[11:43] <ddaa> Meeting in 17 mins
[12:00] <ddaa> lifeless: spiv: SteveA: jamesh: MEETING STARTS
[12:00] <ddaa> == Agenda ==
[12:00] <ddaa> Next meeting 2006-09-25, 10:00-10:45 UTC.
[12:00] <ddaa>  * roll call
[12:00] <ddaa>  * production status
[12:00] <ddaa>  * importd batch progress
[12:00] <ddaa>  * release finder
[12:00] <ddaa>  * Python import
[12:00] <ddaa>  * strategic plan
[12:00] <ddaa>  * bzr-lp features
[12:00] <ddaa>  * interesting bzr list threads
[12:00] <ddaa>  * advertising
[12:00] <ddaa>  * 1.0 targets
[12:00] <ddaa>  * critical bugs
[12:00] <ddaa>  * any other business
[12:00] <ddaa> If you wish to change the time of the meeting or add/remove agenda items, say "bzzzt!".
[12:00] <ddaa> 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.
[12:00] <ddaa> == Roll call ==
[12:00] <ddaa> mpool is on leave until September 19th.
[12:01] <spiv> I'm here.
[12:01] <ddaa> Good morning spiv.
[12:01] <ddaa> Or maybe you wish to be called Ribri?
[12:01] <spiv> lifeless said on #bzr he might be a bit late to this meeting.
[12:02] <ddaa> righty, was chatting with him
[12:02] <jamesh> I'm here
[12:02] <ddaa> Good morning jamesh.
[12:02] <ddaa> SteveA: ahoy Great Launchpad Overlord!
[12:03] <spiv> I misread that as "Overboard", probably because of the "ahoy"...
[12:04] <ddaa> Waters here are nothing you want to be bathing in. They are fully of ravenous zope sharks.
[12:05] <ddaa> When they're done with you your split in at least four different bits: content, interface, browser and template.
[12:05] <SteveA> is it speak like a pirate day already?
[12:05] <jamesh> tomorrow
[12:05] <ddaa> really?
[12:05] <jamesh> http://www.talklikeapirate.com/
[12:06] <ddaa> Anyway... so lifeless late, everybody else on board.
[12:06] <ddaa> == Prodution status ==
[12:06] <ddaa> j-a-meinel has reported sftp mirroring latency problem again, on Friday. I got supermirror privs on vostok, but killing hung branch-puller scripts did not help. Increasing verbosity did not help, since that script appears to have no logging whatsoever.
[12:06] <ddaa> We need to fix that latency problem ASAP, it is harming public confidence in Launchpad hosting.
[12:06] <spiv> (And in six months time, at the opposite end of the year, I recommend everyone observes Pork Like a Tyrant Day...)
[12:07] <ddaa> I'm not quite sure where to start. But I think putting some logging to help debugging what's going on would help.
[12:07] <jamesh> the current puller is single thread
[12:08] <ddaa> I sure hope it's going to stay that way.
[12:08] <spiv> At the moment, a single slow/hung branch pull will block the rest of the queue.
[12:08] <jamesh> would we benefit from going multi-threaded?
[12:08] <ddaa> If we go parallel, I'd like to go multi-process instead of multi-thread.
[12:08] <ddaa> For one thing, that makes it much easier to kill hung tasks.
[12:08] <ddaa> it's also less of a debugging nightmare.
[12:09] <spiv> There's been some discussion on the launchpad list involving myself, lifeless and ddaa about how to improve things so that a) we can do things in parallel, b) so that hosted branch pulling is somewhat isolated from other branch pulling.
[12:09] <ddaa> But I think that if we want to do a quick functional change it should be just running three branch puller, one for external branches, one for imports, one for hosted branches.
[12:10] <spiv> It shouldn't be hard to log each URL before we try mirroring it.
[12:10] <jamesh> for the hosted branches and imports, we should be able to tell which ones need mirroring too
[12:10] <jamesh> if we can pass that info to the puller, then it can pull those classes more frequently
[12:11] <spiv> jamesh: we already pull all hosted branches every run.
[12:11] <ddaa> spiv: I talked a bit about my plan to SteveA, and how that related to importd-ng. It seems you and lifeless were thinking of something else. I'd like if you could help to clarify the disconnect.
[12:11] <spiv> (because there's only ~250 of them and they're local so it's acceptably fast for now)
[12:11] <jamesh> spiv: sure, but it won't be fast forever
[12:12] <spiv> ddaa: lifeless is the best person to articulate the exact idea posted to the list a while back... I'd recommend scheduling some time to talk it over with him.
