[00:10] mars: I don't understand your assertThat suggestion. [00:10] I've not seen it before. [00:10] wgrant, it's something newish that testtools ported from Java. You write a custom matcher and pass it into the assertThat() method along with the object to be asserted [00:12] wgrant, so you would write a matcher that does the "assertEqual(expected, list(object._myMethod()))" dance for you, as if you were writing a helper function or something [00:12] then you pass that matcher and the object under test into assertThat: assertThat(myobject, PassesMyTest(foo)) [00:13] Ah, I see. [00:14] The Java guys came up with it as a hack because assertEquals in Java just prints "True" or "False", which is really lame. So they created the assertThat() method so that it can actually print nice test failure messages. [00:14] However, it also serves as a nice way to pull common custom test conditions together as readable wrapper for helper functions [00:17] wgrant, here's the code: http://bazaar.launchpad.net/~testtools-dev/testtools/trunk/annotate/head:/testtools/matchers.py [00:19] I see it's not used anywhere in LP. [00:19] I'm not sure how to write a matcher. [00:21] wgrant, ok, maybe don't worry about it. I'm just reading the code now, and you have to define a new class to do it. [00:21] mars: Do you object strongly enough to NascentUpload's conditional __init__ that I should change the several dozen callsites which pass in a path? [00:21] wgrant: give me a few minutes and I can help; or you can look at at testtools.matchers, bzrlib.test.matchers, or james' external matcher project [00:21] I suppose it's nicer. [00:22] hi mars, wgrant [00:22] Morning poolie. [00:22] lifeless, I was thinking it was a nice way to wrap a repeated one-line assertEquals(), but I don't think the matchers are light enough to do that yet [00:22] Good morning poolie [00:23] i've been meaning to try joining up multiple related assertions under an aggregating matcher [00:23] so you get a better message when they fail [00:24] A decorator to somehow turn a helper function into a matcher would do the trick, but it's just a fancy way of writing a self.checkThat(foo) method. [00:26] mars: wgrant: whats the MP that this is being discussed on ? [00:26] i think there's a social effect here that we don't normally have tests failing very much [00:26] lifeless: https://code.edge.launchpad.net/~wgrant/launchpad/test-ddeb-matching/+merge/31482 [00:26] so having nicer failures is only a transient benefit [00:26] poolie, here as in Launchpad, or here as in Python? [00:30] in bzr, and probably in lp [00:30] maybe you get more failures in buildbot in lp [00:30] wgrant: ok, matchers will definitely be cleaner. let me sketch you a story [00:31] lifeless: Cleaner than a domain-specific assertion method? [00:31] yeah, they would be [00:31] yes [00:31] Hm. [00:31] definitely cleaner than inventing a per-test way of factoring it out [00:33] wgrant: you also have some dead code AFAICT [00:34] Hm, there's an 'except KeyError' that probably shouldn't be there any more. [00:34] But what else? [00:34] Oh, no, that should be. [00:35] What's dead? [00:39] wgrant: http://paste.ubuntu.com/472863/ [00:39] wgrant: I didn't quite get the intent of your function initially [00:40] wgrant: thats a sketch of a matcher : use it as self.assertThat(self.changes, HasCorrectDebugDebs()) [00:40] it has the following properties; it is itself testable [00:40] lifeless: Um, matchDDEBs is production code. [00:40] Not just for tests. [00:41] wgrant: thats interesting :) this can still fit in that context - matchers are not test-only code. [00:42] anyhow, if it fits - great; if it doesn't, thats fine too. [00:46] lifeless, poolie, in your 'switchable sandbox' workflow, does 'bzr push' still work as normal from within the tree'd branch you are working in? [00:46] wgrant, lifeless, mars: there will be lp.testing.matchers landing very shortly [00:46] or