Date: Wed, 21 Oct 1998 22:47:13 +0800 From: Peter Wemm <peter@netplex.com.au> To: John-Mark Gurney <gurney_j@resnet.uoregon.edu> Cc: Stephen McKay <syssgm@dtir.qld.gov.au>, freebsd-hackers@FreeBSD.ORG Subject: Re: 3.0-R on a amd386dx/40 Message-ID: <199810211447.WAA16977@spinner.netplex.com.au> In-Reply-To: Your message of "Wed, 21 Oct 1998 02:38:48 MST." <19981021023848.61590@hydrogen.nike.efn.org>
next in thread | previous in thread | raw e-mail | index | archive | help
John-Mark Gurney wrote: > Stephen McKay scribbled this message on Oct 21: > > On Wednesday, 21st October 1998, John-Mark Gurney wrote: > > > > >Stephen McKay scribbled this message on Oct 21: > > >> Sounds like the FP emulator isn't working. What do the npx lines say > > >> on boot? > > > > > >well, it doesn't matter... as I told another person, the problem is > > >that the binaries FAIL on a 2.2.1-R system... and this is a system that > > >runs systat perfectly fine... so the problem isn't the floating point.. > > > > > >and the 3.0-R newfs fails on his new 2.2.6-R install that we just did > > >tonight... so I KNOW it's not a math emulation problem... > > > > Oh, I missed the "on 2.2.1" bit. Answer is simple: don't do that. > > > > The only scenario I can think of is that you've built 3.0 from source under > > 2.2.x and then run the 3.0 progs from the obj directories. Again, don't > > do that. > > > > Run 3.0 binaries on a 3.0 system and you'll have a better time of it. > > that's the WHOLE PROBLEM!!! 3.0 binaries on a 3.0-R kernel from the > 3.0-R bin dist don't run!!! > > I'll quote from another message that I sent about this problem: > no, this isn't an upgrade... what we were doing is that he was going to > switch to a new hard drive, but he has files on the old one... so we did > a manual install of 3.0-R from 2.2.1-R... once we had everything up and > running we built a kernel on my 3.0-R system and booted the system... it > failed when it tried to fsck the system... it would fail to newfs and df > would also fail... > > I then tried to run the newfs binary under 2.2.1-R and it would start > exectution perfectly fine, but would crash at the same point that it would > under 3.0-R... Yes, it seems the 3.0 ELF gas encodes some size override prefixes on some instructions that causes math_emulate.c (all the way back to the dark ages) to fall over. It's the new binaries that are tripping up the standard emulator. Damn, I wish the 3.0 floppies had a kld kernel. :-) We could have just told people to load the other emulator which worked (apart from the typo that prevented loading - but this would have been found in better testing). > yes, I know it seems strange to run 3.0-R binaries on such an old release, > but FreeBSD has supported ELF binaries for the longest time, and the newfs > binary is a staticly linked binary, so it requires NO libraries from the > 3.0-R world to run... so it would run perfectly fine.. 2.2.1 should run 3.0 ELF binaries pretty well. Cheers, -Peter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199810211447.WAA16977>