[12:12] <jamesh> it'd probably also be worth looking at making the pull interval for a branch dynamic
[12:12] <jamesh> and suspend pulls of branches that time out
[12:13] <ddaa> Well, for starters. I'd lke to have each branch log at INFO level, for each branch: launchpad page url, source url, destination url. And when it fails, log the error that's put in Launchpad at WARNING level.
[12:13] <spiv> jamesh: Of course.  Just saying the "pull those classes more frequently" for hosted branches is a solved problem, for some values of "solved" :)
[12:13] <ddaa> jamesh: can you do that?
[12:14] <spiv> ddaa: I wonder if DEBUG level would be more appropriate, but otherwise I agree with the extra logging plan.
[12:14] <ddaa> maybe DEBUG and INFO then
[12:14] <jamesh> ddaa: okay.
[12:14] <ddaa> I would rather user DEBUG for INSANELY VERBOSE sort of things
[12:15] <spiv> ddaa: logging each branch will be insanely verbose eventually :)
[12:15] <jamesh> we can always run the script with "-q" to silence the INFO messages once everythings running acceptably
[12:15] <ddaa> and -qq to silence warning, I think
[12:15] <lifeless> hi
[12:15] <ddaa> hi lifeless
[12:16] <ddaa> ACTION: jamesh to put logging into the branch puller
[12:16] <ddaa> I would like to start working on the new architecture for branch puller and importd-ng soon. Like this week or next.
[12:17] <ddaa> Let's move on I think.
[12:17] <ddaa> == Importd batch progress ==
[12:17] <ddaa> Next meeting action:
[12:17] <ddaa>  * ddaa to discuss BatchProgress testing with lifeless
[12:17] <ddaa> Just a reminder. Lifeless asked to postpone that last week because of bzr release.
[12:17] <lifeless> grab me at 2130 AEST
[12:17] <ddaa> lifeless: anything you'd like to say about that right now?
[12:17] <ddaa> lifeless: talk UTC to me please
[12:18] <lifeless> 2130 UTC+1000
[12:18] <ddaa> lifeless: okay
[12:18] <ddaa> == Product release finder ==
[12:18] <ddaa>  * jamesh: report on PRF progress. In particular the outcome of reviewing the product:series:version:tarball table.
[12:18] <ddaa> jamesh: the stage is yours
[12:19] <jamesh> the product release finder can now run to completion, and I posted some results to the list
[12:20] <ddaa> was there noise-looking tarball names in that?
[12:20] <jamesh> there were only a few issues with the output: the code for extracting version numbers from filenames tripped up on "foo-1.0.orig.tar.gz" filenames, and the pattern for redland was too loose
[12:20] <jamesh> so picked up a redland-bindings tarball
[12:21] <jamesh> it might do to use more complex match patterns (modelled after uscan, maybe) to fix these issues
[12:21] <ddaa> What about allowing users to blacklist bad release names post hoc?
[12:21] <jamesh> I also did a productreleasefile/sourcepackagereleasefile cross reference to match productreleases with sourcepackagereleases based on tarball identity, which gave promising results
[12:22] <ddaa> jamesh: I did not see an email about that, that sounds interesting.
[12:22] <jamesh> ddaa: i don't think any of the problem names got through the filename version extraction, but it is something to keep in mind when modifying the code
[12:23] <ddaa> ACTION: ddaa to follow-up on ML about blacklisting bad release names to allow user to fixup bad glob matches.
[12:23] <jamesh> Some of the next steps would be to simplify the UI for entering the release file details to a single field rather than URL+glob
[12:24] <jamesh> and maybe move it to a different form than $series/+source
[12:24] <lifeless> triage:
[12:24] <ddaa> ACTION: jamesh to report on productrelease/sourcepackagerelease cross-checking
[12:24] <lifeless> allow setting the release details
[12:24] <lifeless> ddaa: hes already done that
[12:24] <lifeless> first: allow setting the release details:
[12:24] <lifeless> second: .orig filtering
[12:24] <jamesh> since +source can't be submitted without specifying VCS details
[12:24] <ddaa> lifeless: thanks, did not notice, maybe it's in the week-end email backlog
[12:24] <lifeless> third: deploy
[12:24] <ddaa> jamesh++
[12:25] <ddaa> Well, that or fix +source
[12:25] <jamesh> (we should probably make +source submittable with no VCS details anyway ...
[12:25] <ddaa> since it's broken in a few other ways
[12:26] <ddaa> I'd like to start working on +source soon too. It's incredibly confusing to users now. I'll put a few hours in that this week.