do I have have to cd out and into the branch I am switched to? [00:46] I would encourage lp..testing.matchers or similar too [00:46] james_w, cool, be sure to write the list when it does [00:47] very nice [00:47] I think it's either in PQM's queue, or got lost between ec2 and pqm [00:49] :( [00:49] I want to update the pre-requisite branch of a merge proposal [00:49] * thumper tweaks todo list [00:54] * mwhudson waves at james_w [00:55] Ursinha-afk: doh, I misse dyou [00:55] hi mwhudson [00:56] james_w: i was going to ask how things were going, but maybe we should do that in #linaro [00:58] How is stable ever going to get deployed in the new QA model? [00:58] If it has to be fully QAd... [00:59] where do I find launchpad pqm's web interface? [01:00] james_w, https://pqm.launchpad.net, but it won't help you much [01:00] james_w, you might have to grep the server logs to find your branch [01:00] oh, or not [01:01] the branch is in the queue [01:04] wgrant: a specific rev of stable is fully qa'd, not tip. [01:04] wgrant: I'll make that clearer [01:04] lifeless: Ah, I see. [01:04] mars: yes, it still works. [01:05] lifeless: flacoste's email this morning confused me. [01:05] yes [01:05] > Is there a facility to roll out just the QAd stuff? [01:05] No. Un-qaed revisions blocks deployment. [01:05] what he means is [01:05] It'll roll out an earlier revision. [01:05] we can only rollout qa'd stuff; and that means we can't skip a revision [01:05] if you have ABCDEF and A,C are QA'd we can only deploy A [01:06] B needs to be either rolled back, or QA-ok. [01:07] Right. [01:08] anyone seen ImportError: No module named debian on ec2 recently ? [01:08] That rev was just rolled back. [01:08] But I thought buildbot was broken -- not ec2. [01:09] wgrant, he may have fired up ec2 before the fix landed [01:11] lifeless: You need to merge in a more recent lp:launchpad/devel, which adds me to the list of valid image owners. Though it's not relevant anymore since that revision has indeed been reverted. [01:13] jelmer: into where - the ec2land running branch, my branch being submitted, or the target ? [01:13] lifeless, the ec2land running branch [01:13] ugh [01:13] but ok [01:26] is there a bug about the hot bugs list not updating bug status ? [01:26] lifeless: yes, is memcache misuse bug [01:27] poolie: yes, but whats the bug # [01:27] The milestone view is still broken too :( [01:27] * poolie looks [01:28] ah bug 599614 is the milestone page [01:28] <_mup_> Bug #599614: bad caching in milestone page [01:28] It's a lie. [01:28] Ah, different bug. [01:28] lifeless: i wonder if we should 1- just turn them all off; 2- turn them right down to be unobjectionably low; 3- push expiry; 4- force browser refresh [01:30] Bug #601051 is one. [01:30] <_mup_> Bug #601051: Caching of primary data sources reduces utility [01:30] so this is an interesting case to use approximation [01:30] if you say "about 200 open bugs" it's a bit more tolerant of being out of date, [01:31] but not really, because i could speedily close many of them, and then it would still be wrong [01:31] wgrant: that's a nice description of a general cause [01:31] is it really a registry bug? [01:32] Well, the problematic cases are mostly Registry. [01:32] But possibly not. [01:32] bug 604530 too [01:32] <_mup_> Bug #604530: brower refresh button should refresh TAL caches [02:29] is http://paste.ubuntu.com/472894/ not broken? [02:31] I'd say so. [02:31] It's been BPPH for yeeeears. [02:31] * mwhudson hasn't internalized the soyuz model well enough to say off hand [02:32] there's no distroarchseries on BinaryPackageRelease or its interface [02:32] * thumper has connection issues [02:32] make that three classes that don't implement the interface [02:32] there are people outside messing with wires [02:33] james_w: The idea is right. It's joining BPR to DAS via BPP. [02:33] Except that BPP is BPPH and has been for a long