Wireshark-dev: Re: [Wireshark-dev] Use of MAP_ANON in epan/emem.c
From: Jaap Keuter <jaap.keuter@xxxxxxxxx>
Date: Sun, 18 Jun 2006 08:19:57 +0200 (CEST)
Hi,

Oh I love those UNIX variants.... Yet another compatibility layer
added :(
In this respect would it be wise to have a *BSD buildbot?

Thanx,
Jaap

On Fri, 16 Jun 2006, Guy Harris wrote:

>
> On Jun 16, 2006, at 6:11 PM, Guy Harris wrote:
>
> > The IRIX 6.5 man page for /dev/zero indicates you can mmap it to get
> > anonymous pages.  Are there any UN*Xes we support that *don't* allow
> > you to mmap /dev/zero to get anonymous pages?
>
> Yes - it appears that FreeBSD and OS X don't.
>
> > if so, we might want to use MAP_ANONYMOUS on systems that
> > define it, otherwise use MAP_ANON on systems that define it (OS X
> > 10.4 defines MAP_ANON but *NOT* MAP_ANONYMOUS, and that seems to be a
> > general BSD characteristic - I don't find any definitions of
> > MAP_ANONYMOUS in the sys/sys source directory of fairly recent
> > {Free,Net,Open}BSD), otherwise use "/dev/zero".
>
> I've checked in a change to do that.  It allows Wireshark to build on
> OS X (the change of MAP_ANON to MAP_ANONYMOUS probably broke
> {Free,Open,Net,DragonFly}BSD builds as well), and it includes
> untested code to handle systems that have neither MAP_ANONYMOUS nor
> MAP_ANON - it assumes they have /dev/zero, opens that, and maps it.