[01:37] * mwhudson_ lunches === mwhudson_ is now known as mwhudson [02:35] hello all [02:40] maxb, hi [02:40] maxb, I made changes to your timing branch & tried to land it. [02:40] maxb, but something must have swallowed it (my system has been behaving badly recently) [02:43] mwhudson, btw, could you please take a look at https://code.edge.launchpad.net/~stub/launchpad/kill-harder/+merge/11517 -- istr we do very similar stuff in codehosting and wondered if we should share the love [02:48] allenap, are you around? [02:48] jml: hello [02:48] mwhudson, hi :) [02:49] jml: you know we said that the --stacked stuff in rocketfuel-get was pointless wrt shipit & the other thing? [02:49] jml: i'm not sure that's true in a lightweight checkout [02:49] hmm. [02:49] (it's been getting a standalone branch of shipit for me since before i went to lunch :/) [02:49] mwhudson, that's unfortunate. [02:50] jml: yes [02:50] (it also shows how long it's been since i updated sourcecode -- shipit wasn't there!) [02:50] mwhudson, what should we do about it? [02:50] jml: restore the stacking stuff, maybe? [02:51] mwhudson, hmm. [02:51] mwhudson, so for a light weight checkout, it'd stack on the remote branch? [02:51] jml: hmm [02:51] jml: using launchpad with a remote lightweight checkout sounds pretty masochistic [02:52] jml: i don't think it's a use case we need to cater too *all* that much [02:54] mwhudson, ok, so maybe I don't understand what just happened to you [02:54] jml: maybe i didn't understand what you mean [02:55] mwhudson, so for a light weight checkout, it'd stack on the remote branch? [02:55] ^ in that [02:55] mwhudson, no, before that. [02:55] mwhudson, are you sourcecode branches not kept in a repository that's shared with Launchpad? [02:55] jml: no [02:55] oh actually, that is the problem duh [02:56] i have ~/canonical/lp-sourcecode/sourcecode/ [02:59] * jml wants to stab exetel in the face. [02:59] well, not really. [02:59] what I really want is my karmic updates. [03:00] jml: just think in a few weeks you'll get to experience old world internet [03:00] mwhudson, heh [03:00] mwhudson, a few days, even. [03:00] subsecond ssh connections ! [03:03] jml, mwhudson you guys around? [03:03] barry: yeah, i'm working today as a swap day [03:03] barry: jml is just an addict [03:03] :) [03:03] :( [03:04] i am so close to (trying to) land my bug 417089 branch, i can taste it [03:04] Bug #417089: Conform heading and breadcrumb rules to UI 3.0 [03:04] Malone bug 417089 in launchpad-foundations "Conform heading and breadcrumb rules to UI 3.0" [High,In progress] https://launchpad.net/bugs/417089 [03:04] Bug #417089: Conform heading and breadcrumb rules to UI 3.0 [03:04] Bug #417089: Conform heading and breadcrumb rules to UI 3.0 [03:04] this is going to break everyone's page titles [03:04] i'd better make sure i break ec2test good and proper to help with that then [03:05] yeah, thanks. i've only been working on this damn branch for 2 weeks [03:05] but i think i'm reaching steady state with devel ;) [03:06] i can see how working on the weekend helps there :) [03:06] brb. networking issues. [03:07] i'm having fun installing karmic on my new machine while i wait for the tests to run [03:08] so it's all good. i'll be up at 2am to see how ec2 did [03:08] barry: do you want me to review something? [03:08] mwhudson: actually, the branch has been pretty well reviewed, well except for the 100 tests i've been fixing [03:08] ... and yet again, [03:09] mostly crap stuff like fixing the browser.titles [03:09] ec2test --headless taking so long makes my life less pleasant [03:09] but you can imagine it's a LOT of pages [03:09] jml: i hear you there [03:10] jml: working towards fixing that! [03:10] barry: i guess i'm saying, i'm sure there was a reason you asked if we were around [03:10] barry: do you want help, or just emotional support? :) [03:10] just emotional support :) [03:11] just sayin' hi to my crazy saturday workin' antipodean friends :) [03:15] go ec2, go [03:19] back [03:19] so.... [04:42] the inherent bone-headedness of some of ec2test is wearing me down [04:42] EC2TestRunner should take an instance as a parameter, not construct one, fer 'eavens sake [05:00] also, vals [13:53] Hi. How can I use an object name containing a / in an URL traversal? I thought of mapping the slash to tilde, but this results in a traversal NotFound error from the ParentSet, as the traversal now uses the mapped name for lookup, but the ParentSet doesn't contain an object with this name. [13:55] jmux: What object are you attempting to traverse to that has a / in its name? [13:55] I want to add Debians security components - updates/main, updates/contrib, updates/non-free [13:57] Why? That's probably