05:50:51< abadger1999> mchua: One problem is the organization of the packaging guidelines. I want to work on that at fudcon... unfortunately, that's after november

05:54:04< mchua> abadger1999: ...actually, I just had a good conversation with m_stone on the packaging process that you might be interested in.

05:54:15< mchua> abadger1999: I've been trying to figure out who to share it with for the past, oh... 5 minutes

05:54:21 * mchua --> pastebin

05:54:27< abadger1999> mchua: :-)

05:55:10< mchua> abadger1999: http://fpaste.org/LuUI/

05:55:51 * mchua was just reading https://fedoraproject.org/wiki/Packaging:Committee and wondering whether to ping the members for thoughts on what might be useful to turn into a Draft

05:58:11< abadger1999>

05:58:28< abadger1999> Some of that should just go into the wiki under the Maintainers space... nothing Guidelines related

05:58:39< abadger1999> For instance: http://bugz.fedoraproject.org/sugar-toolkit

05:59:21< abadger1999> that will help with the first problem -- searching bugzilla

06:03:22< abadger1999> mchua: I have an idea but I don't know that it will fly

06:03:28 * mchua looks at Maintainers space to see where this might fit...

06:03:42< mchua> abadger1999: ooh, what is it?

06:04:20< abadger1999> mchua: Most of what m_stone had problems with revolved around authentication. If we could do away with most of it for his use case would that be helpful?

06:04:33< abadger1999> Check out from cvs can be done anonymously.

06:04:50< abadger1999> Searching bugz.fedoraproject.org/PACKAGEcan be done anonymously.

06:05:33< dgilmore> abadger1999: m_stones acls were commit related though

06:06:18< abadger1999> Submitting a patch -- maybe we could do that anonymously (via wwoods 's abrt server ideas)

06:06:30< abadger1999> Or that might be the only place where auth was required.

06:06:31< mchua> abadger1999: I think making the "don't want to bother with auth? here's the limited subset of things you can do?" route clearer would help, yeah; I could try to doc that (maybe this weekend).

06:07:27< abadger1999> mchua: Cool. I think that might be helpful.

06:07:38< mchua> the commit stuff should require auth, imo; take anonymously and invisibly, but if you put something back in it's got to have a name on it.

06:07:49< abadger1999> Another thing that he's pointing out is that we have information available but it's hard to find.

06:07:57< abadger1999>

06:08:19< abadger1999> Yeah -- but commit was the final part of everything he did.

06:08:55< mchua> so up until the "I have a patch and want to commit," he could have avoided auth entirely?

06:09:19< abadger1999> mchua: I think so.

06:09:46< abadger1999> mchua: And we could talk about doing something like this instead of actual commit as well: "fedora-packager-email-patch sugar-toolkit foo.patch m_stone@lp.o"

06:10:20< abadger1999> To work kind of like "git format-patch -1 | git send-email --to foo@bar.com"

06:10:45< mchua> would we need to worry about spam, or about people who haven't signed the ACL committing code?

06:10:58< mchua> or does it all automagically get licensed right when they do, and all?

06:11:06< dgilmore> abadger1999: you can always do cvs diff -u >cvs-diff.patch and attach it to an email/bug

06:11:25< dgilmore> but its not as simple as "git send-email"

06:11:47< abadger1999> mchua: With patches in bugzilla, peole often haven't signed the CLA. However, there's some legal grey in that.

06:12:00< abadger1999> dgilmore: yeah -- if it's a patch to the spec file.

06:12:15< dgilmore> abadger1999: even a patch to code

06:12:47< abadger1999> dgilmore: Really? Code is kept in the lookaside cache... how does cvs know what to base against?

06:12:56< mchua> dgilmore: you need a bz account to attach stuff to a ticket, and if you're not logged into bz you can't see the email addresses of the folks involved with the ticket.

06:13:02< dgilmore> abadger1999: patches are not kept in lookaside cache

06:13:08< abadger1999> mchua: spam we'd have to consider.

