Re: port-i386/42966: kernel built from today's netbsd-5 crashes

[Anne Bennett <anne%porcupine.montreal.qc.ca@localhost>]

>> You have the diagnostic option enabled, which is non-standard; it
>> seems your kernel is hitting an assert that only fires if you've
>> compiled it to do so.

>  With option DIAGNOSTIC I can reproduce this, viz.: [...]
>  Disabling the option in question should allow your kernel to boot
>  normally.

I can certainly try that (probably tomorrow), but I'd like to add
that I didn't just add DIAGNOSTIC when I updated the kernel; that
option was in force on my 5.0.1 kernel and did not cause a problem.

I'm concerned that just turning off DIAGNOSTIC will simply mask a
problem; I hope that the bug can actually be found and squashed.

But thanks for the workaround; it should allow me to run with
the "2010-003" security advisory addressed, which is good.

>> I'm also seeing strange time-keeping events since the last update I
>> made against the netbsd-5 branch, but my experience is much milder than
>> yours, I just see impossible CPU times against particular programs as
>> given by ps(1) and top(1), e.g. a program that's supposedly consumed
>> almost a year of CPU time even though the machine has been up for only
>> three days.

Weird.  I wonder if it's some kind of clue that the oddities I saw
happened with the *old* kernel, but booted by name "/netbsd.old";
I'm somewhat ignorant of how this works, but could something be
being picked up from "/netbsd" (the new kernel) even when this is
not the booted kernel?

That reminds me to report, though I doubt it's relevant: I did make
one change to /etc/sysctl.conf back in January:

  # 2010-01-12 Anne Bennett: need more files, especially when running
  #  a large compilation.  This is previous value * 4:
  kern.maxfiles=26672


Anne.
-- 
Ms. Anne Bennett, as a private citizen:  anne%porcupine.montreal.qc.ca@localhost
Also reachable more officially at work:  anne%encs.concordia.ca@localhost