On Thu, Sep 17, 2015 at 09:33:43AM +0200, Uwe Kleine-König wrote: > The problem is that only 2 of 3 Architecture lines contain the list of > supported architectures, there is still an arch-any line.
Yes I noticed. > The nicer fix would be to make libnuma provide stubs when the respective > syscalls are not there. > > BTW, on arm the culprit is: > > http://git.kernel.org/linus/d80ade7b323152672bf66e74ec11c324332f6d1e > > So something like: > > diff --git a/syscall.c b/syscall.c > index 4589b8547697..2d3a468f1ac1 100644 > --- a/syscall.c > +++ b/syscall.c > @@ -206,7 +206,12 @@ long WEAK set_mempolicy(int mode, const unsigned long > *nmask, > long WEAK migrate_pages(int pid, unsigned long maxnode, > const unsigned long *frommask, const unsigned long *tomask) > { > +#ifdef __NR_migrate_pages > return syscall(__NR_migrate_pages, pid, maxnode, frommask, tomask); > +#else > + errno = ENOSYS; > + return -1 > +#endif > } > > long WEAK move_pages(int pid, unsigned long count, > > should fix the build failure. > > I guess upstream doesn't care about non-linux architectures. To support > them probably the easiest would be to make syscall a stub setting errno > to ENOSYS and returning -1. Well I would love to see it stop failing to build at least. -- Len Sorensen
