[00:54:07] --- broder@mit.edu/dr-wily has become available [00:57:45] Where does the git repo live right now? Is the current one still openafs-git.stanford.edu? [01:08:25] --- Jeffrey Altman has become available [01:35:10] --- Russ has left: Disconnected [02:07:20] Yes. But don't use it for anything 'real'. [02:07:58] Each time we rebuild, we rebuild the whole tree - so all of the SHA1s will change. You won't be able to pull our production tree (when we get there) into any of the previous trees. [02:45:27] --- dev-zero@jabber.org has become available [02:45:31] --- dev-zero@jabber.org has left: offline [04:45:09] --- shadow@gmail.com/owl912329A2 has left [05:12:41] --- shadow@gmail.com/owlFA95C1D6 has become available [05:30:26] --- Jeffrey Altman has left: Replaced by new connection [06:20:01] --- dev-zero@jabber.org has become available [06:20:22] --- dev-zero@jabber.org has left: offline [06:39:07] --- edgester has become available [06:40:27] Simon Wilkinson: I understand that I shouldn't use git for anything real yet. I just wanted to test out the workflow with gerrit as shown in your workshop slides. [07:28:55] edgester: In which case you want to clone from http://git.openafs.inf.ed.ac.uk/git/openafs.git [07:29:08] And use the gerrit that's located at http://gerrit.openafs.inf.ed.ac.uk/ [07:29:46] Ok, so if I want to submit patches I should still just do that through RT and off of CVS? [07:29:58] At the moment, yes. [07:30:09] We'll shout very loudly when things have changed over. [07:30:22] Cool [07:35:22] The list of things that we still need to get done is at http://www.dementia.org/twiki/bin/view/AFSLore/GitTools if anyone has any free time! [07:39:23] Hmm...I should go read some docs on gerrit [07:44:52] *blink* Wow - their diagram for the "default workflow" in gerrit is...complex [07:45:24] Are you guys planning to use Repo as well or just git? [07:58:52] --- edgester has left: Replaced by new connection [07:58:52] --- edgester has become available [08:01:46] I'm looking into sending e-mails to openafs-commits [08:02:35] It looks like users in gerrit can already subscribe to e-mails about when a changeset is "submitted". Could you create a fake account for openafs-commits and subscribe it to that? I don't know if the format is what you want [08:13:41] --- Jeffrey Altman has become available [09:38:10] broder: Just gerrit. We'll use normal git push and git clone, rather than the complexity of repo. [09:38:36] > We'll shout very loudly when things have changed over. [09:38:52] I think with regards to emails, I'd rather have something that hooks a repository we push too. That means that when Derrick commits something behind gerrit's back, the commit list will still get told. [09:38:57] i'd like to trumpet this point. it will be announced. you (for any you) can stop asking every 24 hours. [09:39:41] Sorry Derrick - I've been out of the loop for a few weeks and was confused by reading Simon's workshop presentation [09:39:57] you're hardly the only one. it's been a daily occurance [09:59:39] I'm getting compile errors when compiling a kernel module for 1.5.60 on Linux 2.6.27-14-generic (ubuntu Ibex). [09:59:56] I'm wondering if I'm missing a patch or something [10:00:12] ./configure --enable-supergroups --enable-large-fileserver --enable-bos-new-config --enable-debug --enable-lwp-debug --enable-namei-fileserver --enable-demand-attach-fs --enable-disconnected [10:02:35] instead of telling us that, the errors would be more interesting [10:04:09] CC [M] /home/jwedgeco/openafs/openafs-1.5.60/src/libafs/MODLOAD-2.6.27-14-generic-SP/rxstat.o make[6]: *** No rule to make target `/home/jwedgeco/openafs/openafs-1.5.60/src/libafs/MODLOAD-2.6.27-14-generic-SP/AFS_component_version_number.o', needed by `/home/jwedgeco/openafs/openafs-1.5.60/src/libafs/MODLOAD-2.6.27-14-generic-SP/libafs.o'. Stop. make[5]: *** [_module_/home/jwedgeco/openafs/openafs-1.5.60/src/libafs/MODLOAD-2.6.27-14-generic-SP] Error 2 [10:06:04] get a nightly snapshot. [10:06:10] it was fixed post 1.5.60 [10:06:22] ok, that's what I was wondering [10:23:37] woot! successful compile, thanks! [10:51:55] hmmm, lovely. pinstall breakage in the debian rules file. Gotta swap it with install [10:58:28] --- dev-zero@jabber.org has become available [11:01:59] --- cclausen has become available [11:14:40] edgester: I suspect that's due to the recent build system changes landing from HEAD. I guess Russ hasn't had a chance to update the rules file yet. [11:34:01] --- cclausen has left [11:39:50] --- Russ has become available [11:44:37] The debian rules file doesn't use pinstall. I don't think it ever has... [11:46:05] I already know how to send commit notices to openafs-commits and can set that up trivially as soon as we have a final repository. No one needs to spend a bunch of time on that. [11:50:00] I use a version of git-commit-email that the debian folks came up with and we use it for all of our repositories. I can hook that into the repo as a post-receive hook so it will send out mail whenever anything, including