On Sat, Aug 31, 2013 at 03:47:53PM +0100, Ken Moffat wrote:

 Sorry, this is becoming O/T for LFS itself.

>  Thanks for that confirmation.  I don't install reiser, and I hadn't
> installed raptor at that point,  Don't want to return to the rc1
> system on this box at the moment (just started building LO after
> what I hope is the last error in my "select many languages" code),
> but I've mounted rc1 at /mnt/lfs and I'm surprised to see:
> 
> root@jtm1 ~# ls -ld /mnt/lfs/usr/share/man/
> drwxr-sr-x 38 root root 4096 Aug 27 18:14 /mnt/lfs/usr/share/man/
> 
>  Suid for group (with a group of root) ?  Will need to check the
 sgid of course.
> i686.  The time suggests faad2, libsamplerate, or libupnp (or
> perhaps something at the end of fftw).  I've added some
> instrumentation around those packages but not sure if it will work.
> 
 Weirder and weirder.  On the i686 -rc1 build, again it is sgid.
There the timings suggest faad or libupnp.

 On the current rc2 build I've now gone past that stage and perms
are still:
drwxr-xr-x 38 root root 4096 Aug 31 19:54 /usr/share/man

 but again, note that it was updated recently.  Here, the time
points to  faad2, libupnp, or else cdparanoia or x264.  The only
two of those which install man pages are cdparanoia and faad2.

 I now  suspect a race in faad2 (it was built with -j4 both times
on this box, and -j2 on the other box).

 However, I've now looked at the first part of my backups (I use a
two stage process - rsync to an area on the server, then copy that to
something approaching a set of generation data groups.  Anyway, I see
the sgid on my server itself, both for its current (7.2) and previous
builds, and on most semi-recent desktop builds - including one which
was originally on an older single-processor machine that I copied
over to get me started.  In that case, the date of /usr/share/man is
before I got the new machine, so it isn't a race during parallel
makes.

 Weird.  Of course the thing is that there is no problem with sgid -
a normal user can still read the pages.  I wonder if I'm misreading,
and this is something that happens in the background for 'man' ?
Updatedb is a script and doesn't do it.  I ran mandb but that didn't
alter it either.  Giving up, but will make a note to undo  sgid
whenever I notice it.

ĸen
-- 
das eine Mal als Tragödie, dieses Mal als Farce
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to