/usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/g++-v11/bits/std_function.h:435:145: error: parameter packs not expanded with ‘...’: 435 | function(_Functor&& __f) | ^ /usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/g++-v11/bits/std_function.h:435:145: note: ‘_ArgTypes’ /usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/g++-v11/bits/std_function.h:530:146: error: parameter packs not expanded with ‘...’: 530 | operator=(_Functor&& __f) | ^ /usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/g++-v11/bits/std_function.h:530:146: note: ‘_ArgTypes’ Apparently this specific problem was fixed with >=dev-util/nvidia-cuda-toolkit-1.6.0 however tensorflow-2.8.0 needs cuda 1.5. Changing CC and CXX via package.env results in nvcc not finding compiler (e.g. setting it to gcc-10.3.0 simply appends that to /usr/x86_64-pc-linux-gnu/gcc-bin/11.3.0/). Tensorflow also ignores eselect gcc. Reproducible: Always A workaround so far seems to be to leave eselect gcc as it is, set CC/CXX in package.env for tensorflow to g{cc,++}-10.3.0, soft link g{cc,++}-10.3.0 into the 11.3.0 folder (yes, I know) and recompile dev-libs/protobuf, dev-cpp/abseil-cpp and dev-libs/jsoncpp with gcc10 to avoid linker errors.
Created attachment 782117 [details] emerge --info
Created attachment 782120 [details] tensorflow build.log
Same problem
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=dcba40cc37f24b723915d1c0492aabbdb3bb597f commit dcba40cc37f24b723915d1c0492aabbdb3bb597f Author: Jason Zaman <perfinion@gentoo.org> AuthorDate: 2022-06-25 19:18:04 +0000 Commit: Jason Zaman <perfinion@gentoo.org> CommitDate: 2022-06-26 02:37:38 +0000 sci-libs/tensorflow: add 2.9.0 Closes: https://bugs.gentoo.org/849101 Signed-off-by: Jason Zaman <perfinion@gentoo.org> sci-libs/tensorflow/Manifest | 14 + sci-libs/tensorflow/tensorflow-2.9.0.ebuild | 418 ++++++++++++++++++++++++++++ 2 files changed, 432 insertions(+)