Subject: Re: several messages
To: matthew green , Christos Zoulas <>
From: Todd Vierling <>
List: current-users
Date: 09/30/1998 10:32:08
On Wed, 30 Sep 1998, matthew green wrote:

: i dunno. i recon that netbsd's madvise() should fail if it doesn't work..
: if the emulations rely on it's "success" (as you say), then they should
: be changed to ignore what the *real* madvise() does?

How can a program rely on its success?  It's an _advisory_ call.  All I did
was make the call return `yeah, we got the advice'.  madvise() isn't
guaranteed to do anything beneficial for the program -- that's what mlock()
and munlock() are for.

On Wed, 30 Sep 1998, Christos Zoulas wrote:

: Please revert this and fix it properly (add emulation specific stubs).

Not only is that painful, it's really pointless and adds extra stuff that
would have to be reverted were we to do something "real" with madvise().

Additionally, programs that call it _should_ succeed; we "take it under
advisement" (and currently promptly ignore the advice).  A madvise() call
is only advisory.

I can add "argument sanity checks" to make it return TRT on bad page
arguments, but--especially if we plan on supporting it--it should return
success if handed good arguments.

-- Todd Vierling (Personal; Bus.