| [2010/01/02 00:39:50] @ Log started by gepetto_ | ||
| [2010/01/02 00:39:50] @ github joined channel #puppet-dev | ||
| [2010/01/02 00:39:50] <github> puppet: Ricky Zhou 0.25.x SHA1-fd631b9 | ||
| [2010/01/02 00:39:50] <github> Do not close stdout or stderr in execute. | ||
| [2010/01/02 00:39:51] <github> http://bit.ly/6KZpDu | ||
| [2010/01/02 00:39:53] @ github left channel #puppet-dev () | ||
| [2010/01/02 01:23:41] <jamesturnbull> bugs ... I blame the users ... if no one used Puppet we'd never have any bugs | ||
| [2010/01/02 01:36:30] <mqr_> jamesturnbull? | ||
| [2010/01/02 01:38:50] <jamesturnbull> mqr_: here | ||
| [2010/01/02 01:39:17] <mqr_> Looks like #2994 isn't as resolved as I'd like. | ||
| [2010/01/02 01:39:18] <gepetto_> mqr_: #2994 is http://projects.reductivelabs.com/issues/show/2994 "Puppet - Bug #2994: Should not attempt to manage content without checksum - ReductiveLabs.com" | ||
| [2010/01/02 01:39:52] <mqr_> The changes I made to make it always pass the tests make it sometimes fail in the real world. | ||
| [2010/01/02 01:40:18] <jamesturnbull> that's an issue | ||
| [2010/01/02 01:40:25] <jamesturnbull> tests pass at least | ||
| [2010/01/02 01:40:27] <jamesturnbull> :) | ||
| [2010/01/02 01:40:35] <mqr_> It appears to be a design problem, and I should have changed the tests to conform to the code rather than visa versa. | ||
| [2010/01/02 01:41:17] <mqr_> The thing that concerns me, a little, is that this will mean a behavior change: links will always be checksumed. | ||
| [2010/01/02 01:41:19] <jamesturnbull> so ... | ||
| [2010/01/02 01:41:26] <jamesturnbull> ah | ||
| [2010/01/02 01:41:43] <jamesturnbull> performance hit too? | ||
| [2010/01/02 01:41:52] <mqr_> Possibly a performance issue, but I'm not seeing a way around it. | ||
| [2010/01/02 01:42:20] <jamesturnbull> hmmm ugly | ||
| [2010/01/02 01:42:30] <mqr_> We carefully loose the distinction between links => manage and links => follow | ||
| [2010/01/02 01:42:47] <mqr_> And if we're following, we need the checksum. | ||
| [2010/01/02 01:43:46] <mqr_> I thought we could tell, and only checksum when needed, but in the real world that isn't always the case. | ||
| [2010/01/02 01:44:16] <jamesturnbull> no I see that | ||
| [2010/01/02 01:44:43] <mqr_> I'm logged on to a users test box (Mark Plaksin, who reported the issue) and I"ve been trying to find a way to get clever and tell if its needed. | ||
| [2010/01/02 01:44:58] <mqr_> For every trick I can come up with a case that breaks it. | ||
| [2010/01/02 01:46:00] <jamesturnbull> yeah I can see how that'd go in circles | ||
| [2010/01/02 01:46:29] <mqr_> So I'm thinking we need to partially roll back my last few tweeks (and adjust the tests) for the final. | ||
| [2010/01/02 01:47:35] <mqr_> After some digging, I believe that the only impact will be performance; you apparently can't manage links that you couldn't checksum anyway. | ||
| [2010/01/02 01:47:48] <mqr_> Does this sound reasonable? | ||
| [2010/01/02 01:49:10] <jamesturnbull> so looking back on tweak #2 - the last fix - what's the issue there? from the code it appears it should only checksum if @links == follow | ||
| [2010/01/02 01:49:10] <gepetto_> jamesturnbull: #2 is http://projects.reductivelabs.com/issues/show/2 "Puppet - Feature #2: Add 'link' ability to File - ReductiveLabs.com" | ||
| [2010/01/02 01:49:38] <jamesturnbull> gepetto_: thanks botly | ||
| [2010/01/02 01:50:48] <mqr_> Yes, that's the source of the problem. @links gets set to :manage when we want the client to think that it's a regular file and just copy it over because @links _had_been_ set to :follow. | ||
| [2010/01/02 01:51:57] <jamesturnbull> ah crapola | ||
| [2010/01/02 01:52:25] <jamesturnbull> yeah I think your approach is right short of a massive refactor | ||
| [2010/01/02 01:52:27] <mqr_> I think the idea was we don't want the client to see any difference between data coming from a file and data coming from a link with links => follow. | ||
