Klemens Nanni <k...@openbsd.org> writes: > On Tue, Feb 08, 2022 at 01:11:23AM +0300, Andrew Krasavin wrote: >> I apologize for the long answer, it's been quite a busy day.
sorry for the late reply, building the packages too longer than what i thought >> > Another thing, upstream won't release further versions and suggest to >> > build from the latest commit instead... so I'd fetch from a specific >> > commit instead of keeping a long list of PATCHFILES. >> > >> > cf https://github.com/google/googletest/releases/tag/release-1.11.0 >> > >> > > GoogleTest will not accept patches of new features to v1.11.0. >> > > We recommend building GoogleTest from the latest commit instead. >> > > Exceptional critical bug fixes may be considered. >> > >> > FWIW i had the following diff locally (which has to be updated to the >> > latest commit) and I can confirm that devel/msgpack and proj still >> > builds fine with this and test are passing. >> >> Thanks for testing that! >> Yes, your version of the port that uses the actual gtest commit is >> much better than the one I suggested. I would be glad if such changes >> would be integrated into the port tree. >> > >> > I'll update it to 14aa11db02d9851d957f93ef9fddb110c1aafdc6 (latest >> > commit as of now), cherry pick you fix-gtest-help-test (which is the >> > only test failing for me here), do a mini bulk overnight and report back >> > how it goes :) >> > >> >> Please clarify about the fix-gtest-help-test. I had this test failed >> without this patch. According to the code, I have enabled the same >> logic for OpenBSD as was already done for Linux, >> GNU/kFreeBSD, and hurd. >> >> If you get an error during the test after applying my patch then >> please tell me more about it. Maybe I should cancel my pull request >> before it is already merged. sorry for the confusion, I meant to say that your pr fixes the issue and that _without_ it the help-test was the only failing one. Thanks for fixing it :) > Your patch works for me on amd64 and sparc64 and is required to get 100% > tests working. > > Here's the diff I just tried. > It also records the missing python dependency and zaps SEPARATE_BUILD > (default with cmake). > base-gcc is too old for recent c++, so zap that (not used anyway). agreed on all changes I just finished a mini bulk build with `show-reverse-deps devel/gtest` in a clean proot and successfully built 1872 packages, so I'm quite sure it won't brings things down in real bulks. if it helps, I've uploaded the logs here. https://tmp.omarpolo.com/amd64/ only graphics/openscenegraph failed repeatedly to build, but it doesn't seem to be related to gtest? https://tmp.omarpolo.com/amd64/paths/graphics/openscenegraph.log Note that I've tested with the previous commit (14aa11d), but the next one (the one in your diff) shouldn't bring in differences: https://github.com/google/googletest/commit/43efa0a4efd40c78b9210d15373112081899a97c so OK for gtest as long as you don't forget to fix msgpack too :) (a revision bump shouldn't be needed in this case, since gtest is bumping SHARED_LIBS too, but maybe to play on the safe side a bump is required?) thanks Andrew for the msgpack fix Index: Makefile =================================================================== RCS file: /home/cvs/ports/devel/msgpack/Makefile,v retrieving revision 1.12 diff -u -p -r1.12 Makefile --- Makefile 14 Feb 2020 11:11:32 -0000 1.12 +++ Makefile 5 Feb 2022 21:30:27 -0000 @@ -23,6 +23,8 @@ MODULES = devel/cmake BUILD_DEPENDS = devel/gtest TEST_DEPENDS = devel/gtest +CONFIGURE_ARGS += -DCMAKE_CXX_STANDARD=17 + pre-configure: cd ${WRKSRC} && sed -i 's,-Werror -g -O3,,' CMakeLists.txt \ example/c/CMakeLists.txt \
