[MacPorts] #34231: ImageMagick: should not depend on pango + fix openMP

[MacPorts]

#34231: ImageMagick: should not depend on pango + fix openMP
---------------------------------------+------------------------------------
 Reporter:  frederic.devernay@…        |       Owner:  ryandesign@…           
     Type:  defect                     |      Status:  new                    
 Priority:  Normal                     |   Milestone:                         
Component:  ports                      |     Version:  2.0.4                  
 Keywords:  haspatch                   |        Port:  ImageMagick            
---------------------------------------+------------------------------------

Comment(by ryandesign@…):

 Replying to [comment:2 frederic.devernay@…]:
 > Here's the proper reference for the gcc 4.2/4.3 bug with OpenMP:
 > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36242#c4 [[br]]
 > [[br]]
 > to cite Jakup, GCC OpenMP guru: [[br]]
 > "Jakub Jelinek 2008-06-11 20:29:19 UTC [[br]]
 > [[br]]
 > Using OpenMP pragmas in multiple pthread_create created threads
 concurrently
 > isn't supported in 4.2/4.3.
 > There is a support for it on the trunk since a few days ago though."

 And does ImageMagick use OpenMP pragmas in multiple pthread_create-created
 threads concurrently? What is the consequence—is there a crash? If so, can
 you attach a crash log? How can I reproduce the problem on my own system?
 I am on Snow Leopard with Xcode 3.2.6 so my ImageMagick is compiled with
 Apple gcc 4.2.1 so it should exhibit the problem.

 > gcc 4.2 is for old geezers anyway (gcc 4.2.1 was released almost five
 years ago!)

 Apple's version of gcc 4.2.1 is the default compiler in Xcode 3.2.6, the
 last free version of Xcode for Snow Leopard. It's also the last version of
 gcc Apple will ever ship in Xcode. For these reasons it's still a relevant
 version of gcc for OS X users.

 > here's another report about that OpenMP bug in another context:
 > http://lists.apple.com/archives/coreaudio-api/2009/May/msg00035.html

 I don't think that's the same problem. They're talking about problems
 using OpenMP from a "time constraint thread" which "is very different from
 a normal thread context" and the guess is "that OpenMP has some baseline
 incompatibility when running in this context". The thread also says gcc
 4.4 still exhibits the problem, whereas the problem you're describing
 should be fixed by that version.

 If there is really a problem with OpenMP when compiled by gcc 4.2.1 and
 earlier, then in light of the requests in #15945 and #24944 to provide
 options to enable or disable OpenMP at install time, what we should do is
 add an openmp variant, make it on by default when the compiler is clang or
 llvm-gcc-4.2, and off by default otherwise. That should enable OpenMP on
 Lion and later, and on Snow Leopard with Xcode 4, and disable it
 otherwise. Furthermore perhaps we could offer Xcode 3 users on Leopard and
 Snow Leopard the option of turning on OpenMP support by changing the
 compiler to llvm-gcc-4.2 or clang, if that works—the version of clang in
 Xcode 3 is old so I'm a little skeptical of that anyway.

-- 
Ticket URL: <https://trac.macports.org/ticket/34231#comment:3>
MacPorts <http://www.macports.org/>
Ports system for Mac OS