2012/06/13/#launchpad-yellow.txt

[12:05] <gary_poster> benji, did you see william & my comments on https://bugs.launchpad.net/launchpad/+bug/996729 ?
[12:05] <_mup_> Bug #996729: zope.testing --subunit allows bad output on stdout, which can break subunit processing and tests <paralleltest> <qa-needstesting> <Launchpad itself:In Progress by gmb> < https://launchpad.net/bugs/996729 >
[12:05] <gary_poster> summary: we've got unknown and mysterious work to do
[12:05] <benji> gary_poster: not yet; taking a look
[12:06] <benji> "In a world, where unknown and mysterious work must be done..."
[12:06] <gary_poster> heh
[12:08] <gary_poster> bac benji frankban gmb call in 2
[12:08] <gmb> ack
[12:08] <bac> rt
[12:10] <gary_poster> bac benji frankban gmb https://plus.google.com/hangouts/_/5e18fb3d6f696ec0a22278e5d5f27e055d2f4ef4?authuser=1&hl=en-US
[12:11] <gmb> argh, conneciton errors...
[12:22] <gary_poster> https://bugs.launchpad.net/launchpad/+bug/996729
[12:22] <_mup_> Bug #996729: zope.testing --subunit allows bad output on stdout, which can break subunit processing and tests <paralleltest> <qa-needstesting> <Launchpad itself:In Progress by gmb> < https://launchpad.net/bugs/996729 >
[12:31] <benji> https://pastebin.canonical.com/67982/
[12:34] <bac> rejoining
[12:43]  * gmb lunches
[12:43] <bac> so benji i've already "released" my changes on top of yours as p10.  are you going to revert your change and make p11?
[12:43] <benji> bac: sure
[13:52] <bac> benji: please ping me when p11 is available
[13:53] <bac> benji: or i can do it if you're busy
[13:53] <benji> bac: that would be great, I'm neck deep in this at the moment
[13:54] <bac> benji: okey doke
[14:13] <frankban> gary_poster: for my little plugin, should/could I use email@canonical and AGPL?
[14:14] <gary_poster> frankban, yeah, your canonical email would be good since it was slack time project.  AGPL...shrug I guess that's the right thing to do, yes.  Is it a problem?
[14:16] <frankban> No, I don't think.
[14:22] <gary_poster> cool
[14:22] <gary_poster> benji http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/waterfall
[14:36] <bac> benji: in the future would you do work on shared branches (like zope.testing) in dev branches so there is one checkin-per-revision bump?  the reverse cherry pick worked fine but p9 spanned about 5 branch revisions.</picky>
[14:37] <benji> bac: that's a good idea
[14:39] <bac> gary_poster, i've made my changes and updated zope.testing.  i'm going to do a full ec2 run to ensure the zope.testing drop doesn't munge anything.
[14:40] <gary_poster> great thanks bac.  I suggest you further manually examine the results before you land
[14:40] <gary_poster> in particular, make sure test counts don't change
[14:40] <bac> gary_poster: oh, ok.  let me go kill that 'ec2 land' then
[14:40] <gary_poster> bac, I can run it on the 32 core machine if you'd like, in a bit
[14:41] <bac> gary_poster: you may if you want.  seems like overkill since we are going back to what should be a known version.  my changes shouldn't cause any problems wrt test counts.
[14:42] <gary_poster> bac, neither should ours have :-P
[14:42] <bac> true enough
[14:50] <gary_poster> benji, maybe you didn't commit the change to p6 to your branch?  it isn't here
[15:46] <gary_poster> benji http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/4/steps/shell_9/logs/subunit
[16:02] <gary_poster> benji http://pastebin.ubuntu.com/1039294/
[16:05] <gary_poster> benji lp.testing.tests.test_zope_test_in_subprocess.TestZopeTestInSubProcessLayer
[16:09] <gmb> benji, gary_poster, Quick question: what would be the sanest way for zope.testing to issue a warning about duplicate tests? An exception is simple, but the bug talks about warning for now (and I don't want to break any more test runs just yet).
[16:09] <gmb> It would seem that writing to stdout or stderr would be a bad idea, especially if --subunit is switched on... Any ideas?
