[buildd-tools-devel] testers wanted: sbuild and build-dependencies

[Ian Jackson]

Roger Leigh writes ("Re: [buildd-tools-devel] testers wanted: sbuild and build-dependencies"):
> The existing approach to determinism is not to support alternatives
> at all, which is clearly not ideal.  While I don't think we should
> be encouraging the use of alternative build-deps, this is one of the
> most commonly reported bugs in sbuild--there are valid use cases for
> them.

We want better than mere determinism.  After all, apt is deterministic
(since it does not contain a random number generator).

What we want is predictability.  In particular, if one writes
   Build-Depends: lib-new-name | lib-old-name
as Goswin suggests, I specifically want that in a suite where
lib-new-name is available, it is used.  If lib-new-name is available
but cannot be installed for some other reason, the build should be
blocked or fail.

That semantic expectation is violated by a general-purpose heuristic
optimiser such as the one in apt/aptitude.

Ian.