[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: NEW processing

Thomas Viehmann wrote:
> The particular pass through NEW that has been used to demonstrate the
> deficiency of NEW processing was necessitated by the rename
> iceweasel-l10n-hi to iceweasel-l10n-hi-in introduced in the previous
> upload (and processed in 3-4 days or so). This rename took place after
> uninstallability of iceweasel-l10n-all had been pointed out to the
> maintainer after he asked for an unblock on release. In essence, this
> whole trip through NEW would not have happened if the maintainer would
> actually routinely install his packages before uploading. I am all in
> favor of fast-tracking urgent stuff, but the deal should involve the
> maintainer making extra-sure to get things right, too.
Because we all know that uninstabillity problems can only occur when you
change the package name...

Or undistributable/illegal stuff. Or lintian errors.
In that sense, you should perform quality checks and moderate _every_
upload as well.

I'm all for doing legal checks and semi-automated quality/sanity checks
on the *first* upload.
But what's the point on doing it on binary package renames?
Crap can be introduced into the archive just as easily without it /ever/
processing NEW.

If the source package existed in the archive before it shouldn't pass
through NEW!

Quality assurance of existing packages is a job for the QA team.
The two teams should definitely cooperate, share common patterns and
possibly scripts.
In that way, all packages can be automatically or semi-automatically
checked, even those that don't ever change their package names.

Am I the only one that finds all of the above obvious?


Reply to: