On Fri, 2010-02-26 at 12:37 -0800, Matt Helsley wrote: > [ Depends on recent series of non-RFC Makefile patches ] > > SUBARCH is gross. We shouldn't need to detect the arch to compile to > using uname. It's bad practice because we could be using a cross compiler, > or someone may (gag) autoconfiscate this tree. If we must select a true > subarch, we should just let builders set -mXX in CFLAGS. > > This patch tries to build all of the clone_*.[cSs] files. Any builds > that fail are quietly ignored and don't get used. One that succeeds > gets hardlinked to clone.a and subsequently linked into libeclone.a Hmm, I think detecting the architecture via uname is fine as long as the user is able to override it (make SUBARCH=foo). I don't think trying to compile all of clone_* is going to be a good practice. For one thing, gcc -m64 on powerpc happily accepted the 32-bit clone wrapper the last time I tried it, hence the #error directives in clone_ppc*.S. _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
