Michael A. Peters wrote:
1 complete pass of memtest shows no errors, I'll just let it go for
several, but could a mis-behaving kernel mod cause something like this?
Maybe I'll boot into an older kernel and see if things go better
older kernel made no difference.
However, running mock with a heavy load on server made it work much
better. Not 100% but nearly. I really don't understand it. Somethings
not right.
Oddly, I seem remember the same thing happening (improper configuration
generation) when I build 5.2.4 or 5.2.5 way back - I blamed it on nvidia
because I removed that module and both arches then built next attempt,
but that wasn't it this time.
I don't know for that heavy load is what made it work almost
consistently in mock. I may do some testing and see - this really is
quite odd.
3 passes of memtest86 no failures, so it's probably not a bad bit in
memory. That doesn't make sense anyway because if configure is properly
regenerated is always goes to completion. Some race condition in the
auto tools?