[12:26] <ddaa> ACTION: ddaa to start fixing +source
[12:26] <jamesh> ddaa: what do you think of making +source only about VCS details?
[12:26] <jamesh> the source package bit is already available on another form
[12:27] <ddaa> jamesh: there are good and bad things to it. The bad thing is that it would multiply the number of page loads for users setting up new products to fill in all the details.
[12:27] <ddaa> so I'm not quite sure yet
[12:27] <lifeless> focus gentlemen, the meeting is 50% done, and this is design that can be done on list or a bug
[12:27] <ddaa> yeah, let's move on
[12:28] <lifeless> do you agree with the priorities I suggested ? if so move on
[12:28] <jamesh> lifeless: sounds good.
[12:28] <ddaa> == Python import ==
[12:28] <ddaa> https://launchpad.net/products/launchpad-bazaar/+bug/56360
[12:28] <ddaa>  * ddaa: report on bzr testament encoding bug, and maybe progress of Python import.
[12:29] <ddaa> So, using bzr-0.9, plus a couple of fixes to importd and cscvs seems to fix the testament encoding problem.
[12:29] <ddaa> But the python import is still failing because after a few thousand revisions the svn server eventually reset the connection.
[12:29] <ddaa> and pysvn does not give us good exceptions to detect that sort of error.
[12:29] <ddaa> So I'll just keep restarting the import until it works.
[12:30] <ddaa> SteveA: talk about that later if you wish.
[12:30] <ddaa> == strategic plan ==
[12:30] <ddaa> Last meeting action:
[12:30] <ddaa>  * SteveA: jamesh: review 32/Bazaar
[12:30] <SteveA> hi
[12:31] <jamesh> I forgot about this.  I'll send some stuff to mbp tomorrow
[12:31] <SteveA> i haven't been paying attention -- didn't realize it was a meeting
[12:31] <SteveA> thought it was just high seas piracy
[12:33] <ddaa> Bah, let's move on.
[12:33] <ddaa> == bzr-lp features ==
[12:33] <ddaa>  * mpool: report on progress for bzr-lp features
[12:33] <ddaa> mpool is still on leave, so unless somebody else has something relevant to say, we'll move forward.
[12:33] <lifeless> move on
[12:33] <ddaa> == Interesting bzr list threads ==
[12:33] <ddaa> Follow up to last meeting. Do you guys have keywords for outstanding bzr threads from last week?
[12:34] <ddaa> Did not have time to read any of the ML last week, so that would help.
[12:34] <lifeless> nope, nothing interesting happened
[12:34] <lifeless> actually, thats a lie
[12:35] <lifeless> A bunch of interesting things happened, and they are now in my long term memory. Something will trigger an associative lookup but I'm *terrible* at date-based recall - its why history was really annoying for me
[12:36] <ddaa> spiv: jamesh: anything in particular your remember?
[12:36] <spiv> ddaa: Btw, there's a daily snapshot of the Python SVN repo at http://svn.python.org/snapshots/projects-svn-tarball.tar.bz2 -- perhaps doing the initial import from a local copy of that would be better.
[12:36] <jamesh> ddaa: not really.
[12:36] <SteveA> lifeless: I found history interesting after I left school.  Schools teach history wrong, in general.
[12:37] <SteveA> svn servers will reset connections.  it happens.
[12:37] <ddaa> spiv: good suggestion.
[12:37] <ddaa> ACTION: ddaa to look at tarball-based import of python
[12:37] <spiv> ddaa: j-a-meinel did a huge bunch of reviews of the smart server branches, I don't remember much else :)
[12:37] <lifeless> SteveA: oh I found it very interesting. And I'll occasionally make comments about william of orange at the right point in a conversation
[12:37] <SteveA> so, we should be able to deal with this. (eventually.)  do we have some kind of plan for that?
[12:37] <lifeless> SteveA: but asking me what year WWI started, and I'm screwed.
[12:38] <ddaa> SteveA: move away from pysvn and use python-subversion, presumably that would give more helpful exceptions.
[12:38] <SteveA> lifeless: the year isn't important.  its place in the flow of events is, though.  and that's what school teaching of history often gets wrong.
[12:38] <SteveA> ddaa: is there a better pysvn upstream we can use?
[12:39] <lifeless> SteveA: well, the flow I'm fine on ;)
[12:39] <jamesh> so they should teach students the partial ordering of events rather than the cardinality?
[12:39] <lifeless> jamesh: rotfl
[12:39] <SteveA> jamesh: exactly.  history as a digraph
[12:39] <ddaa> SteveA: no idea, but it's custom bindings for a GUI app, so it does not have the same requirements as cscvs.
