[02:39] wgrant: https://code.launchpad.net/~stevenk/launchpad/bulk-eid-to-obj/+merge/171705 should address one of your concerns with export-pu-auditor [03:31] StevenK: Looks pretty good, but needs a few fixes [03:32] StevenK: Also, person merges [03:38] wgrant: Person merges? [03:39] StevenK: Person merges [03:40] . [03:43] wgrant: Repeating yourself does not help with more context/ [03:43] s/\//./ [03:46] StevenK: Person merges! [03:46] How do they interact with your branches? [03:47] wgrant: In terms of the eid branch, we can't do anything. If a person is converted to an eid and stored in auditor, and then merged into person, we will return the first person because we store by id [03:48] Why can't you do anything? [03:48] There's a well-defined successor to a merged person [03:48] person.merged [03:49] Ugh [03:49] Why can't we have simple things [03:50] This function already hurt my brain, and then I had to twist it to the side to understand your issues about its duplication, and now I have to deal with person merges. [03:50] What duplication? [03:51] "These are redundant. type_ids[foo] is just obj_id_to_eid[foo].keys()." [03:51] Oh, that thing [03:51] Do you understand now? [03:51] Yes, I've fixed it. [03:51] They are equivalent, except that type_ids[foo] has duplicates [03:51] Right [03:51] Tests even pass. [03:52] wgrant: http://pastebin.ubuntu.com/5803604/ [03:53] Ah [03:53] Now that has a slight behaviour change [03:53] Which you probably want a test for [03:53] What happens when you ask for an ID that doesn't exist? [03:53] In the current version on LP it will give you None [03:53] In your new version it will omit the key [03:53] In another version it may crash [03:54] Yeah, that was probably why I had it at the top, but you complained. [03:54] Also, scheme is annoying [03:54] I'd say something like "instance, cls, id = eid.split(':')" [03:54] I think that's what they are [03:55] But I can't tell right now, because scheme is rather opaque :) [03:55] StevenK: I think omitting them from the response is probably best, so your new behaviour is correct [03:55] But we likely want a test for that [03:55] So we don't accidentally start crashing if an object disappears [03:56] But then auditor client will ask for them and then give KeyError [03:56] Isn't this on the other side of auditorclient? [03:57] We're parsing its responses [03:57] Not giving it things to ask for [03:57] Yes, we're parsing its responses [03:58] So, we get back lp:PackageUpload:6 as the eid for obj [03:58] Then the auditorclient calls enterpriseids_to_objects() and it omits lp:PackageUpload:6 because it doesn't exist. It then tries to map it to an object and crashes [04:00] Yes [04:00] Because enterpriseids_to_objects cannot know what behaviour the callsite wants [04:00] It has to be the callsite's responsibility to handle a missing object [04:00] Oh, I can use get anyway [04:01] Of course [04:01] That's probably the correct course of action for that particular case [04:04] wgrant: http://pastebin.ubuntu.com/5803623/ [04:06] In terms of person merges, I have NFI [04:08] StevenK: existant isn't a word, and that assertEqual formatting is odd, but otherwise that looks good [04:08] Person merges don't belong in this branch === tasdomas_afk is now known as tasdomas [04:19] wgrant: That MP is updated. [04:23] StevenK: k [04:24] wgrant: http://pastebin.ubuntu.com/5803648/ should resolve another of your export-pu-auditor issues [04:25] Oh, I didn't drop scheme [04:26] I'd also prefer those two maps to be renamed as I said in my first comment, but it's at least a bit less confusing now [04:28] wgrant: scheme death, and maps renamed: http://pastebin.ubuntu.com/5803653/ [04:30] Right [04:39] wgrant: One more with feeling for that MP. [04:41] GAH [04:41] YUI 3.10.3 was released 2 days after 3.10.2 [04:42] Any notable fixes? [04:42] We don't care about tracking the x.x.1 increments unless they have relevant fixes [04:42] wgrant: The .swf vuln fixed in 3.10.1 snuck back into .2 [04:43] So they fixed it again for .3 [04:46] Hah [04:47] I don't think we serve any of the two swfs in our build tree [04:47] But convoy will probably return them if asked [04:48] 3.11 looks to have some nice speedups, too [04:48] * StevenK waits for YUI 3.11 JavaScript for Workgroups [05:35] wgrant: Do you approve of the auditorclient diff? [05:43] StevenK: Sounds reasonable [09:42] wgrant: applied your requested fixes, except the use of ISpecificationTarget, see my comment at https://code.launchpad.net/~ttx/launchpad/lp1193389/+merge/171491 === BradCrittenden is now known as bac === BradCrittenden is now known as bac === benji__ is now known as benji === benji is now known as Guest32120 === wedgwood_away is now known as wedgwood === tasdomas is now known as tasdomas_afk === matsubara is now known as matsubara-lunch === matsubara-lunch is now known as matsubara === wedgwood is now known as wedgwood_away