not the right way to go about things. [13:59] Because I want to import the debian security archive into Launchpad as it is. [14:00] I see. [14:00] Where are you seeing the traversal problem? It should only manifest itself in the API, AFAIK. [14:01] I added an uri_name property to the component interface, which does the mapping [14:02] This works as expected, so I get the updates~main link to manage the component in the ComponentSet View, but the traversal doesn't work and I don't know. how to tell Zope / Launchpad to map updates~main back to updates main for the ComponentSet lookup [14:03] updates/main [14:03] You'll have to adjust ComponentSet to do it itself. There's nothing really magical. [14:05] I added the reverse mapping to __getitem__, and this works, but it seems some other function is used for traversal and I couldn't figure out which one [14:05] Odd. [14:06] Although I wonder why you want to do this -- that's not how Soyuz models updates/security. [14:07] Current Soyus DB model doesn't allow this construct of sub-components [14:07] No. It uses pockets for that purpose. [14:08] What exactly are you trying to achieve? [14:10] Just to keep the Debian structure for import. I know I can set the pocket and override the component for gina imports, but I would like to use the Debian component names. [14:12] But more importantly I have series names, which also contain a slash, which will have the same problem [14:12] that will cause utter chaos. [14:12] Why do you have such things? [14:14] It has been like this the last 4 years - names like halut/2.0.0 where halut is our codename for Debian Etch and theh rest is the Version string [14:14] Oooh dear. [14:14] That's not going to go well. [14:17] I would just need a mapping that tells Launchpad to map series and component names from "x/y" to "x~y" for travsersal... [14:17] More stuff than traversals will break if you mangle series names like that. [14:18] In fact, PostgreSQL will probably tell you that you are being silly. [14:19] Why? I don't think anything in PostgreSQL depends on not using '/' [14:19] No, but distroseries.name should have a valid_name constraint on it. [14:19] And that should forbid '/'. [14:22] Ok, but that's not really hard to change. [14:23] No. But more stuff than traversals depends on that constraint. [14:23] It's probably a very bad idea to attempt to change it. [14:32] So is there a known way to do this mapping for traversal? [15:09] My solution: one can use the traverse function of the Navigation class of the parent set to traverse the uri part directly. I missed that when I read the Navigation class. [15:52] * maxb is scared by the preceding conversation and wonders if mapping to pockets wouldn't be entirely saner in the long run [15:52] There is a discussion in the reprepro manpage about how the debian slash-in-component stuff is so broken [15:52] search for FakeComponent === mwhudson_ is now known as mwhudson [23:25] maxb: It does seem like a much saner approach to import them as pockets and perhaps alter archivepublisher to do them as evil nested components. [23:25] Do we need to publish them that way? [23:26] Let that insanity stay with security.debian.org :-) [23:26] maxb: If they have users, yes. [23:27] This sort of thing will come into the picture if Debian PPAs end up happening. [23:27] I don't understand.... given LP hosts no debian archives currently, can't it decide what components to allow [23:28] Well sure, but the main debian archive doesn't have nested components, that insanity is limited to s.d.o [23:28] maxb: Does it not seem like jmux is planning to do exactly that? [23:28] (not Debian, but a derivative?) [23:29] How's that going to work then? Isn't that going to require a second public instance of Launchpad, or the derivatives buildds in canonical's datacentre? [23:30] Or something similarly tricky to arrange [23:30] Probably. [23:30] Anyway, the solution is for the derivative not to copy debian's weird naming for security updates [23:30] :-) [23:30] Right. [23:31] But if they need to for compatibility, it's to map pockets differently on the output. [23:32] * maxb mutters something about symlinks :-) [23:33] Ah, you're right, that will work. I was thinking that Release would have the full name, but it just lives within updates/. [23:34] That makes everything muuuch easier. [23:37] Oh, it *does* actually use the full names, but only at the top. [23:37] How confusing. [23:44] It's more than a bit weird, but it mainly seems to be that you generate the "Components:" line in the Release file including extra path components which shouldn't really be there [23:45] Yes. [23:45] Which is substantially easier than doing what I suspected might be necessary.