Fix endianness traits about mui HOT 3 CLOSED

Apologies, a bug report should have been included before closing.

Including this patch broke an existing implementation of MUI + OpenFOAM, compiled with GCC 7.3.0 on a Debian based system. The compilation error is due to there being multiple definitions of specialisations found in endian_traits.h.

e.g.:

"In function mui::detail::endian_converter<2ul>::htobe()': (.text+0x0): multiple definition of mui::detail::endian_converter<2ul>::htobe()'
(.text+0x0): first defined here
In function mui::detail::endian_converter<2ul>::betoh()': (.text+0x10): multiple definition of mui::detail::endian_converter<2ul>::betoh()'
(.text+0x10): first defined here
In function mui::detail::endian_converter<4ul>::htobe()': (.text+0x20): multiple definition of mui::detail::endian_converter<4ul>::htobe()'
(.text+0x20): first defined here
In function mui::detail::endian_converter<4ul>::betoh()': (.text+0x30): multiple definition of mui::detail::endian_converter<4ul>::betoh()'
(.text+0x30): first defined here
In function mui::detail::endian_converter<8ul>::htobe()': (.text+0x40): multiple definition of mui::detail::endian_converter<8ul>::htobe()'
(.text+0x40): first defined here
In function mui::detail::endian_converter<8ul>::betoh()': (.text+0x50): multiple definition of mui::detail::endian_converter<8ul>::betoh()'
(.text+0x50): first defined here....."

I suspect this is down to the fact that MUI is embedded as a class that is then utilised by a good majority of other OpenFOAM classes, rather than the more traditional use-case of simply including mui.h at the top of a not heavily-OO design. However, as the original code worked without error on the same system, I reverted the changes to avoid it being broken for any other user of the master branch.

More details will be provided in any future bug reports.

from mui.

Thanks Stephen! With that clue I reproduced the issue in LAMMPS (using MUI in main() and in a custom fix for the communication ). I'd forgotten that explicit specialisations are not implicitly inline. I've got a fix on my traits branch - I'll make a PR

from mui.

I have integrated the fix and all seems well this time, so I'll leave the patch applied unless somebody else comes across a problem. I'm going to close this bug for now.

from mui.