Re: lib/39684 (RB_REMOVE trashes link to tree, when requested to remove object that isn't in tree)

[christos%zoulas.com@localhost (Christos Zoulas)]

The following reply was made to PR lib/39684; it has been noted by GNATS.

From: christos%zoulas.com@localhost (Christos Zoulas)
To: gnats-bugs%NetBSD.org@localhost, lib-bug-people%netbsd.org@localhost, 
        gnats-admin%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost, 
        m-Matti-a.Lehtonen%IKI.Fi@localhost
Cc: 
Subject: Re: lib/39684 (RB_REMOVE trashes link to tree, when requested to 
remove object that isn't in tree)
Date: Tue, 7 Oct 2008 14:13:58 -0400

 On Oct 7, 11:50am, cube%cubidou.net@localhost (Quentin Garnier) wrote:
 -- Subject: Re: lib/39684 (RB_REMOVE trashes link to tree, when requested to 
 
 |  You know, the fact that you're the one letting your code logic do
 |  something that it should not do puts you in a very bad position to make
 |  such comments.  If code of yours runs into any of the devices you list,
 |  could you tell us brands and exact models?  I feel kind of worried.
 
 Well, this is really not productive. I think that the submitter here
 has a point. The manual page states:
 
      Accordingly, RB_REMOVE() and SPLAY_REMOVE() return the pointer to the
      removed element, otherwise they return NULL to indicate an error.
 
 Looking at the red-black code this is a lie. It never returns NULL and it
 should really either:
        1. not trash the tree or
        2. at least document that it will trash the tree
 
 christos