06:13:32< mchua> dgilmore: you could take their names (which are listed to not-logged-in-bz-users) and google, but... that's iffy, and a potentially large extra step that might not find anything.

06:13:36< abadger1999> dgilmore: Right. I'm saying you can't do " cvs diff -u >cvs-diff.patch " with changes to the code.

06:13:39< dgilmore> mchua: there is -owner@fedoraproject.org aliases

06:14:05< dgilmore> abadger1999: you can if you make a patch and put it in cvs

06:14:13< dgilmore> i think you can do a cvs add locally

06:14:18< dgilmore> but maybe not

06:14:24< abadger1999> dgilmore: Yes, but you can't put it into cvs if you don't have commit acls.

06:14:28< dgilmore> abadger1999: but you can make a patch

06:14:38< abadger1999> dgilmore: Nope, not locally with cvs.

06:14:45< dgilmore> "make prep" is anonymous

06:14:48< abadger1999> dgilmore: svn yes to locally.

06:15:12< dgilmore> abadger1999: so you send code patch + cvs patch for spec

06:15:13< mchua> dgilmore: hm. I wonder if there's a way to make the -owner@fp.o alias visible from bz... some way to autogenerate it in the html by grabbing the contents of the component field?

06:15:26 * mchua actually didn't know about that alias

06:15:41< abadger1999> dgilmore: Right. And you'd have to make the patch for fedora-packager-send-email foo.patch to work

06:15:43< dgilmore> mchua: not sure

06:15:56< dgilmore> abadger1999: right.

06:16:52< abadger1999> mchua: In the past we have talked about adding the alias to the bugzilla cc list instead of all of the comaintainers explicitly. But the alias is not perfect yet.

06:17:02< dgilmore> mchua: we could have a tool like fedora-packager-send-email that from a checkout will send patches to the owner without needing any auth at all

06:17:10< abadger1999> So we've held off on committing too much to it.

06:17:16< mchua> abadger1999: not perfect yet = not all packages have aliases, not all aliases go to the right place... ?

06:17:42< abadger1999> mchua: let's see.... There's a sync time between creating a new package and the alias being created.

06:17:55< mchua> dgilmore: ooh, if all aliases get to the right place, I think that would pretty much solve it.

06:18:09< abadger1999> The alias currently goes to owner + cclist + committer + watchcommits

06:18:11< dgilmore> abadger1999: fedora-packager-send-email would probably need the same syntax as "git send-email" with the same options

06:21:04< abadger1999> dgilmore: git seems to have a format-patch command that sends email. We probably want a little different

06:21:23< mchua> abadger1999: a long sync time? are the final creation times of a package, its email alias, its bz component, etc. far enough apart that "package is up" shouldn't wait for all of them to be finished?

06:21:28< abadger1999> Since we will only be generating the patch in a few instances.

06:21:35< mchua> abadger1999: ...maybe a better question would be "when is a package 'done'"?

06:21:43< dgilmore> abadger1999: yeah. it has options we would need, smtp-auth, mail-server etc

06:21:50< abadger1999>

06:22:18< mchua> (and should "alias created & working" be one of the criteria for "done"?)

06:22:39< abadger1999> mchua: That's an interesting question :-) Once the cvsadmin request is processed, everything else is automatic.

06:22:58< abadger1999> mchua: pkgdb and n empty cvs directory happen immediately.

06:23:05 * mchua nods

06:23:28< abadger1999> mchua: cvs acls, koji, and aliases should happen more than once an hour.

06:23:39< abadger1999> mchua: bugzilla I believe is hourly.

06:24:00< abadger1999> mchua: And then I think we're completely done.

06:24:17< abadger1999> As soon as cvs acls are done, the packager can start working.

06:24:28< mchua> abadger1999: so basically, the maximum lag is an hour.

