So it seems that py-lief is using pybind11 for the python bindings. mamba is also using pybind11 to bind to libmamba.
However, mamba is using the pybind11 from conda-forge and I think that py-lief is either vendoring or ad-hoc downloading pybind11 because I couldn't find it in the dependencies of the meta.yaml.
There is a weird segfault happening in rare cases when using mambabuild on Windows (cf. https://github.com/mamba-org/boa/issues/235 and https://github.com/mamba-org/boa/issues/197) and I think it might be related to two different versions of pybind11. It fails during the packaging stage. I added a bunch of print statements to the conda-build code and I am pretty sure it's in the inspect linkages step.
I'll try to modify this recipe to patch out the internal pybind11 and use the proper conda-forge one + the pybind11_abi package.
So it seems that
py-lief
is usingpybind11
for the python bindings.mamba
is also usingpybind11
to bind tolibmamba
. However,mamba
is using thepybind11
fromconda-forge
and I think thatpy-lief
is either vendoring or ad-hoc downloadingpybind11
because I couldn't find it in the dependencies of the meta.yaml.There is a weird segfault happening in rare cases when using
mambabuild
on Windows (cf. https://github.com/mamba-org/boa/issues/235 and https://github.com/mamba-org/boa/issues/197) and I think it might be related to two different versions ofpybind11
. It fails during the packaging stage. I added a bunch ofprint
statements to the conda-build code and I am pretty sure it's in the inspect linkages step.I'll try to modify this recipe to patch out the internal pybind11 and use the proper conda-forge one + the
pybind11_abi
package.