[00:49:15] --- dev-zero@jabber.org has become available [00:53:55] --- haba has left [01:39:33] --- haba has become available [03:00:11] --- abo has left [03:00:27] --- abo has become available [03:32:39] --- sxw mobile has become available [05:10:45] --- sxw mobile has left [05:13:53] --- sxw mobile has become available [05:49:07] --- sxw mobile has left [05:49:08] --- sxw mobile has become available [06:41:16] --- Jeffrey Altman has become available [06:47:17] --- sxw mobile has left [06:58:31] --- deason has become available [07:02:47] --- Simon Wilkinson has become available [07:06:59] --- sxw mobile has become available [07:17:51] --- deason has left [07:18:23] --- deason has become available [07:39:09] --- reuteras has left [07:53:49] --- sxw mobile has left [08:53:44] --- sxw mobile has become available [09:09:06] --- sxw mobile has left [09:26:07] --- Rrrrred has left [09:26:34] --- RedBear has become available [09:28:30] --- haba has left [09:41:12] --- sxw mobile has become available [09:57:47] --- dev-zero@jabber.org has left [10:09:05] --- sxw mobile has left [10:10:38] --- sxw mobile has become available [10:46:24] --- Russ has become available [11:01:01] --- dev-zero@jabber.org has become available [11:25:00] --- cclausen has become available [11:26:03] --- sxw mobile has left [11:29:11] Should the make_release script also try to push a documentation bundle out to the git-ified web site? [11:31:31] Yes. [11:32:13] --- sxw mobile has become available [11:32:51] I'm increasingly incline to turn make_release into a web application. [11:33:01] Not least of which so that jaltman can also make releases. [11:35:18] I suspect that Jeff could, if we bribed him with good enough beer, log on with ssh and run a UNIX command. :) But yeah, a web application would be kind of neat. There is the problem that you need AFS creds to write into grand.central.org, though, and I'm not sure that using a keytab for that accessible to an OpenID-authenticated web page is a great idea. [11:35:43] Yeh. I think that there would have to be some kind of a human step in the process. [11:35:57] At the very least, the tag needs to be PGP signed. [11:36:49] But at the moment you have to be very careful. There's nothing to stop you from signing a revision on a topic branch and pushing that revision into git. [11:42:06] Well, but tags are fairly ephemeral in Git. You can just delete them again. [11:42:35] Yeh. As long as they haven't escaped. [11:42:48] Things get exciting if someone has pulled a tag that you then delete. [11:51:35] Doesn't it just stick around locally forever but not really cause any problems, particularly if you reuse the same tag? [11:51:46] Last I checked, if the tag is moved in the repo, git pull will just move your local copy without any fuss. [11:54:56] The man page for 'git tag' suggests that the behaviour is a little more complex - basically upstream can't move a tag behind your back. [12:00:03] --- asedeno has left: Lost connection [12:00:03] --- kaduk@mit.edu/owl has left: Lost connection [12:03:02] it looks from the manpage that there's no technical reason it couldn't be done, but someone (i.e. linus) was really really against doing it [12:04:26] Russ: you could setup a script with remctl for Jeff, right? There is a window binary, so no ssh involved [12:06:21] Same issue with keytabs for access to grand, sadly. [12:07:07] Also, I think there are Kerberos realm issues. We don't have a 'OPENAFS.ORG' realm. [12:12:41] Okay, so that's exciting. If you push to the wrong git instance, it will happily take all of your code. [12:12:44] Ooops. [12:13:45] that doesn't sound good [12:13:56] Not really, no. [12:14:27] If nothing else, abandoning all of these commits is going to be really dull., [12:17:59] At least I didn't just push the Linux kernel into gerrit.openafs.org. That would be really, really boring. [12:20:41] --- sxw mobile has left [12:22:00] --- cclausen has left [12:28:13] --- sxw mobile has become available [12:29:38] --- cclausen has become available [12:32:03] I keep thinking I'm going to accidentally push to the wrong branch; I know it's just a matter of time [12:34:33] --- asedeno has become available [12:36:14] --- sxw