Re: [GR] DD should be allowed to perform binary-only uploads
Steve Langasek <firstname.lastname@example.org> wrote:
> The error rate on requeue requests that reach me is significant, even from
> people who are well-informed and involved in the process (e.g., fellow
> release-team members). Maybe they're less cautious because they know I vet
> all requests, but I would expect that opening dep-waits/requeues up to the
> general dev population would result in a *lot* of unnecessary rebuild tries,
> stuck packages holding up transitions, etc., because the average developer
> simply isn't clued in on this stuff.
> Heck, before m68k was dropped as a factor in package propagation into
> testing, I was routinely finding bogus dep-waits set by the m68k buildd
> maintainers themselves, and that's only about a half-dozen people.
I can name you the reasons why I don't have much of a clue about this (I
- it wasn't part of the NM T&S process, IIRC; all the technical details
of testing propagation, freeze/unblock/, and general release team work
were a bit meager maybe.
- it's not easy to see what's going on there, and why. For example, I
don't know where I can read what dep-wait means and why and how a
package is put in this state. I think I know what it means and why it
needs to be put there (manually), but that's just because it seems
logical. And although I'm not the most involved developer, at least I
once setup a buildd and read about wanna-build (about two years ago,
- What's the contact point for asking for dep-wait or requeue? I guess
it's that famous bunch of addresses that's also known for getting no
response, and when you want to learn something, an occasional "thanks,
it seems you've grasped the principle" or "Thanks, but you missed the
following" is very helpful.
Of course it's difficult to change that. Someone should write a nice
page about it, and "someone" is, as usual, a synonym for "not me".
> Aurélien got the first
> part of this back in December by keeping tabs on missing builds on arm and
> feeding this to James, but then he proceeded to set up a rogue autobuilder
> when he decided things weren't working to his satisfaction, so it's kinda
> hard to look at that and say yes, this is someone we should bring in to
> collaborate directly on w-b stuff for that arch.
I won't express an opinion here on what Aurélien did, but maybe he is
> (No, I have no idea what this has to do with -vote.)
For my part, it's because a) I'm lazy, b) I'm currently not subscribed
to -project or -devel (so please Cc me if you move).
Dr. Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX/TeXLive)