libutil in Debian

Jilles Tjoelker jilles at stack.nl
Sat Jul 13 21:40:42 UTC 2013


On Thu, Jul 11, 2013 at 10:42:18AM +0200, Robert Millan wrote:
> 2013/7/9 Justin Hibbits <jhibbits at freebsd.org>:
> > I was thinking more in terms of adding the functions to the debian local
> > patch set.  I don't know how intrusive it would be, but it may be worth
> > looking into.

> You may not believe this, but it is even worse. Can you believe they
> even refused to add trivial syscall stubs, such as nlm_syscall()? They
> say this "belongs elsewhere" even though it's -lc in FreeBSD like most
> (all?) syscall stubs.

> Look at the kind of workarounds we have to endure:

> http://anonscm.debian.org/viewvc/glibc-bsd/trunk/freebsd-utils/debian/patches/036_nfs_glibc.diff?revision=4047&view=markup

glibc upstream appears indecisive about whether to add stubs for
special-use syscalls. There are more packages with their own stubs.

Note that FreeBSD libc has nlm_syscall() under the FBSDprivate_1.0
symbol version, reflecting the lower level of guarantees about its ABI
because it is only supposed to be used by rpc.lockd.

> Heck, even the trivial update to <sys/queue.h>, which was *already* of
> BSD origin since ancient times, was restricted to only apply on our
> port, so that the new macros (e.g. LIST_FOREACH_SAFE) were not
> available on Debian GNU/Linux.

I would say that a Debian patch is indeed not the first thing to try for
a <sys/queue.h> update, since it makes things incompatible between Linux
distributions. Rather, that should be done in glibc upstream. About
copyright assignment, note that the original file is already not
copyright FSF and was added to glibc anyway.

-- 
Jilles Tjoelker


More information about the freebsd-arch mailing list