Date: Tue, 4 Apr 1995 11:00:47 +0200 (MET DST) From: Mikael Hybsch <micke@dynas.se> To: freebsd-hackers@FreeBSD.org Subject: (at least) 2 problems with 950322-SNAP Message-ID: <Pine.SOL.3.91.950404095036.6713A-100000@spirit.dynas.se>
next in thread | raw e-mail | index | archive | help
I has running 2.0 at home and felt that I should try 950322-SNAP. I have now spent 2 evenings trying to install it without success. Day 1: I downloaded the distribution on a spare UFS partition and made the 3 floppy disks. The installation newfs'ed the / and /usr partitions and loaded the cpio floppy. After that the problems starts. The generic kernel won't boot. The matcd driver says something about "trying to fix" and then goes into a endless (> 2 min) loop printing: "matcd0: after reading status byte, buss didn't go idle" No problem. I just use the -c options to disable the matcd driver. In the configuration I type "disable matcd0". It says that there is no such device. I type "ls". The line for the matcd driver looks funny, something like "match int0x77777777 ...". In short: I cannot boot the generic kernel and I cannot disable the matcd driver. Next, I boot the fixit floppy and copy the fixit kernel to the hard disk. (It boots (from floppy) and I only need the wd driver to install the system). Now when I boot the fixit kernel, the machine resets itself before the kernel loading is complete. After printing the information about text and data sizes, there is a new field inside brackets that wasn't there in 2.0. Just after printing one of these numbers is fails. This just happens with the fixit kernel if I try to boot it from wd0. Day 2: At work I download the 950322-SNAP source on a 2.0-RELEASE machine and build myself a generic kernel without the matcd driver. At home I successfully boot the homemade kernel, and the installation goes on. After installing the bindist it complains that the checksums doesn't match for the manpages. I drop out into a shell just to find that the checksum failed because after installing bindist the do_cksum.sh script find cksum in /usr/bin instead of in /stand. The cksum in /usr/bin terminates with signal 10 bus error - core dumped. A closer investigation finds that all dynamically linked programs fail in the same way. The only diagnostics I was able to do was to find out that if I rename ld.so, the programs correctly say "No ld.so". If I leave ld.so and remove all shared libraries it still fails with signal 10. I tried to mount /usr on the fixit floppy it case there where some problems with my homemade kernel. I also tried to reinstall the system from scratch, but the "bus error" problem won't go away. Tonite I will try the 2.0-RELEASE ld.so. I need some ideas of what could be wrong or I'll have to go back to 2.0-RELEASE. My machine is a 486-33 VLB with 2 ide drives, a Sony CDU33A, one SB16 MCD and 16 M RAM. All (but SB16) found by the generic kernel before matcd0 failed. -- Mikael Hybsch Email: micke@dynas.se DynaSoft, Dynamic Software AB Phone: +46-8-615 84 00 Hammarby Fabriksv 13, Box 92058 Fax: +46-8-641 92 00 S-120 06 STOCKHOLM, SWEDEN
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SOL.3.91.950404095036.6713A-100000>