exarkunAnd another one, http://buildbot.twistedmatrix.com/builders/lucid32-py2.6-select/builds/1091/steps/bzr/logs/stdio00:04
glyphpoolie: hi, exarkun found some bugs00:26
lifelesshi nigelb00:28
nigelbhey lifeless00:28
nigelbhmm, I thought this was #launchpad-dev for some reason.00:29
lifelessnigelb: nup :P00:29
nigelblifeless: you're everywhere, which confused me :P00:30
poolieglyph, nigelb, lifeless, hi02:38
=== krow_ is now known as krow
nigelbhello poolie07:17
jamhey poolie, I'm going to try to be back in time for the stand-up, but we're getting a late start at the house today. So don't wait on it for me.07:22
pooliehi poolie, hi nigelb07:37
pooliehi jam07:38
poolielifeless, hi07:43
pooliere bug 81960407:43
ubot5Launchpad bug 819604 in Bazaar "when an idle ssh transport is interrupted, bzrlib errors" [Critical,Confirmed] https://launchpad.net/bugs/81960407:43
poolieyour last comment there is correct as far as it goes but it's not really where i thought we ended up07:47
pooliemaybe i'll just reply07:48
poolieactually my question is really: why are you unable to deploy to codehosting again?07:51
pooliehow is this any worse than before/07:51
poolielifeless: ^^?07:53
lifelesspoolie: we've stopped scheduling monthly downtime07:56
lifelesspoolie: and in terms of  being worse than before, we had to kill the service every deploy, before.07:57
poolieso, if there is a disruption, it will disrupt them more frequently?07:57
lifelessif we want to be able to experiment effectively we need a graceful option, otherwise each test is downtime.07:57
pooliebut, for other reasons, every deploy is downtime at the moment07:58
poolieeg the xmlrpc error that comes back during a deploy07:58
poolieunless that's recently fixed07:58
lifelessthat ticket is in progress with hloeung at the moment07:59
lifelessshould be fixed in a day or two I hope07:59
pooliethat's great07:59
pooliewhat will happen to connections that are busy for a long time and not idle?08:00
hloeungyeah, it should be done within the next couple of days.08:00
hloeungpoolie, I believe the hard timeout is 2mins08:00
pooliewhich hard timeout?08:01
lifelesspoolie: appserver connections? or bzr connections?08:03
lifelessso, I would like to interrupt them at the next verb08:04
lifelessand if we have that, we could set a decent (say 1 hour) period after which we would interrupt anyway08:04
poolieok, but what will happen today if we fix this particular bug?08:04
lifelesswe'd reset the timeouts08:04
lifelessand have a fallback limit of 30minutes after which we would interrupt anyway08:05
lifelessthe lower backend timeouts would cause a timeout after (N, to be determined - say 3) minutes of inactivity, which would likely be low enough to get the CI servers and PQM and the like to migrate over to the new service instance.08:06
lifelesswe will need to address both bugs before we can deploy reliably with minimal interruption to users.08:07
jelmervila, do you have a pointer to the testtools-related failure on babune?08:10
poolielifeless: so, you'll leave the old back ends running for up to an hour?08:11
vilajelmer: http://babune.ladeuil.net:24842/view/%20High/job/selftest-freebsd/lastFailedBuild/#showFailuresLink08:12
lifelesspoolie: something-like, once we've got all the friction sorted08:12
lifelesspoolie: at the moment no timeout is high enough - we've still got a back end running from when we cut over to the haproxy setup last week, I believe.08:13
jelmervila, poolie, jam: http://bugs.python.org/issue1254408:21
gourfor which date is 2.4 scheduled?08:29
vilagour: we're talking about it right now, it depends on https://bugs.launchpad.net/bzr/+bugs?search=Search&field.importance=Critical&field.status=New&field.status=Incomplete&field.status=Confirmed&field.status=Triaged&field.status=In+Progress&field.status=Fix+Committed08:30
pooliegour: sometime between this thursday and two weeks after that08:30
vilagour: mostly bug #771184 and bug #80904808:30
ubot5Launchpad bug 771184 in Bazaar "option to disable/enable fetching of all tags" [Critical,Confirmed] https://launchpad.net/bugs/77118408:30
ubot5Launchpad bug 809048 in Bazaar "bzr crashed with AttributeError in stopTest(): '_TypeEqualityDict' object has no attribute 'clear' - breaks bzr selftest on oneiric" [Critical,Confirmed] https://launchpad.net/bugs/80904808:30
gourcool...it's on time08:30
gourwhat's with colo branches in 2.4? they are in core?08:32
poolienot in core; very good in bzr-colo08:34
gourthank you08:36
ubot5Ubuntu bug 806348 in Ubuntu Distributed Development "BzrCheckError: Internal check failed: Cannot add revision(s) to repository: missing chk node(s) for id_to_entry maps" [High,In progress]08:37
