powerpc (non-64 variant) 11.0-CURRENT r279514->r279813 update's installworld failed to complete

[Mark Millard]

A user experience report: I tried to take my recently constructed SSD for booting

$ freebsd-version -ku; uname -a
11.0-CURRENT
11.0-CURRENT
FreeBSD FBSDG3C0 11.0-CURRENT FreeBSD 11.0-CURRENT #0 r279514M: Mon Mar  9 22:24:27 PDT 2015     root at FBSDG4S0:/usr/obj/usr/srcC/sys/GENERICvtsc-NODEBUG  powerpc

on PowerMacs and update a copy of it to -r279813 via:

cd /usr/srcC/
svnlite update -r279813
make -j 8 buildworld kernel KERNCONF=GENERICvtsc-NODEBUG INSTKERNNAME=kernel11C
make installworld

But the installworld eventually failed with a Bus Error notice reported after the line indicating /libexec/ld_elf.so.1 had been established...

=>> libexec/rtld-elf/tests/target (install)
...
/usr/libexec/ld-elf.so.1 -> /libexec/ld-elf.so.1
Bus error (core dumped)
*** Error code 138

After that any attempt to load a program (say by typing ls) also got a bus error. Already loaded programs continued to work from what I could tell (such as the shell I was typing to). (This tracks when ld-elf.so.1 would be involved vs. not.)

Another attempt was to buildworld and buildkernel on one r279514 boot SSD and to then installkernel and installworld via DESTDIR=/mnt to another SSD. But /sbin/init got a Bus Error when booting the updated SSD.



I ended up reverting to an SSD where the -r279514 11.0-CURRENT kernel and world had not been updated. I did not ever try substituting ld-elf.so.1.old .



This effort was made on a PowerMac G5 despite the non-64 status of the build. (I have yet to build and install a powerpc64 variant of a fairly recent 11.0-CURRENT snapshot.)



I'm not likely to retry the experiment anytime soon. So I do not directly have evidence of repeatability.

===
Mark Millard
markmi at dsl-only.net