Open Source Meeting/2009-05-21

[14:04] Khamon Fate: Oh Philip, you'll know, is Scott Linden a real person or just kinda Turing bot that holds officehours to test AI?
[14:05] Soft Linden: That can mean lots of things, though. For example, figuring out how to do a traffic simulation would be working with government. He's got office hours in his profile if you want to learn more about what he does.
[14:05] Khamon Fate: and why is everybody wearing banjos? I didn't see that memo.
[14:05] Rob Linden: let's try this again...
[14:05] Khamon Fate: wb Rob
[14:05] Morgaine Dinova: Hiya again Rob :-)
[14:05] Merov Linden: doesn't wear a banjo...
[14:05] Rob Linden: is also banjo free
[14:06] Morgaine Dinova: I play MIDI on my whiskers.
[14:06] Khamon Fate: I was at his office hour Soft; it was embarrasing to say the least and what does that have to do with opensource so shutup Khamon!
[14:06] Khamon Fate: Is this opensource or JIRA?
[14:06] Rob Linden: well, per my mail to sldev, I put one thing on the agenda: https://wiki.secondlife.com/wiki/Open_Source_Meeting/Agenda
[14:07] Rob Linden: https://lists.secondlife.com/pipermail/sldev/2009-May/013970.html
[14:08] Rob Linden: anyone have thoughts about what sort of QA we should have before calling each release done?
[14:08] koen3 Bing: sorry all need to go offline its here in holland 23.08
[14:08] Khamon Fate: night koen
[14:09] Khamon Fate: before client or server releases?
[14:09] Morgaine Dinova: if (regression_test(FAILS) == 0) { release();} sounds about right :-)
[14:09] Rob Linden: Snowglobe viewer releases specifically
[14:10] Merov Linden: the problem becomes: define regression_test() Morgaine
[14:10] Techwolf Lupindo: My only request is a web page somewhere where one can get all the info they need to test a build. Not building instructions, that allready covered, a page of how to go about QAing a build.
[14:10] Philip Linden: how sad, i crashed. :)
[14:10] Rob Linden: note: we don't have dedicated QA on Snowglobe, and we're not trying to wrangle those resources at Linden Lab, other than what the people you see here can do
[14:10] Khamon Fate: wb Philip
[14:10] Digit Ditko: is there any type of test plan?
[14:10] Morgaine Dinova: Merov: apply the rule recursively, and you end up with TDD :-)
[14:10] Philip Linden: in snowglobe
[14:10] Melinda Latynina: I think that having everyone use the latest version with nothing but shallow bug fixes, for some period of time without important problems should be a good criterion.
[14:10] Techwolf Lupindo: Philip, got a crashlog?
[14:11] Soft Linden: Rob posted this a couple pages back: https://wiki.secondlife.com/wiki/Category:Test_Scripts
[14:11] Techwolf Lupindo: :-)
[14:11] Soft Linden: Those are test plans
[14:11] Dzonatas Sol: Just a note, don't want to start a new discussion: as for QA, I would like to see the separation between UI and network code in order to QA the network side on its own.
[14:12] Rob Linden: Let me ask the question I asked in my email:
[14:12] Rob Linden: who here is interested in chipping in on QA
[14:12] Rob Linden: ?
[14:12] Soft Linden: Dz - you might look at some of the prior unit testing discussions. That would fall into that area, I think, before retreading any ground.
[14:13] Robin Cornelius: Well without dedicated QA resources, using the viewer as much as possible for all usual activities is one way and the simplest. Although thats probably no where near enough
[14:13] Melinda Latynina: rob: i'm happy to use new builds as my everyday client and to report bugs, if that's what you mean.
[14:13] Dzonatas Sol: Soft - thanks for the hint
[14:14] Khamon Fate: Rob do y'all bot any of these test such as camming, group IM, landmarking? Some of these items could be exhaustively tested (burned) for days with botted clients and perhaps give a more realistic picture of normal usage.
[14:14] Robin Cornelius: but that combined with with crash feedback and trying to get that to minimise is a positive step
[14:14] Philip Linden: we'd love help building automation tests.
[14:14] Rob Linden: is anyone interested in doing anything more formal than just using it day-to-day?
[14:14] Digit Ditko: sure
[14:14] Khamon Fate: that was my train of thought, recruiting some people to help script those.
[14:14] Rob Linden: and of course, on the related point, anyone interested in helping to automate the process?
[14:15] Robin Cornelius: Khamon, does not help so much with client testing
[14:15] Khamon Fate: not as much true
[14:15] Rob Linden: Digit: great!
[14:15] Digit Ditko: I'll go through specific tests on the wiki test plans
[14:15] Techwolf Lupindo: I don't know where to start. But if there is a problem area, I could hammer on that part and find a few bugs.
[14:15] Merov Linden: going through the whole https://wiki.secondlife.com/wiki/Category:Test_Scripts would be a lot
[14:16] Techwolf Lupindo: I just glance at those tests. A lot of them are incomplete.
[14:16] Morgaine Dinova: If you're looking for dedicated human resources for QA, won't the project get logjammed on QA the same as the other work? Humans generally don't scale (although we're trying to get Nyx cloned).
[14:16] Philip Linden: yeah what would be awesome would be people willing to execute the tests in the wiki and report back results.
[14:16] Rob Linden: I think the scope of testing depends on the number of volunteers willing to help
[14:16] Merov Linden: One thing I've seen in other project was to do "test sprints"
[14:16] Soft Linden: Yup. Beefing up these scripts would be helpful too.