jelmerjam, vila: ^08:37
gourthanks to jelmer & {hg,git} plugins, bzr is the only dvcs we use now :-)08:37
vilagour: \o/08:37
gour(after spending many years with darcs, played with mtn, fossil, hg...)08:38
=== med_out is now known as medberry
gouri've asked web2py to upgrade repo format to 2a..in order to provide correct info, since when is bzr using 2a as default?08:49
lifelesssince 1.19 or something - its in NEWS08:49
lifelesspoolie: did you want to talk about these codehosting bugs?08:50
poolieon the phone at the moment08:50
pooliewhich ones?08:50
gourlifeless: ta08:50
fullermdThink it only because default in 2.0.  It existed in 1.18...08:51
gourfullermd: ahh...that's what we need. thanks08:52
fullermdSounds like 1.17 has support for it too...08:53
jamvila: http://babune.ladeuil.net:24842/job/selftest-chroot-oneiric/lastBuild/testReport/junit/bzrlib.tests.test_selftest/TestTestResult/test_verbose_report_known_failure/09:11
jamcan't we just do a getattr() to handle different testtools/python versions?09:12
jelmerjam: instead of checking the version you mean?09:15
jelmervila: is babune perhaps running some other pre-0.9.12 snapshot of testtools?09:15
jamjelmer: I think checking versions is going to be hard to get right09:15
jelmerjam: 0.9.12 introduces the fixes, 0.9.11 doesn't have them09:16
jamjelmer: that specific bug is in bzrlib because of attribute issues with python's runner, right?09:16
jelmerjam: ah, sorry09:16
jelmerjam: There are multiple issues here09:16
jelmerthat one is unrelated to testtools, I've just put up a proposal for it that uses getattr09:17
jelmerjam: https://code.launchpad.net/~jelmer/bzr/2.4-809048-workaround/+merge/7083409:17
lifelesspoolie: the two: gracefully handle tcp disconnects during idle periods of bzr+ssh connetions, and secondly a way to tell the smart server 'disconnect after serving the next verb or now if you are already idle'09:33
pooliei agree they'd be useful09:36
pooliei see you mentioned them in bug 795025 but they seem a bit distinct from the main bug in the bzr-sftp front end, so i might file two separate bugs09:37
ubot5Launchpad bug 795025 in Launchpad itself "Codehosting service didn't stop in a timely manner" [Critical,Triaged] https://launchpad.net/bugs/79502509:37
pooliei'm going to downgrade bug 819604 because it doesn't meet our definition of critical09:37
ubot5Launchpad bug 819604 in Bazaar "when an idle ssh transport is interrupted, bzrlib errors" [Critical,Confirmed] https://launchpad.net/bugs/81960409:37
poolieie it's not a release blocker09:37
poolieand create series tasks09:37
poolieit should be fixed though09:38
lifelesspoolie: so bug 795025 is about not shutting down; the *cause* for that is active sessions having no way to interupt09:42
ubot5Launchpad bug 795025 in Launchpad itself "Codehosting service didn't stop in a timely manner" [Critical,Triaged] https://launchpad.net/bugs/79502509:42
lifelesspoolie: I don't particularly care if you want to split out a bug asking for a way to interrupt, but it seems unnecessary09:43
lifelesspoolie: these are critical from LP's side, in the sense of queue jumping bugs; and (as I think we've established) we cannot sensibly move forward on the forking-daemon until we have 795025 addressed in some fashion09:44
poolieso for 795025 the only change needed to fix it is for bzr to close the connections, either on request or after a 5m timeout?09:45
poolie(or, whatever idle timeout?)09:45
lifelesson request09:46
pooliei guess idleness can be done sufficiently well from haproxy,because the socket will just be idle09:46
pooliewell, as long as the timeout is long enough we're sure it's not just that one side is busy09:46
lifelessa timeout in the server would be a nice bonus, but unless the work just fell out while looking at on-request, I would do it separately.09:47
pooliei only ask because you suggested a timeout09:47
poolieso as far as deployments09:47
lifelessah, I was describing the overall situation before, was probably unclear09:48
lifelessyes, I think haproxy is reasonable for timeouts at least in the medium term. We will have to fiddle the timeout counter in it to find a sweet spot eventually.09:49
=== gour1 is now known as gour
pooliethe issue is basically that if you upgrade/downgrade one of the back ends, you will need to choose between leaving the back end live indefinitely, or perhaps making long-lived clients error out09:50
poolieis it intolerable to make them error?09:50
pooliedoesn't tarmac cope with this already?09:50
