[16:14] moin [16:16] lifeless: hi [16:16] hiya jml [16:17] lifeless: just polishing off the last of the testtools/twisted-related failures [16:17] \o/ [16:18] lifeless: looks like iterating the reactor is necessary for many tests because of bugs in Twisted (either core or APIs) [16:18] lifeless: I'm thinking of giving AsynchronousDeferredRunTest some kind of option that controls whether we do that iterating [16:18] jml: don't measure, don't get. [e.g. thats kindof predictable] [16:18] jml: I noticed you deleted the mention of iteration from the docstring [16:19] lifeless: I actually had deleted iteration at one point, because I thought we could do without it. (I fixed the motivating use case at the root) [16:19] jml: of a helper function; I assume you had good reason but I wanted to mention that it feels kindof lacking - that its not very clear whats going on. [16:19] lifeless: but now I've restored it along with the docstring [16:20] jml: an option feels nice; then we can report on the failing code periodically [16:20] lifeless: anyway, I'm thinking maybe a 'strict' optional parameter to ADRT that defaults to True. [16:21] lifeless: more hassle now, fewer backwards compatibility problems later [16:21] I think thats swings and roundabouts [16:21] folk will learn to always put strict=False in [16:21] it will never be able to go away. [16:22] lifeless: good point. [16:22] perhaps a parallel class [16:22] or something like that [16:22] lifeless: suggested names? [16:22] ADRTIterating [16:22] ADRTBrokenTwisted [16:22] uhm [16:23] lifeless: what advantage does that have over an option? [16:23] jml: the non iterating codepath doesn't need conditionals [16:24] lifeless: hmm [16:25] * jml gives it a shot [16:25] lifeless: btw, would appreciate your thoughts on https://bugs.launchpad.net/testtools/+bug/654474 [16:25] <_mup_> Bug #654474: An unexpectedSuccess is like a failure not a success [16:28] they are shallow, but added [16:28] jml: I'd like us to do a release this week. [16:28] And get it up to debian unstable [16:28] is that reasonable? [16:29] lifeless: yes, but as mentioned before, I would like to get this Launchpad branch that actually uses the deferred stuff landed before we do a release. [16:29] jml: yes, I'm asking 'all things considered' [16:29] lifeless: however, that seems likely to happen early this week. [16:29] lifeless: and after that, a release would be very welcome. [16:30] ok, please let me know if that is likely to slip, so that I can put 0.9.5 up - maxb is keen to help bzr in debian build sanely [16:31] I've been thinking about the datamapping layer vs test parallelisation [16:31] I'm still of the opinion that test parallelisation is past the really hard bit [16:32] if I did nothing but it until the epic, it might be working by then [16:32] but I don't think so - latency is going to be a killer on getting it all figured out and sorted. [16:32] and I only have 3 work weeks between now and then [16:32] lifeless: *nod* [16:33] lifeless: I think parallelization needs someone to own it, coordinate, report on progress etc. [16:33] lifeless: that is if we actually want to make progress on it [16:33] So, balanced consideration - a data mapping layer, with test double - thats our most urgent infrastructural need with the largest payoff both short and long term. [16:34] jml: I've found a small bug in the canonical.lp.host glue in my databasefixture branch, else that would have landed. [16:34] and immediately after it my librarian branch. [16:35] lifeless: what's your plan w/ data mapping layer then? [16:35] jml: spend 9-10 days sketching and bootstrapping it [16:35] jml: talk a lot [16:36] jml: and at the epic recruit a task force for it [16:36] lifeless: ok. [16:36] lifeless: makes sense. [16:36] oh, and do plenty of literature searches [16:36] we *may* find something precanned that we can 'fix up'. [16:37] lifeless: sounds a little big & risky, but a risk worth taking. [16:37] jml: indeed [16:38] lifeless: any feelings on diverging from stdlib unittest behaviour re unexpected success? [16:38] I think we need to remove the 'please report bugs in launchpad' link from the footer. [16:38] jml: I don't think the stdlib unittest