time. [02:34] does this imply that the "packages" attribute isn't used anywhere? [02:34] I would hope that was the case. [02:35] james_w: is it in the interface? [02:35] mwhudson: yes [02:36] sigh [02:36] delete it or write tests for it? [02:37] I'm inclined to delete it [02:37] delete it [02:37] if it's not used [02:37] which it seems like it can't possibly be [02:38] thumper: telecomnz seems to have had an outage [02:45] someone let that wet string dry out again? [02:51] :P [02:58] the first set of changes to wean SoyuzTestPublisher off the sampledata is now in ec2 to generate a list of fallout for me to work through [02:58] That could be amusing. [02:58] I'm doing it piecemeal [02:59] I'm considering demolishing parts of builder.txt. [03:00] this doesn't just make it ignore sampledata. There's one pipe to make the SoyuzTestPublisher back on to the factory, and another to give an API to avoid sampledata, with minimal changes to existing test paths. [03:00] Soyuz tests seem to be getting a bit of a makeover at the moment. [03:00] Ah, excellent! [03:00] that will allow us to gradually migrate [03:01] if it turns out there isn't much fallout then I'll just switch the existing code paths. The branch I just sent to ec2 does this for all the tests in test_publisher, just to get a feel for the fallout. [03:04] wgrant: so, I think I've captured what you wanted in bug https://bugs.launchpad.net/launchpad/+bug/601051 [03:04] <_mup_> Bug #601051: memcache is used as a bandaid for responsiveness [03:04] wgrant: if instead you meant to say 'page X is cached and shouldn't be', please change it to say that and put it on the right product. [03:05] lifeless: Indeed, thanks. [03:05] There are a few around now, so your changes sound good. [03:13] OMFG, my code is actually working [03:14] thumper: that's the beauty of statistics; eventually that million to one chance comes up! [03:14] haha [03:14] (I am *so* dead at the next allhands...) [03:14] fcuking aussies think they are better than the rest of us [03:15] s/think/know/ ;-) [03:15] I was amused to read an article on the sydney morning herald website [03:15] about how australia was going to beat the all blacks last weekend [03:16] one of his 10 reasons was "because australians are better than kiwis at everything" [03:16] ah yes. The SMH. Purveyor of Kwality Journalism. [03:16] Excellent reasoning. [03:16] another was "because it is about time" [03:16] and "we have too good a coach to lose to the all blacks 8 TIMES IN A ROW" [03:17] I gather the AllBlacks won then? [03:17] heh, by lots [03:17] ha! nice. [03:17] 49 to something small like 14 [03:17] orsum! [03:17] was it a decent match depsite the score diff? or just embarrassing? [03:17] wallabies got someone red carded for the first time ever against the ABs [03:17] Ahh. Biased ref eh? [03:17] no [03:18] stupid player [03:18] (zing. trolled!) [03:18] ABs had a player yellow carded for not using arms in a tackle (read as shoulder charge) [03:18] oof [03:18] Wallabies got a player yellowed for exactly the same reason with 5 minutes [03:19] ref warned both captains to stop slowing the ball down and he would move to yellow cards for anyone who continues [03:19] I wonder [03:19] for tag clouds [03:19] hrm. ok, It's been about 25 years since I last played rugby; but we were taught to more or less hit with the shoulder and wrap the arms around to pull 'em down. that no longer the case? [03:19] if using just the last 10K modified or something to summarise would be good [03:19] rather than 'all open' [03:19] wallaby player gets yellow carded for slowing the ball down, but it was the same guy, 2 yellows = red [03:19] (which we don't even do today) [03:20] spm: that's kinda OK as long as the arms are up and out at the time of shoulder contact :) [03:20] lifeless: I'd suspect it would be - if only for 'most recent ~= most relevant' [03:20] lifeless: Don't we use 'all open' at the moment? [03:20] lifeless: