Re: Using buildds only (was: Results of the meeting...)
On Mon, August 22, 2005 10:45, W. Borgert wrote:
> Fortunately, Martin Krafft came up with the idea of
> allowing source-only uploads only together with a signed test protocol.
> The test protocol would have to include the output of lintian, linda,
> and piuparts - warnings allowed, errors not.
I dislike this idea: it is way overengineered. For starters I don't
understand why you would want to run both lintian and linda, since those
tools are intended to be replacements of eachother, not complements. If
they would complement eachother, then why are the vast majority of their
tests present in both programs? I'll just talk about lintian below, but
you could substitute linda or any other testing tool aswell if you like
that tool better.
The whole certificates construction implies that people would want to
deliberately circumvent running lintian on their packages. I don't believe
there are a significant number of people doing that, and if they really
want to, they can do so anyway.
If you really want to enforce lintian checks, you should run them
server-side. O wait, we already do that. IMO, that works just fine as it
does now; the results are visible for everyone. You could extend this to
running lintian for new uploads and rejecting them on errors if you would
so wish, but I do not see indications that this is actually necessary.