Re: [Fink-devel] boost1.57 test packaging
Hanspeter, Your current packaging builds fine here. FYI, I have created a patch which fixes all the -std=c++11 compilation issues in rstudio so it can be built that way on 10.9 and later. I also have split-off versioned variants for 10.7/10.8 of rstudio-desktop/rstudio-server so that the main packages for 10.9 and later can use boost 1.58 rather than the older boost 1.55. Jack On Mon, Apr 20, 2015 at 10:36 AM, Hanspeter Niederstrasser f...@snaggledworks.com wrote: I've updated boost1.58.info in my experimental to use the just released boost-1_58_0 from upstream. Please test and report back any problems. Upstream screwed up their (lack of) use of -install_name and c_v even more, so there's way more games being played in the .info file now to make sure the libraries are all properly named and versioned. I think I got them all. http://cvs.snaggledworks.com/viewvc.cgi/fink/10.7/main/finkinfo/libs/ Hanspeter -- -- BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15utm_medium=emailutm_campaign=VA_SF ___ 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
Re: [Fink-devel] boost1.55.info missing libicu depends
Hanspeter, It is unfortunate that https://svn.boost.org/trac/boost/ticket/10785 wasn't followed up and fixed in 1.58. The issue apparently impacts a number of packages... https://github.com/scantailor/scantailor/issues/125 https://github.com/kylelutz/compute/issues/6 https://github.com/lballabio/quantlib/issues/163 and the only known workaround is to force the builds to -std=c++11 (which introduces its own set of issues). Jack On Mon, Apr 20, 2015 at 1:14 AM, Jack Howarth howarth.at.f...@gmail.com wrote: I also did a test with your current boost packaging reverted back to build 1.57 and rstudio-desktop gets further but runs into the error described here... https://svn.boost.org/trac/boost/ticket/10785 which appears to be unresolved in 1.58 as well. The only known workaround is to force -std=c++11 or -std=c++14 rather than the default -std=c++03. Of course jumping up to -std=c++11 triggers a different set of issues. /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:38:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:172:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ On Sun, Apr 19, 2015 at 10:29 PM, Jack Howarth howarth.at.f...@gmail.com wrote: There may not need to be a rush. It looks like there might be issues with clang vs the new 1.58 release... In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/nullary_function.hpp:14: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/memory.hpp:17: In file included from /sw/opt/boost-1_58/include/boost/thread/csbl/memory/pointer_traits.hpp:16: /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: error: use of undeclared identifier 'check' BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE(has_member_function_callable_with_pointer_to, pointer_to) ^ /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:386:36: note: expanded from macro 'BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE' static const bool value = sizeof(checkU(0)) == sizeof(::boost::intrusive::detail::yes_type); \ ^ In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/nullary_function.hpp:14: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/memory.hpp:17: In file included from /sw/opt/boost-1_58/include/boost/thread/csbl/memory/pointer_traits.hpp:16: /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: error: 'U' does not refer to a value BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE(has_member_function_callable_with_pointer_to, pointer_to) ^ /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:386:42: note: expanded from macro 'BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE' static const bool value = sizeof(checkU(0)) == sizeof(::boost::intrusive::detail::yes_type); \ ^ /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: note: declared here /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:377:22: note: expanded from macro
Re: [Fink-devel] boost1.55.info missing libicu depends
