Possible macOS ABI breakage in boost-python when boost-python is compiled with clang 13 and downstream packages with clang 14? about boost-feedstock HOT 13 CLOSED

I believe I hit this too in the vigra feedstock.

To avoid this, I've recompiled the new boost version 1.80 with the new compilers just now. Maybe we can start a new migration "soon"

from boost-feedstock.

This is boost, and not boost-cpp.
I've tried so far here with EigenPy, and it does not work apparently: conda-forge/eigenpy-feedstock#102

from boost-feedstock.

Sent PRs to rebuild 1.80 & 1.78:

• conda-forge/boost-cpp-feedstock#129
• conda-forge/boost-cpp-feedstock#130

from boost-feedstock.

Moved this issue from the obsolete boost-cpp feedstock to here.

from boost-feedstock.

Since this was about a change between clang 13 -> 14 (possible ABI break...?), and we're now on clang 16, I'm going to close this issue as obsolete. Please let us know if there are still outstanding problems!

from boost-feedstock.

We've rebuilt a few versions of boost several times very recently. We've also migrated to a more recent version of boost. Think this issue is likely fixed, but it would be great if someone could check 🙂

from boost-feedstock.

We've rebuilt a few versions of boost several times very recently. We've also migrated to a more recent version of boost. Think this issue is likely fixed, but it would be great if someone could check 🙂

@jcarpent it seems that also boost 1.78 was rebuilt. Probably it could make sense to remove the workaround of https://github.com/conda-forge/pinocchio-feedstock/blob/fca148b434c087b0e4956044bb94962e84480f1d/recipe/conda_build_config.yaml#L4 from all the libraries in which we added it? In any case, I opened a fake PR in pinocchio to check if pinocchio built with clang 13 is working fine with boost-python built with clang 14 (I doubt it): conda-forge/pinocchio-feedstock#84 .

from boost-feedstock.

It might not work as boost-cpp 1.78 has not been rebuilt since last May

from boost-feedstock.

Thanks @traversaro for investigating it. It might be worth testing first with eigenpy as it is at the root of the building tree.

from boost-feedstock.

It might not work as boost-cpp 1.78 has not been rebuilt since last May

I may be missing something but I think #149 rebuilt it.

from boost-feedstock.

It might not work as boost-cpp 1.78 has not been rebuilt since last May

I may be missing something but I think conda-forge/boost-feedstock#149 rebuilt it.

That rebuilt Boost.Python, which is what the boost package is. Since it sounds like this is about boost-cpp, have transferred to that feedstock.

from boost-feedstock.

@isuruf asked for additional investigation into this comment ( conda-forge/boost-cpp-feedstock#130 (comment) ). @traversaro, is that something you have time to do? Should add no worries if not

from boost-feedstock.

@isuruf asked for additional investigation into this comment ( #130 (comment) ). @traversaro, is that something you have time to do? Should add no worries if not

I am out of the lab until the next 9th of November, so I can't reliably investigate that at leat until that date.

from boost-feedstock.