Testing

[Hal Murray]

There is a discussion on the TZ list about a screwup in medical software that 
can't handle DST changes.  It's also on RISKS.  (But not on their archives 
yet.)
  http://catless.ncl.ac.uk/Risks/

So I'm thinking about testing.

What can we do to help people test NTP?
Does anybody have a list of rough edges?

What should we test before a release to make sure our code is doing what we 
expect?
What should end users test to make sure our code is working in their setup?

Are those the same lists?  Or would priorities change?  What does priority 
mean?

Should we collect directions (or a script) to semi-automate making leap 
seconds?

One possibility would be a build option to enable a config option for 
leap-tonight.  That's a kludge, but simple for testers who aren't NTP geeks to 
understand.  Plan B would be to hack the leap seconds file.  Cleaner for the 
ntpd code, but more complicated overall.



-- 
These are my opinions.  I hate spam.