=== ddaa [n=ddaa@nor75-18-82-241-238-155.fbx.proxad.net] has joined #launchpad-meeting === ddaa [n=ddaa@nor75-18-82-241-238-155.fbx.proxad.net] has joined #launchpad-meeting === poolie [n=mbp@ozlabs.org] has joined #launchpad-meeting [12:01] hi hi [12:03] oho, I'm late [12:03] jamesh: lifeless: poolie: spiv: SteveA: MEETING STARTS [12:03] ding ding ding [12:04] == Agenda == [12:04] Next meeting 2006-10-02, 10:00-10:45 UTC. [12:04] * roll call [12:04] * production status [12:04] * release finder [12:04] * Python import [12:04] * strategic plan [12:04] * bzr-lp features [12:04] * interesting bzr list threads [12:04] hi [12:04] * advertising [12:04] hi [12:04] * 1.0 targets [12:04] * critical bugs [12:04] * any other business [12:04] If you wish to change the time of the meeting or add/remove agenda items, say "bzzzt!". [12:04] If we are short on time, the "any other business" item will be automatically, dropped. So if you ''want'' to discuss something more, speak up now. [12:04] hi [12:04] == Roll call == [12:04] here [12:04] No excuse was given. [12:04] hi [12:04] ddaa: bw that did not hightlight the nick, if you want attendion I suggest one nic per line [12:04] lifeless: ack [12:04] sounds like an irssi bug to me [12:05] ACTION: lifeless fix irssi nick hilight bug for next week ;) [12:05] or at the least, a counter-intuitive policy decision [12:05] So, anybody has any extra agenda item for this meeting? [12:06] SteveA: poolie: maybe about the maybe-ex-Rotterdam sprint? [12:06] ddaa: everyone to use irssi for next week, so people are aware of its highlighting policies [12:06] I want to note that the rotterdam sprint will be in london [12:06] xchat should highlight things okay [12:07] let's do that first [12:07] ddaa: will that be OK with you? [12:07] there are several bzr meetings coming up: [12:07] SteveA: well, okay, it's going to cost some money in train because of the shorter notice and because I bought non-exchangeable tickets for Rotterdam. [12:07] ddaa: like 100EUR? [12:07] ddaa: yes. do the best you can. same dates, just london. [12:07] otherwise, the main issue with me is that I'm going to loose an opportunity to enjoy some .nl delicacy, but it's not really a work issue [12:07] presumably not so much compared to flying to sgp [12:08] poolie: all taking into account, more probably like 250 euro, not much a biggie though. [12:08] tim penhey will be joining Canonical working on launchpad-bazaar, starting on the 2nd october [12:08] please make him welcome [12:08] he'll be meeting in london with ddaa and steve to get into it [12:08] then will be coming to singapore to meet with mark, robert, stub, and me [12:09] Should we get him into next week's meeting? [12:09] then during november he'll be moving to NZ, [12:09] he will resurface in early december - this implies he will not be in SF [12:09] ddaa: good idea [12:09] ACTION: mbp mail tim about next launchpad-bazaar meeting [12:09] ddaa: one other item, 32/Bazaar === _thumper_ [n=tim@host86-141-71-114.range86-141.btcentralplus.com] has joined #launchpad-meeting [12:10] poolie: that's in the agenda already [12:10] okay, let's move on [12:10] oh my mistake [12:10] _thumper_: hi! [12:10] hi _thumper_ [12:10] _thumper_: welcome to the bzr-launchpad coordination meeting, Tim. [12:10] <_thumper_> hi all [12:10] thumper is Tim [12:10] the meeting is chaired by ddaa [12:10] Welcome Tim [12:10] <_thumper_> ddaa, Hi [12:10] So, this is our weekly meeting. [12:10] maybe some introductions are in order? [12:11] please do [12:11] ddaa: you're the chair... [12:11] ddaa: so I'm the guy currently going crazy trying to drive the launchpad-bazaar work [12:11] SteveA is my manager, one of the two launchpad project managers with kiko [12:11] _thumper_: hi there, welcome onboard [12:11] ddaa: arguably, you were crazy before you got involved in launchpad-bazaar [12:12] SteveA: induitably [12:12] well, he is french [12:12] lifeless is test-driven development god and knows more than you ever want to about http and weird corner cases in VCS [12:12] jamesh: I was thinking that, but I wasn't