Fix endianness traits

[rupertnash]

Hi @SLongshaw

I saw that your reverted the endian traits patch in #14 . I'd like to get this accepted because without it I can't compile MUI on my laptop.

Can you please give me a bug report? Then I have a chance to fix it!

I tested initially with clang (Apple 10.0 / upstream 6) and gcc/6.3.0, but have just tried on Cirrus (http://www.cirrus.ac.uk) with all the GCC versions available and not had problems. (gcc/6.2.0 gcc/6.3.0(default) gcc/7.2.0 gcc/8.2.0)

Cheers

[SLongshaw]

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.

[rupertnash]

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

[SLongshaw]

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.