[00:41:49] --- manfred furuholmen has left [01:03:08] --- dev-zero@jabber.org has become available [01:25:57] --- manfred furuholmen has become available [03:21:37] --- dragos.tatulea has become available [03:21:44] howdy [03:21:50] hi [03:22:00] hi Simon [03:22:12] what's up? [03:22:27] Plodding along. Knee deep in theatre at the moment :) [03:23:26] How are you getting on? [03:24:08] Working working working. [03:24:26] (also procrastinating a bit :) ) [03:24:43] I'm good at the last part... [03:25:08] and oafs wise...I'm hoping to be able to start pinning soon [03:25:23] Cool. [03:25:33] now that I managed to fix some bugs [03:25:49] I've got some big cache manager changes so we can do vcache disk storage. Need to get them working on a few more platforms. [03:25:55] but it's going really slow because of little devel time [03:26:09] ohh, that's awesome [03:26:55] The problem is that I'm changing struct vcache so that everything that goes to disk is in one block (makes saving and loading much quicker), but it means that all of the structure accessors change. [03:27:20] that's big [03:27:38] Yup [03:29:59] --- stevegt has left: Replaced by new connection [03:31:23] any news about transition to git since the hackathon? [03:32:36] Not that I've heard. [03:33:22] I was pushing quite hard then that we should set a deadline, and if we weren't done by then, we should just use the standard CVS import tools, and rebase if changes in delta form eventually become available. [03:33:33] There's lots of things we want to do that are now blocked on the git conversion. [03:34:02] what things? [03:34:28] gerrit for code review. [03:35:11] oh, what's with gerrit? that's the tool that google released for android scm [03:35:12] right? [03:35:18] Yes. [03:35:31] and why do we need it? [03:35:41] http://openafs-gerrit-demo.appspot.com/ [03:36:21] It provides a way of automating the code review process. As well as a good patch browser (you can see changes in as much context as you like), it lets you annotate those changes with review comments. [03:36:48] Once a patch has been approved by the required number of people, you can apply it to the tree by just hitting an 'apply patch' button. [03:39:13] I suspect any answers about git conversion won't come today though, unless people get very bored of turkey. [03:39:25] lol [03:39:57] yeah, the app demo looks nice [03:41:29] "standard CVS import tools" for importing into git? [03:41:37] git cvsimport [03:41:49] Which is what that demonstration gerrit instance is built on. [03:43:23] The problem is that it doesn't do the right thing with our deltas, so we inevitably lose history. And, we'd need to have two sets of 'delta' tools on the website - one for pre-transition, and one for after. [03:44:00] --- manfred furuholmen has left [03:45:55] hmm [03:50:07] --- abo has become available [04:37:32] --- Simon Wilkinson has left [04:54:46] --- Simon Wilkinson has become available [05:39:58] --- edgester has become available [06:30:20] --- reuteras has left [06:30:25] --- reuteras has become available [06:44:53] happy turquoise day [06:45:41] :) [06:45:55] Happy "No quite December yet, but let's have a party anyway" day. [06:46:12] we are hosting this year. lots of food [06:46:18] need more mouths [06:46:26] Excellent. Much turkey being consumed? [06:46:36] at 5pm [06:46:59] speaking of turkeys [06:47:00] http://blog.wired.com/wiredscience/2008/11/turkeytech.html [06:47:37] ours is 20lbs [06:47:45] I think you've got the right idea. I've got to wait another month for turkey goodness. [06:50:31] what is your usual christmas meat? (it sounds a bit funny, but you get my point) [06:51:00] ham, roast beef, [06:51:07] sometimes turkey [06:51:33] we haven't done xmas ribs in a while. they are just too messy for 70 [06:51:38] in my country it's pork [06:52:08] --- reuteras has left [06:52:41] have a great day. back to the kitchen [06:53:26] bye [06:53:37] how about UK? [06:56:00] Turkey generally. [07:03:52] hi simon, the gerrit code review thing looks interesting [07:10:26] Yeh. It's quite neat. [07:10:35] It's really stalled on the git conversion at present, though. [07:13:56] hmmm, I was wondering how well it would work with git [07:14:29] It's designed to work with git. It only works with git. [07:14:49] ah, I guess I'm curious how the workflow works with git [07:15:44] I guessing it's like this: I clone openafs git, make changes in my git repo. push to gerrit, which pushed to openafs git. [07:19:01] Pretty much, yes. [07:19:16] There's a tool 'repo' which some of that from the user. [07:19:24] You tell repo to sync with OpenAFS [07:19:46] You tell repo you're about to start work on a new set of changes (similar to starting a new patch with quilt) [07:19:46] hmmm, does the approval and comment data get pushed into openafs git repo? [07:19:59] The approval data does, the comment data doesn't seem to. [07:20:05] bummer [07:20:32] We'd probably change that, either to push it into git with everything else, or to create closed RT tickets with the comment information in them. [07:23:16] sounds good [07:24:22] I feel a little bad about not helping with the CVS2git transition, but I had no clue how to use git. I've recently started using git for a personal project, so I'm learning about it [07:24:28] At some point, I'm going to write something detailing how all of this could work, but I've been wanting to have a 'real' version of the git repository to build