Peter Dimov wrote:
The tests for sched_yield and nanosleep now fail as expected on g++/Sun. :-) clock_gettime also fails to link across the board.
I think that we ought to fix this by adding -lrt even in <threading>single instead of #undef-ing the feature macros, since these calls aren't technically tied to threads.
How about this one?
These aren't really threading API's, so I guess they're either present or not right? In which case adding -lrt on platforms that require seems to be the right thing to do, the following patch should do the job, but I don't have a machine handy to test it on, can you check this on Linux at least? Cheers, John. Index: build/v2/tools/gcc.jam =================================================================== --- build/v2/tools/gcc.jam (revision 44056) +++ build/v2/tools/gcc.jam (working copy) @@ -712,7 +712,7 @@ case SunOS* : { flags gcc OPTIONS <threading>multi : -pthreads ; - flags gcc FINDLIBS-SA <threading>multi : rt ; + flags gcc FINDLIBS-SA : rt ; } case BeOS : { @@ -740,7 +740,7 @@ case * : { flags gcc OPTIONS <threading>multi : -pthread ; - flags gcc FINDLIBS-SA <threading>multi : rt ; + flags gcc FINDLIBS-SA : rt ; } } }