Primary Architectures: Another Proposal (RFC)

[Josh Boyer]

On Mon, Apr 9, 2012 at 9:55 PM, Kevin Kofler <kevin.kofler at chello.at> wrote:
> Josh Boyer wrote:
>> Because you, as the maintainer of the package in question, aren't
>> building and testing the package locally before you submit the official
>> build?  Or you aren't leveraging scratch builds to do the same if your
>> local hardware is prohibitively slow?
>
> The local or scratch build is just a wasted build, then I have to redo the
> build as an official build, so I have to do an extra build for no reason. If
> I do all the builds as official and stick with the one that builds, that
> saves the final extra build.

"It builds" doesn't necessarily mean it works.  In the rawhide case,
you might very well submit something that builds but is totally
broken.  By the time you download the official build and test it
locally, it can already be in the day's compose and now you've just
inflicted a broken build on all the users.  And no, "rawhide eats
babies" is not a valid excuse IMHO.

In a stable or branched release, your method would work as the way
we use koji tags with -candidate prevents it from going directly to
users in a compose.

>> I don't think we need to encourage more "throw it at the wall and see
>> if it sticks" maintainership.
>
> Why? It saves a build that'd otherwise be wasted, saving my time. It also
> doesn't preclude downloading and testing the output once it builds.

An ounce of prevention.  Personally, I'd rather "waste" a small bit of
my time to save inflicting broken stuff on users.  I can always work
on something else while that particular build is running.

josh