going to be the one to say it ;) [12:12] ddaa: also, I went to uni with _thumper_ [12:12] 'with' meaning about 2 years behind [12:13] <_thumper_> ah, ok [12:13] jameh is our polyvalent mr fixit, with an uncanny talent for getting his brain around complicated problem [12:13] <_thumper_> now that might be why a face isn't leaping to mind [12:13] _thumper_: I think you met me, ddaa and SteveA in London [12:13] _thumper_: I dont have a clear picture of your face either. We'll fix that in singapore ;) [12:13] spiv does many thing, but here he's the guy who wrote the bazaar.launchpad.net sftp server, and is currently writing the bzr smart server [12:14] poolie you already know, he's the bzr Overmind. [12:14] jamesh is James Henstridge. spiv is Andrew Bennetts, in case the irc nicks aren't clear. [12:15] lifeless is Robert Collins. [12:15] <_thumper_> SteveA, I've been using the mouseover in Konversation :) [12:16] nice [12:16] I suggest we move on so for once this meeting might end not late. [12:16] please do [12:16] though it looks already compromised... [12:16] we ended roughly on time last week I think [12:17] == Prodution status == [12:17] Nothing new to report on production. [12:17] Tomorrow is LP production rollout. Will require a importd rollout because of the ProductSeries.import_branch schema change. [12:17] * jamesh: put logging into the branch puller [12:17] jamesh: got news? [12:17] this is in the review queue now [12:17] _thumper_: http://www.mega-nerd.com/erikd/Img/codecon06-people.png [12:17] _thumper_: thats a recent (if scruffy ;)) picture of me [12:17] jamesh: do you think it's cherrypickable in production? [12:17] I can review that branch if you want [12:18] ddaa: should be. [12:18] <_thumper_> lifeless, which one? [12:18] _thumper_: haha [12:18] SteveA: thanks. It isn't very complicated. [12:18] jamesh: can you arrange with stub to get it rolled out tomorrow, then? [12:18] _thumper_: didn't realise it was one big picture - the left most is me [12:18] after SteveA reviewed it, since he so helpfully volunteered. [12:19] ddaa: I'll make sure it is listed as a cherrypick request [12:19] thanks lots [12:19] == Product release finder == [12:19] * jamesh: report on PRF progress. [12:19] * ddaa: start fixing +source [12:19] No progress on +source fixage last week. [12:19] no progress on product-release-finder this week. [12:20] jamesh: is that blocked on +source? [12:21] ddaa: the main issue is making it easier to enter release file globs. If we leave the entry on +source, then it might be blocked. If we move it to the product series +edit page then it isn't. [12:21] So, yeah, it's blocked on some UI fix. [12:21] I'm fine with you hacking what's needed into the UI [12:21] I think it should be a separate form, so we can make it clear what it's here for. [12:21] we can fix it later [12:21] the UI shoiuld never block things [12:22] I've got a discussion about 1.0 UI pending with mpt [12:22] jamesh: lets merge it [12:22] jamesh: as SteveA says, UI should not block [12:22] ddaa: I think it would be good to focus the $series/+source on VCS imports, so I'd lean towards moving the item over to the main edit form [12:22] so I think we can just do whatever work, we'll need to shuffle things around quite a bit anyway [12:22] which also means I'm not blocked on your work [12:23] okay [12:23] == Python import == [12:23] Ran tarball based-import of Python last week. It failed because of a logic error in the handling of svn copy simultaneous with a delete of a part of the copied directory. [12:23] This problem (and maybe a few variations on the theme) is also the cause of most of the svn import failures. I started working on that as a top priority. [12:23] Already put up for review some test cleanups that will make it easier to use separate test repositories. [12:23] ACTION: fix critical issue blocking python, and check that it allows the import to complete. [12:24] could say more things about how python-subversion is horribly broken pile of crud, but it would be offtopic [12:24] are we still waiting on the bzr update in rocketfuel [12:24] ? [12:25] hhu [12:25] good question [12:25] hmm, I thought I tried to send that in [12:25] bzr upgrade blocked by test failure in supermirror scripts [12:25] in addition to the importd-smallfixes branch [12:26] so some additional fixage is needed before bzr>=0.9 can be landed [12:26] Who think he can tackle that? [12:26] More recent versions cause more failures, I remember at least bzr.dev (0.11) causes failures in cscvs as well. [12:26] I think it needs a simultaneous update of bzr and launchpad, doesn't it? [12:26] jamesh: yes [12:27] that's the importd-smallfixes branch [12:27] but it also need more fixes [12:27] ok [12:27] I'm getting confused [12:27] david, please summarize clearly the steps needed to get a new bzr into RF [12:28] 1. fix the failures in the launchpad test suite with bzr-0.9 [12:28] that is, in addition to the fixes that are currently in ddaa/launchpad-importd [12:28] hu [12:28] currently in ddaa/launchpad/importd-smallfixes [12:29] I suggest just doing bzr 0.11 FWIW. we're at rc1 [12:29] +1 [12:29] then 2. fix new failures caused by 0.11 in cscvs [12:30] the launchpad test suite fixes use the bzrlib.urlutils module, which isn't available with the current version of bzr, so the simultaneous update is needed [12:30] 3. land new bzr simultaneously with ddaa/launchpad/importd-smallfixes, plus maybe other whatershed fixes [12:31] That's it. [12:31] in point 2., by "failures" you mean "failed imports" or "test failures2 ? [12:31] test failures [12:31] does that mean you want us to go to 0.9 and then to 0.11 ? [12:31] or can we go straight to 0.11 at step 3? [12:32] we can go straight to 0.11, I can work with bzr-0.9 in importd until then [12:32] ok [12:32] lifelses will do step 3 [12:32] who will do steps 1 and 2? [12:33] !lifeless [12:33] I'd be interested in doing them, so I can get a bit more familiar with the supermirror code. [12:33] twelve minutes left in this meeting [12:33] thanks lifeless [12:33] let's take this to mail [12:33] ? [12:34] ddaa: ok, let's talk about your workload after this meeting [12:34] ack [12:34] thanks, I now understand the steps involved [12:34] == strategic plan == [12:34] * mpool owns that agenda item [12:34] thanks [12:34] https://wiki.canonical.com/32/Bazaar [12:34] this is almost approved by mark [12:34] thanks very much to ddaa and robert for their comments [12:35] I added some small comments today. Sorry for the delay [12:35] i think i'm still waiting for some others here to read it, or tell me they did [12:35] jamesh: thanks [12:35] it looks good [12:35] spiv said he read it and had nothing to add [12:35] poolie: you already have my comments [12:35] _thumper_: you haven't been asked before, but i'd appreciate if you read that page [12:35] ddaa: right [12:35] <_thumper_> poolie, just clicked on it [12:36] i realize it references many things you probably won't be familiar with, you can ask me about them if you need to [12:36] thanks [12:36] spiv: thanks [12:36] ok then that's everyone i think, thanks v much [12:36] == bzr-lp features == [12:36] * mpool: report on bzr-lp features [12:36] poolie: it's your again! [12:37] poolie: what bzr-lp features are currently worked, what are the short term plans, what is blocking? [12:38] spiv: can you tell us about supermirror smartserver? [12:38] that's in the 1.0 targets [12:38] poolie: No work directly on that yet -- work should start on that shortly, after HTTP smart server/client is working. [12:39] poolie: Probably will discuss it with you and lifeless in person tomorrow. [12:39] ok [12:40] poolie: I guess you do not have much to talk about that this week [12:40] moving on [12:40] not really [12:40] jamesh: i think i saw something from you about fixing branch redirections [12:40] ? [12:40] T -5 [12:40] spiv/poolie: please mail a brief update to the launchpad list after you talk tomorrow. [12:40] skip it [12:40] == Interesting bzr list threads == [12:40] Let's skip that this week. Look for the minutes of the past couple of meeting for details of what this is about. [12:40] SteveA: ok. [12:40] == Advertising == [12:40] * spiv: blog about similarities between SVN and bzr checkouts, in relation to Launchpad. [12:41] spiv: can you please update the SmartServer spec to have the current protocol and design decisions [12:41] spiv: I do not think you posted the draft last week [12:41] spiv: its aged. [12:41] Right, it's here: https://devpad.canonical.com/~andrew/paste/fileywRJCw.html [12:41] I'll also send mail about it after the meeting. [12:41] lifeless: ok [12:41] ok [12:41] poolie: there was some talk about how lp: URIs should be handled on the launchpad list. Also, tomorrow's rollout should have all the database changes needed to identify branches by product and series [12:42] Comments welcome. [12:42] I have a few nitpicks in my bag left. I'll give you some. [12:42] == 1.0 targets == [12:42] supermirror-smart-server: spiv: still looking on track for october 8th? [12:42] importd-bzr-native: ddaa: doing final cleanups and hct cruft collection. Currently prioritised out by python import fixage. [12:42] bzr-roundtrip-svn: not for 1.0 [12:42] Pending action: [12:42] * mpool: read up/tick off svn roundtripping discussion [12:43] i did read it, will send mail [12:43] ddaa: I think so, yes, although there's still lots to do. [12:43] poolie: I'd love your feedback on those discussion, they touche many important subjects like integrity checking, bzr copy, and forward compatibility for svn imports. [12:44] IMO just get the tranport-level rpc thingy up, with the lp glue would already be a remarkable achievement. [12:44] considering the extremely short timeline you had [12:44] ddaa: right, that's what I'm aiming for. [12:45] I have to go and be ready for the next meeting [12:45] its now T-0 [12:45] another late meeting [12:45] moving on [12:45] ddaa: I'm very interested in the hct stuff, I was hoping to update that for looms rather than us discarding it [12:45] lifeless: I've asked for your feedback about that some time ago [12:46] anyway, it has just been bitrotting since March [12:46] for the smart-server, I'm confident andrew and I can deliver, if we keep pairing : progress has been fast while pairing [12:46] the hct stuff is still in the source tree, in the not-used place [12:46] lifeless: the hctapi stuff is very out of date, and we'd want to do it differently now [12:46] so, it is easy to see the code and revitalize it, if you want to [12:46] but it is out of the way of general development [12:46] then, there is the sourcecode/hct tree, but that's another discussion [12:46] e.g. use xmlrpc.launchpad.net rather than the separate trebuchet server [12:46] SteveA: sure, I dont know enough about where scott put *what* to know whats important at the moment [12:47] all I know for sure is that sourcerer et all are bitrotting [12:47] hctapi is a small bit, and AIUI not significant in the grand scheme [12:48] Moving on, then. [12:48] == Important bugs == [12:48] Last week, Steve asked me to pick a few important bugs to highlight. Here's my selection of the week. [12:48] * Related to +source [12:48] bug 2649: CVS branch details should not be editable or displayed. [12:48] https://launchpad.net/products/launchpad-bazaar/+bug/2649 [12:48] bug 46240: posting $series/+source yields a confusing warning [12:48] https://launchpad.net/products/launchpad-bazaar/+bug/46240 [12:48] bug 50569: the product series page does not allow entering source or ftp details for projects without SVN or VCS [12:48] https://launchpad.net/products/launchpad-bazaar/+bug/50569 [12:48] * bug: 48813: Efficiently mirroring sftp hosted branches with minimal latency [12:48] https://launchpad.net/products/launchpad-bazaar/+bug/48813 [12:48] * bug: 58889: Merged and abandoned branch should not appear in main branch listings [12:48] https://launchpad.net/products/launchpad-bazaar/+bug/58889 [12:49] jamesh: I take it you'll do 50569 this week to complete the PRF work. [12:49] ddaa: okay [12:49] I'll talk about which, if any, of those bugs I'll be working on this week later with SteveA. [12:50] That said, I think the agenda is complete. [12:50] running late, so no "any other