From owner-freebsd-current Wed Oct 2 12:34:30 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id MAA00644 for current-outgoing; Wed, 2 Oct 1996 12:34:30 -0700 (PDT) Received: from spinner.DIALix.COM (root@spinner.DIALix.COM [192.203.228.67]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id MAA00633 for ; Wed, 2 Oct 1996 12:34:25 -0700 (PDT) Received: from spinner.DIALix.COM (peter@localhost.DIALix.oz.au [127.0.0.1]) by spinner.DIALix.COM (8.8.0/8.8.0) with ESMTP id DAA00393 for ; Thu, 3 Oct 1996 03:34:21 +0800 (WST) Message-Id: <199610021934.DAA00393@spinner.DIALix.COM> To: current@freebsd.org Subject: ETA for libg++-2.7.2 between 12 and 24 hours Date: Thu, 03 Oct 1996 03:34:20 +0800 From: Peter Wemm Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk I think I've finally finished it. I've not tested many canonical C++ programs to make sure they link, but it pretty well builds OK now. Some "highlights" of the experience: - the exception code in g++ appears very buggy. It generates invalid assembler code in -fpic mode... (how can you have something in the Global Offset Table(GOT) if it's not global?) - gas attempted to deal with this, but got it wrong. - ld dumped core when faced with gas's workaround. - bsd.lib.mk is driving me *insane*. libstdc++ needs a lot of different compiler options for different files, bsd.lib.mk doesn't handle this so you have to copy out the default rules three times over (one normal, one pic, one profiled) - the standard configure scripts seem to freak out when running with --srcdir (which I was using to seperate the sources from the test build I did to see how it worked), and since I was using the generated file lists and Makefiles from the output of configure I got burned by only being told about a subset of the files and had to start some of it again. - some of the libstdc++ files are compiled several times with different compiler options each time. Sigh. At least it compiles cleanly, unlike the old libg++ in the tree. BTW, does anybody use genclass? The old version in the tree was missing it, is it worth getting it going? I'll be doing this sometime tomorrow (my time) after I've had some sleep and tested a 'make world' (and implicitly, groff which is a C++ beast). It's pretty big.. I do not know that I can trim the disk space down too much without risking breaking the stand-alone compile: (ie: cd src/contrib/libg++; ./configure).. Perhaps the saved disk space would be worth it though. -Peter