The issue wouldn't be so annoying if clang supported a pragma to enable -std=c++11 on the source file level so that only the offending files can be compiled with that option. The packages with cmake builds don't lend themselves to limited application of flags. On Mon, Apr 20, 2015 at 8:26 AM, Jack Howarth howarth.at.f...@gmail.com wrote: Hanspeter, It is unfortunate that https://svn.boost.org/trac/boost/ticket/10785 wasn't followed up and fixed in 1.58. The issue apparently impacts a number of packages... https://github.com/scantailor/scantailor/issues/125 https://github.com/kylelutz/compute/issues/6 https://github.com/lballabio/quantlib/issues/163 and the only known workaround is to force the builds to -std=c++11 (which introduces its own set of issues). Jack On Mon, Apr 20, 2015 at 1:14 AM, Jack Howarth howarth.at.f...@gmail.com wrote: I also did a test with your current boost packaging reverted back to build 1.57 and rstudio-desktop gets further but runs into the error described here... https://svn.boost.org/trac/boost/ticket/10785 which appears to be unresolved in 1.58 as well. The only known workaround is to force -std=c++11 or -std=c++14 rather than the default -std=c++03. Of course jumping up to -std=c++11 triggers a different set of issues. /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:38:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:172:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ On Sun, Apr 19, 2015 at 10:29 PM, Jack Howarth howarth.at.f...@gmail.com wrote: There may not need to be a rush. It looks like there might be issues with clang vs the new 1.58 release... In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/nullary_function.hpp:14: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/memory.hpp:17: In file included from /sw/opt/boost-1_58/include/boost/thread/csbl/memory/pointer_traits.hpp:16: /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: error: use of undeclared identifier 'check' BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE(has_member_function_callable_with_pointer_to, pointer_to) ^ /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:386:36: note: expanded from macro 'BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE' static const bool value = sizeof(checkU(0)) == sizeof(::boost::intrusive::detail::yes_type); \ ^ In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/nullary_function.hpp:14: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/memory.hpp:17: In file included from /sw/opt/boost-1_58/include/boost/thread/csbl/memory/pointer_traits.hpp:16: /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: error: 'U' does not refer to a value BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE(has_member_function_callable_with_pointer_to, pointer_to) ^ /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:386:42: note: expanded from macro 'BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE' static const
Re: [Fink-devel] boost1.55.info missing libicu depends
FYI, I am able to build rstudio-desktop against boost 1.57 if I limit the invocation of -std=c++11 to just the src/cpp/session/modules/SessionGit.cpp and src/cpp/session/modules/SessionSVN.cpp source files. Unfortunately the cmake build doesn't seem amiable to selectively compiling those files with that option and no clang pragma to apply it seems to exist. I've opened https://llvm.org/bugs/show_bug.cgi?id=23295 to get some attention from the clang developers in case they can recommend a fix for boost 1.57/1.58. On Mon, Apr 20, 2015 at 8:34 AM, Jack Howarth howarth.at.f...@gmail.com wrote: The issue wouldn't be so annoying if clang supported a pragma to enable -std=c++11 on the source file level so that only the offending files can be compiled with that option. The packages with cmake builds don't lend themselves to limited application of flags. On Mon, Apr 20, 2015 at 8:26 AM, Jack Howarth howarth.at.f...@gmail.com wrote: Hanspeter, It is unfortunate that https://svn.boost.org/trac/boost/ticket/10785 wasn't followed up and fixed in 1.58. The issue apparently impacts a number of packages... https://github.com/scantailor/scantailor/issues/125 https://github.com/kylelutz/compute/issues/6 https://github.com/lballabio/quantlib/issues/163 and the only known workaround is to force the builds to -std=c++11 (which introduces its own set of issues). Jack On Mon, Apr 20, 2015 at 1:14 AM, Jack Howarth howarth.at.f...