* Stefano Zacchiroli <[EMAIL PROTECTED]> [080215 11:46]:
> On Fri, Feb 15, 2008 at 09:42:47AM +0900, Charles Plessy wrote:
> > This is a very good idea, but the reason why source-only uploads are not
> > allowed is that there are concerns that if the binary package is not
> > used for real, the quality of the source package will drop. Within this
> > hypothesis, there is no incentive for the laxist developper to use the
> > valuable feedback that you propose.
>
> I personally consider this argument bogus as well. Let's imagine we can
> split DDs and DMs into "good" and "bad" uploaders. Good uploaders
> nowadays use a clean p/cowbuilder environment, test their packages, yada
> yada, and then upload.
Maintainers only doing builds in clean environments are not good
maintainers, as they miss testing their build system in unclean
environments. Current buildds are in avarage far too clean to test this.
A good maintainer therefore should (at least sometimes) build in an
unclean environment.
> Good and bad uploaders exist now with binary uploads and will exist with
> source uploads.
What you miss is that while the .deb does not directly require the built
package to be tested, it at least makes non-testing more visible: If you
upload a source package that fails to build proper files, you can far to
easy claim something in the buildds was different and when you tested it
is worked. When you upload a .deb, either that works or it doesn't.
Everyone can look at those and clearly see that the developer is to
blame.
> The question is how much we think that requiring a deb for the upload is
> an incentive for pushing the bar of a random uploader nearer to the
> status of good uploader than to the bad one.
Whether this is an incentive to better test it it hard to say.
But what can be said is that there is a test-case of source only uploads
(though with other parameters changed as well), and that fails horribly
w.r.t. package quality.
> I think it is indeed an
> incentive, but the current drawbacks are far worst than this benefit.
What drawbacks do you see?
> Hence I think we should push for source upload. Other technical
> incentives can then be found and I've already suggested some of them,
> e.g. tuning our upload tools so that they indeed require the existence
> of a .deb, not necessarily uploading it later on.
So why don't you start with implementing them first? Building a
framework for doing unclean builds (a.k.a. buildd from hell), checking
for missing build-conflicts, tools for comparing different .debs of the
same version and stuff like that can all be implemented without having
to switch to source-only first and are all worth even if that never
comes.
Hochachtungsvoll,
Bernhard R. Link
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
