From owner-freebsd-hackers Wed Oct 25 12:54:30 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id MAA09453 for hackers-outgoing; Wed, 25 Oct 1995 12:54:30 -0700 Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id MAA09442 for ; Wed, 25 Oct 1995 12:54:26 -0700 Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id MAA19300; Wed, 25 Oct 1995 12:46:01 -0700 From: Terry Lambert Message-Id: <199510251946.MAA19300@phaeton.artisoft.com> Subject: Re: Netscape puzzle To: peter@haywire.dialix.com (Peter Wemm) Date: Wed, 25 Oct 1995 12:46:00 -0700 (MST) Cc: freebsd-hackers@FreeBSD.ORG In-Reply-To: <46l1tu$km$1@haywire.DIALix.COM> from "Peter Wemm" at Oct 25, 95 06:01:34 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 1879 Sender: owner-hackers@FreeBSD.ORG Precedence: bulk > >> > We are all aware of the bind/sendmail static data initialization shared > >> > library changes, right? > >> > >> Refreh my memory. > > >They changed the initializations to something that would allow copy on > >write shared library data to be used as well. > > >Basically, it's getting rid of local data for crappy shared library > >implementations. > > Uhh... Am I missing something? This thread is about netscape, yes? > Netscape is linked *static*, and has it's own self-contained resolver. > Shared libraries dont even come into the picture, because BSDI (1.1) > doesn't *have* them... Yeah. You're missing that it introduced a general bind incompatability that you must code around to make it work, and perhaps the NetScape people didn't code around it. You can hack it by modifying the default contents for the statically initialized data (basically, blowing instructions in the binary), or by hacking the bind check call. > Our implementation of the resolver is not an issue with netscape, > because our resolver is not loaded into netscape. Sendmail had the > problem, yes. The implementation of the resolver to which the NetScape binary is statically linked is the issue. > If you mean the RES_INIT and res_init() patch, that's already in > sendmail, but isn't a 'netscape puzzle' issue... From memory > somebody's already mentioned that there is an issue of the resolver > code inside the statically linked netscape/bsdi executable using > select for a timeout, while netscape is using setitimer which is > disturbing the method that the resolver is using to detect a timeout > on the select. (or something like that.. :-) I mentioned it -- the system call restart issue on the select is only one possibility. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.