the demo system on. [07:24:46] that's understandable [07:25:31] My personal view (as I said earlier, not sure if you were around) is that at some point we need to give up on doing git 'right' and just get it done. There's so much that's backed up behind it now. [07:25:32] I haven't submitted my test code for inclusion in cvs, maybe I will for git [07:26:04] ah, so move to git and fix some history later? [07:26:46] Yup. Move to git with the naively generated history. Then, should a better history become available, we can then just rebase the tree onto that one at a later date. [07:27:12] hmmm, will that play havoc with any cloned trees? [07:27:34] It would mean that the cloner has to update their clone, yes. [07:27:51] k [07:28:43] have we used CVS in an odd way that makes the conversion difficult? [07:28:59] Yes. We use deltas. [07:29:10] ah [07:29:27] Wanting to keep delta information (so you have approximately one git changeset per delta) is what makes the whole thing hard. [07:29:40] ah [07:29:57] > Wanting to keep delta information (so you have approximately one git changeset per delta) is what makes the whole thing hard. [07:30:02] Why is it that hard? [07:30:09] I don't know about importing tools [07:30:16] in git [07:30:28] It's hard because of the way we've used deltas. [07:30:44] but shouldn't it be smth easy like get one commit, and put it in git? one by one? [07:30:47] A delta isn't necessarily all committed at the same time. [07:31:11] huh? [07:32:04] what do you mean? [07:32:05] So, say you've got fileA. delta 1 gets committed, fixing an issue at the start of that file. delta 2 then gets committed, adding some code to the end of fileA. A problem is then found with delta 1's fix, and a revised version is produced. That gets committed, but is flagged as being part of delta 1. [07:32:32] sheesh [07:32:32] Our current delta tools will give you both changes when you request the patchset for delta 1. [07:33:00] nasty [07:33:20] Now magnify that across a large codebase, and with a large number of other fixes coming between the original delta 1 and the fix. [07:35:03] That's why all of this is a hard problem. Max and Mike made some progress at the Hackathon (although maybe they were just saying that, as we were threatening not to let them eat if they didn't ;-) ), but it wasn't done by the end of our time at Google, and I'm not sure how much has happened since. [07:43:27] --- reuteras has become available [08:19:16] --- reuteras has left [08:40:22] --- manfred furuholmen has become available [08:40:24] I always found it strange that so many have pork as the traditional Xmas meat given that Jesus was most likely kosher and wouldn't have eaten a honey glazed ham [08:41:02] Ah, but perhaps you are mistaking midwinter as a Christian festival. [08:42:29] (many of the European 'Christmas' traditions predate Christianity) [09:02:10] i have a experimental cvs-git conv on http://www.beolink.org/downloads/it/ [09:06:51] --- dev-zero@jabber.org has left [09:07:01] http://www.beolink.org/downloads/git/ [09:36:36] --- dragos.tatulea has left [10:36:18] --- manfred furuholmen has left [10:57:03] --- dev-zero@jabber.org has become available [11:24:28] --- dragos.tatulea has become available [13:20:52] --- manfred furuholmen has become available [13:22:58] --- RedZBear has left [13:32:30] if someone want check a git repo.. you can find it at http://www.beolink.org/cgi-bin/gitweb.cgi [13:52:50] --- dragos.tatulea has left [14:03:11] --- stevegt has become available [14:08:18] --- edgester has left [14:12:33] --- edgester has become available [14:31:31] --- manfred furuholmen has left [14:44:22] That's pretty neat. Does it solve the double commit -> one delta problem? [15:23:48] --- dev-zero@jabber.org has left [16:04:06] --- tkeiser@sinenomine.net/owl has become available [17:02:08] --- Marc Dionne has become available [17:07:34] In the git repo I have (manfred's version from nov. 11), deltas with multiple commits show up as separate git commits [17:08:49] are they supposed to show as one commit? [17:20:41] sounds like maybe that's what we want. i don't see why that would be a big issue though [18:35:19] --- Marc Dionne has left [19:40:47] deltas should show up as one commit. that is the history we want to keep [19:43:26] cool, then it sounds like Manfred's git repo is what we want [19:48:41] that is not how I interpretted Marc's comment. [19:49:10] Manfred's git repo as I understand it is just a cvs2git import. deltas are not preserved. [20:35:32] --- edgester has left [21:31:10] i should confirm that. it will be easy to confirm. if so, that's unfortunate as i could have (and did) already do that myself. several times. [21:32:55] yeah ok. "deltas with multiple commits show up as separate git commits" is not what we want [21:45:31] as to "giving up" the thing we work with, at this point, should be the thing mike and max have now and not just git cvsimport, since it already does something better than the baseline. but max claimed he was going to do a little more on it, and since i haven't had time to get any infrastructure set up to deal, i wan't going to force the issue. [22:12:36] --- dev-zero@jabber.org has become available [23:27:29] --- reuteras has become available [23:29:35] --- dev-zero@jabber.org has left [23:35:11] --- stevegt has left: Replaced by new connection [23:35:12] --- stevegt has become available [23:53:17] --- manfred furuholmen has become available