[16:09] <gary_poster> gmb, I'l be curious to see what benji thinks, but since bac has fixed the existing tests, I think it is OK to raise an (informative) error
[16:10] <gary_poster> gmb, IOW, there should be nothing to warn about at this time because bac has fixed it
[16:10] <gary_poster> So let's jump straight to the error
[16:10] <gmb> gary_poster, Something like ("Duplicate test id found: %s" % test.id)? Or are you thinking more info than that.
[16:10] <gmb> But okay, I'm happy to do that.
[16:10] <benji> +1
[16:10] <gary_poster> gmb, yeah, sounds like a good start :-)
[16:10] <gmb> Wahey.
[16:11] <gmb> Okay, branch coming shortly.
[16:11] <gmb> s/branch/mp
[16:11] <gmb> zt
[16:11] <gmb> Er.
[16:11] <gmb> Note to self: x-chat != vim
[16:25] <gmb> Ahaha.
[16:25] <gmb> Hmm.
[16:25] <gmb> So, my fix for bug 682711 breaks the test suite for zope.testing quite comprehensively.
[16:25] <gmb> Hmm.
[16:31] <gary_poster> benji https://bugs.launchpad.net/launchpad/+bug/986429
[16:31] <_mup_> Bug #986429: lp.testing.tests.test_zope_test_in_subprocess.TestZopeTestInSubProcess.test deposits global "zope:layer" tag in subunit stream <paralleltest> <qa-untestable> <Launchpad itself:Fix Released by gmb> < https://launchpad.net/bugs/986429 >
[16:33] <gmb> Ah, wait, no it doesn't. My assumptions were broken.
[16:47] <gmb> gary_poster, So, interesting meta-problem: the tests for zope.testing contain a doctest called src/zope/testing/testrunner/testrunner-debugging-layer-setup.test, which writes a python file, then loads it and runs it. The file it writes out contains two layers and a function with a doctest, and when that doctest gets loaded, of course, it blows up because the doctest gets registered twice. Any ideas about how I should handle that?
[17:25] <gary_poster> gmb, sorry, was lunching
[17:25] <bac> gmb, yeah i saw repetitive test ids in my change in the zope.testing tests.
[17:26] <gary_poster> gmb, make it a flag?
[17:26] <gary_poster> and then our bin test can always turn the flag on, he said, waving his hands?
[17:43] <gary_poster> benji https://plus.google.com/hangouts/_/afdae2431954bae99aae1d5075449031334b6481?authuser=1&hl=en-US when you are ready
[17:57] <gmb> gary_poster, Yeah, that sounds like the sanest solution.
[17:57] <gary_poster> cool
[18:43] <bac> gary_poster: my branch has failed twice in ec2 with hung tests.
[18:44] <gary_poster> bac :-(
[18:45] <gary_poster> bac, you want me to run it in parallel machine for you?  I could do it in about 40 minutes
[18:45] <bac> sure
[18:46] <bac> gary_poster: it is  lp:~bac/launchpad/bug-682772
[18:46] <gary_poster> bac ok will get it set up for you and ping you when done
[18:46] <bac> gary_poster: do you mean you can get to it in 40 minutes or it will take 40 minutes?
[18:47] <gary_poster> bac, I meant #1 but I changed my mind.  It will take 40 minutes
[18:55] <gary_poster> bac, fwiw it is running here if you want to follow along http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/9
[18:56] <gary_poster> bac, which tests hung, out of curiosity?
[18:58] <gary_poster> bac, it looks like list-tests is broken http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/9/steps/shell_9/logs/stdio
[18:59]  * bac looks
[19:00] <bac> gary_poster: where does that show brokenness?
[19:00] <gary_poster> bac, it is actually supposed to run tests after it runs --list-tests
[19:00] <gary_poster> bac, it didn't get any tests to run
[19:00] <bac> gotcha
[19:01] <bac> gary_poster: the two ec2 runs ended like http://paste.ubuntu.com/1039543/
[19:02] <gary_poster> bac, that looks like it in includes an incomplete version of what benji used
[19:03] <gary_poster> bac, have you verified that what you have differs from the p6 version in only the way you expect?
[19:03] <gary_poster> in zope.testing
[19:03] <bac> gary_poster: i have not done that.  i examined the diff from the reverse cherry pick and it looked reasonable
[19:04] <gary_poster> bac, your p11 revert I think went wrong
[19:04] <gary_poster> understandably
[19:04] <gary_poster> since there were a number of commits
[19:04] <gary_poster> bac, I'm running p10 now
[19:05] <gary_poster> because we think we fixed our issue
[19:05] <gary_poster> and the fix was in Launchpad
[19:05] <gary_poster> bac, and it is working so far (http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/10/steps/shell_9/logs/stdio fwiw)
[19:05] <bac> gary_poster: do you see something fishy from p11?