gerrit, pushes to it. [12:12:50] Sounds good. [12:13:32] Ok. I'll find something else to work on next time I get bored [12:13:52] I guess that means I just need to make sure that gerrit can do pushes automatically. I think I can configure that, but I haven't tested it. [12:26:37] broder: Yeah, sorry, I didn't realize that was on the project list. [12:26:53] I should have looked. [12:27:07] I'll annotate that list a little further ... [12:49:17] I'll ask this question here first, and take it to the mailing lists later... How much of wdelta's current functionality do we want to preserve in the git world. Do we just want to have the 'URL to fetch a delta' bit, or is browsing also required? [12:50:49] I use browsing to figure out what deltas have been applied since the last stable release to figure out what to pull up to Debian. [12:51:15] But with Git, I might be able to get that other ways. [12:53:56] `git log openafs_stable_1_4_9..HEAD` should get you that, right (roughly) [12:56:09] Although that'll probably only work within a single branch, since branches are usually cherry-picked, not merged [13:09:27] I think with git you'll definitely be able to get that in other ways. [13:11:57] Yeah, with one commit per delta, it's a lot easier. [13:12:06] Given that, I don't think we really need anything other than getting a diff for a delta. [13:12:33] Which I could also get from Git, but which some of our users might prefer to get from the web site when we point people at a particular delta. [13:13:58] We also need to be able to do that for 'old' deltas, where we need to join multiple commits together. [13:14:32] I'll hunt down the wdelta source and see how simple it'll be to teach it about git. If it's tricky, I'll probably just end up writing something hacky from scratch. [13:15:15] Hrm...why didn't I get any of the delta references when I cloned the repo? Do I have to do something special to suck those down, too? [13:15:23] Yes. [13:16:01] Either clone with --mirror, or specifically add a refspec that includes refs/deltas/* to your remote specification. [13:16:11] *nods* [13:16:38] Somewhere in the archives for this chatroom, there is the exact incantation. [13:19:33] I...think that `git config --add remote.origin.fetch +refs/deltas/*:refs/deltas/*` will roughly do the right thing, although I can't tell if I'd still get updates to actual branches :) [13:20:02] From memory, I think you will. [13:20:47] One thing to bear in mind is that pushing into a repository which has a huge number of references is very slow. It's likely that the gerrit repository will not contain the delta information for that reason. [13:23:51] Hmm...looks like you're not preserving committer information alongside author? [13:24:43] committer should be there. [13:24:51] If it's not, it's a bug. [13:25:34] Oh, never mind - git log and git show just don't display committer by default [13:25:42] Yeh. [13:25:55] I thought I remembered seeing them before - my bad [13:26:07] committer is whoever CVS says made the commit, author is whoever the CVS log information says wrote the delta. [13:26:15] Right, that makes perfect sense [13:26:20] Apart from the 1st commit, where the author is (now) hardcoded to be IBM. [14:10:00] Russ: you're right, the debian config doesn't use pinstall, but something is using pinstall and breaking my attempts to update the debian package files [14:15:23] * edgester smacks head. I used the 1.5.60-doc.tar.gz with today's snapshot. ugh [15:34:31] --- cclausen has become available [15:37:38] --- dev-zero@jabber.org has left [15:40:22] --- deason has become available [17:22:03] * edgester sings "I've been working on the deb files, all the live long day" [17:46:06] --- Russ has left: Disconnected [17:48:19] --- Russ has become available [19:21:11] --- edgester has left [20:02:26] I use the browsing capability of wdelta extensively to review the history of the project on various branches [20:03:45] I use it to look at changes that may be useful to me [20:46:16] Wouldn't git log be better for both now that we have a useful log command in the VCS? [20:46:31] I never use cvs log, but that's just because cvs log is horribly broken. [20:47:53] --- dev-zero@jabber.org has become available [20:47:57] --- dev-zero@jabber.org has left: offline [20:58:51] probably [21:17:28] --- deason has left [21:29:53] I can use the wdelta browser from my mobile phone. I can't do that with git log. [21:31:30] i bet there's a gitweb function you can use from your phone [21:31:42] http://git.eyrie.org/?p=kerberos/remctl.git;a=log;h=refs/heads/etch [21:40:00] that interface is sufficient for me [21:55:49] does that interface work on a phone? [21:55:55] it appears to load the entire changelog [22:15:18] All of it for that branch. [22:32:30] --- asedeno has left [22:32:30] --- asedeno has become available [22:32:33] --- asedeno has left [22:32:46] --- asedeno has become available [22:39:34] --- asedeno has left [22:39:39] --- asedeno has become available [22:48:18] --- cclausen has left [22:57:47] --- cclausen has become available