Apart from the dupe bug. [03:20] thumper: roight [03:20] wgrant: bug 404*** - tags from closed bug sshow in the cloud [03:20] wgrant: that might be dupe related, I dunno [03:20] Hmm. [03:20] wgrant: check your bug mail ~ 2am [03:21] * thumper is making bzr push lp:project work when there isn't yet a trunk [03:21] thumper: orsum [03:23] lifeless: Ah. [03:26] losa ping [03:26] yo [03:26] whats the unit on https://lpstats.canonical.com/graphs/MemcachedServersAll/ [03:26] requests per ??? ? [03:26] tribble [03:26] * lifeless trouts spm [03:26] spm: there is something I'm supposed to be talking to you about... [03:26] * thumper thinks [03:27] lifeless: I think it's seconds tbh. [03:27] spm: and https://lpstats.canonical.com/graphs/MemcachedTotal/ [03:27] spm: does that mean, as I think it does, that we're getting 20% hit rate? [03:27] lifeless: 5 min average, or possibly 1 min avg; per second. [03:28] lifeless: I guess so - I'd suggest verifying with stub tho; he's more across the fine details of those than me [03:29] meh, I'll shoot my foot off, its fine. [03:29] fwiw, the serversall is a tad busy, but somewhat by (my) choice. was thinking that if we note server X is showing numbers wildly different to A, B & C, then there is "A Problem™" [03:30] but by and large, they all track fairly close together, which is a good thing [03:31] mwhudson: please tell me that if the xmlrpc server returns a fault it aborts the transaction [03:33] thumper: i think so, you should be able to confirm by looking at the publication code [03:33] mwhudson: where is the xmlrpc publication code? [03:33] do you know? [03:40] thumper: otp [03:43] ack [03:45] :( [03:45] lp.codehosting.inmemory is going to make my life hard [03:47] * lifeless sets up a flamefest on lp-dev [03:48] lifeless: about? [03:49] memcached [03:49] is the set up you need to pqm-submit launchpad branches documented somewhere? [03:49] i know it used to be on launchpad.canonical.com [03:50] if this is wgrant's stuff, use ec2land ;) [03:50] it's just generally trying to recover my set up [03:50] i didn't lose much data, but i hadn't backed up lots of configuration :( [03:51] here is what I used: [03:51] echo "star-merge bzr+ssh://bazaar.launchpad.net/~lifeless/launchpad/malone bzr+ssh://bazaar.launchpad.net/~launchpad-pqm/launchpad/devel" | gnome-gpg --cl | mail launchpad@pqm.canonical.com -s "[r=mwhudson,thumper][ui=none][bug=607776] Further trade off ideal dup detection for speed, with an eye to bring back in ideal detection soon." [03:51] when I had something that I needed to do directly :P [03:52] I just use the pqm-submit plugin, and have an alias for db-submit [03:52] mwhudson: you need (or should have) the canonical smtp server set up for outgoing email [03:53] mwhudson: [/home/tim/repo/canonical/launchpad] [03:53] pqm_email = Launchpad PQM [03:53] submit_branch = /home/tim/repo/canonical/launchpad/devel [03:53] mwhudson: and set a public location for the submit branch [03:54] thumper: ta [03:56] hmm [03:57] my memcached mail seems to have gone awol ? [03:58] either that or lp list handling has gone boom ? [03:58] spm: ^? [03:58] I've got your email. ? [03:58] "So, we've got a partial deployment of memcache in place..." [03:59] ok cool. just gmail being stupid I guess [03:59] not showing me what I sent ... [04:00] thumper: the publication code is in canonical.launchpad.webapp.servers [04:00] mwhudson: please add to https://dev.launchpad.net/LandingChanges [04:00] thumper: and yeah, inmemory doesn't support transactions [04:02] james_w: actually maybe i am just being very antiquated in wanting to use pqm-submit [04:04] lifeless: One concern: as of 10.08 the front page makes requests to the blog to retrieve the feed. That's stored in memcached. [04:04] I'm not sure if that's going to be a problem, though. [04:05] well [04:05] we can keep that one page I guess. [04:05] (that said, the blog is local in the DC