behaviour has had a lot of user feedback and polish [16:38] jml: remember that it was done in situ, not by merging an established implementation. [16:39] *nod* [16:39] lifeless: we need to have some way for users who find problems w/ Launchpad to report those problems. [16:40] jml: agreed [16:40] jml: we're getting primarily ubuntu bugs at the moment [16:40] lifeless: fsvo "primarily" [16:40] poolie filed a bug about the page we're directing them to and how it could help with the confusion [16:40] jml: its about equal to genuine Launchpad bugs. [16:41] jml: we're not dying, but none of those bug reports will get good treatment in Ubuntu because they aren't apport filed [16:41] and many should be questions in the first place. [16:41] s/q/Q/ [16:44] wtf with bug 606296 [16:44] <_mup_> Bug #606296: Fix codehosting acceptance tests to not use sample data

[17:02] shrug [17:04] jml: new user confused by UI I guess [17:13] jml: the link in the footer used to be edge only [17:13] jml: I made them the same during the convergence part of unifying edge [17:37] lifeless: ahh right [17:37] jml: so I think rolling it back to what we had on prod before (nothing) will stop confusing users who are new to the environment. [17:37] lifeless: indeed. [17:37] jml: until we have a better UI for /launchpad/+filebug [17:38] - at which point I'd like to bring the link back, because I really value user feedback === Guest35900 is now known as jelmer [17:38] jml: you agree? [17:38] lifeless: yeah [17:39] lifeless: I also think we need to think more broadly about user feedback / beta stuff anyway [17:39] agreed [17:40] ok. that's all the tests running locally [17:40] not sure how the ForBrokenTwisted experiment will pan out with actual use, but worth a shot [17:41] well, how does the code look ? [17:41] and how will we/users know when they can use the unbroken one? [17:43] lifeless: they'll use the unbroken one by default. the trick is how do you know that the unclean warning you get is from your buggy tests or twisted's buggy APIs [17:43] would it be feasible to whitelist? [17:43] lifeless: I don't think so. [17:44] jml: by twisted release? [17:44] jml: I mean, release a new testtools in a few weeks when twisted is fixed [17:44] lifeless: sorry, I don't understand what would actually be on the white list. [17:45] jml: versions of twisted that are not broken. [17:45] lifeless: oh right. but you can use it with current Twisted for many things [17:56] lifeless: anyway, I'm doing a full test run now. mildly optimistic. [17:58] * jml off. [18:03] gl === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away [22:24] lifeless: failures. looks like I'm going to have to practically rewrite the distributionmirror_prober tests. [22:25] jml: can we land what you do have working ? [22:25] lifeless: yeah, we could actually. Wouldn't be able to delete the old layers, of course [22:26] lifeless: funny thing is, the failures in the distributionmirror tests only showed up in completely different tests. I wouldn't have known it was distributionmirror unless I had accidentally left DelayedCall debugging enabled [22:27] lifeless: I'd like to poke a little to see if I can get the errors better lined up with the tests that caused them. [22:27] that would be nice [22:38] Time Out Counts by Page ID [22:38] Hard Soft Page ID [22:38] 3555 0 LoginToken:+accountmerge [22:38] 266 54 Person:+commentedbugs [22:38] 203 6295 Archive:+index [22:38] 77 280 BugTask:+index [22:38] 12 93 ProjectGroupSet:CollectionResource:#project_groups [22:38] 10 245 Distribution:+bugtarget-portlet-bugfilters-stats [22:38] 10 26 DistroSeries:+queue [22:38] 10 9 Person:+bugs [22:38] 8 1 BinaryPackageBuild:+retry [22:38] 6 38 Milestone:+index [22:38] 3553 https://launchpad.net/token/.../+accountmerge [22:38] one url [22:42] will get to both of those later, if I can. [22:42] wgrant: do you know where sense Hofstede hangs out on IRC? [22:42] bed for now [22:42] jml: gnight [22:45] lifeless: he's 'sense' i think [22:45] I think he has merge on autopost with wget ;) [22:46] hm? [22:46] oh [22:47] mwhudson: https://devpad.canonical.com/~lpqateam/lpnet-oops.html - 3555 oopses, one url ;) [22:47] ah [22:47] that's a bit rude