@gmail.com wrote: I also did a test with your current boost packaging reverted back to build 1.57 and rstudio-desktop gets further but runs into the error described here... https://svn.boost.org/trac/boost/ticket/10785 which appears to be unresolved in 1.58 as well. The only known workaround is to force -std=c++11 or -std=c++14 rather than the default -std=c++03. Of course jumping up to -std=c++11 triggers a different set of issues. /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:38:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:172:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ On Sun, Apr 19, 2015 at 10:29 PM, Jack Howarth howarth.at.f...@gmail.com wrote: There may not need to be a rush. It looks like there might be issues with clang vs the new 1.58 release... In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/nullary_function.hpp:14: In file included from /sw/opt/boost-1_58/include/boost/thread/detail/memory.hpp:17: In file included from /sw/opt/boost-1_58/include/boost/thread/csbl/memory/pointer_traits.hpp:16: /sw/opt/boost-1_58/include/boost/intrusive/pointer_traits.hpp:36:1: error: use of undeclared identifier 'check' BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE(has_member_function_callable_with_pointer_to, pointer_to) ^ /sw/opt/boost-1_58/include/boost/intrusive/detail/mpl.hpp:386:36: note: expanded from macro 'BOOST_INTRUSIVE_HAS_STATIC_MEMBER_FUNC_SIGNATURE' static const bool value = sizeof(checkU(0)) == sizeof(::boost::intrusive::detail::yes_type); \ ^ In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from /sw/opt/boost-1_58/include/boost/thread.hpp:23: In file included from /sw/opt/boost-1_58/include/boost/thread/barrier.hpp:21: In file included from
Re: [Fink-devel] boost1.55.info missing libicu depends
The fix wasn't so hard to implement in cmake as I feared. The additional patch... diff -uNr rstudio-0.98.1103.orig/src/cpp/core/CMakeLists.txt rstudio-0.98.1103/src/cpp/core/CMakeLists.txt --- rstudio-0.98.1103.orig/src/cpp/core/CMakeLists.txt 2015-03-06 17:12:51.0 -0500 +++ rstudio-0.98.1103/src/cpp/core/CMakeLists.txt 2015-04-20 09:33:29.0 -0400 @@ -242,6 +242,9 @@ endif() +# avoid https://svn.boost.org/trac/boost/ticket/10785 +set_source_files_properties(system/file_monitor/MacFileMonitor.cpp PROPERTIES COMPILE_FLAGS -std=c++11) + # define include dirs set(CORE_INCLUDE_DIRS ${CORE_INCLUDE_DIRS} include) diff -uNr rstudio-0.98.1103.orig/src/cpp/session/CMakeLists.txt rstudio-0.98.1103/src/cpp/session/CMakeLists.txt --- rstudio-0.98.1103.orig/src/cpp/session/CMakeLists.txt 2015-03-06 17:12:51.0 -0500 +++ rstudio-0.98.1103/src/cpp/session/CMakeLists.txt2015-04-20 09:32:06.0 -0400 @@ -136,6 +136,10 @@ ${CMAKE_CURRENT_BINARY_DIR}/SessionAddins.cpp ) +# avoid https://svn.boost.org/trac/boost/ticket/10785 +set_source_files_properties(modules/SessionGit.cpp PROPERTIES COMPILE_FLAGS -std=c++11) +set_source_files_properties(modules/SessionSVN.cpp PROPERTIES COMPILE_FLAGS -std=c++11) + # platform specific source files if(UNIX) set(SESSION_SOURCE_FILES ${SESSION_SOURCE_FILES} allows rstudio-desktop to build against boost 1.58 and the resulting application runs fine. On Mon, Apr 20, 2015 at 8:50 AM, Jack Howarth howarth.at.f...@gmail.com wrote: FYI, I am able to build rstudio-desktop against boost 1.57 if I limit the invocation of -std=c++11 to just the src/cpp/session/modules/SessionGit.cpp and src/cpp/session/modules/SessionSVN.cpp source files. Unfortunately the cmake build doesn't seem amiable to selectively compiling those files with that option and no clang pragma to apply it seems to exist. I've opened https://llvm.org/bugs/show_bug.cgi?id=23295 to get some attention from the clang developers in case they can recommend a fix for boost 1.57/1.58. On Mon, Apr 20, 2015 at 8:34 AM, Jack Howarth howarth.at.f...@gmail.com wrote: The issue wouldn't be so annoying if clang supported a pragma to enable -std=c++11 on the source file level so that only the offending files can be compiled with that option. The packages with cmake builds don't lend themselves to limited application of flags. On Mon, Apr 20, 2015 at 8:26 AM, Jack Howarth howarth.at.f...