Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-11-19
21:15:00 sean-k-mooney cfriesen: i did fine it weird too
21:15:19 sean-k-mooney its there to differenceate between the url or the log
21:15:45 sean-k-mooney but i would have gon with opesntack server console show and openstack server log show
21:16:48 sean-k-mooney cfriesen: sorry tehy are console log show and console url show
21:17:19 sean-k-mooney but yes openstack server console show and openstack server log show would have made more sense to me
21:19:21 mriedem cfriesen: likely a question for dtroyer in #openstack-sdks
21:20:27 mriedem cdent: so we're intermittently hitting that in py36 jobs?
21:20:34 cdent yes
21:21:02 cdent seems to depend on whether the node has a recently updated python. 3.6.7 was released to ubuntu not long ago
21:21:13 sean-k-mooney cdent: is that happening in nova or in the hacking repo?
21:21:18 cdent nova
21:21:44 cdent it's just the one test
21:22:02 cdent the root is that having a newline token being added on the end is causing pep8 to think it needs to do a "logical check"
21:22:11 sean-k-mooney can you past the test name again ill try and run it on python 3.6 on centos
21:22:18 cdent which means it bypasses the physical checks that happen otherwise
21:22:29 cdent sean-k-mooney: nova.tests.unit.test_hacking.HackingTestCase.test_check_doubled_words
21:22:40 cdent it ought to pass fine in <3.6.7
21:22:46 mriedem http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22testtools.matchers._impl.MismatchError%3A%20%5B(1%2C%200%2C%20'N343')%5D%20!%3D%20%5B%5D%5C%22%20AND%20tags%3A%5C%22console%5C%22%20AND%20build_name%3A%5C%22openstack-tox-py36%5C%22&from=7d
21:23:31 cdent yeah, so always bionic
21:23:36 mriedem yup
21:23:41 mriedem do we want to skip the test for now?
21:23:58 cdent if you're happy with that, that's probably the easiest way to go
21:24:47 mriedem or at least skip on py36
21:25:09 sean-k-mooney python 3.6 will be the defult on rhel8 also as per doug hellmann's mail earilier today
21:26:02 cdent yeah, I'll do a skipIf
21:26:09 cdent one moment please
21:26:42 sean-k-mooney im testing with Python 3.6.5 on centos 7 and it passed
21:27:56 cdent yeah, that's expected
21:28:31 sean-k-mooney oh it needs to be 3.6.7+ to fail
21:43:14 openstackgerrit Chris Dent proposed openstack/nova master: Skip double word hacking test https://review.openstack.org/618843
21:43:21 cdent mriedem, efried, sean-k-mooney ^
21:44:24 sean-k-mooney is that really a nova bug or a hacking bug?
21:45:06 cdent the failing test is because of changes to the tokenize stdlib. The hacking check is in nova itself, not in hacking
21:45:20 sean-k-mooney i guess we shoudl make the test support both the old and new api
21:45:24 cdent at least the way nova is experiencing the problem
21:45:46 cdent sean-k-mooney: I tried and failed to do that because it is in the guts of pep8 where the problem happens
21:45:50 cdent all hacking has is a regex
21:46:24 cdent the presence of the added NEWLINE token causes pep8 to believe it needs to make a logical instead of physical check
21:47:16 sean-k-mooney not sure i follow but then again i have not looked at the python bug yet
21:47:24 cdent there's little we can do from our side of the world to make it not add the NEWLINE
21:47:41 cdent looking at the python bug (and the related tracker for api breakages it causes) will make things much more clear
21:49:13 sean-k-mooney so just to be clear on what line in test_chesck_double_words is it addint the new line
21:49:17 sean-k-mooney the + operator
21:49:34 cdent at the end of the entire string
21:49:37 cdent and it isn't adding a new line
21:49:45 cdent it is adding a NEWLINE token when tokenizing
21:49:55 sean-k-mooney oh
21:50:13 sean-k-mooney that seams like a think it should not be doing... but ok
21:50:21 sean-k-mooney anyway you temp fix makes sense
21:50:33 cdent yeah, you'd think so, but someone decided that the method should behave more like C
21:50:48 cdent the comments on the python bug are amusing
21:50:49 sean-k-mooney c ends with \0
21:50:56 sean-k-mooney e.g. a 0 byte
21:50:58 sean-k-mooney not a new line
21:51:20 cdent right, but it tokenizes a NEWLINE
21:51:21 cdent apparently
21:51:25 sean-k-mooney so it does not match c either in terms of string termination of const char* s
21:51:30 cdent it's not about string termination
21:51:34 cdent it is about code tokenizing
21:51:59 cdent please, read the bug
21:52:05 cdent it will all become clear
21:52:30 sean-k-mooney yes i was as we were chatting "As was pointed out in https://bugs.python.org/issue33766 there is an edge case in the tokenizer whereby it will implicitly treat the end of input as a newline."
