Date: Fri, 15 Nov 1996 00:15:37 -0800 From: "Jordan K. Hubbard" <jkh@time.cdrom.com> To: Bruce Evans <bde@zeta.org.au> Cc: cvs-all@freefall.freebsd.org, CVS-committers@freefall.freebsd.org, cvs-sys@freefall.freebsd.org, jkh@freefall.freebsd.org Subject: Re: cvs commit: src/sys/i386/i386 locore.s Message-ID: <11301.848045737@time.cdrom.com> In-Reply-To: Your message of "Fri, 15 Nov 1996 18:37:48 %2B1100." <199611150737.SAA03853@godzilla.zeta.org.au>
next in thread | previous in thread | raw e-mail | index | archive | help
> > Modified: sys/i386/i386 locore.s > > Log: > > movl instruction should have been lea (this is why userconfig didn't > > work in 2.1). > > Erm, the movl looks OK to me, and the lea isn't an i386 instruction so > it doesn't compile. [scratches his head] Why did this compile for me? And why did that single fix (suggested by Don Lewis, BTW, who knows far more about the x86 instruction set than I) result in USERCONFIG_BOOT working where it did not before? Perhaps this was something which should have stayed out of 2.2 and remained confined to 2.1.6 due to differences in the compiler tech? After all the shenatigans of the last week, you can bet that I tested this very very carefully with a before-and-after check of just that one line changed, and the movl version blows up under 2.6.3 whereas the lea version works fine - I'm actually booting 2.1.6 floppies now where I never could before. I'm not saying you're wrong, since what I know about x86 assembly could fit into a thimble (the only instruction set I ever learned backwards and forwards was the for the NS32K architecture, and it spoiled me for other processor architectures), I'm simply saying that the empirical evidence runs very much counter to what you're saying. Any theories? Jordan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?11301.848045737>