Proposed udpates policy change

[David Malcolm]

On Mon, 2010-03-08 at 22:09 +0000, Richard W.M. Jones wrote:
> On Mon, Mar 08, 2010 at 09:59:29PM +0000, Matthew Garrett wrote:
> > We assume the following axioms:
> [..]
> > 2) It is impossible to ensure that functionality will not be reduced 
> > without sufficient testing.
> 
> Your axioms are obviously wrong.  An update which simply bumped a
> release number would have the same functionality.  Since you claim
> these are axioms -- self-evident truths that form the basis for
> further argument, anything else you wrote is unsound.  Sorry :-!

Richard: I'm afraid, from bitter experience, I can dispute that
argument :(

The essence of my counterargument is that in the meantime an update
might have been pushed to the build root tag that affects the rebuild,
and you might "silently" lose functionality.  

The most common example I can think of is when a test in a configure
script starts failing for some reason beyond the control of the person
building the update: a package can be rebuilt without changing its
source, but all functionality relating to the configuration test will
"silently" go away.  (youch!)

Or if you want a more direct but hopefully less common example, what
happens if a bug has been introduced into the compiler since you last
rebuilt?  All you're doing is bumping a release number, without changing
your sources, but you could end up in a world of (new) pain.

Hope this is helpful; FWIW I think we need better automated testing
around our updates process.

Dave