Re: kern/46463: netbsd-6: panic and filesystem corruption running tmux

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

The following reply was made to PR kern/46463; it has been noted by GNATS.

From: christos%zoulas.com@localhost (Christos Zoulas)
To: gnats-bugs%NetBSD.org@localhost, martin%NetBSD.org@localhost, 
gnats-admin%netbsd.org@localhost, 
        netbsd-bugs%netbsd.org@localhost, rhansen%bbn.com@localhost
Cc: 
Subject: Re: kern/46463: netbsd-6:  panic and filesystem corruption running tmux
Date: Thu, 31 May 2012 11:28:18 -0400

 On May 31,  3:10pm, martin%duskware.de@localhost (Martin Husemann) wrote:
 -- Subject: Re: kern/46463: netbsd-6:  panic and filesystem corruption runnin
 
 |  On Thu, May 31, 2012 at 02:50:06PM +0000, Christos Zoulas wrote:
 |  >  Yes, and we can remove the close-on-fork part too I guess.
 |  
 |  I'm not sure we should do that by default, but we could add a flag for
 |  kqueue1() to make it inheritable. Otherwise it might be a suprise with
 |  unclear consequences on existing apps (like now suddenly running into
 |  RLIMIT_NOFILE).
 
 I don't think that this will happen in reality; there are very few programs
 using kqueue, and I think that most people don't treat the kqueue file
 descriptors specially. I.e. if you ask jill-random if kqueue descriptors are
 inherited on fork, she will probably say yes.
 
 |  However, is this realy worth the effort? All other OSes seem to not inherit
 |  this descriptors as well.
 
 I don't know. I just dislike the special treatment. I think that the power
 of unix comes from the fact that file-descriptors have all treated equally
 for common operations. Yes, I know we have added all kinds of warts through
 the years, but...
 
 christos