lifelessno, tarmac errors out, pqm errors out, bzr-eclipse errors, buildbot errors.09:50
poolieand doesn't recover?09:51
pooliehow has it possibly coped with our monthly roll outs so far?09:51
lifelessI'm sure it does when it opens a new transport09:52
lifelessusers seem to find hour long test runs being aborted a bit of a pain09:52
pooliei think i understand the situation09:54
lifelessperhaps we can touch base tomorrow; its getting late here :)09:56
pooliehere too09:56
pooliei don't want to have a long discussion about whether to fix it or not09:56
pooliewe should fix it09:56
pooliei just want to have clear meanings for bug metadata09:56
poolieas do you09:56
poolieunfortunately the meanings are a bit different :)09:57
lifelessthats a beer topic I think09:57
gouris there something one would miss for not using colo-branches when simply using shared repo and each branch in its own dir, iow. should i bother with colo-branches?09:57
pooliethe main difference is that it's more biased towards having one working tree09:59
poolieyou can also also use colo-sync-from and colo-sync-to10:00
gouryeah, but i do not get if having one working tree is so important10:00
poolieit's good if the tree is large relative to your disk (or other resources) or if it takes a long time to compile10:00
poolieor if your tools like staying in one location10:01
poolieor if your working methods do10:01
gouri believe that using feature-branches is then good-enough for me10:01
gourwe'll stay with python for all development (qt & web2py), so compile is not really issue here (except some cython)10:02
pooliejam, could you maybe give francesco a bit more of a tip where to add the test or where to look for inspiration?10:14
jamvila, your recent mp: https://code.launchpad.net/~vila/bzr/migrate-config-options/+merge/7076712:36
vilajam: yes ?12:37
jamYou mention taking the register key from the option object12:37
jambut I don't see that in the patch12:37
vilajam: different mp12:37
jamvila: read what you wrote on that one12:37
jambullet point 212:37
jamI guess you were saying these are things that aren't part of this but would be done next?12:38
jamI thought it was the list of things in this patch12:38
vilajam: read what I wrote in introduction: 'follow-up'12:38
vilajam: yup12:38
jamvila: I tend to skim and focus on bullet points12:38
vilano, the introduction was just one line12:38
jamputting things that aren't part of the patch into the MP... can be confusing12:39
jamanyway, looks good12:39
jamjust confused12:39
jamvila: I have a big concern about your config stack work13:10
jambooleans don't come back as booleans...13:10
jamI'm looking at how Martin is using "repository.fdatasync" and I'm pretty sure it is wrong13:11
jamhe is doing:         self.write_stream.close(13:11
jam            want_fdatasync=self._pack_collection.config_stack.get('repository.fdatasync'))13:11
jamwhich is sort of what I would expect. But you can't set it to "False" or "F" or ... all of those evaluate to True in python13:11
jam(because they are strings)13:11
jamI don't see any tests from poolie that the config item can actually disable fdatasync13:12
vilajam: I'm working on the boolean config options just now13:14
vilajam: you're right that there is a bug right now, my next submission will fix it though13:14
jamvila: I'm looking to do something with it for 2.4. Should I just use "get_user_option_as_bool" for now?13:16
jamI was also a bit disturbed that just adding that suddenly made an HPSS ratchet test fail because it went from 14 HPSS calls to *34*13:16
vilajam: yup13:16
jambut that could have been my fault.13:16
jamI haven't worked through all the failing tests yet13:16
vilajam: what means "something" ? You're backporting the fdatasync stuff ?13:17
jamvila: I was using fdatasync to figure out how to do "no fetch tags"13:18
jamonly to find out it didn't work13:18
vilaha right, adding a new boolean option then ?13:18
jam(repository.fdatasync as an example to crib from)13:18
jamalso, oddly the HPSS verbs are different13:18
jambranch.get_config() calls 'Branch.get_config_file'13:19
jamBranchStack() calls 'get'13:19
jamnot even 'get_bytes' from what I can tell13:19
vilaStore calls get_bytes, Stack doesn't know about which files are involved13:21
vilabut fdatasync shouldn't care about remote repositories (unless there is a verb for that) or it's a bug no ?13:23
vilas/care about/apply to/13:23
vilajam: be aware that remote branches doesn't obey bazaar.conf nor locations.conf, they need a specific stack (TBD)13:24
jamvila: Remote cares about fetch_tags13:25
jamI didn't test anything for fdatasync in that area13:25
