On Thu, Jun 23, 2011 at 11:42:18PM +0200, Joerg Sonnenberger wrote: > > >> Does anyone feel like committing this (or fixing this funny > > >> behaviour in some other way)? Diff is appended below. > > > > > > OK, let's ask the other way 'round: > > > Any objections to committing the patch below? > > > > Assuming the -P option of pwd means what it does elsewhere, that > > meaning can be summarized as "please work properly" so that seems like > > a fine thing. > > I disagree with that assessment. The problem is that it is too general > and can break some natural things. E.g. > test $(pwd -P) = $(cd $(pwd -P)/foo && cd .. && pwd -P) > is not invariant any longer. That's not invariant anyway. There are plenty of systems where the pwd -P behavior is the default, such as NetBSD. The ksh behavior of lying to you via $PWD is a bug, plain and simple. > > not sure if it's a good idea during the freeze though. > > Certainly not. Well, as I recall it's wanted to fix build problems. -- David A. Holland dholland%netbsd.org@localhost