Re: ATF tests still failing to complete within 2 hours on amd64

[David Holland <dholland-current%netbsd.org@localhost>]

On Thu, Dec 01, 2011 at 09:54:23PM -0500, Thor Lancelot Simon wrote:
 > > Only happened once, but given that other people have reported this
 > > panic in chat I think that test may need its tolerance loosened. :(
 > 
 > The risk of a false positive is 3 in 10,000.  I suspect that, in fact,
 > the data being emitted by the RNG are in fact badly biased at boot
 > time under some circumstances: a much larger problem than a test that
 > falses positive and reboots the machine immediately at startup one
 > in 3333 tries!
 > 
 > I have recently committed changes that allow the i386 port to load
 > saved entropy from the boot blocks such that any positive result
 > from this test would indicate either a bug in the test (unlikely;
 > it is very widely used) or a bug in the generator itself.  I'd be
 > very interested in hearing from anyone who gets a reboot with the
 > boot block entropy loading turned on.

Well, if it's broken we should certainly try to fix it. That said,
even if it's not broken, I think three failures in 10,000 is too high
for production use... at least if the failure causes a panic at boot
time. Particularly a panic at boot time when new users are trying to
boot the installer for the first time.

-- 
David A. Holland
dholland%netbsd.org@localhost