Subject: Re: builtin.mk for bind9
To: Jeremy C. Reed <firstname.lastname@example.org>
From: Mark Davies <email@example.com>
Date: 07/18/2004 12:02:52
On Sunday 18 July 2004 11:08, Jeremy C. Reed wrote:
> > There aren't any additional header files/libraries but as I said in the
> > original message:
> I don't understand. The other software needs some "threadsafe resolver
> routines" but they aren't provided any library.
No they aren't provided by any _additional_ library/header than what was used
by earlier versions of the resolver, ie libc or libresolv and resolv.h and I
cant see any UNIX system not having those so testing their existance seems
> All I am saying is that the builtin.mk should check that the library
> exists instead of assuming there is a library if just named is the
> correct version. (Because maybe a system using pkgsrc has a
> /usr/sbin/named that doesn't use that library, i.e. static, or the library
> can't be found by the pkgsrc because in some odd spot.)
I'll concede that that is a possibility (and did in the original message) but
does it happen in reality? It would certainly be better to do the test based
on the contents of resolv.h but I don't know any easy way to get from that to
a particular bind version number.
> But the recent cvs comment said "our base OS", so maybe should check for
> "NetBSD" too and then assume.
Actually it said "the base OS" to allow for other than NetBSD.