Date: Wed, 21 Oct 1998 06:01:38 -0700 From: Robert Nordier <rnordier@FreeBSD.ORG> To: Stephen McKay <syssgm@dtir.qld.gov.au>, John-Mark Gurney <gurney_j@resnet.uoregon.edu> Cc: freebsd-hackers@FreeBSD.ORG Subject: Re: 3.0-R on a amd386dx/40 Message-ID: <19981021060138.A10093@freebsd.org> In-Reply-To: <199810211020.UAA09162@nymph.dtir.qld.gov.au>; from Stephen McKay on Wed, Oct 21, 1998 at 08:20:59PM %2B1000 References: <19981021004537.00515@hydrogen.nike.efn.org> <199810210847.SAA06623@nymph.dtir.qld.gov.au> <19981021022458.12727@hydrogen.nike.efn.org> <199810210933.TAA07949@nymph.dtir.qld.gov.au> <19981021023848.61590@hydrogen.nike.efn.org> <199810210955.TAA08445@nymph.dtir.qld.gov.au> <19981021030658.41265@hydrogen.nike.efn.org> <199810211020.UAA09162@nymph.dtir.qld.gov.au>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Oct 21, 1998 at 08:20:59PM +1000, Stephen McKay wrote: > On Wednesday, 21st October 1998, John-Mark Gurney wrote: > > >Stephen McKay scribbled this message on Oct 21: > >> Try "awk", for example. It uses floating point, and I think it will die. > >> "df" uses double. "newfs" uses float. "ls" doesn't, and it works. It's > >> a floating point problem, or I'll eat my copy of the Pointy Hat! > >> > >> I still think the "npx" lines from boot are relevant. > > > >ok, here are the lines: > >npx0 on motherboard > >npx0: 387 emulator > > > >these are the lines from a 2.2.6-R system that has problems running > >the binaries... > > I assume they are the same when booting 3.0-R? I am still uncomfortable > with the idea of debugging a (possibly complex) problem by running 3.0 > binaries on a 2.2.x system. It's not designed to work. Maybe it does. > Maybe it doesn't. I'd rather you reported the behaviour of 3.0 binaries > under a 3.0 kernel and 2.2.6 binaries under 2.2.6 kernel. Call me > old fashioned. ;-) > > >I can't see how newfs would fail (elf) but systat (a.out) > >would work perfectly fine under the same kernel if the emulator was > >broken... Because not just the object format is different: different toolchain as well. > > Well, I kinda agree. Bummer. "The signs are clear, but point into the > darkness", or some such. > > >oh, and fsck (a.out), newfs (a.out) and df (a.out) also > >work perfectly fine... > > But, awk (elf) fails? And say, date (elf) works? That would point at FP > probs even if, somehow, awk (a.out) still works. > > And now Robert Nordier thinks FP is involved. I hope he can do better > than me, because I think I'm going back to the shallow end of the pool. > The problem is that the ELF assembler is emitting an operand-size prefix for the i387 fnstsw instruction: 0x66 0xdf 0xe0 rather than just 0xdf 0xe0 (as was the case for a.out). The default FP emulator isn't expecting operand-size prefixes, and adjusts the eip register by 2 instead of 3. -- Robert Nordier 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?19981021060138.A10093>