@gmail.com wrote: Hanspeter, It is unfortunate that https://svn.boost.org/trac/boost/ticket/10785 wasn't followed up and fixed in 1.58. The issue apparently impacts a number of packages... https://github.com/scantailor/scantailor/issues/125 https://github.com/kylelutz/compute/issues/6 https://github.com/lballabio/quantlib/issues/163 and the only known workaround is to force the builds to -std=c++11 (which introduces its own set of issues). Jack On Mon, Apr 20, 2015 at 1:14 AM, Jack Howarth howarth.at.f...@gmail.com wrote: I also did a test with your current boost packaging reverted back to build 1.57 and rstudio-desktop gets further but runs into the error described here... https://svn.boost.org/trac/boost/ticket/10785 which appears to be unresolved in 1.58 as well. The only known workaround is to force -std=c++11 or -std=c++14 rather than the default -std=c++03. Of course jumping up to -std=c++11 triggers a different set of issues. /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:38:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/BoostErrors.cpp:172:16: error: exception specification of overriding function is more lax than base version const char* name() const ; ^ /sw/opt/boost-1_57/include/boost/system/error_code.hpp:190:32: note: overridden virtual function is here virtual const char * name() const BOOST_SYSTEM_NOEXCEPT = 0; ^ On Sun, Apr 19, 2015 at 10:29 PM, Jack Howarth howarth.at.f...@gmail.com wrote: There may not need to be a rush. It looks like there might be issues with clang vs the new 1.58 release... In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/system/file_monitor/MacFileMonitor.cpp:27: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/Thread.hpp:25: In file included from /sw/src/fink.build/rstudio-desktop-0.98.1103-4/rstudio-0.98.1103/src/cpp/core/include/core/BoostThread.hpp:26: In file included from
Re: [Fink-devel] boost1.57 test packaging
I've updated boost1.58.info in my experimental to use the just released boost-1_58_0 from upstream. Please test and report back any problems. Upstream screwed up their (lack of) use of -install_name and c_v even more, so there's way more games being played in the .info file now to make sure the libraries are all properly named and versioned. I think I got them all. http://cvs.snaggledworks.com/viewvc.cgi/fink/10.7/main/finkinfo/libs/ Hanspeter -- -- BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15utm_medium=emailutm_campaign=VA_SF ___ 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
Re: [Fink-devel] boost1.57 test packaging
Hanspeter, I see that you are limiting the build of boost 1.58 to just 10.9/10.10. The MacPorts boost 5.18 packaging uses the patch at... https://trac.macports.org/browser/trunk/dports/devel/boost/files/patch-apple-clang-no-libcxx.diff with the comment... # error that occurs on OS X 10.7 and 10.8 due to the assumption # that if clang is the compiler in use it must be using libc++. # Apple Clang uses libstdc++ by default on these OS versions. # The patch adds an additional BOOST_* configuration flag # that is set if Apple clang is being used but libc++ # is not. This flag is then used to prevent boost or a # dependent package from using functions such as std::forward that # are only available in libc++. Fixes build of libcdr on these # OS versions without affecting build on 10.6 and less (where clang is not # the default compiler) or 10.9 and up (where libc++ is the default). It may be worth supporting 10.7/10.8 with this patch provided that the test case from https://svn.boost.org/trac/boost/ticket/10785 compiles okay under the clang++/libstdc++ combination. The bug report is silent about which c++ library was used for the tested but I assume it was likely libc++. Hopefully the bug is libc++-specific, otherwise there isn't much reason to add boost 1.58 for 10.7/10.8 since we don't have the option of the -std=c++11 workaround with libstdc++. Jack On Mon, Apr 20, 2015 at 10:36 AM, Hanspeter Niederstrasser f...@snaggledworks.com wrote: I've updated boost1.58.info in my experimental to use the just released boost-1_58_0 from upstream. Please test and report back any problems. Upstream screwed up their (lack of) use of -install_name and c_v even more, so there's way more games being played in the .info file now to make sure the libraries are all properly named and versioned. I think I got them all. http://cvs.snaggledworks.com/viewvc.cgi/fink/10.7/main/finkinfo/libs/ Hanspeter -- -- BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15utm_medium=emailutm_campaign=VA_SF ___ 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
