[00:05:10] --- dragos.tatulea has left [02:14:06] --- tkeiser@sinenomine.net/owl has become available [05:03:12] --- Simon Wilkinson has become available [05:05:42] --- Simon Wilkinson has left [05:44:25] --- dev-zero@jabber.org has left [06:16:43] --- matt has become available [06:46:52] --- matt has left [07:45:59] --- edgester has become available [08:01:10] --- matt has become available [08:12:37] --- summatusmentis has become available [08:20:24] good morning gentlemen [08:20:46] hi [08:21:23] I'm working on the per testing framewok for openafs [08:21:35] s/per/perl/ [08:22:06] are you familiar with perl? [08:22:20] only minimally. I tried teaching myself perl once [08:22:27] ah [08:23:00] I remember it being odd, syntactically, coming from C, but enjoyable [08:23:07] k [08:23:24] jaltman asked for a config file to handle the test config [08:23:45] using a Perl module seems to be the best choice [08:24:36] you can use perl code in the config file to be flexible and all variables are in a separate namespace by default [08:25:29] example test code header: use lib '..'; # include the parent folder in lib search [08:26:18] use AfsTestConfig; $AfsTestConfig::working_dir # configured scratch directory for tests [08:28:02] opinions? [08:32:15] a perl module was how i did it [08:32:44] --- summatusmentis has left [08:32:58] sounds good. Are you talking about how the original test code did it? [08:33:19] i am, but i only used a module in a very limited manner, to configure one thing [08:33:24] hi shadow [08:33:44] good morning [08:34:04] ah, I want every test program to include the config module to keep all config in one place [08:34:14] yes [08:34:16] good idea [08:35:01] is it preferable that I make a new testing framework instead of fixing the old one? [08:36:13] your call. whatever is going to be easier for you [08:36:22] there's no "investment" in the old one [08:36:30] ok [08:36:30] so in this case as a contributor you get to decide [08:36:53] I think I'll start new and copy in stuff from the old stuff as needed [08:37:11] eventually, the new stuff will cover all of the old [08:37:16] openafs for windows: what bison is preferred (I had to hack my 2.52 bison.simple for rxk5 compile_et) [08:37:17] hopefully ;) [08:37:51] yuck, bison on windows, ouch [08:38:50] if one is preferred it's probably documented somewhere [08:40:41] shadow: how is the git migration going? [08:40:54] using "Win32 ports of flex and bison" is mentioned in README-NT, that's right, just not which; I'm using Monmouth 2.5.2 [08:41:03] mike and max have something. i will look at it monday. today i am in the middle of release crap [08:41:14] ah [08:41:25] istr instructions on the mailing list explaining exactly which at some point? [08:43:44] don't know; now that I've changed the parser template, need to keep it around :( [08:44:46] --- Simon Wilkinson has become available [08:44:53] hi Simon [08:54:14] --- Simon Wilkinson has left [09:01:53] win32 has not used flex or bison since IBM days [09:02:36] mit kerberos relies on there being a cygwin installation. [09:03:28] ActiveState Perl is what we are currently using. Although according to the git folks, ActiveState Perl utterly fails to work with git [09:05:49] SecureEndpoints: I'm glad you're here. Does using a Perl module for test configuration sound good? [09:07:36] perl module for test is fine. although you might want to look at one of the test frameworks built around scons [09:07:43] I believe that is perl based. [09:07:52] I'm not really here. just between vb games [09:08:12] <- sees scons, pokes self in eye [09:08:22] could you provide a pointer to such a framework? I'm using test::more because it's included with perl [09:08:24] SecureEndpoints: comerr [09:08:45] --- dev-zero@jabber.org has become available [09:11:11] shadow@gmail.com/owlE16C8B4F: I haven't heard about scons before, can yo shed some light on it? [09:11:49] YAMR. sadly i gotta go. looka t google. [09:12:00] google shows me that scons deals with python, huh? I thought the tests were to be dine in perl [09:12:04] if we're speaking of scons, might also look at cmake [09:12:06] see ya [09:12:17] * edgester shudders at cmake [09:12:41] I tried using cmake once to compile paraview and couldn't get the stupid thing compiled [09:12:57] my one experience with cmake was a bad one [09:13:16] Not the experience other people apparently have, though sorry. [09:13:42] People were talking it up as a build+test frame work at gsoc mentor summit [09:14:17] * edgester think he's been a victim of the fad train again [09:14:45] i'd punt scons. no pyhton thanks [09:15:16] shadow@gmail.com/owlE16C8B4F: thanks [09:16:21] I'm not against python, but all testing code should be in the same language (possibly with a little c code) and I know perl much better than python [09:16:49] I almost kicked perl for python when dealing with unicode filenames under windows. [09:17:13] I'm still not sure how to do unicode dirname in perl on windows [09:49:09] matt: re bison and comerr, you are correct [09:50:09] but bison is not in the PATH used to build openafs on my machine. [09:50:11] hmm.... [09:52:07] and that is because the C output of