Eric Bélanger wrote:
On Tue, 9 Dec 2008, Dusty Phillips wrote:
2008/12/9 Pierre Schmitz <pierre@archlinux.de>:
Am Dienstag 09 Dezember 2008 19:36:14 schrieb Aaron Griffin:
On this note: Dusty added web-based signoffs to our dev site a while back https://dev.archlinux.org/packages/signoffs/
Could we please take a look at this to determine if we need anything else to begin using this? If you guys want, I'm sure I could even automate something to move signed off packages to the real repo automatically...
I don't think its very usable atm. Some points to improve this:
* Only show up core packages (atm every package in testing is shown)
I can do this and try to add an additional interface where people can request a specific package from extra also be signed off when required.
There a feature request to make the signoff list sortable by repo (FS#11325). Once that's implemented (if it's implementable), you sort by repo and you get the all core signoff at top of list.
* Send a message to the ml once a new package is added to testing (or updated)
This should be done manually by the dev who puts the package in testing. This way he can list what changes were done to the package. This will prevent unwanted changes to go unnoticed.
* Send a message to the ml once a package has enough sign-offs
I don't think it's necessary. The person who started the signoff can check the dashboard after a couple of days. And do that on a daily basis afterwards.
This is silly. What's the point of web based signoffs if you're going to fill your inbox anyway? I can add a summary to the dashboard if you aren't accustomed to actually looking at the signoff page. If you don't look at the dashboard then... well my work here is done.
* don't move automaticaly; there might be good reasons to keep packages longer
Maybe a one click move action in the signoffs page?
That's unecessary. As Pierre pointed out in another email, using testing2core is easy enough.
One problem with the current signoff system is that there are packages which are harder to get signoffs for because they need special hardware (e.g. wireless drivers) or setup (e.g. raid, lvm) to test efficiently. I was thinking about having a list in the wiki of those low-usage packages along with the name of the devs who can potentially signoff these packages for a given arch. This way we'll know what to expect in terms of dev signoffs and won't have packages stuck in testing for weeks waiting for signoffs that will never come because no dev uses them. We could treat these packages as special cases. We could automatically send the signoff email to the arch-general ML for user signoff. We could also set a rule that after X days in testing and no bug reports, the package is cleared to be moved to core.
I am willing to setup the list if you think it's a good idea. Such a list might even create more interest in the signoffs as it will remove the "another dev will signoff, why bother?" way of thinking.
I think that is a good idea. I started making a list of hardware related packages in [core] to do this once but never got around to doing anything with it.... Allan