business?" [12:50] MEETING CLOSED [12:51] Thanks you everybody for attending. [12:53] Offtopic ranting [12:54] This week-end I started some experimental branch to cleanup the libsvn binding of cscvs [12:55] I came to the conclusion that the alledgedly "complete" Python bindings of subversion were not only undocumented, but largely unusable because they do not seem to have any support for callback methods. [12:55] Which are pretty essential to many things, like setting log messages from the svn_client API [12:55] the cscvs and bzr-svn code seem to avoid the issue by talking repository-level operations directly [12:55] which involves hundreds of line of duplicated logic [12:56] ddaa: which python library were you playing with ? [12:56] python-svn or python-subversion ? [12:56] svn [12:56] python-subversion [12:56] IIRC that thats the swig bindings [12:56] pysvn is nice, but it's quite limited [12:56] yes [12:56] and its horrendously incomplete. [12:56] Ugh, swig. [12:56] when I did the SVN module for cscvs, it did not even support doing a *commit* [12:56] The good news is that I finally managed to get it to commit with a python-provided log message callback [12:57] lifeless: it still does not [12:57] I'm considering the option of setting up a alternate binding (maybe called python-aversion?) [12:58] which is a basically an experimental fork of python-subversion, and hammer on that until it does all that cscvs needs [12:58] with feedback to the svn guys [12:58] so I'd suggest a different tack [12:58] It looks like swig would support that approach nicely. [12:58] pyrex [12:58] clean start [12:59] lifeless: I'm very much looking for something that could be merged upstream into subversion [12:59] swig is a good way to go blind, and much less satisfying than masturbation [12:59] and there is still a lot of work done in python-subversion, I am not keen at starting up from scratch [01:00] the biggest issue seems to be the lack of documentation, so it's very hard to figure out what is really unsupported, from what is just normal libsvn pain. [01:00] subversion has significant investment in swig, it's used for the ruby and perl bindings too. [01:00] so I think there's no choice here [01:01] there are good arguments for and against each of the options [01:02] lifeless: btw, python-subversion has much saner memory management now [01:02] you can essentially just ignore all the apr pool crap [01:02] we should come up with a short experiement to help us estimate what is the best course of action [01:02] <_thumper_> using swig for all bindings would also mean that python bindings are not likely to be pythonic [01:02] for example, for someone to spend a day writing throw-away pyrex bindings for it [01:02] I guess using it explicitly can still give performance improvements, but it's significantly better than it used to be. [01:03] and to spend a day extending the swig bindings in some way that we need [01:03] and comparing progress [01:03] SteveA: hu, libsvn is very large [01:03] this test would be very biased, because starting from scratch would give fast progress anyway [01:03] _thumper_: they have different swig rules for each language [01:04] _thumper_: I believe libsvn is made of the same stuff as the souls of the damned in hell. [01:04] _thumper_: I dont think they influence the pythonicness of the bindings much, beyond the swig-suck factor [01:04] <_thumper_> lifeless: so a different mapping file for each language? [01:04] so it's going to be painful either way, unless you put the work into thick bindings like pysvn, but then it would probably be better to extend pysvn. [01:04] <_thumper_> I have used swig before but for python and java [01:05] _thumper_: IIRC, yes. I haven't looked in 2 years though, so.. [01:07] how many SVN calls does CSCVS need? [01:07] we don't need complete bindings. we need bindings that cover what we need. [01:07] not many many [01:07] but the current code is probably much less efficient that it could be [01:08] so, depending how many different calls we need [01:08] pyrex may be a good solution [01:08] as we can do just