[19:05] <bac> or just that it broke the world?
[19:05] <gary_poster> bac, the result you just showed me from the pastebin looked exactly like ou had a naive version of the stdout replacement code
[19:06] <bac> ok
[19:06] <bac> ok, so just toss p11?
[19:07] <gary_poster> bac, benji, if build 10 (which uses p10) passes, yeah, I will be tempted to toss p11
[19:07] <bac> but i've got to the branch back to a good state
[19:07] <gary_poster> and revert the revert in zope.testing
[19:07] <gary_poster> yeah
[19:07] <bac> forks are fun
[19:07] <gary_poster> IIRC it is bad form to uncommit and push, isn't it
[19:08] <gary_poster> I say, let's wait for this rest result
[19:08] <bac> i think if you're guaranteed no one is working on the version you're wiping out it can be done
[19:08] <bac> but probably better to avoid
[19:08] <gary_poster> I will be on team lead call by the time it finishes
[19:08] <gary_poster> but if it succeeds
[19:08] <gary_poster> then we land p10
[19:08] <bac> ok, well, i'll change my LP branch to use p10 and submit to ec2
[19:09] <bac> unless it looks like my change to zope testing broke --list
[19:09] <gary_poster> and we merge the reverse of the last two changes to zope.testing
[19:09] <gary_poster> bac, p10 --list works fine
[19:09] <bac> oh, yay
[19:09] <gary_poster> I am pretty sure that this is, again, just the fact that you didn't revert completely back to p6 when you added your code in (which was very understandable)
[19:11] <bac> well, i couldn't as it wasn't yet bad.
[19:11] <gary_poster> :-)
[19:11] <bac> r 37 was a good check in
[19:11] <bac> of my stuff
[19:11] <gary_poster> ack
[19:12] <bac> so here's the pedantic question
[19:12] <bac> if i revert the last change, do we change setup.py to call it p12?
[19:12] <gary_poster> so bac, benji...I have team lead call...bac, why don't you join us briefly  in https://plus.google.com/hangouts/_/afdae2431954bae99aae1d5075449031334b6481?authuser=1&hl=en-US
[19:12] <bac> otherwise it'll say p10 and the next guy will re-use p11
[19:37] <gary_poster> bac, benji, yay, I think!
[19:37] <bac> gary_poster, benji:  http://paste.ubuntu.com/1039594/
[19:37] <bac> p10 nominally == p12
[19:38] <gary_poster> bac, awesome
[19:38] <gary_poster> bac, benji we have a full, green test run http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/waterfall
[19:39] <bac> for julian:
[19:39] <bac> bzr ci -m "Update zope.testing to p12 which reverts p11 to be the inaccurately maligned p10.  Dont ask."
[19:39] <gary_poster> lol
[19:44] <gary_poster> bac or benji lemme know if I should re-run any of your branches
[19:44] <gary_poster> on the buildbot
[19:45] <benji> gary_poster: ok, I think I'll be good
[19:45] <bac> gary_poster: benji and i decided to let him pqm-submit his branch.  i'll merge, update to use p12, and then get you to run it
[19:45] <gary_poster> great benji, bac
[19:58] <benji> bac: any help using pqm-submit; I've only used it once or twice and it's been quite a while
[19:58] <benji> this command line: bzr pqm-submit -m 'Use version p10 of our zope.testing fork for its redirection of non-subunit output away from stdout when --subunit is specified'  lp:~/launchpad/bug-996729
[19:59] <benji> gives me this error: bzr: ERROR: There is no public branch set for "bzr+ssh://bazaar.launchpad.net/~benji/launchpad/bug-996729/".
[19:59] <bac> benji: actually if you have a MP you can just use 'bzr lp-land'
[19:59] <benji> bac: I do, but it's already been merged; is setting it back to approved kosher?
[20:00] <bac> how about: bzr pqm-submit -m "[r=gary][bug=996729][no-qa] blah"
[20:01] <bac> i don't need to specify the branch b/c i've got my .bazaar/locations.conf set up correctly.  i recall you do not, right?