isn't it ?) [04:05] I think so. [04:06] remember that right now memcache isn't 'production available' - its not rated as a sev 1 when its down. [04:07] poolie: you said the other day that we could talk feature flags [04:07] poolie: did we? I forget [04:07] hey [04:08] i did say that; we didn't talk; talking now could be good [04:08] this is a good time for me actually [04:08] skype? [04:08] sure [04:19] poolie: https://bugs.edge.launchpad.net/ubuntu/+filebug-show-similar?title=lifeless%20is%20excited%20by%20the%20speed [04:20] hi [04:21] is there any doc howto install and configure launchpad dev? [04:22] j24: dev.launchpad.net [04:25] I've read it but not clear [04:25] I've install it [04:28] poolie: I've install by apt-get [04:28] and now want to run it [04:28] but it never start [04:29] what did you install? [04:32] launchpad-integration [04:32] $ co [04:32] & co [04:33] and what are you trying to accomplish [04:33] I want a local launchpad [04:33] to manage our own projects [04:33] ok [04:34] the "launchpad-integration" package isn't the launchpad source [04:34] I've install this [04:35] https://dev.launchpad.net/Running [04:35] http://paste.ubuntu.com/472926/ [04:35] yeah, that's not actually launchpad [04:35] as the name says, they're the dependencies and integration code [04:37] how can I install it so? [04:37] I want to install how own launchpad server [04:37] read that wiki page [04:38] is there any package that can do it on ubuntu? [04:38] with have to install the source [04:38] I've read the wiki but they always talk about manual settings [04:39] not howto install your own launchpad [04:47] poolie: if I download the source ok [04:47] it work if I run make run [04:47] but to setup a server it not good [05:03] poolie: how can people will use bazar for them self when it's not friendly at all [05:03] and ditto for launchpad [05:09] j24: Launchpad's not designed for you to install it yourself. [05:09] What is the prolbem that you have with Bazaar? [05:09] * mwhudson can't remember if he used to use gnome-gpg or seahorse [05:12] j24: I use bzr locally without launchpad for most of my work, what is it you are trying to do? [05:14] StevenK: ping [05:22] spm: yo! OSError: [Errno 2] No such file or directory: '/var/lock/launchpad-codeimportdispatcher.lock' [05:42] thumper: Pong [05:43] Is this how adapters are intended to be used?: http://www.muthukadan.net/docs/zca.html#adapters [05:45] cody-somerville: not necessarily how we do it [05:46] cody-somerville: you might want to take a look at BranchTarget as an example of how lp.code does it [05:49] Does launchpad use invariants, handlers, or subscription adapters? [05:50] i don't know what a subscription adapter is [05:50] i don't think we use invariants [05:50] what do you mean by handlers? [05:50] we do use zope.event a bit [05:50] lifeless: huh. so I see. two of 'em on diff servers. curious.... [05:50] mwhudson, handlers are subscription adapter factories that don't produce anything. [05:51] mwhudson, http://www.muthukadan.net/docs/zca.html#subscription-adapter [05:51] cody-somerville: no, we use a somewhat simpler thing [05:51] notify(event) [05:51] even implements some IFoo [05:51] and things that have register(IFoo) get called back [05:54] cody-somerville: ok, no i don't think we use subscription adapters [05:56] they look like a solution looking for a problem, to me ;) [05:59] We probably use them for publication hooks, but that's all I know of. [05:59] lifeless: O hai, no review? [05:59] grah [05:59] url me up [05:59] I shall do [05:59] lifeless: https://code.edge.launchpad.net/~stevenk/launchpad/move-ifp-from-idistroseries/+merge/31520 [06:01] StevenK: so [06:01] INSERT INTO is one case [06:02] I've since converted these calls to store.execute(), by the by [06:03] StevenK: I'm looking at store.execute on the merge page :P [06:03] * StevenK kicks his browser [06:03] _copy_lucille_config can be done with the