Re: Expected?: "failed: Miscompare for aalgo=hmac/sha1 ealgo=camellia-cb" 12.0-ALPHA10 kernel with pre-openssl update -r339076 head?

On 2018-Oct-18, at 12:12 PM, Mark Millard <marklmi_at_yahoo.com> wrote:

> As part of helping to track down a powerpc64 system
> crash when "kyua test -k /usr/tests/Kyuafile" is run
> I substituted into a -r339076 context official kernel
> materials from:
> 
> https://download.freebsd.org/ftp/snapshots/powerpc/powerpc64/12.0-ALPHA10/kernel*.txz
> 
> This does lead to kyua reporting:
> 
> sys/geom/class/eli/init_test:init_a  ->  failed: Miscompare for aalgo=hmac/sha1 ealgo=camellia-cbc keylen=192 sec=8192  [275.805s]
> 
> that had been passing with just my buildworld buildkernel
> materials for -r339076 . (So far in the run it is the only
> such report.)
> 
> Is this difference in this odd context expected/reasonable?
> Should I ignore it?
> 
> (I have no reason to normally run such a odd mix of vintages
> of world vs. kernel materials. I'm just checking if the
> crashing is somehow specific to my builds or not.)

For reference:

My normal kernel build is via devel/powerpc64-xtoolchain-gcc
but I've built a -r339076 based kernel from my sources via
gcc 4.2.1 and the system binutils ( world still at -r339076
via devel/powerpc-xtoolchain-gcc ). Under that kernel kyua
reported:

sys/geom/class/eli/init_test:init_a  ->  failed: Miscompare for aalgo=hmac/ripemd160 ealgo=camellia-cbc keylen=128 sec=1024  [465.515s]

(A different, far later step for the failure?)

So apparently compiler/toolchains matter even when the
sources are the same.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)