vilajam: hehe, indeed :)13:25
blackarchonsomeone here involved in qbzr or explorer?13:27
blackarchoni'm wondering if there is indeed any code to get the window positions and window dimensions from qbzr.conf?13:28
jamblackarchon: I haven't poked at that code recently, but ISTR that there are window positions that are remembered13:31
jamI think it involves deriving from a particular class13:31
jamlike QBZRWindow or somethign13:32
blackarchonwell I'm asking because this doesn't seem to be working at all13:33
blackarchonok, I will try to find the exact place for this restoreSize function which is called13:34
vilablackarchon: AFAICS, there are options for *size* but not for *positions*14:23
hikikoI would like some help: I ve created a branch of a project and I want to commit to it14:44
hikiko(a launchpad branch)14:44
hikikowhen I use the command: bzr commit -m "blah blah"14:45
hikikoI get the following error:14:45
hikikobzr: ERROR: Cannot lock LockDir(lp-86944400:///%2Bbranch/stellarium/.bzr/branchlock): Transport operation not possible: readonly transport14:45
hikikoalthough i am logged in in launchpad14:46
hikikodo you have any idea on what I am doing wrong?14:46
vilahikiko: you probably don't have *write* access on this branch (actually, you're probably using a checkout which is bound to the remote branch instead of a local branch that you then push somewhere else)14:52
hikikohow can I change this? I mean what I did was: checkout the project code14:54
hikikothen create my launchpad user and branch14:54
hikikoand then attempt to commit14:54
hikikowhat step is missing?14:54
hikikoI see14:55
hikikoyou mean i have to do sth like:14:55
hikikobzr branch lp:~<link to my branch>14:56
hikikoand then change that code14:56
hikikoand bzr push lp: ...14:56
blackarchonyes, this would work.14:57
fullermdIf you checked out _before_ creating your LP user, the checkout is connected over a _non-logged-in_ transport.14:57
hikikoyes I checked before14:57
hikikoand if I want to update15:01
hikikobzr up15:01
hikikowill give me the new code from the branch or from the project?15:01
hikikook I found it :)15:04
hikikothank you very very much for your help all!15:04
blackarchonum... where can I find the default ignore file, which is installed with bzr?15:11
LeoNerd~/.bazaar/ignore  perhaps?15:12
blackarchonno, I edited it - and left no backup of it :(15:12
fullermd'm pretty sure bzr writes it if it doesn't exist...15:13
blackarchonoh yes? let me try...15:13
LeoNerdmv it away and start again?15:13
fullermdSemirelatedly, it's probably rather discouraged to be editing it, as a rule.  Since it would no longer match what anybody else has, you can wind up with surprises in branches multiple people touch.15:15
blackarchonwell no, it doesn't get magically recreated :(15:15
fullermdIf I mv mine out of the way and run a 'bzr stat', it writes out a new one...15:16
blackarchonah yes, but only if this bzr command succeeds15:17
blackarchonmy bzr stat wasn't inside a bzr directory, so it hasn't created the file - thx! :)15:18
blackarchonI want to fix this bug: https://bugs.launchpad.net/qbzr/+bug/57893515:29
ubot5Ubuntu bug 578935 in QBzr "qbrowse crashes when run in repository: 'NoneType' object has no attribute 'last_revision_info'" [High,Confirmed]15:29
blackarchonso I added a check if the argument isn't a branch or a wt15:30
blackarchonbut I still need an idea of how to proceed, I want to disable all objects in qbrowse15:31
blackarchonbut I don't have the right hought on how to do that15:32
blackarchonany hints?15:35
=== beuno is now known as beuno-lunch
vilamore than 3 hours to land on pqm, is it me or it's getting worse ?16:44
jelmervila, it's not just you16:48
vilajelmer: :)16:49
=== deryck is now known as deryck[lunch]
jelmervila: can you submit my 2.4 branch? I don't seem to have permission to submit to release branches17:08
vilajelmer: 2.4-809048-workaround ?17:09
jelmervila: thanks17:12
vilajelmer: nothing showing up on pqm for now :-/17:13
vilajelmer: ok, here it is17:16
vilajelmer: thanks for that ! Merge it into trunk when you can !17:17
jelmerwill do :)17:18
vilaok, I'm off, see you online on Thursday ;)17:21
=== beuno-lunch is now known as beuno
=== deryck[lunch] is now known as deryck
santagadajelmer: dots don't work on bzr 2.1, only none works19:51
santagadajelmer: and I think that for deployment we can't wait for launchpad so we will just tar everything up to the clients19:52
=== r0bby is now known as robbyoconnor
lifelessmorning poolie23:02
pooliehi thar23:28
=== yofel_ is now known as yofel

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!