Created attachment 249234 [details] Patch for libdjinterop * Backport upstream patch deeb55cc8cf9abed30629507565343e4dbd32e97 to remove hardcoding of -stdlib=libc++ * Backport upstream patch to remove hardcoding of -O3 optimization Porters Handbook, https://docs.freebsd.org/en/books/porters-handbook/book/#dads-cflags * Define Boost library as a dependency for "make test" * Sort USES= and add missing USES= pathfix (reported by framework) * Add USE_LDCONFIG * Disable build of unit tests unless requested even if Boost libs are available Unit tests still fails, is that intentional? If not please put a note somewhere about in commit msg or Makefile and report upstream
Comment on attachment 249234 [details] Patch for libdjinterop Unit tests failing are intentional. Upstream don't even run them properly (as in, at all) in their own CI, so their "passes" are illusions. I'm not moving the boost dependency out of BUILD_DEPENDS. CMakeLists.txt is unfortunately specifically structured such that test programs are only built and available when boost is available at configure time, and it's not worth the extra burden to hack our Makefile to avoid it.
At least when I briefly tested it on my dev box it seemed to pick up boost during do-test, it might be a side-effect due to how do-test is defined with USES= cmake:testing is used. I haven't looked at upstreams Makefile though if it works by accident.