lex is committed to the repository and used directly [09:52:42] see error_table_nt.c and et_lex.lex_nt.c [09:53:16] scons does use python. I misremembered [09:53:22] so no scons [09:54:09] SecureEndpoints: ok [10:00:27] matt: from src/commer/NTMakefile prep: $(DESTDIR)\bin\touch et_lex.lex_nt.c $(DESTDIR)\bin\touch error_table_nt.c @echo If you update et_lex.lex.l, to build correctly, you should delete error_table_nt.c [10:26:12] yes, it's building; with the bison i'm using, checked mscvrt halts due to error in bison template supplied; this is rather trivial, I was just noting it, in case you have a preferred bison kit, where this cannot occur [10:28:32] again, having fixed it for this bison, there may be nothing much to be concerned with [10:31:34] cl.exe in vs2003 doesn't like to compile decorated nls function prototypes in cm_nls.h... [10:34:13] (and link.exe is prone to report bogus unresolved symbols when linking with many libraies, so far has always proved can be worked around, oh well) [10:35:33] matt: its only building for you because you are changing the source files and you must be deleting the output files. [10:35:43] pardon? [10:36:03] the *_nt.c files? [10:36:27] must delete, because we have new .l, .y files [10:39:13] now, this comple_et change may be one of those that should be factored out of rxk5 [10:39:25] but atm, it's in m64 [10:41:24] (tbutc, yuk) [11:19:22] --- dev-zero@jabber.org has left [11:54:20] --- matt has left [12:38:02] --- manfred furuholmen has become available [12:49:13] --- Simon Wilkinson has become available [13:15:33] --- matt has become available [13:20:04] --- Simon Wilkinson has left [13:36:49] --- Simon Wilkinson has become available [14:14:02] 1.4.8 builds for fedora-7, fedora-8, fedora-9, rhel4 and rhel5 are done. [14:14:04] I'm just kicking off fedora-10 [14:14:41] Simon Wilkinson: is this because of kernel updates? [14:15:16] Because of new release. [14:15:39] fedora 10? [14:16:10] All those builds are because of Fedora 8. [14:16:18] Fedora 8? I'm going nuts. [14:16:19] 1.4.8 [14:16:31] ouch [14:16:32] Fedora 10 is in beta, so I'm doing builds for that too. [14:18:08] Hopefully that made some kind of sense. [14:18:26] it did [14:18:28] some [14:20:43] You don't get more than some at this time of night. [14:21:29] Simon Wilkinson: did your patch for maemo file manager go anywhere? [14:21:41] It's a bit more complicated than it first appears. [14:21:48] The maemo file manager is pretty messed up. [14:22:03] You can fix it so that it knows about AFS, and knows that AFS is an 'expensive' file system. [14:22:10] that doesn't surprise me [14:22:26] Doing so fixes some of the problems. [14:23:16] But, it still insists on stating the children of every directory node that gets examined. Which means that when you go into /afs, it stats /afs/*/* [14:23:35] * edgester jaw hits the floor [14:23:39] ouch! [14:23:50] That wouldn't be quite so much of a problem, if it wasn't for the fact that its doing all of this in the same thread as the UI. [14:23:55] So when a stat blocks, you lose. [14:24:10] double ouch! [14:24:40] There are problems at a number of abstraction levels. The stating is actually being done by the gtkfilesystem model, not by hildon fm. [14:24:43] have you discussed this with the maemo folks and looked at getting a chsnge in freemantle? [14:24:52] I haven't discussed any of this with them yet. [14:25:22] I think Jeff has had some conversations, and that they're receptive to submissions. [14:25:47] I think the key is probably to use the 'gnomevfs' filesystem model, not the 'unix' one that they're currently using. [14:26:18] I haven't yet looked at the implications of making that change - I ran out of time at the Hackathon, and I've been doing other things since. [14:26:49] My priority at the moment is to get my code cleanup changes out of my tree, and to get disked-backed vcaches done, as that's blocking production use of the disconnected code. [14:27:20] --- haba has left [14:27:38] --- haba has become available [14:27:58] Simon Wilkinson: sounds good. [14:28:08] --- tkeiser@sinenomine.net/owl has left [14:31:39] > now, this comple_et change may be one of those that should be factored > out of rxk5 since rxk5 can't possibly need to change how error tables are generated... [14:31:40] Grrrr. Normal SRPM doesn't build on Fedora-10, as they've changed the patch fuzz rules. [14:33:20] shadow: yeah [14:50:55] --- manfred furuholmen has left [15:34:17] --- summatusmentis has become available [15:53:41] which is better, copying 10 lines into all test files, or hard-coding the perl include path so that the test must always be run form the same folder relative to the base of the test code tree? [15:54:01] s/form/from [15:58:59] --- marc.c.dionne has become available [16:28:48] --- marc.c.dionne has left [16:41:50] --- summatusmentis has left [17:27:05] --- Marc Dionne has become available [17:52:09] --- summatusmentis has become available [18:25:21] * edgester drops some testing code *plunk* [18:25:31] * edgester yells code drop! [18:26:54] oh no! Now look what you did! You've broken it! I spent $100 on that code! [18:27:11] This is why we can't have