Re: [Fink-devel] boost1.57 test packaging
Hanspeter, I verified that the boost 1.58 packaging does indeed build on MacPorts under darwin12 (but my first attempts to recreate this from your info file and patch file including their patch and suppression of -std=cxx=+ fails on darwin12 under fink). In any case, using the MacPorts boost 1.58, I was able to confirm that https://svn.boost.org/trac/boost/ticket/10785 also exists for the darwin12 build of boost 1.58 against libstdc++ and that -std=c++11 has to be invoked to suppress it. Since obviously we don't have functional c++11 support with libstdc++, the boost 1.58 build on darwin12 and earlier will be crippled. Jack On Mon, Apr 20, 2015 at 12:32 PM, Jack Howarth howarth.at.f...@gmail.com wrote: Hanspeter, I see that you are limiting the build of boost 1.58 to just 10.9/10.10. The MacPorts boost 5.18 packaging uses the patch at... https://trac.macports.org/browser/trunk/dports/devel/boost/files/patch-apple-clang-no-libcxx.diff with the comment... # error that occurs on OS X 10.7 and 10.8 due to the assumption # that if clang is the compiler in use it must be using libc++. # Apple Clang uses libstdc++ by default on these OS versions. # The patch adds an additional BOOST_* configuration flag # that is set if Apple clang is being used but libc++ # is not. This flag is then used to prevent boost or a # dependent package from using functions such as std::forward that # are only available in libc++. Fixes build of libcdr on these # OS versions without affecting build on 10.6 and less (where clang is not # the default compiler) or 10.9 and up (where libc++ is the default). It may be worth supporting 10.7/10.8 with this patch provided that the test case from https://svn.boost.org/trac/boost/ticket/10785 compiles okay under the clang++/libstdc++ combination. The bug report is silent about which c++ library was used for the tested but I assume it was likely libc++. Hopefully the bug is libc++-specific, otherwise there isn't much reason to add boost 1.58 for 10.7/10.8 since we don't have the option of the -std=c++11 workaround with libstdc++. Jack On Mon, Apr 20, 2015 at 10:36 AM, Hanspeter Niederstrasser f...@snaggledworks.com wrote: I've updated boost1.58.info in my experimental to use the just released boost-1_58_0 from upstream. Please test and report back any problems. Upstream screwed up their (lack of) use of -install_name and c_v even more, so there's way more games being played in the .info file now to make sure the libraries are all properly named and versioned. I think I got them all. http://cvs.snaggledworks.com/viewvc.cgi/fink/10.7/main/finkinfo/libs/ Hanspeter -- boost1.58-nolibcxx.patch Description: Binary data -- BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15utm_medium=emailutm_campaign=VA_SF___ 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
Re: [Fink-devel] boost1.57 test packaging
On Mon, April 20, 2015 2:32 pm, Jack Howarth wrote: Hanspeter, I verified that the boost 1.58 packaging does indeed build on MacPorts under darwin12 (but my first attempts to recreate this from your info file and patch file including their patch and suppression of -std=cxx=+ fails on darwin12 under fink). In any case, using the MacPorts boost 1.58, I was able to confirm that https://svn.boost.org/trac/boost/ticket/10785 also exists for the darwin12 build of boost 1.58 against libstdc++ and that -std=c++11 has to be invoked to suppress it. Since obviously we don't have functional c++11 support with libstdc++, the boost 1.58 build on darwin12 and earlier will be crippled. Jack I have no interest in maintaining boost 1.58 on Mountain Lion or earlier. Hanspeter -- -- BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15utm_medium=emailutm_campaign=VA_SF ___ 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