Tuesday, 2019-02-19

*** chandankumar has quit IRC00:28
*** apetrich has quit IRC03:14
*** irclogbot_1 has quit IRC04:57
*** jrist has quit IRC05:06
*** jrist has joined #openstack-mistral05:10
*** quiquell|off is now known as quiquell|rover06:29
*** apetrich has joined #openstack-mistral06:38
*** jtomasek has joined #openstack-mistral06:38
openstackgerritAdriano Petrich proposed openstack/mistral master: Return empty list as tuple  https://review.openstack.org/63750707:13
apetrichrakhmerov, I added the tests from the comments ^^ but that is failing07:16
*** quiquell|rover is now known as quique|rover|r--07:26
*** akovi has joined #openstack-mistral07:52
*** pgaxatte has joined #openstack-mistral08:14
*** shardy has joined #openstack-mistral08:15
*** shardy has quit IRC08:15
*** shardy has joined #openstack-mistral08:31
*** jtomasek has quit IRC08:36
*** quique|rover|r-- is now known as quiquell|rover08:46
d0ugalapetrich: Yeah, it'll never pass :)08:49
d0ugalapetrich: the patch handles empty lists only, that is why it isn't a correct fix08:49
apetrichd0ugal, I know but I wanted the regression tests there08:57
d0ugalah, cool08:58
d0ugalmakes sense08:58
* d0ugal is still waking up08:58
apetrichI wonder if we should just remove that way of comparing08:59
d0ugalHow? I'm not sure we can really08:59
apetrichbecause bool() would work for the empty lists09:00
apetrichoh.09:00
d0ugalit should also be deprecated really09:00
apetrichnot remove just not use it09:00
d0ugalRight, that is the solution in tripleo anyway09:00
d0ugalDid you see the fix merged?09:00
d0ugalhttps://github.com/openstack/tripleo-common/commit/573f67df3d6216f62467eb526a15be3120f9a01a09:00
apetrichd0ugal, no I didn't09:02
apetrichd0ugal, nice. that means we are not on the critical issues09:03
quiquell|roverd0ugal: I am going to try to continue with the list fix09:03
d0ugalquiquell|rover: Great, I think rakhmerov is also looking into it. So make sure you don't duplicate effort :)09:03
d0ugalapetrich: Yeah, now it is just a gotcha09:04
d0ugalbut one that wont come up often, I hope09:04
quiquell|roverd0ugal: Yep stuff is working now at tripleo09:04
quiquell|roverd0ugal: maybe is good to document that though09:04
quiquell|roverd0ugal: like "don't use = []"09:04
apetrichquiquell|rover, great work09:04
d0ugalMaybe, but I don't know where I would document it :)09:04
d0ugalI assume nobody would read the docs anyway lol09:04
quiquell|roverd0ugal: pragmatic guy you are09:05
d0ugalbut maybe a mistral bug we can reference would be wise09:05
quiquell|roverOk let's try to sort out the comments to have the review there in good shape09:05
d0ugalSounds good, thanks!09:05
quiquell|roverd0ugal: so the review is has a legit test failure and we have to find a way to cover the cases09:12
quiquell|roverd0ugal: is that it ?09:13
d0ugalquiquell|rover: Yes, I believe so09:13
d0ugalFix those tests and don't break any others :)09:13
quiquell|roverd0ugal: ack tdd !!!09:14
quiquell|roverd0ugal, apetrich: It's not quite clear when do we have to return a tuple and when a list09:19
d0ugalAgreed. I don't fully understand either. I think you might need rakhmerov to understand the logic09:19
apetrich+109:19
quiquell|roverd0ugal: looks complex stuff like, really understand yaql asumptions09:19
quiquell|roverapetrich, d0ugal: will wait to rakhmerov is this is ok09:20
d0ugalYup, that is fine09:25
rakhmerovd0ugal: hi, what's up?09:28
rakhmerovI stepped out and just got back09:28
rakhmerovso missed what you were discussing09:28
d0ugalrakhmerov: Nothing much, quiquell|rover was just looking how to fix the YAQL issue correctly09:28
rakhmerovaah..09:28
d0ugalbut I think the conclusion was to let you handle it as you have more insight09:28
rakhmerovI spent a sleepless night thinking about it :)09:28
quiquell|roverrakhmerov: Yep is more than a quick fix09:28
rakhmerovso, I don't have a good fix in mind yet09:29
d0ugalIt is tricky09:29
quiquell|roverbtw why don't mistral just use tuples ? is the "str" issue ?09:29
rakhmerovand moreover, I don't think there's a way to satisfy both things, the 'str' function and this comparison09:29
rakhmerovquiquell|rover: str($.my_list) gives "(1,3,4,)"09:30
rakhmerovwhich is confusing09:30
rakhmerovand not only confusing but it was a regression made in Rocky09:30
rakhmerovbefore Rocky it would return "[1,2,3]"09:31
quiquell|rovermaybe upstream fix a yaql ? or this is crazy ?09:31
d0ugalYou could use the JSON function to create that result09:32
rakhmerovd0ugal: the question is in compatibility09:36
d0ugalSure09:36
rakhmerovit used to work09:36
d0ugalrakhmerov: What changed in Rocky?09:36
rakhmerovwe added this pre-processing09:37
rakhmerovof data context09:37
rakhmerovwe did it because we couldn't have sets of dicts09:37
d0ugalRight09:37
quiquell|roverd0ugal: It could have change yaql rpm update09:37
quiquell|roverwe were using python2-yaql-1.1.3-2.el7.noarch09:38
d0ugalrakhmerov: did you consider changing your dicts to frozendicts?09:39
rakhmerovquiquell|rover: fixing yaql is an option yes, but it'll take a lot of time09:40
rakhmerovd0ugal: maybe, I don't remember all the details of that now09:41
quiquell|roverrakhmerov: I can compare yaql versions between working and not working09:41
quiquell|roverrakhmerov: if this can help you guys09:41
rakhmerovit's the same09:41
quiquell|roverack09:41
quiquell|rover:-/09:41
rakhmerovyaql version is the same, it hasn't changed for 2-3 years09:41
d0ugalrakhmerov: np, just an idea. That is how I would normally do sets of dicts09:41
rakhmerovI guess no one really supports it09:41
rakhmerovd0ugal: got it, yes09:42
rakhmerovd0ugal: the thing is that part of conversions happen internally, and there are some YAQL options btw that affect that that we (as far as I remember) had to change09:45
d0ugalwow, there really hasn't been a YAQL release for some time09:45
d0ugalhttps://github.com/openstack/mistral/blob/master/mistral/expressions/yaql_expression.py#L39-L4409:45
d0ugal^ some of those engine options are not even released btw09:45
d0ugalso when there is a release, I hope it doesn't change things09:45
d0ugalbrb09:45
* quiquell|rover takes the backpack and goes back to #tripleo09:46
quiquell|rover:-)09:46
rakhmerovd0ugal: as far as I can see, all of these options are released09:47
rakhmerovthey are all present in our config09:47
d0ugaloh, maybe they just got added to the CLI interface then09:50
d0ugalI was just looking at https://github.com/openstack/yaql/compare/1.1.3...master09:51
rakhmerovok09:59
rakhmerovone simple fix that I have in mind is to fix standard YAQL functions so that they coerce collections before comparing them10:00
rakhmerovbut not sure what else it can break :)10:00
d0ugalThat would be a big change in YAQL10:06
rakhmerovyes10:08
rakhmerovI would probably suggest we revert my patch for now, breaking list comparisons seems a worse problem to me than breaking str()10:12
rakhmerovI don't expect to find a good solution within 2-3 days (have to now switch to other things)10:13
d0ugalok, that sounds like a sensible plan to me10:13
rakhmerovbut I'll see what I can do with str() function, maybe there's a way to fix it w/o breaking these comparisons10:13
rakhmerovI spent 2-3 hours yesterday evening debugging YAQL when it evaluates this expression10:14
rakhmerovthe logic there is terribly difficult and these internal conversions of list into tuples works in a very implicit way10:14
rakhmerovI found where it happens but it's not configurable in any way from outside10:14
rakhmerovseems like we can only fix YAQL itself10:15
rakhmerovso I guess dealing with "str" would be a better option10:15
d0ugalyup10:15
rakhmerovapetrich, akovi: do you guys agree?10:15
rakhmerovany other ideas?10:16
rakhmerovd0ugal: btw, would you have some time to propose a release of oslo.cache?10:16
d0ugalI guess I can10:17
rakhmerovthanks10:18
d0ugalrakhmerov: it is already released10:24
rakhmerovreally?10:24
rakhmerovhave a link?10:24
rakhmerovI just checked and didn't find it10:24
rakhmerovmaybe overlooked10:24
d0ugalhttps://github.com/openstack/oslo.cache/tree/1.33.010:24
d0ugal1.33.010:24
d0ugal9 days ago10:24
d0ugalthere are no changes in master not in that tag10:25
d0ugalI'm a bit confused10:25
d0ugalit was dogpile we had the issue with, right?10:25
rakhmerovyeah...10:27
rakhmerovwait10:27
rakhmerovbut it was uncapped later than 9 days ago, wasn't it?10:27
d0ugalThat is why I'm confused10:28
d0ugalhttps://review.openstack.org/#/c/636037/10:28
d0ugalYeah, Feb 18th10:28
d0ugaloh, sorry, the commit was 9 days ago10:28
d0ugalbut it was only merged on the 18th10:29
d0ugalThe release was made yesterday only10:29
d0ugalhttps://github.com/openstack/releases/commit/62688fdb54ec434e55333cadf7c30909b5b3de9e10:29
d0ugalhttps://review.openstack.org/#/c/637609/10:30
d0ugalMerged just over 12 hours ago10:30
rakhmerovd0ugal: ok10:32
rakhmerovso now our issues with dogpile.cache should go away? :)10:33
d0ugalI hope so!10:33
rakhmerov:)10:33
rakhmerovlet's try to recheck then10:33
rakhmerovbtw, https://review.openstack.org/#/c/63717010:33
rakhmerovrevert10:33
d0ugal+W, thanks10:34
openstackgerritRenat Akhmerov proposed openstack/mistral master: Stop using deprecated keystone_authtoken/auth_uri  https://review.openstack.org/59418710:37
*** jtomasek has joined #openstack-mistral11:32
rakhmerovd0ugal: https://review.openstack.org/#/c/637145/ :)12:24
rakhmerovit passed now12:24
rakhmerovlet's merge it12:24
rakhmerovapetrich: Adriano, can you please review and approve it?12:25
d0ugaloops, I just did12:25
d0ugallol12:25
d0ugalphew apetrich agrees ;)12:26
apetrich:)12:26
rakhmerovok :)12:26
rakhmerovnp12:26
rakhmerovthanks12:26
openstackgerritMerged openstack/mistral master: Revert "Fix how Mistral prepares data for evaluating a YAQL expression"  https://review.openstack.org/63717012:35
rakhmerovaah, https://review.openstack.org/594187 passed too, merging it )12:41
rakhmerovso our CI is finally unblocked12:41
rakhmerovgood12:41
apetrichgreat13:10
d0ugalunblocked... but for how long? :)13:16
*** smrcascao has joined #openstack-mistral13:44
*** smrcascao has quit IRC13:56
*** smrcascao has joined #openstack-mistral14:06
*** jtomasek has quit IRC14:17
*** smrcascao has quit IRC14:36
openstackgerritMerged openstack/mistral master: Update dogpile.cache to match global requirements  https://review.openstack.org/63714516:25
*** quiquell|rover is now known as quiquell|off16:38
*** smrcascao has joined #openstack-mistral16:49
*** akovi has quit IRC16:49
*** smrcascao has quit IRC16:49
*** pgaxatte has quit IRC16:50
*** smrcascao has joined #openstack-mistral16:51
*** akovi has joined #openstack-mistral16:56
*** akovi has quit IRC17:01
*** smrcascao has quit IRC18:10
*** jtomasek has joined #openstack-mistral18:44
*** openstackgerrit has quit IRC20:09
*** openstackgerrit has joined #openstack-mistral21:09
openstackgerritMerged openstack/mistral master: Stop using deprecated keystone_authtoken/auth_uri  https://review.openstack.org/59418721:09
*** jtomasek has quit IRC21:57
*** d0ugal has quit IRC22:03
*** d0ugal has joined #openstack-mistral22:20

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!