[12:39] <SteveA> anyway, when you learn relativity too, you find it really is a digraph
[12:39] <SteveA> or many digraphs
[12:40] <ddaa> It looks like the bzr-lp threads thing is not a success. I suggest to drop it.
[12:40] <ddaa> == Advertising ==
[12:40] <ddaa> Last meeting actions:
[12:40] <ddaa>  * spiv: blog about similarities between SVN and bzr checkouts, in relation to Launchpad.
[12:40] <SteveA> threads?
[12:40] <SteveA> oh, right, asking people to note interesting happenings
[12:40] <ddaa> SteveA: your suggestion to asks the folks here for stuff they found interesting in the bzr mailingh list
[12:40] <SteveA> I would say that this week, perhaps nothing too interesting happened
[12:40] <SteveA> keep trying it
[12:40] <ddaa> SteveA: ok
[12:40] <ddaa> spiv: news?
[12:41] <spiv> ddaa: I have a draft.  I'll post it to the list shortly.
[12:41] <ddaa> Way cool.
[12:41] <ddaa> == 1.0 targets ==
[12:41] <ddaa> supermirror-smart-server: spiv: still looking on track for october 8th?
[12:41] <ddaa> importd-bzr-native: removal of Arch support code almost complete. Missing one importd patch, and the database patch. After the launchpad patch lands, I believe we will be able to delete pybaz, gnarly and bzrtools from the dists tree.
[12:41] <ddaa> bzr-roundtrip-svn: not for 1.0
[12:41] <ddaa> Pending action:
[12:41] <ddaa>  * mpool: read up/tick off svn roundtripping discussion
[12:42] <ddaa> spiv: how the supermirror-smart-server looking?
[12:42] <spiv> Very good.
[12:42] <spiv> Most of the work to date is in the 0.11 release candidate branch.
[12:43] <spiv> I have a branch where "bzr+ssh://" urls work.
[12:43] <jamesh> spiv: how easy will it be to integrate with twisted/conch?
[12:43] <spiv> (which we'll ask to merge into 0.11)
[12:43] <lifeless> nit: bzr 0.11 is frozen; rc1 is next monday
[12:44] <spiv> jamesh: Worst case, just let it spawn a bzr process.  But I think in-proc will be fairly straightforward.
[12:44] <ddaa> Anyway, I do not think the release mgmt would be a blocker to the lp feature.
[12:44] <spiv> jamesh: the serialisation logic is farily cleanly seperated.
[12:44] <lifeless> ddaa: it is if the protocols are not compatible
[12:45] <ddaa> lifeless: mh, I thought it was an entirely new transport.
[12:45] <spiv> Next steps: smart server-over-http, and supermirror integration.
[12:45] <ddaa> So, the supermirror-smart-server is on track according to spiv.
[12:46] <ddaa> == Critical bugs ==
[12:46] <ddaa> https://launchpad.net/bugs/31308 Cannot set branch associated to a product series. Fix commited.
[12:46] <ddaa> No new critical bugs. Will remove this section on the next meeting unless somebody really enjoys me saying "no critical bug" every week.
[12:46] <spiv> Yes.  lifeless has been a big help this last week.
[12:46] <SteveA> well
[12:46] <jamesh> I've also got a branch to add Product.development_focus in the PQM queue
[12:46] <SteveA> lower the importance
[12:46] <SteveA> so we get the top most important bugs
[12:46] <SteveA> it is great that we have few critical bugs
[12:47] <SteveA> do we have any highly important bugs?
[12:47] <lifeless> I have to go prepare the next meeting
[12:47] <jamesh> which will allow us to do the "lp:/python" style URIs (picking the default series)
[12:47] <ddaa> SteveA: many many
[12:47] <SteveA> so, oldest critical + oldest high
[12:47] <SteveA> max 7
[12:47] <ddaa> SteveA: that's why I asked for your feedback on bug triaging last week
[12:47] <SteveA> that's what we do in the launchpa dmeeting, anyway
[12:48] <ddaa> So. I think that meeting is done.
[12:48] <SteveA> ok, thanks david
[12:49] <lifeless> ddaa: did you deal with A
[12:49] <lifeless> https://lists.ubuntu.com/archives/launchpad-users/2006-September/000608.html
[01:21] <ddaa> lifeless: done on tuesday last week
[01:21] <lifeless> sweet, I can forget about it then
[01:23] <ddaa> look for emails with Yate in the Subject in launchpad-users
[01:28] <ddaa> lifeless: so, talk about BatchReport?
[01:28] <ddaa> hu, BatchProgress I mean
[01:28] <lifeless> -> #launchpad