[00:13] Joshua Harlow proposed stackforge/cloud-init: Add importlib for python 2.6 *only* https://review.openstack.org/197691 === rangerpb is now known as rangerpbzzzz [00:39] ok seems good to go ^ smoser ^ [00:39] all fixed up === zz_natorious is now known as natorious === natorious is now known as zz_natorious [12:17] Odd_Bloke, do you have some time today that we could look at the reporting stuff ? i'd like to make some progress on that. [12:17] smoser: Potentially; looking at relatively important partner stuff at the moment. [12:21] Odd_Bloke, how do i make jenkins re-test something ? [12:25] smoser: Pushing up a new patchset should do it automatically. [12:25] yeah, but i think theres a way to say "re-run". [12:25] ohwell. [12:25] harlowja will probably know, when he wakes up in 3 days or so. :p [12:26] smoser: Why do you need to rerun without any changes? [12:26] i wasnt sure how the py26 stuff was fixed. thought it was actually not related. [12:35] Scott Moser proposed stackforge/cloud-init: pip2distro: work in progress: build tox env for specific targets https://review.openstack.org/197671 [12:56] smoser: Ah, good point about using setup in timeit. === rangerpbzzzz is now known as rangerpb [12:56] OTOH, we _aren't_ using the regex more than a handful of times, so I think my examples were more valid than they appeared. [12:56] But it was a minor point; I don't particularly mind which we do. :) [13:59] harlowja, please ping when you wake up. [13:59] really want to resolve the py26 stuff today. [14:00] https://review.openstack.org/#/c/197691/ [14:01] Odd_Bloke, i have no objection to using unittest2 [14:01] i dont fully understand: [14:01] This would mean we couldn't use testtools' TestCase class, but we would still be able to use their matchers in our assertions, if we so wished (http://testtools.readthedocs.org/en/latest/for-test-authors.html#assert-that-function). [14:02] claudiupopa, do you have any objection to unittest2 ? [14:02] does it have assertRaises as a context manager? ;-) [14:02] No, no objection. [14:05] smoser: So we're using testtools.TestCase, which inherits directly from unittest.TestCase. [14:05] smoser: It is therefore incompatible with using unittest2.TestCase as our root test case. [14:05] ah. ok. [14:05] Why are we using testtools? What benefits does it bring us? [14:06] ^ I'm curious about this as well. [14:06] (I think testtools has useful stuff, I just don't think we're using any of it ATM :p) [14:07] i have no attachment to it [14:07] it blames to [14:07] http://paste.ubuntu.com/11810295/ [14:08] is setUp and tearDown stuff it provides ? [14:09] guess not [14:10] Nope, that's in unittest as well. [14:33] ok, we really need to resolve this today, as all tests fail on on the py26 gate right now. [14:54] Odd_Bloke, did you see my comment at https://review.openstack.org/#/c/170257/4/cloudinit/templater.py ? [14:54] smoser: Re: timeit? [14:54] yeah. [14:54] am i wrong ? [14:55] there definitely *is* a pentalty for compiling, but it does make up for itself. [14:55] (assuming it gets used) [14:55] smoser: If you use the regex a lot, yeah. [14:55] So the truth is somewhere between our two benchmarks. [14:55] :p [14:56] Mine tests the worst case (we use it zero or one times). [14:56] Yours tests the best case (we use it 1000s of times). [14:56] well, but yours has so much noise [14:56] re.compile() seems to take about as much time as 'import re' [14:57] The actual truth: we've spent more time discussing/thinking about this than is likely to be saved by selecting the correct option across all instances forever. ;) [14:57] correct === zz_natorious is now known as natorious [14:58] now lets get thsi stupid py26 stuff settled. [14:58] * smoser waits for harlowja [15:03] That guy. [15:03] Does he ever wake up. [15:24] harlowja: What does testtools actually get us over unittest2? [15:24] (Except for ugly exception testing :p) [15:26] fixtures integration, having to figure out what to do about py3 (unittest2 only works on py2.x), so we'd end up making a mini testtools anyway, testtools also heavily used throughout openstack so its common ground.... the author is helpful and involved in openstack to [15:26] same author owns unittest2 :-P [15:27] a bunch of other matchers, https://github.com/testing-cabal/testtools/tree/master/testtools/matchers [15:27] integration with https://github.com/testing-cabal/fixtures [15:27] ... [15:28] so thats my 2 cents [15:29] testtols is using unittest2 anyway and its providing the py2/py3 compat [15:30] need me to ask the author to jump in here (he's in new zealand) so he's probably not awake yet [15:30] harlowja_at_home: lifeless? [15:30] yes [15:30] harlowja_at_home: So you can use the matchers outside of the TestCase. [15:31] (I know this because I wrote the function that supports doing it ;) [15:31] ok [15:31] I thought unittest2 worked on 2 and 3. [15:31] https://hg.python.org/unittest2/file/d091f0086b03/setup.py#l40 [15:32] so seems not [15:32] harlowja_at_home: The description at https://pypi.python.org/pypi/unittest2 says "It is tested to run on Python 2.6, 2.7, 3.2, 3.3, 3.4 and pypy." [15:32] *shrugs* [15:33] SHURG [15:33] lol [15:34] guess the classifiers haven't been updated [15:35] harlowja_at_home: Maybe because it doesn't make sense to use for later versions? [15:35] (Or maybe just olde) [15:36] probably [15:38] * harlowja_at_home shrugs [15:38] lol [16:17] anyways, will be off most of today, for some reason they gave us 2nd & 3rd off === natorious is now known as zz_natorious [16:25] dagummit. [16:25] harlowja_at_home, want to get through this py2.6 thing [16:25] i dont have strong feelings on fixtures/unittest2/testtools [16:25] but do have strong feelings on jenkins gate job failing :) [16:25] and also my feeling is that less dependencies is better. [16:26] smoser, ya, only feeling i have is that this is what the rest of openstack uses, it has some nice helpers, losing a context manager, meh [16:26] also, harlowja_at_home right now onyour branch, if i 'tox -e py27' it doesn't work. [16:26] its a test-time-dependency [16:26] complains of the importlib;python_ver==2.6 [16:26] or whatever that is. [16:26] yup, upgrade pip [16:26] :-( [16:26] upgrade pip is not too easy though. [16:27] ya, not much i can do about that one, this change for all these in pbr is forcing this throughout all of openstack [16:27] but i just remembered system i'm running it on is 14.04 where usually i do wily [16:27] for better or worse [16:27] ya, i've also got my 14.04 vm and now have to uprade pip, sorta blows [16:27] what version do you need ? [16:28] https://launchpad.net/ubuntu/+source/python-pip [16:28] 6+ i think [16:28] wily is 1.5.6 [16:28] afaik only pip 6+ understands version markers [16:29] http://lists.openstack.org/pipermail/openstack-dev/2015-June/067823.html [16:29] and such from robert [16:30] i'm pretty sure that these changes are gonna force many distros to have a massive pip upgrade [16:30] are these version numbers the same ? 1.5.6 is in wily, but i need 6? [16:30] 6.0+ [16:30] they did a huge version number bump, from 1.5 -> 6.0 [16:31] now its 7.0 i think, lol [16:31] * https://pypi.python.org/pypi/pip (7.1) [16:31] ok. [16:31] so i'd kind of like to avoid this. [16:31] the only thing we're needing it for is python2.6 needing importlib [16:31] (at least at this point) [16:31] right ? [16:32] ya [16:32] in our tox environment we can easily enough just use a different py26 requirements environment then everything else. [16:32] that is possible to [16:32] i think i like that better at the moment. [16:33] than making people (including me) install pip from outside of distro [16:33] it'd be more acceptable if at least wily had a suitable pip [16:33] ya, i'm pretty sure now that openstack is using those markers, alot of people are gonna have to upgrade [16:34] *for better or worse* [16:35] ok. so my take out of the above is [16:35] a.) use testtools unless Odd_Bloke cries [16:35] b.) remove requirement of newer pip as smoser is a luddite [16:36] :-P [16:36] if you're on holiday i can try to do that. [16:36] i get it done, but probably heading out soonish [16:36] i just need some sane path to running this on ubuntu, and "download pip from the intertubes" isn't really sane [16:36] time to get an upgrade :-P [16:36] harlowja_at_home, you are allowed to not do this today :) its ok for you to celebrate your countries independence from the queen [16:37] country's even. [16:37] lol [16:37] (if you only have the one) [16:37] obama the queen [16:37] smoser: FWIW, using a virtualenv on Ubuntu solves the pip problem. [16:37] ya, its still a pita though :-P [16:38] create virtualenv, enter virtualenv, install newer pip [16:38] using virutalenv to run tox which runs creates a virtualenv [16:38] i really like the simplicity of 'tox' [16:41] Joshua Harlow proposed stackforge/cloud-init: Add importlib for python 2.6 *only* https://review.openstack.org/197691 [16:41] ok let's see what happens [16:46] what'd you do ? [16:46] you took importlib out ? [16:47] oh. isee. [16:47] but the py26-gate thing wont figure that out, will it ? [16:51] seems to have figured it out; [16:51] https://jenkins02.openstack.org/job/gate-cloud-init-python26/1/ [17:04] i have a feeling that we should have a template engine that uses python formatting. [17:05] as probably immediately more powerful and better