink does by default.
>>>
>>> The fix is always of the form...
>>>
>>>
>>> diff -uNr mpfr-3.1.2.orig/configure mpfr-3.1.2/configure
>>>
>>> --- mpfr-3.1.2.orig/configure 2013-03-13 11:38:20.000000000 -0400
>>>
>>> +++ mpfr-3.1.2/configure 2014-10-26 01:00:37.000000000 -0400
>>>
>>> @@ -10401,7 +10401,7 @@
>>>
>>> case ${MACOSX_DEPLOYMENT_TARGET-10.0},$host in
>>>
>>> 10.0,*86*-darwin8*|10.0,*-darwin[91]*)
>>>
>>> _lt_dar_allow_undefined='${wl}-undefined ${wl}dynamic_lookup' ;;
>>>
>>> - 10.[012]*)
>>>
>>> + 10.[012][,.]*)
>>>
>>> _lt_dar_allow_undefined='${wl}-flat_namespace ${wl}-undefined
>>> ${wl}suppress' ;;
>>>
>>> 10.*)
>>>
>>> _lt_dar_allow_undefined='${wl}-undefined ${wl}dynamic_lookup' ;;
>>>
>>> diff -uNr mpfr-3.1.2.orig/m4/libtool.m4 mpfr-3.1.2/m4/libtool.m4
>>>
>>> --- mpfr-3.1.2.orig/m4/libtool.m4 2013-03-13 11:38:16.000000000
>>> -0400
>>>
>>> +++ mpfr-3.1.2/m4/libtool.m4 2014-10-26 01:00:37.000000000 -0400
>>>
>>> @@ -1052,7 +1052,7 @@
>>>
>>> case ${MACOSX_DEPLOYMENT_TARGET-10.0},$host in
>>>
>>> 10.0,*86*-darwin8*|10.0,*-darwin[[91]]*)
>>>
>>> _lt_dar_allow_undefined='${wl}-undefined ${wl}dynamic_lookup' ;;
>>>
>>> - 10.[[012]]*)
>>>
>>> + 10.[[012]][[,.]]*)
>>>
>>> _lt_dar_allow_undefined='${wl}-flat_namespace ${wl}-undefined
>>> ${wl}suppress' ;;
>>>
>>> 10.*)
>>>
>>> _lt_dar_allow_undefined='${wl}-undefined ${wl}dynamic_lookup' ;;
>>>
>>> Without this hot-fix shared libraries are being linked as if the
>>> target were Puma instead
>>>
>>> of Yosemite. The issue is very wide spread as libtool needs to be updated
>>> in all the
>>>
>>> packages using it. It also can cause nasty side-effects as witnessed by
>>> the large number
>>>
>>> of gcc49 test suite regressions when linked in this way. I will push a
>>> fix to gcc49-4.9.1 later
>>>
>>> today.
>>>
>>> Attached is the fixes for libmpfr4 and gmp5. Note that I also added
>>> the missing
>>>
>>> upstream accumulative patches to take libmpfr to 3.1.2-P10.
>>>
>>> Jack
>>> l
>>>
>>
>>
>
--
David Fang
http://www.csl.cornell.edu/~fang/
------------------------------------------------------------------------------
_______________________________________________
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
List archive:
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel
