probabilistic approach to tinderboxing

[Bjoern Michaelsen]

On Thu, Jun 14, 2012 at 09:57:48AM -0500, Norbert Thiebaud wrote:
> When you apply a series in sequence, at any point the result must be
> buildable...
> 1/ it is good practice
> 2/ if that is not the case bisection is much harder

... but it has a lower importance than having the tip of master buildable all
the time. Which is why a tinderbox should leave testing intermediate states
aside as long as there are still things to test that ensure the tip of master
to stay buildable.

Best,

Bjoern