Re: Tinderbox.

[Havoc Pennington <hp redhat com>]

Michael Meeks <michael ximian com> writes:
> On 11 Mar 2001, Havoc Pennington wrote:
> > This should be a step in the "work plan" in our proposal, between "set
> > up branches" and "set up tinderbox."
> 
>         Useful though a tinderbox is, I'm taking a preemptive strike on   
> the issue - I do not believe that random people should be allowed to
> commit unapproved crud to CVS that happens to make it build on the
> tinderbox [1].

The usual way it works is that if anyone commits to the module, they
are "on the hook" and have to hang around until Tinderbox comes up
green. If it turns red, they have to fix the commit they just made. So
usually the commit would be a fix to their previous commit.

>         I tend to think that if you are not competant to fix the trivial
> sort of build problems that you get from time to time with CVS HEAD
> (anything) that you should not be building from CVS.
> 
>         Of course, one way to solve this would be to have the tinderbox
> tag the CVS at its last successful build of a package: tinderbox_head or
> somesuch.
> 

The motivation for tinderbox is that breaking the build really hurts
the productivity of all the dozens of other people hacking on the
tree. So breaking the build is highly discouraged. Thus the "on the
hook" policy after commiting.

Havoc