On Sun, 2006-07-23 at 17:44 +0200, Enrico Scholz wrote: > rc040203@xxxxxxxxxx (Ralf Corsepius) writes: > > >> This will create a huge amount of variations: > >> > >> * soft-float/hard-float > >> * little/big-endianess > >> * cpu optimized libs (e.g. ARM XScale, EP9301, Thumb/non-thumb); multi-lib > >> support would be probably too much overkill for embedded platforms > > The contrary is true. Multilibs initially have been invented for > > embedded targets and have a long history there, predating using them on > > "non-embedded" OSes. > > I do not see sense for multilib here because binary packages must be > built per architecture (e.g. soft/hard-float are ABI incompatible, > kernel assumes a certain endianess, optimized programs should be used > on embedded platforms due to the limited resources). => multilib'ed kernels/OS runtime libs. > Enabling the multilib bits adds just unneeded complexity (both in > packaging, bootstrapping and performance+size aspects). Yes, they impact bootstrapping + size and toolchain size, but the rest isn't. >From a user's perspective, multilibs are very convenient. For example they enable you to switch the HW without having to recompile the toolchain/OS. All you need to recompile is your application with different flags. Ralf -- fedora-extras-list mailing list fedora-extras-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-extras-list