the calls we need [01:08] and ensure they work well [01:08] but then it's an increased maintenance burden [01:08] SteveA: I agree there are good arguments for pyrex [01:09] maybe it would be a reduced maintenance burden [01:09] we need concrete figures [01:09] ddaa: can you think of a good way of finding out how many SVN API calls we use? [01:09] grep [01:09] :) [01:09] or instrumentation [01:10] grep is good enough, should be one hour of work at most [01:12] SteveA: if you have some time, we can talk about my tasks for this week now [01:12] then I have to lunch, then get a haircut so I do not look like a terrorrist on my USA-compliant passport. [01:12] ok [01:13] how's the arch removal stuff? [01:13] anything at all left to do there? [01:14] SteveA: it's just in the same status as when we last talked about [01:14] the ProductSeries.targetarch* patch needs to DBA-reviewed [01:15] then patch your looked at can be landed, with some additional cleanups to allow removal of pybaz friends from the dists config [01:15] duh, my grammar sucks [01:17] SteveA: that's all there is about arch removal. [01:17] then, there's hct, which is the big unknown [01:19] So my plan for this week is 1. work on python-blocker as soon as the current branches are reviewed 2. extra garbage collection in remove-gnuarch 3. bzr-0.10 compatibility fixes [01:20] BTW, I started using loom when working on cscvs, it's really cool. [01:20] oops [01:21] SteveA: do you have anything you would like to change in that priority list? [01:22] for python, my plan is to check the import completes as soon as I have the minimal fix, using a custom cscvs, and do my best to have rename support ASAP, to redo the import from scratch quickly [01:28] sorry -- call with mark [01:28] I'm back [01:28] maybe ping stu to see if he can do that DB review today [01:29] "extra garbage collection" means what? [01:29] removing references to pybaz etc. from lib/ symlinks and Makefiles [01:29] ok [01:29] not implementing a GC [01:29] moving hctapi.py out of the way [01:29] ROTFL [01:30] Do I have THAT much of a reputation for being sidetracked??? [01:30] no comment [01:30] I take that as a yes. [01:30] ok, looks good [01:31] are there any bzr-0.11 compatibility things you can ask jamesh to do? [01:32] there can certainly be, but 1. I'd like to actually poke the supermirror code a bit, because I'm not familiar enough with it 2. the cscvs fix are better done by me because I am currently actively working on that code base [01:32] how about asking james to do that and have you review it? [01:33] watching is not same as doing [01:33] there are only so many hours in the week [01:33] If you think that's critical enough to warrant that, I'll do it. [01:34] but I do not see the urgency [01:34] I want you to be using the same bzr for importing as we're using in RF [01:35] otherwise... [01:35] well [01:35] well, if I have to choose between fixing bzr-0.10 compatibility and implementing svn rename support... [01:35] okay, so it means you can get on with a buildbot replacement sooner [01:35] the other thing jamesh can do is to look into what parts of SVN API we use [01:35] and try the experiment with pyrex [01:35] this would be interesting, but it was just a little skunk project [01:36] I do not mean to divert workforce on that, the current svn binding situation is not very aesthetic, but it's good enough [01:37] if it turns out you use just 10 API calls [01:37] then custom bindings would be a clear win [01:38] I'd like you to be able to start replacing buildbot sooner [01:38] rather than later [01:38] You have all my support for that plan. [01:38] and, getting jamesh to help with bzr updates will help make that sooner [01:38] then we have a deal [01:39] I'm looking forward to more deep-immersion coding [01:39] I think after all I may not have that much of a future in mgmt :) [01:39] ok. please email james cc list about what's needed for bzr 0.11 compatibility [01:40] will do [01:41] and also, book your travel to london today [01:41] how will you travel? [01:41] train, as usual [01:41] ok [01:41]