nice things [18:27:19] lol! [18:27:39] >_> [18:36:23] --- tkeiser@sinenomine.net/owl has become available [19:31:54] --- Marc Dionne has left [19:46:53] --- edgester has left [19:54:01] 8am EST meeting to discuss openafs website. I should prepare for it. [19:54:17] what sorts of prep? [19:54:44] sleep [19:54:48] I don't know. some sort of plan for you to execute on might be nice [19:55:10] derrick: that's valid prep [19:55:32] I pulled together a list of potential questions for the questionnaire, didn't get much further than that this weekend [19:55:58] that's a start, but there will be more. [19:56:07] I was thinking that I should put together a site map describing the things that I know I want [19:56:51] well, as a vehicle for discussion [19:58:12] that's reasonable [19:59:56] should I pull anything together? I wasn't sure where to go from the list of Qs [20:04:09] i need to pull it up again. [20:04:16] are you planning to use surveymonkey or what? [20:05:00] haven't thought about it. I was just getting ready to check out survey sites. I'm entirely unfamiliar with them [20:05:24] Jeff's email said the survey would be on openafs.org [http://openafs.org], but that doesn't necessarily need to hold true [20:07:48] there will be a link from openafs.org. the survey doesn't have to be on that site [20:08:00] that's what I assumed [20:10:11] surveymonkey looks easy and complete, objections? [20:10:36] you might want to speak with Esther about her experiences. [20:11:17] or ask Kevin Walsh. He was doing some work with surveymonkey [20:11:36] I have a meetup.com account for the newyork openafs users group. I think they may have surveys [20:11:53] who is Kevin Walsh? [20:12:11] workshop organizer from NJIT [20:12:20] ah [20:13:36] my ex-cow-orkers at CMU have used it for some things; i can ask tomorrow if you want other opinions on it [20:14:17] sure, if you get around to it, not terribly pressing if not [20:14:38] --- haba has left [20:14:43] neither of the people i could ask are people i particularly mind taking a few minutes to talk to [20:15:09] Creedon joined the new york openafs user's group [20:15:16] --- haba has become available [20:15:23] derrick, ok, either way [20:15:48] he was looking for my multihoming slides [20:18:10] meetup.com polls will not meet our needs [20:19:46] surveymonkey looks to be the de facto standard [20:20:00] for free online surveying [20:20:16] there is currently no backend for openafs.org [http://openafs.org] correct? [20:21:18] anyone know anything about polldaddy or questionpro or surveygizmo? [20:21:36] I've never heard of any of them [20:22:29] although polldaddy has conditional branching [20:22:51] surveygizmo is really powerful, but really expensive [20:23:58] do we really need enough of a feature-set to pay for service? [20:24:08] Kevin's comment about surveymonkey is that the free stuff is not sufficient for anything serious [20:24:17] ah [20:24:38] we can afford to pay for a service [20:24:59] we're not a registered non-profit yet are we? [20:25:08] for example, a survey service that permits graphics to be uploaded as part of the survey is important [20:25:12] no [20:26:30] doesn't look like polldaddy hosts images, but they allow embedded html [20:28:25] http://www.youtube.com/watch?v=VEaFK2v6PbM [20:28:47] my thought is that we will want to provide mockups for people to vote on [20:29:44] just layout? or actual design mockups? [20:30:10] layout [20:30:21] also, one of the writers for my section at the campus newspaper talked about that video for the weekly "best of the tube" article :) [20:38:11] I don't think poll daddy is the answer [20:39:27] I'm not sure what criteria you're using, questionpro looks similar to polldaddy [20:39:59] questionpro is much too expensive [20:40:44] ok, how long are we expecting to need polling functionality? [20:41:12] several months and its probably a good thing for us to have to survey folks about new releases, features, etc. [20:41:21] that's reasonable [20:41:45] surveymonkey is $200 / year [20:43:02] which is tenable [20:44:47] there's limesurvey, an open source solution [20:44:48] polldaddy is $200/year as well. [20:45:08] which would mean installing it somewhere [20:45:26] I assume the backend is DB-based thogh [20:45:28] though* [20:46:14] limeservice.com is a company that hosts surveys built with limesurvey [20:46:42] i bet limesurvey is packaged for debian [20:46:44] eh [20:46:52] i think i will ditch for the night [20:46:57] gn [20:47:02] 'night shadow [20:50:33] there's no limesurvey in the ubuntu repos, but it's possibly in debian [20:56:53] limeservice.com [http://limeservice.com] is usage based paymet [20:56:55] payment* [20:57:20] we might want to prototype the survey with a couple of services [20:57:59] we can't easily do that without paying all the services, unless we want to be limited to their free features [21:04:10] I'm gonna head to bed also, I'll be back tomorrow, bright and early [21:07:32] gn [22:24:40] --- floh has become available [22:45:08] --- haba has left [23:08:30] --- Simon Wilkinson has left [23:14:14] --- reuteras has become available [23:56:51] --- manfred furuholmen has become available