documented than 'builtin' [17:05] and no dependencies [17:06] but it might not be sufficent for how we want to expose objects to templating. [17:10] The new formatting syntax is pretty comprehensive. [17:10] cloud-init-formatter-v1000 [17:10] lol [17:10] Though we should investigate if you can do anything untoward using it. :p [17:11] Odd_Bloke, i dont think i can access object.property [17:11] can i ? [17:11] smoser: [17:11] smoser: Sure. [17:11] smoser: http://paste.ubuntu.com/11811141/ [17:12] awesome. [17:12] * harlowja_at_home pretty sure i made that possible :-P [17:12] ok. so yeah, i like this. [17:12] harlowja_at_home, you did. you rock. [17:12] oh, i think thats the built-in python formatting, nothing i did :-P [17:12] right. [17:13] thats what i'm saying we should offer as a templating option [17:13] ah [17:13] except afaik it only got good on 3.0 + [17:13] https://www.python.org/dev/peps/pep-3101/ [17:14] Oh, Python 2.6 might be screwed. [17:14] and 2.7? [17:14] That example is on 2.7. [17:14] k [17:14] guess they must of backported some of it [17:15] or something [17:15] Just tested it, and the above works on 2.6. [17:16] cool, up to u guys [17:22] i think makes sense to add it [17:22] and will just fail on py2.6 . [17:22] https://review.openstack.org/#/c/197691/5/cloudinit/sources/openstack/httpopenstack.py [17:22] as seen there, you had to make that change as the .format didnt' work on py26 right? [17:23] Odd_Bloke, could you review https://review.openstack.org/#/c/197691/ [17:23] it is sane to me. [17:26] the one thing ihave left is that we added the py26 environment, which means 'tox' will fail now on any recent ubuntu. where as previously it would pass ( as we didn't need py26) [17:29] u running all the tox envs? [17:29] for fun [17:32] well, i want to run 27 and 34 for sure by default === zz_natorious is now known as natorious === natorious is now known as zz_natorious [17:51] Merged stackforge/cloud-init: Add importlib for python 2.6 *only* https://review.openstack.org/197691 [17:52] Joshua Harlow proposed stackforge/cloud-init: Bring over the 'templater' from bzr https://review.openstack.org/170257 [17:53] Joshua Harlow proposed stackforge/cloud-init: pip2distro: work in progress: build tox env for specific targets https://review.openstack.org/197671 [17:53] Joshua Harlow proposed stackforge/cloud-init: tools/tox-venv: support running other than ./tools/tox-venv https://review.openstack.org/195631 [17:54] Joshua Harlow proposed stackforge/cloud-init: Expose api response properties and cache buffer decoding https://review.openstack.org/195800 [17:54] why does tox docs not work [17:54] :-( [17:55] http://paste.ubuntu.com/11811416/ [17:57] https://bugs.launchpad.net/oslotest/+bug/1379998 [17:59] no idea, jump on #openstack-oslo ask pbr folks? [18:03] harlowja_at_home, well. this fixes it. [18:03] but i dont know how that bug reports fixed in oslotest [18:04] ya, not sure [18:04] http://paste.ubuntu.com/11811486/ [18:04] that fixes it. [18:04] but will break again when 1.3b4 is released unless its fixed there. [18:04] ya [18:05] ok, i'm off [18:05] will be on maybe at night or something, ha [18:06] ok. [18:09] Scott Moser proposed stackforge/cloud-init: fix 'tox -e docs' by limiting sphinx versions https://review.openstack.org/198080 [18:13] Merged stackforge/cloud-init: Bring over the 'templater' from bzr https://review.openstack.org/170257 [18:14] if someone else could check my work on https://review.openstack.org/#/c/198080/ [18:14] i'd appreciate it. Odd_Bloke claudiupopa ^ [18:16] Merged stackforge/cloud-init: Bring over the 'safeyaml' from bzr https://review.openstack.org/170252 [18:16] Merged stackforge/cloud-init: tools/tox-venv: support running other than ./tools/tox-venv https://review.openstack.org/195631 [21:18] Scott Moser proposed stackforge/cloud-init: fix 'tox -e docs' by limiting sphinx versions https://review.openstack.org/198080 === rangerpb is now known as rangerpbzzzz [21:37] Scott Moser proposed stackforge/cloud-init: Expose api response properties and cache buffer decoding https://review.openstack.org/195800 [21:38] Merged stackforge/cloud-init: fix 'tox -e docs' by limiting sphinx versions https://review.openstack.org/198080 [21:42] Merged stackforge/cloud-init: Expose api response properties and cache buffer decoding https://review.openstack.org/195800 [21:44] why can't anything be easy. [21:44] attempt at running http://paste.ubuntu.com/11812524/ on vivid [21:44] results in http://paste.ubuntu.com/11812519/ [21:45] and, oh joy. py34-coverage and py27-coverage now fail. [21:45] fail as in we dont have 90% coverage. [21:45] Odd_Bloke, ^ claudiupopa ^ feel free to fix that. :)