On Sat 2015-05-23 04:37:40 -0400, Niels Möller wrote:
The issue is that there are some differences in generated nettle header files depending on architecture and compiler version.
If the differences are significant (e.g. if there are structs, functions, or #defines that are declared differently based on architecture and compiler version) then we can also just ship the .h files in /usr/include/${ARCH_TRIPLE}/nettle/ instead of /usr/include/nettle/. This can make the nettle-dev package co-installable across architectures. (e.g. nettle-dev:i386 could be installed alongside nettle-dev:amd64 without the files colliding, since they're at different places in the filesystem)
Take a look at the libgpg-error-dev package in debian, which ships its header files in the multi-arch include directories.
since nettle is using pkg-config instead of some kind of /usr/bin/nettle-config script, and pkg-config is already multi-arched, nettle-dev could do even better at being multi-arch compatible than libgpg-error-dev.
That said, if there are no significant differences in any of the architectures that debian supports, i don't think that embedding the exact version of the compiler used to generate the .h files is actually useful, and it would be great to reduce superfluous variance.
--dkg