21:53:45 efried cdent: What is [1] in the commit message?
21:54:02 cdent that link that sean-k-mooney just pointed at, did I forget to add it? le sigh
21:54:23 efried not https://bugs.python.org/issue33899 ?
21:54:39 openstackgerrit Chris Dent proposed openstack/nova master: Skip double word hacking test https://review.openstack.org/618843
21:54:52 cdent argh too many bugs
21:55:39 sean-k-mooney efried: the openstack bug points to the python bug
21:55:48 openstackgerrit Chris Dent proposed openstack/nova master: Skip double word hacking test https://review.openstack.org/618843
21:55:51 sean-k-mooney but yes the commit does not have a bug link
21:56:09 sean-k-mooney hehe ok now it does
21:56:12 efried so none of those is https://bugs.python.org/issue33766
21:56:14 cdent there we go, now it links to the two python bugs I wanted to link to, neither of which are the one that sean pointed to (which was an originating bug)
21:56:16 cdent indeed
21:57:05 efried Not having dug any deeper, I'll point out that ``echo 'print("a");print("b")' > test.py`` does indeed end with a newline.
21:57:16 sean-k-mooney yes https://bugs.python.org/issue35107 is tracking the fallout of fixing https://bugs.python.org/issue33899 which ended up breaking backward compatiblity
21:57:33 efried 0000026
21:57:33 efried 0000020 ( " b " ) \n
21:57:33 efried 0000000 p r i n t ( " a " ) ; p r i n t
21:57:33 efried efried@efried-ThinkPad-W520:~$ od -c /tmp/test.py
21:57:33 efried efried@efried-ThinkPad-W520:~$ echo 'print("a");print("b")' > /tmp/test.py
21:57:47 cdent i don't think that's the point efried
21:57:52 efried I'm sure it's not.
21:57:59 efried hence the 'not having dug any deeper'
21:59:11 cdent it's the same deal: discussion on that bug revealed the different between a newline character and a NEWLINE token and ways in which tokenizer failed to produce NEWLINE in some cirtical areas
21:59:16 cdent the fix, however, was too hasty
22:00:11 sean-k-mooney ya and they should not have backporte it to python 2.7 luckally they reverted it on the 2.7 branch
22:01:12 efried cdent: Did this start appearing because we *just* started testing with py3.6.7?
22:01:18 cdent yes
22:01:38 cdent ubuntu bionic got it within the past few days
22:01:43 sean-k-mooney looking at https://bugs.python.org/issue35107 they are still discusiing what the behavior should be but i agreee with them that you should be able to round trip
22:01:46 cdent and our images updated recently
22:02:56 sean-k-mooney am i missrembereing or did ubunut ship with a beta python version at lauchch or was that just for 3.7
22:03:28 sean-k-mooney i remembere there was somting about an unreleased python version at one point on ubuntu but dont know if that was 3.6 or 3.7
22:03:29 efried right, cuz we've been testing with 3.7 for a while now, so if it was >=3.6.7 we should have hit it.
22:04:26 sean-k-mooney efried: well no the original bugfix was merged 2018-07-06 10:23
22:05:00 mriedem efried: i think the 3.6 jobs switched to bionic in the last 24 hours
22:05:03 mriedem according to logstash
22:05:23 sean-k-mooney so it proably took a while to get in the ubuntu package repos and then for use to update the image with it
22:05:42 sean-k-mooney or what mriedem said

Earlier   Later