GenericCollection stuff [06:03] its a set-wide mutation, which it supports. [06:04] right [06:04] so - INSERT INTO needs a storm feature, please file one (it needs INSERT INTO support, which AFAIK is not yet available) [06:05] the UPDATE can be done already without writing SQL. [06:06] I'm suprised Storm can't do INSERT [06:06] INSERT yes [06:06] INSERT INTO no [06:07] sorry, I should be clear INSERT INTO ... SELECT [06:07] lifeless: So it's an insert with a sub-select? [06:08] yes [06:08] it performs a query and inserts the result into a table [06:10] lifeless: Bug filed, #613300 [06:10] <_mup_> Bug #613300: Storm needs to support INSERT INTO ... SELECT [06:36] setting up the functional layer takes 25 seconds on this machine :( [06:37] hah [06:37] :( [06:38] lifeless: Anything else, re: that MP? [06:38] not at the level you were asking about [06:38] mwhudson: What is it, a coal-fired 486? [06:38] which was just 'why the raw sql' [06:39] StevenK: netbook [06:39] lifeless: Well, the other question was is the ducking underneath and all that still necessary? [06:40] what do you mean 'ducking underneath' ? [06:40] lifeless: In regards to the comment at line 108 in the MP diff [06:42] ah [06:42] well while INSERT INTO...SELECT is used; yes. [06:54] TWENTY FIVE SECONDS [06:54] spm: how about rt 40685 - is that blocked ? [06:56] mwhudson: What sort of hardware is that? [06:56] interesting [06:56] wgrant: it's an aspire one, some atom thing at 1.66 ghz [06:57] https://code.edge.launchpad.net/django <-- the first branch lists two series, one of the series doesn't even belong to the django project - is that a feature or a bug? [06:58] lifeless: only on resourcing afaict [06:59] spm: ok. [06:59] spm: wanna bikkit! [06:59] spm: you might like to eyeball https://dev.launchpad.net/LEP/ReleaseFeaturesWhenTheyAreDone [07:00] mwhudson: Haha. [07:01] * mwhudson having defeated zope again for the moment, retires for the evening [07:02] mwhudson: Did you get around to sending those branches off, or did the Atom defeat you? [07:02] wgrant: i didn't send your branches, sorry [07:02] mwhudson: No worries. [07:05] wgrant: Are they all totally good to go / [07:06] * StevenK kicks IArchiveSet.getByDistroPurpose() [07:08] lifeless: All except test-ddeb-matching are reviewed. I should probably get that one re-reviewed, since there were some changes made. [07:08] StevenK: Why? [07:08] StevenK: ArchiveCollection may interest you. [07:08] wgrant: thumper: I want a local bazarr management tool [07:08] and take a look on launchpad [07:08] it look good [07:09] but it's seems impossible to use for local projects management [07:09] wgrant: Because I'm calling it from a doctest, and it's printing a security proxied IArchive, and it gets called from the script, and it's None [07:09] StevenK: Awesome. [07:09] wgrant: FSVO [07:20] spm: any concerns about that wiki page from your perspective ? [07:24] lifeless: in stage 1, it might be worth raising the idea of switching app servers in/out of the pool of available services on the LB's [07:24] spm: isn't that whats done ? [07:24] nope [07:24] ... [07:25] tbh, short of config file changes and reloads of the LB's, it may not even be possible. [07:25] ok [07:25] I'm going to ostrich this one [07:26] heh [07:26] if it fuads [07:26] we'll stop moving through the process and address it [07:26] @ $job-1; we were able to switch servers on/off in the web ui; but as we only had 3 app servers to worry about; and rollouts were light; it wasnt hard to do so. [07:29] heh, and the idea of automated config changes to the running config of the LB's gives me the willies. switching set A in; out; B in etc. It may work; but *wince*. [07:30] it'd be nice if haproxy (it may, i just don't know) had some sort of remote push notify that a service was going down. vs it's current detect that service is down. === almaisan-away is now known as al-maisan [08:32] good morning [09:15]