Bump of required compiler versions following EOL of manylinux_2_24 - move to GCC 8?
Dear all I've been trying to keep https://scipy.github.io/devdocs/dev/toolchain.html up to date with changes around compiler constraints and corresponding C/C++ capabilities for a while already. Up until basically today, the hardest constraints were either Windows (due to ABI compatibility with CPython 2.7 / 3.{x<7}) or the compilers available in the various manylinux images. With the EOL of the Debian-based "outlier" manylinux_2_24 (in a string of otherwise RHEL-derived manylinuxes), we're now in the relatively new situation that all 3 main platforms have very recent & capable compilers. I've therefore tried to update the rationale to include more niche platforms and other relevant usage patterns in https://github.com/scipy/scipy/pull/16589, but the long story short is that - depending on how the various constraints are weighted - we could bump the minimum GCC version up to anywhere between 7-10 (inclusive). I'm proposing to move to GCC 8 as a lower bound. This would allow us to use C++17 and fix various issues e.g. around boost. For LLVM I'm not sure we want to do that, but for now I've included it as well (based on what we run in CI). Feedback about any aspect of that PR is very welcome! Best regards, H.
On 15/7/22 09:43, h.vetinari@gmx.com wrote:
Dear all
...
I'm proposing to move to GCC 8 as a lower bound. This would allow us to use C++17 and fix various issues e.g. around boost. For LLVM I'm not sure we want to do that, but for now I've included it as well (based on what we run in CI).
Best regards, H. _______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: matti.picus@gmail.com
I think it makes sense to move to gcc8. gcc8.1 was released May 2, 2018, over 4 years ago. Matti
On Sat, Jul 16, 2022 at 7:16 PM Matti Picus <matti.picus@gmail.com> wrote:
On 15/7/22 09:43, h.vetinari@gmx.com wrote:
Dear all
...
I'm proposing to move to GCC 8 as a lower bound. This would allow us to use C++17 and fix various issues e.g. around boost. For LLVM I'm not sure we want to do that, but for now I've included it as well (based on what we run in CI).
Best regards, H. _______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: matti.picus@gmail.com
I think it makes sense to move to gcc8. gcc8.1 was released May 2, 2018, over 4 years ago.
This sounds good to me too. I'm aiming to merge the PR to update the toolchain roadmap with this constraint soon. Cheers, Ralf
participants (3)
-
h.vetinari@gmx.com
-
Matti Picus
-
Ralf Gommers