From owner-freebsd-current Wed Jul 30 00:10:16 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id AAA14956 for current-outgoing; Wed, 30 Jul 1997 00:10:16 -0700 (PDT) Received: (from hsu@localhost) by hub.freebsd.org (8.8.5/8.8.5) id AAA14949 for current; Wed, 30 Jul 1997 00:10:13 -0700 (PDT) Date: Wed, 30 Jul 1997 00:10:13 -0700 (PDT) From: Jeffrey Hsu Message-Id: <199707300710.AAA14949@hub.freebsd.org> To: current Subject: Re: pccard and -current; a long way to go. :-( Sender: owner-freebsd-current@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > The people who recognize the benefits of moving to ELF are the ones > who need to do the PR. The question is not "would ELF be better?" > but rather "would ELF be _enough_ better to make it worth the pain > of the transition?" Focusing on the language tools as I have, I see > certain benefits: I work on a lot of language tools and the first thing I have to do is modify them to understand our ad-hoc a.out format. This argues strongly against using a non-standard binary format when the wealth of language tools and utilities are all ELF compatible. People argue that we don't need ELF because all the tools we have work fine with a.out. But, they are neglecting to consider all the tools we don't have because we don't use ELF. Now that we have more complete support in our system for ELF (I silently cheered when jdp committed the code in discussion), I intend to make full use of it. > * Our linker has major design faults in its handling of shared > libraries. They're probably never going to be fixed without > a rewrite of the whole thing or at least the shared library > code. The ELF linker doesn't have these faults. I see this problem too in conjunction with language tools. My next JDK port which I'm working on now (not sure if I should do a stable 1.1.4 port or just skip ahead to 1.2, neither of which the Linux people have yet) would happily use ELF if it solves the dlsym(RTLD_NEXT) problem (namely, we don't have it) and it also eliminates a bunch of kludges in the Java interpreter code that I have to make because we prepend underscores and the rest of the world doesn't.