stuck in loop with restore_ports.tcl migrating to macOS Sierra (10.12.1)

Murray Eisenberg murrayeisenberg at gmail.com
Thu Nov 3 15:43:53 CET 2016


OK, and in 

  /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.12.sdk/usr/include/Availability.h

I find at lines 341-355:

/* for use marking APIs unavailable for swift */
#if defined(__has_feature)
  #if __has_feature(attribute_availability_swift)
    #define __SWIFT_UNAVAILABLE                   __OS_AVAILABILITY(swift,unavailable)
    #define __SWIFT_UNAVAILABLE_MSG(_msg)         __OS_AVAILABILITY_MSG(swift,unavailable,_msg)
  #endif
#endif

#ifndef __SWIFT_UNAVAILABLE
  #define __SWIFT_UNAVAILABLE
#endif

#ifndef __SWIFT_UNAVAILABLE_MSG
  #define __SWIFT_UNAVAILABLE_MSG(_msg)
#endif


Is there once again a problem in Availability.h with 'defined(__has_feature)’  which should really be 'defined(__has_feature)’ 
again, as discussed in ticket #48471?

If so, the workaround for that was to make a local copy of Availability.h in /opt/local/include and edit it there.


> On Nov 3, 2016, at 10:29 AM, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
> 
> Look at this ticket <https://trac.macports.org/ticket/52776> comment 3
> 
> pthreads is not picking up the definition from Availability.h for some reason.
> 
> Ken
> 
> 
> On 2016-11-03, at 7:26 AM, Murray Eisenberg wrote:
> 
>> I looked at the tickets #46589, 51971, and 52326 about gmp, and I don’t see mention of availability.h there.  
>> 
>> I thought the issue with availability.h concerned gcc48 and was resolved somehow (with newer Xcode? with patched port?) some time ago.
>> 
>> The logs seem to indicate that the current issue with gmp involves pthread.h.
>> 
>> 
>>> On Nov 3, 2016, at 10:10 AM, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
>>> 
>>> Murray, that is very specific now
>>> 
>>> I have that port installed.
>>> 
>>> $ port -v installed gmp
>>> The following ports are currently installed:
>>> gmp @6.1.1_0 (active) platform='darwin 16' archs='x86_64'
>>> 
>>> I just rebuilt it right now from source without trouble.
>>> 
>>> so it's something on your machine. Jerermy points to a possibly corrupt Availability.h file in the trac ticket.
>>> 
>>> So you might look at that file, or just reinstall Xcode and the command line tools.
>>> 
>>> (Why is this not coming to you as a prebuilt binary from the buldbots, I wonder?)
>>> 
>>> 
>>> Best,
>>> 
>>> Ken
>>> 
>>> 
>>> 
>>> 
>>> On 2016-11-03, at 7:01 AM, Murray Eisenberg wrote:
>>> 
>>>> After the re-install script (from the migration instructions) got into an infinite loop, I started to reinstall ports manually, starting with the first one on my “myports.txt” list: analitza 
>>>> 
>>>> The failure came when installing that failed during the automatic installation of dependencies, in that case gmp.
>>>> 
>>>> Today, looking at the dependencies for gmp, I see that all build and library dependencies for that are already installed _except_ kdelibs.
>>>> 
>>>> So I tried reinstalling kdelibs, and that in turn choked at trying to install its dependency gmp.
>>>> 
>>>> So everything pretty much comes down to failure to configure gmp. 
>>>> 
>>>> Configuring gmp (specifically, @6.1.1_0) fails with what appears in main.log as:
>>>> 
>>>>  :info:configure configure: error: C++ compiler not available, see config.log for details
>>>> 
>>>> In turn, config.log reports:
>>>> 
>>>>  /usr/include/pthread.h:423:1: error: C++ requires a type specifier for all declarations
>>>> __SWIFT_UNAVAILABLE_MSG("Use lazily initialized globals instead”)
>>>> 
>>>> And that seems to reduce to the issue of the problem with /usr/include/pthread.h, namely:
>>>> 
>>>>  /usr/include/pthread.h:423:1: error: C++ requires a type specifier for all declarations
>>>> __SWIFT_UNAVAILABLE_MSG("Use lazily initialized globals instead")
>>>> ^
>>>> /usr/include/pthread.h:423:66: error: expected ';' after top level declarator
>>>> __SWIFT_UNAVAILABLE_MSG("Use lazily initialized globals instead")
>>>>                                                              ^
>>>> 2 errors generated.
>>>> configure:10556: $? = 1
>>>> failed program was:
>>>> /* This test rejects g++ 2.7.2 which doesn't have <iostream>, only a
>>>>     pre-standard iostream.h. */
>>>> #include <iostream>
>>>> 
>>>> I just was about to try to do that  
>>>>> On Nov 2, 2016, at 11:46 PM, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
>>>>> 
>>>>> can you remind me the name of a port that triggers the error so I can test it (hopefully not clang-3.8 which would take all night to build ) ;>
>>>>> 
>>>>> K
>>>> 
>>>> ---
>>>> Murray Eisenberg			murrayeisenberg at gmail.com
>>>> 503 King Farm Blvd #101	Home (240)-246-7240
>>>> Rockville, MD 20850-6667	Mobile (413)-427-5334
>>>> 
>>>> 
>>> 
>> 
>> ---
>> Murray Eisenberg			murrayeisenberg at gmail.com
>> 503 King Farm Blvd #101	Home (240)-246-7240
>> Rockville, MD 20850-6667	Mobile (413)-427-5334
>> 
>> 
> 

---
Murray Eisenberg			murrayeisenberg at gmail.com
503 King Farm Blvd #101	Home (240)-246-7240
Rockville, MD 20850-6667	Mobile (413)-427-5334


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.macports.org/pipermail/macports-users/attachments/20161103/c5faf83a/attachment-0002.html>


More information about the macports-users mailing list