From owner-freebsd-hackers Fri May 17 07:35:27 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id HAA20746 for hackers-outgoing; Fri, 17 May 1996 07:35:27 -0700 (PDT) Received: from rocky.sri.MT.net (rocky.sri.MT.net [204.182.243.10]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id HAA20738 for ; Fri, 17 May 1996 07:35:21 -0700 (PDT) Received: (from nate@localhost) by rocky.sri.MT.net (8.6.12/8.6.12) id IAA27102; Fri, 17 May 1996 08:35:15 -0600 Date: Fri, 17 May 1996 08:35:15 -0600 From: Nate Williams Message-Id: <199605171435.IAA27102@rocky.sri.MT.net> To: "Michael L. VanLoon -- HeadCandy.com" Cc: Peter Mutsaers , freebsd-hackers@freebsd.org Subject: Re: internal compiler error In-Reply-To: <199605170751.AAA16653@MindBender.HeadCandy.com> References: <87loisb709.fsf@plm.simplex.nl> <199605170751.AAA16653@MindBender.HeadCandy.com> Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk > ago. GCC 1.4.5 was used by NetBSD looong after GCC 2 was released. Until PK's shlib scheme, yes. NetBSD actually completely switched to gcc2 at that point, where some folks in the FreeBSD camp (ie; me) whined badly enough that FreeBSD had two compilers GCC1 and GCC2 for quite a while, but I finally gave up since maintenance was a nightware. > It was very quick, easy on memory, and very stable (read not very > buggy). GCC 2.x.x in the 2.5.x time-frame was a buggy mess, from the > many accounts I've heard. But the shlib hacks didn't work in GCC 1. > The two biggest reasons they finally switched to 2.x.x were because 1) > GCC 2.7 finally fixed enough of the bugs that it was deemed usable, > and some of the non-i386 ports (like the 68K) didn't produce very good > code in 1.4.5, and 2) there were just some features they needed to > move forward (many of them having to do with obscure non-Intel > processors, from what I remember). The *biggest* reason was the shlibs worked in GCC 2 and not in GCC 1. Now, the above reasons may be why they switched from Gcc 2.5.3 -> 2.7.2, since the GCC in NetBSD was 2.5.3 until recently. Did NetBSD *ever* run any variant of gcc 2.6? > The core group has said more than once, however, that if they could > find another ANSI-compliant compiler, source-distributable, not under > the GPL (well, under a Berkeley-style license), that was much smaller > and simpler, and worked well with all the architectures, they'd jump > to it without a complaint. Lcc is one that has come up several times, > but has been deemed "not quite ready, yet". It has been deemed 'not using a valid copyright', and to top it off the shlib scheme used in both camps won't work with it. :( Nate