06:25:02< abadger1999> mchua: yep. if I just miss the current sync, an hour.

06:25:17< abadger1999> cvs acls sync every ten minutes currently.

06:25:54< mchua> abadger1999: that's not bad at all. the chances that someone will want to anonymously submit a patch in the first hour is probably low enough that it's okay for the alias to break during that time.

06:26:04< abadger1999>

06:26:08< abadger1999> That's true.

06:26:42< mchua> abadger1999: and I'm guessing we could also set things up so that if you did fedora-packager-send-email to an invalid address, it could send you back a nice note with "that's not a valid package name, if you're trying to commit to a new package, wait an hour"

06:28:04< abadger1999> mchua: Yep. If spambecomes a problem, I suppose we could tie into the app that wwoods is thinking of writing for abrt.

06:28:34< abadger1999> mchua: But I don't think that spam will be a problem at first. The aliases already exist. This isn't publishing them any further.

06:31:08< mchua> abadger1999, dgilmore: sounds like there's two things to be done then; (1) improve docs for auth-less workflow, which takes care of everything except commit and should be relatively fast/easy/not-hit-blockers, and (2) make fedora-packager-send-email, which... is possible, though a more involved project

06:31:21< mchua> to take care of the 'commit' part

06:31:56< abadger1999> mchua: And we can document that people can send email to package-owner@fp.o aswell.

06:32:50< abadger1999> mchua, dgilmore: Maybe fedora-packager submit-patch that can send without auth to -owner or with auth to bugzilla would be good.

06:33:46< abadger1999> then it's one command but those that want tracking of the issue can get it via bugzilla; those that don't want to track down their user/pass can just send email.

06:34:02< mchua> abadger1999: aye. ok, throwing up docs for auth-less workflow + alias Somewhere On Wiki, and then piping up here and on the fedora-packaging mailing list as to where to best put/publicize the content will take care of the first, then

06:34:29< mchua> as for starting the second... *looks up fedora-packager*

06:34:45< abadger1999> mchua: Yep. And fedora-devel-list probably. fedora-packaging is for Guidelines changes.

06:35:03< abadger1999> mchua: fedora-packager == dgilmore

06:35:59< mchua> dgilmore: would filing an enhancement request on https://fedorahosted.org/fedora-packager/newticket with notes from this chat be the best way to start?

06:36:06< abadger1999> Somehow Mr Fedora-packager doesn't roll off the tongue like Mr Yum, though.

06:36:18< dgilmore> mchua: :) sounds good to me

06:38:00 * mchua thinks of last edge-case - if you for some reason can't install fedora-packager

06:39:44< mchua> dgilmore, abadger1999: actually... how bad an idea would it be to have a send-a-patch@fedoraproject.org alias that only accepts emails with a valid package name as the subject line, and forwards any attachments that are patches to the relevant alias?

06:40:23< mchua> s/alias/-owner@fp.o

06:40:44< abadger1999> mchua: I think it would be fine.... We'd need implementation though.

06:41:56< dgilmore> mchua: what abadger1999 said

06:42:59< mchua> abadger1999, dgilmore: ok, that's option #3 for me to chase down then; a fedora-packager enhancement would fix things for everyone who can install fedora-packager, an email alias would fix things for everyone who has email.

06:43:32< mchua> abadger1999, dgilmore: thanks - I'll putter off and see what I can do... meeting with some students from pcalarco's school (Notre Dame) on Monday, maybe I'll be able to persuade someone that this would be a good project.

06:44:42< abadger1999> mchua: No problem. Thanks for bringing energy to get things done!

06:44:53 * dgilmore likes energy

06:44:55< mchua> (in any case, if I can persuade any of them to stick around and do some packaging with me, those docs will get written; I tend to write docs by teaching other people how to do stuff.)

06:45:20< mchua> (...not that I really know how to package, but I at least know how to stumble around in it somewhat productively by now.)

06:45:32 * mchua has too much energy. plz haz some.