| [2010/01/02 01:52:37] <jamesturnbull> agreed | ||
| [2010/01/02 01:53:36] <jamesturnbull> okay off to go shopping and eat but agree you approach is reasonable albeit annoying | ||
| [2010/01/02 01:54:09] <mqr_> If I put up a patch with those changes in about 45 minutes, can you push it to 0.25.x and I'll ping Mark to try it as sort of a special edition rc3x (the directors cut)? | ||
| [2010/01/02 01:55:01] <jamesturnbull> yeah might be a couple of hours | ||
| [2010/01/02 01:55:16] <jamesturnbull> but yes | ||
| [2010/01/02 01:55:38] <mqr_> No worries. He won't be getting to it for 7 hrs or so, if I understand his schedule. | ||
| [2010/01/02 01:56:00] <mqr_> I'll ping you when the branch is up, and then be turning in myself. | ||
| [2010/01/02 02:46:10] <mqr_> It's up, I'm going down. | ||
| [2010/01/02 03:51:54] @ github joined channel #puppet-dev | ||
| [2010/01/02 03:51:54] <github> puppet: Markus Roberts 0.25.x SHA1-cbc2ef0 | ||
| [2010/01/02 03:51:54] <github> Partial rollback of refinements to fix for #2994 | ||
| [2010/01/02 03:51:55] <gepetto_> github: #2994 is http://projects.reductivelabs.com/issues/show/2994 "Puppet - Bug #2994: Should not attempt to manage content without checksum - ReductiveLabs.com" | ||
| [2010/01/02 03:51:55] <github> http://bit.ly/4JBPht | ||
| [2010/01/02 03:51:57] @ github left channel #puppet-dev () | ||
| [2010/01/02 03:52:00] <jamesturnbull> mqr_: pushed and merged | ||
| [2010/01/02 03:57:12] <jamesturnbull> mqr_: there is still #2999 though | ||
| [2010/01/02 03:57:12] <gepetto_> jamesturnbull: mqr_: #2999 is http://projects.reductivelabs.com/issues/show/2999 "Puppet - Bug #2999: Absent package problem on Solaris - ReductiveLabs.com" | ||
| [2010/01/02 11:54:18] <mqr_> Yes, that's what I'm having for breakfast. | ||
| [2010/01/02 13:32:06] <mqr_> Fix for #2999 up, larger issues deferred to #3000 and 0.25.3 | ||
| [2010/01/02 13:32:07] <gepetto_> mqr_: #2999 is http://projects.reductivelabs.com/issues/show/2999 "Puppet - Bug #2999: Absent package problem on Solaris - ReductiveLabs.com" | ||
| [2010/01/02 13:32:08] <gepetto_> mqr_: #3000 is http://projects.reductivelabs.com/issues/show/3000 "Puppet - Refactor #3000: Normalize package providers' treatment of absent packages - ReductiveLabs.com" | ||
| [2010/01/02 18:18:47] <jamesturnbull> mqr_: consistency ... we don't need no stinkin' consistency | ||
| [2010/01/02 18:19:13] <jamesturnbull> mqr_: is Mark testing the fix? | ||
| [2010/01/02 18:19:16] <mqr_> I'm assuming sarcasm | ||
| [2010/01/02 18:19:31] <mqr_> Haven't heard from him. | ||
| [2010/01/02 18:21:12] <mqr_> His logon on the testing box has been idle for just under 20 hours, | ||
| [2010/01/02 18:21:45] <mqr_> So what is the right thing for the providers to do? | ||
| [2010/01/02 18:22:08] <mqr_> I'm assuming it should be consistent, right? | ||
| [2010/01/02 18:25:30] <jamesturnbull> yeah | ||
| [2010/01/02 18:25:36] <jamesturnbull> mqr_: I was being sarcastic | ||
| [2010/01/02 18:26:16] <mqr_> Is {:ensure => :absent} the Right Thing To Do? | ||
| [2010/01/02 18:26:26] <jamesturnbull> mqr_: I've been saying for a while that we should cleanup the type/provider code overall - there are two distinct "eras" of them | ||
| [2010/01/02 18:26:53] <jamesturnbull> I am not sure because I wonder why someone wouldn't assume that was the Right Thing To Do | ||
| [2010/01/02 18:27:00] <mqr_> So we need a third. | ||
| [2010/01/02 18:27:22] <jamesturnbull> which suggests some logic at work that is clouded to me | ||
| [2010/01/02 18:27:50] <mqr_> My second guess was "{:ensure => :purged}" | ||
| [2010/01/02 18:27:56] <jamesturnbull> I think a cleanup too would make the code base much more approachable to newcomers too | ||
| [2010/01/02 18:28:42] <jamesturnbull> I'd lean to absent over purged | ||
