ade at FreeBSD.org
Sat Nov 26 15:42:02 EST 2005
On Nov 26, 2005, at 02:59 , Edwin Groothuis wrote:
> If it are these kind of errors, shouldn't phase 2 then not only be
> started when phase 1 somehow fails? Doing it a second time for "just
> in case something went wrong in the first run" sounds like such a
> waste of time then if it went right the first time.
Detecting when the first phase failed (for whatever reason) and
passing that back up through a myriad of co-operating processes back
to tinderbuild is decidedly non-trivial.
It's considerably easier to simply try things twice, to pick up on
transient cases, most of which I've found are due to issues in the
NFS server side of things under high load.
> Unless of course Phase 2 does do something else than phase 1.
Nope. There's no difference between the two phases.
More information about the tinderbox-list