mobile has left [12:37:48] --- sxw mobile has become available [12:40:18] pushing the wrong branch shouldn't be too disasterous as gerrit already has all those objects. [12:42:41] --- kaduk@mit.edu/owl has become available [12:45:34] yeah, I get little confused as to what actually happens if you push something to master when you're not actually on that branch [12:45:50] I guess the $branch in refs/for/$branch is a bit superficial from git's point of view, when you're pushing? [12:46:14] as in, it doesn't have any real relevance to the actual branch, until (I assume) when you cherry-pick it chooses which branch to cherry-pick onto [12:55:00] The problem with pushing to the wrong branch, such as pushing unstable to the stable branch, is that every commit from unstable that isn't on stable would show up in Gerrit for review, I think. [12:55:12] i believe so [12:55:15] Which would suck. [12:55:29] * Russ wonders if there's some way to tell Gerrit not to accept more than N changes as part of a single push. [13:02:16] I wonder if it really does that though; I thought gerrit makes changes when and only when you send an object, and it wouldn't send an object that already exists on the receiving end [13:02:52] err, hmm, well, git itself wouldn't; but I suppose gerrit's git server thing is magical, so I dunno [13:14:37] --- haba has become available [13:55:07] Using gerrit for Prometheus, are we? [13:58:13] --- cclausen has left: Replaced by new connection [14:00:23] --- dev-zero@jabber.org has left: Replaced by new connection [14:00:24] --- dev-zero@jabber.org has become available [14:02:24] --- cclausen has become available [14:02:44] --- Jeffrey Altman has left: Replaced by new connection [14:33:15] pushing stable to unstable won't do as Russ suggests. [14:34:27] remember that the parent of an object is an integral part of that objects hash. [14:35:22] git will only push those objects whose hash isn't already upstream, and the objects pushed will remain rooted at the same location. [14:36:24] what will go wrong is when/if the submit button is pressed, the objects will get cherry picked to the wrong branch [14:37:18] or if we were running in merge mode, you'd have a merge commit between stable and unstable. [14:41:52] okay, so the branch you specify in refs/for/branch is basically just to tell gerrit what branch to cherry-pick to when it's approved? [14:42:32] I thought conceivably it could 'hide' objects not in that branch when it negotiates with clients what objects to send, but I guess it doesn't [14:53:49] the thing is that the client just knows its pushing to a git server, and nothing about gerrit. [15:04:33] --- Jeffrey Altman has become available [15:04:50] yes, but gerrit still has to communicate what objects it has so the client knows what to send, right? [15:12:25] --- deason has left [15:20:38] --- deason has become available [15:30:39] --- dev-zero@jabber.org has left [15:35:16] So, what I understand happens is that when you do a push of A:B what git does is check to see all of the objects that are reachable from A, but not reachable from B. It then sends that set of objects to the server. [15:35:35] So 'hiding' objects on the server side would just result in more data being sent. [15:36:30] The 'B' you specify determines this reachability set, and also where that particular branch head ends up pointing. (so, a push to refs/heads/master will move the 'master' branch to point at the tip of the tree you are pushing) [15:42:17] Having spoken to Shawn, when I get a spare moment, I'm going to see about adding a common ancestor check to gerrit, so I can't do this in future ... [17:26:40] --- Russ has left: Disconnected [17:43:54] --- Russ has become available [18:20:01] --- stevenjenkins has left [18:20:05] --- stevenjenkins has become available [18:26:01] --- Jeffrey Altman has left [18:37:26] --- Jeffrey Altman has become available [19:25:42] --- dwbotsch has left [19:26:11] --- Rrrrred has become available [19:26:20] --- sxw mobile has left [21:42:35] --- deason has left [22:46:49] --- reuteras has become available [23:33:19] --- brantgurga has become available [23:41:48] --- dev-zero@jabber.org has become available