| [2010/01/02 18:28:55] <jamesturnbull> purged isn't supported on all providers | ||
| [2010/01/02 18:29:01] <mqr_> I think (maybe, this is just a guess) that the nil comes from cut & paste from elsewhere and the raising is just a philosophical bent; we couldn't get the version, o its an error. | ||
| [2010/01/02 18:29:33] <jamesturnbull> hmmm yeah | ||
| [2010/01/02 18:29:33] <mqr_> Also (and this is the reason I didn't go with it) we don't really know that it's purged. | ||
| [2010/01/02 18:30:03] <mqr_> But it could just as well be "by design" | ||
| [2010/01/02 18:31:26] <jamesturnbull> nods | ||
| [2010/01/02 18:33:40] <mqr_> I'm thinking rc3 with the #2994 and #2999 patches is looking real good for a 0.25.2 | ||
| [2010/01/02 18:33:41] <gepetto_> mqr_: #2994 is http://projects.reductivelabs.com/issues/show/2994 "Puppet - Bug #2994: Should not attempt to manage content without checksum - ReductiveLabs.com" | ||
| [2010/01/02 18:33:41] <gepetto_> mqr_: #2999 is http://projects.reductivelabs.com/issues/show/2999 "Puppet - Bug #2999: Absent package problem on Solaris - ReductiveLabs.com" | ||
| [2010/01/02 18:43:33] <Volcane> jamesturnbull: sorted out my ec2 pains | ||
| [2010/01/02 18:46:24] <jamesturnbull> Volcane: oh? | ||
| [2010/01/02 18:46:39] <jamesturnbull> mqr_: ditto | ||
| [2010/01/02 18:47:00] <Volcane> yeah somehow i managed to create my bucked in us and trying to bundle up in eu (assumed thats why you twittered?) | ||
| [2010/01/02 18:47:31] <jamesturnbull> Volcane: nah - if you wanted to write some articles I can point you at some people | ||
| [2010/01/02 18:47:45] <Volcane> oh lol | ||
| [2010/01/02 18:47:53] <Volcane> yeah let me get version 1 done i guess | ||
| [2010/01/02 18:48:36] <jamesturnbull> tell them you are Dutch - that seems to help - there is a whole mafia of Dutch journos in the Linux world | ||
| [2010/01/02 18:48:44] <Volcane> hehe | ||
| [2010/01/02 18:54:41] <jamesturnbull> mqr_: as asoon as mark confirms he's tested it and it's kosher we'll go then | ||
| [2010/01/02 18:55:02] <mqr_> Cool. | ||
| [2010/01/02 18:55:08] <jamesturnbull> mqr_: I haven't heard from Luke but he was going to brief marketing boffins and we were aiming for monday/tuesday ish | ||
| [2010/01/02 18:55:20] <mqr_> He had me do that. | ||
| [2010/01/02 18:55:35] <jamesturnbull> ok | ||
| [2010/01/02 18:55:44] <mqr_> I passed the release note along and answered a few questions. | ||
| [2010/01/02 18:56:15] <mqr_> Also set up the "tickets closed for 0.25.2" query for them | ||
| [2010/01/02 18:56:25] <jamesturnbull> ok | ||
| [2010/01/02 21:04:28] <jamesturnbull> mqr_: you see #2675? | ||
| [2010/01/02 21:04:29] <gepetto_> jamesturnbull: mqr_: #2675 is http://projects.reductivelabs.com/issues/show/2675 "Puppet - Bug #2675: ending slash in directory name gets stripped off - ReductiveLabs.com" | ||
| [2010/01/02 21:21:10] @ github joined channel #puppet-dev | ||
| [2010/01/02 21:21:10] <github> puppet: Jeroen van Meeuwen 0.25.x SHA1-87136b4 | ||
| [2010/01/02 21:21:10] <github> Fix typo in documentation: wheter vs. whether. | ||
| [2010/01/02 21:21:11] <github> http://bit.ly/6gfPdE | ||
| [2010/01/02 21:21:13] @ github left channel #puppet-dev () | ||
| [2010/01/02 21:39:52] <mqr_> jamesturnbull: He want's to reintroduce the bug? | ||
| [2010/01/02 21:44:00] <Volcane> he wants trailing / to work the same as witout it | ||
| [2010/01/02 21:47:03] <mqr_> It does now. Trailing slashes are stripped and (AFAKCT) do not affect anything. The patch would restore the distinction by retaining the trailing slash. | ||
| [2010/01/02 21:50:02] <Volcane> dont think he is hot on ruby, he did seem to just poke around till it worked for him, but before poking around it didnt work for him | ||
| [2010/01/02 21:53:04] <jamesturnbull> mqr_: he's on the -user channel if you wanted to talk throygh withhim |
Generated by irclog2html.py 2.6 by Marius Gedminas - find it at mg.pov.lt!