Re: [David Madore <madore@quatramaran.ens.fr>] libc/1638: O_NOLINK documented but not implemented

[David Madore <madore at clipper dot ens dot fr>]

On Thu, Mar 09, 2000 at 11:16:58AM -0500, Thomas Bushnell, BSG wrote:
> O_NOLINK and O_NOFOLLOW are different on the Hurd;
> libc/sysdeps/mach/hurd/fcntl.h says:
> 
> # define O_NOLINK	0x0040	/* No name mappings on final component.  */
> # define O_NOFOLLOW	0x00100000 /* Produce ENOENT if file is a symlink.  */
> 
> which are notably different.

IMHO, both should be documented, then.  I didn't find any trace of
O_NOFOLLOW in the libc doc.

> The original "bug report" does not sound like an actual confused user,
> however, but someone trolling for a political fight.  

Sorry, you lose (or, if this was some kind of meta-troll, maybe you
win, but I don't commend meta-trolls).  In the first place, I have no
objection to the name ``GNU/Linux'', quite the contrary.  Second, I
don't have time, want or patience for politics, let alone political
fights.  (A political troller would have written something like ``The
FSF considers that the Linux OS is in fact the GNU/Linux system.'':
can you spot the difference?)

And I *was* confused.  I looked up the behavior under OpenBSD and
Solaris, and found that when O_EXCL|O_CREAT is used, symlinks are not
followed.  I wanted to know what was the case under GNU/Linux: knowing
that the man pages are deprecated and that, indeed, I have always
found the libc info doc considerably more reliable and well-written
than the man pages, I looked up the glibc doc, which mentioned only
O_NOLINK and O_NOFOLLOW.  So I tried and found no trace of O_NOLINK.

If you think that I have nothing better to do than submit a bogus bug
report for the mere pleasure of trolling about names like some
immature hacker wannabee, I would be justified in finding this
offending.  But I won't, because there's no point in being offended
over this.  I will merely point out that there's no point in having a
bug report system if people who report bugs get isulted.  (All right,
*that* sentence is a troll.)


On Thu, Mar 09, 2000 at 11:19:00AM -0500, Thomas Bushnell, BSG wrote:
> The SuSE, Debian, Red Hat, et. al., systems are GNU/Linux systems.

I agree.

> They are GNU systems in which one component has been replaced with the
> Linux kernel.  Making that replacement means that some of the features
> of the GNU system are no longer available.  

I also agree.  But I fail to see in what way it is relevant to the
question under discussion.  The documentation is the documentation for
the libc, not for the system.  It should document the way the glibc
works on the systems on which the glibc works, shouldn't it?


To make everyone happy, how about writing something like ``On the GNU
system (running on the original GNU kernel, Hurd, that is), the
following flags are defined [...].  On the GNU/Linux system, out of
the flags described above, O_NOFOLLOW is defined but the others
aren't.''?

-- 
     David A. Madore
    (david.madore@ens.fr,
     http://www.eleves.ens.fr:8080/home/madore/ )