[20:01] <benji> I guess not.  Any refrence for setting it up?
[20:04] <bac> benji: i'm looking
[20:04] <benji> bac: I'm inclined to just make another MP (or reset the one I have)
[20:06] <bac> benji: here is the relevant part of my .bazaar/locations.conf
[20:06] <bac> http://paste.ubuntu.com/1039635/
[20:06] <bac> i'm sure it has some cruft in it due to the way things evolved
[20:08] <benji> bac: it worked!
[20:08] <benji> (hopefully PQM will actuall accept it)
[20:08] <bac> whee ha
[20:08] <bac> how did you make it this far in life without a locations.conf?
[20:10] <benji> I have one but since I use a lpsetup-created setup now the paths were different, once fixed (plus another tweak) it worked
[20:30] <gary_poster> benji, looks like it hasn't landed yet?
[20:30] <benji> gary_poster: hmm it should have
[20:30]  * benji looks
[20:32] <benji> gary_poster: hmm, this doesn't look like my fault: "Conflicts during merge: Text conflict in lib/lp/soyuz/model/archive.py"
[20:32]  * benji looks at subsequent PQM emails.
[20:32] <gary_poster> benji, that's to dbdevel, I think you'll find
[20:32] <benji> mmm
[20:33] <gary_poster> benji, we have two recent lands to devel, within the past 14 minutes
[20:33] <gary_poster> https://code.launchpad.net/~launchpad-pqm/launchpad/devel
[20:33] <benji> ooh, it is
[20:34] <benji> arg! Sender not authorised to commit to branch bzr+ssh://bazaar.launchpad.net/~benji/launchpad/devel
[20:34] <benji> I hate PQM.
[20:34] <gary_poster> :-)
[20:35] <benji> any idea how I'm supposed to make it think I'm authorized?
[20:35] <benji> wait! that says ~benji
[20:35] <gary_poster> benji, look at that error message
[20:35] <gary_poster> right :-)
[20:35] <benji> I hate PQM more now.
[20:35] <gary_poster> heh
[20:37] <benji> is this what I need in my .bazzar/locations?
[20:37] <benji> public_branch = bzr+ssh://bazaar.launchpad.net/~launchpad-pqm/launchpad/devel
[20:37] <benji> submit_branch = bzr+ssh://bazaar.launchpad.net/~launchpad-pqm/launchpad/devel
[20:38] <gary_poster> eh, I don't use locations.  I used the raw pqm options for awhile and now just use bzr lp-land when I need to do this sort of thing
[20:38] <gary_poster> I'll see if I can help...
[20:39] <gary_poster> benji, you want this kind of thing:
[20:39] <gary_poster> submit_branch = bzr+ssh://bazaar.launchpad.net/~launchpad-pqm/launchpad/devel
[20:39] <gary_poster> public_branch = bzr+ssh://bazaar.launchpad.net/~gary/launchpad
[20:39] <gary_poster> public_branch:policy = appendpath
[20:39] <gary_poster> push_location = lp:~gary/launchpad
[20:39] <gary_poster> push_location:policy = appendpath
[20:39] <gary_poster> s/gary/benji obv
[20:41]  * benji resubmits.
[20:42]  * benji watches https://pqm.launchpad.net/ while reciting the good luck charm of his people:  "come on baby!"
[20:46] <gary_poster> that worked, benji, cool.
[20:46] <gary_poster> bac ^^
[20:47] <benji> yep, it looks good
[20:47] <bac> ok
[20:50] <gary_poster> benji, you are sure about that logs dir removal?
[20:50] <gary_poster> http://bazaar.launchpad.net/~launchpad-pqm/launchpad/devel/revision/15409
[20:52] <bac> gary_poster: can you run my branch through your parallel buildbot?  lp:~bac/launchpad/bug-682772
[20:52] <bac> it has my changes, merged with the most recent trunk, using p12
[20:53] <gary_poster> bac, sure
[20:58] <gary_poster> bac http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/13fwiw
[20:58] <gary_poster> http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/builders/lucid_lp/builds/13 that is
[20:58] <bac> thanks
[20:59] <gary_poster> welcome
[21:35] <bac> gary_poster: http://ec2-50-16-151-255.compute-1.amazonaws.com:8010/waterfall
[21:35] <bac> Mr. Green Jeans
[21:35]  * bac -> lands
[21:35] <gary_poster> bac, yay!