Am 29.08.2015 um 10:59 schrieb Matthias Klose: > Package: src:spring > Version: 100.0+dfsg-1 > Severity: serious > Tags: sid stretch > > spring ftbfs in unstable,
[...] > /usr/include/c++/5/bits/atomic_base.h:355: error: undefined reference to > '__atomic_is_lock_free' > collect2: error: ld returned 1 exit status > test/CMakeFiles/test_ThreadPool.dir/build.make:443: recipe for target > 'test/test_ThreadPool' failed > make[5]: *** [test/test_ThreadPool] Error 1 Hi, apparently one test case fails due to a recent change in GCC 5 which requires to link against -latomic because test/engine/System/testThreadPool.cpp uses std::atomic<int> cnt(0); BOOST_CHECK(cnt.is_lock_free()); However there is an open GCC bug for this issue too. Judging from the bug log they intend to revert this change in GCC 5.3 or later. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65913 Quote: "Does that mean for all the future versions (5.2, 6.0, etc.) -latomic flag is necessary if atomic<T>::is_lock_free() is used in my program?" "From my reading of the comments, the patch and current svn sources, the answer is "no"." The patch is still pending but I wonder if a bug report should be filed against GCC-5 because we are dealing with a regression here. What is your stance on this? Regards, Markus
signature.asc
Description: OpenPGP digital signature