Hurd socket ioctls

[Mark Kettenis]

Last january Marcus posted a patch to revise the socket ioctls.  The
changes to the hurd package got applied, but the glibc part of this
patch was never integrated.  At the time Roland raised some objections
about adding interfaces to libhurduser before an RPC versioning scheme
is in place.  Is this really enough reason to keep us from applying
Marcus' patch?  His patch only adds a single new interface to
libhurduser.  This means that programs linked against this new
libhurduser that use this new interface won't run on older systems.
But I really don't think that there will be programs that will invoke
pfinet_siocgifconf() directly.  That would be a really stupid thing to
do when there is the if_nameindex() function and the SIOCGIFCONF
ioctl.  And even if there are programs that somehow pick up a
reference to pfinet_siocgifconf(), it would not be particularly
diffucult to find out what's the problem.

Moreover, the Debian glibc package has been including Marcus' patch
for quite some time now.  So there probably are quite a few
libhurdusers that include pfinet_siocgifconf.

So I'd vote for adding the changes to the official glibc too.  And yes
I volunteer to do the work :-).