On 5/18/2005 11:23 AM, Li-Lun Wang (Leland Wang) wrote: > On Wed, May 18, 2005 at 01:17:53AM -0500, Jonathan Noack wrote: >>>On 05/18/05 00:11, Li-Lun Wang (Leland Wang) wrote: >>>>I just upgraded my pentium-m laptop from 5.4-STABLE to the newest >>>>6.0-CURRENT. However, the newest loader does not seem to work for >>>>me. BTX reboots right after it detects the internal keyboard/console >>>>and dumps the registers. I have tried to set CPUTYPE?=p-m, CPUTYPE=p3, >>>>and even not setting CPUTYPE at all, but none of them solves my >>>>problem. The old loader from 5.4 works perfectly. Any ideas? >>> >>>If you are building/installing just the loader, do a full build/install >>>world without setting CPUTYPE. From my experience it seems like it is >>>gcc that is being affected (when it is compiled with CPUTYPE set) and >>>producing incorrect code for the loader, so a build/install world is >>>necessary. This may not help, but it narrows the problem. > > I rebuilt the world without setting CPUTYPE, and the loader works. > I wonder if setting CPUTYPE=p3 when building the world will also work. > It is strange how gcc could be affected... It is strange, but there's a way to confirm it is gcc: with your world built without setting CPUTYPE (as you have it now), compile and install *JUST* the loader with CPUTYPE defined. The steps to do this (the 'grep' just shows that CPUTYPE is set): # grep CPUTYPE /etc/make.conf CPUTYPE?=p-m # cd /usr/src/sys/boot/ # make depend && make && make install && make clean && make cleandepend When I tried the same thing on my athlon-xp machine, the new loader worked. So it is only when gcc is compiled with CPUTYPE set that there is a problem. I now run 5.4 on the machine and no longer experience the issue, so who knows... -- Jonathan Noack | noackjr_at_alumni.rice.edu | OpenPGP: 0x991D8195
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:34 UTC