[14:16] Techwolf Lupindo: example: "Test textures for the following can be found here... when I figure out how to upload TGA's. "
[14:16] Rob Linden: by all means
[14:16] Philip Linden: there are a lot more OS people than we have QA
[14:16] Merov Linden: sort of a gathering of all willing for a couple of hours and pound through as much test cases as possible
[14:16] Rob Linden: if we had a lot of volunteers, we could divide and conquer
[14:17] Khamon Fate: given specs, I'm willing to execute a series of tests and report back, at least the ones I can do intelligently.
[14:17] Philip Linden: so both are good. We need both automation and people willing to run tests.
[14:17] Khamon Fate: You also need some type of standard reporting mechanism or you'll never compile the data.
[14:17] Philip Linden: ideally, yes, we automated as many of these tests as possible but in the meantime we need people.
[14:18] Digit Ditko: it can be very diffucult to generate "realisric" automated tests, humans still need to be in the loop for ad hoc type testing
[14:18] Techwolf Lupindo: A methoid of assigning out tests intellentilly would be needed if thre are many vonenteers.
[14:18] Khamon Fate: well yes, you can't just pop a bot out and check back three days later
[14:18] Khamon Fate: it'll have likely been sitting there for most of the time with some type of error
[14:19] Khamon Fate: but you can operate several of them at a time and complete more tests individually.
[14:19] Merov Linden: for the moment Techwolf, it doesn't seem we're having that problem... :)
[14:19] Rob Linden: On the tracking front: we could use wiki or jira for that (or something else I'm not thinking of)
[14:19] Melinda Latynina: dividing up and running through tests could be a lot more fun if done in-world at the same places and times. it would remove a lot of the lonlyness of the task and help when people need to test stuff on other avatars.
[14:19] Robin Cornelius: Sorry i missed some of the conversation.. what was the idea to have a list of tests to perform and split this up?
[14:19] Morgaine Dinova: I suggest you add a Test menu up top with a range of tests that can be run, or even a nice big red TEST button for everyone to press. Then even Joe Blogs can play at QA.
[14:19] Philip Linden: That is a good point melinda, we used to have an area where we did exactly that.
[14:19] Merov Linden: +1 Melinda, that's what "tests sprints" are about
[14:20] Soft Linden: These tests are also a way to learn a bit about the viewer - they likely exercise some features you don't use often, or at all
[14:20] Soft Linden: If you can get a group, and have different people running a given test in different cycles, it's kind of fun.
[14:20] Rob Linden: cool...well, let's ask this another way: who would be willing to show up for a test sprint if we held one?
[14:21] Data Landau: y
[14:21] Rob Linden: maybe we should make the next one of these meetings a test sprint?
[14:21] Digit Ditko: me
[14:21] Harleen Gretzky: I would
[14:21] Melinda Latynina: gee, when u put it that way! i'd be interested
[14:21] Khamon Fate: Rob you should host a mini QA track at SLCC
[14:21] Aimee Trescothick: yep
[14:21] Merov Linden: oui
[14:22] Rob Linden: excellent! that's starting to look like critical mass
[14:22] Soft Linden: Rob - do you know if QA uses the plans out of the pWiki? I'm wondering if there's any updating to be done.
[14:22] Philip Linden: having a group of people trying to run as many scripts as possible, while together - that is a great idea.
[14:22] Morgaine Dinova: Rob: I think you'll find that pretty much every techie that frequents this place will attend. Just add it to Office Hours and the HUD will announce it automatically :-)
[14:23] Thickbrick Sleaford: I think a table of links to test scripts in the wiki specific to snowglobe where people can insert a "run on date, pass/fail" will help direct attention
[14:23] Rob Linden: cool....should we try this for next week's meeting, or should we schedule separately?
[14:23] Philip Linden: Rob, do we want to gather test script results into a common jira entry in the meantime?
[14:23] Philip Linden: for those who would like to try running one?
[14:23] Rob Linden: Philip: yup....
[14:24] Philip Linden: which jira?
[14:24] Rob Linden: looks
[14:25] Rob Linden: http://jira.secondlife.com/browse/VWR-13640
[14:25] Rob Linden: alright folks...set yourselves up as watchers on that issue
[14:25] Melinda Latynina: if each script is in a separate sub-task, it'd be easy to update each one as they're run
[14:25] Khamon Fate: gotta run, will check back for this
[14:25] Craig80 Nightfire: ty
[14:25] Rob Linden: I'll put more information there, and yes, as Melinda suggests, one script per subtask
[14:25] Techwolf Lupindo: returns "In reference to the test sprints, i'll be there is scheduled ahead or happen to cache me online when its starts.
[14:26] Rob Linden: if someone wants to start filling that in now during this meeting, you'll have my admiration and gratitude....make sure you coordinate here in chat though so we don't get a lot of dupes
[14:26] Philip Linden: In any case, it will be great to test how quickly a group of folks together can thoroughly test a build.
[14:26] Techwolf Lupindo: watches it. :-)
[14:27] Melinda Latynina: so everyone, bring your favorite griefing tools! :-)
[14:27] Techwolf Lupindo: lol
[14:27] Philip Linden: As I mentioned on sldev, we've gotta test this because it is going to go up on secondlife.com/download
[14:27] Philip Linden: which is a big change!
[14:27] Techwolf Lupindo: Isn't that for the pile on test on the beta grid Melinda? :-)
[14:29] Rob Linden: cool...well, are we pretty much finished on this issue?
[14:29] Philip Linden: one more thing... if you report on a test script, also include how long it took you to run it.
[14:29] Philip Linden: that will be useful in terms of looking at things to automate first.
[14:30] Melinda Latynina: great idea!
[14:30] Morgaine Dinova: Philip: just put some strong disclaimers around it (mentioning community effort and community tested) and it'll be fine. Remember that people are very used to running flakey 3rd party viewers, and that's not just the techies. :-)
[14:30] Techwolf Lupindo: That would be a good newbie test figure. How long did it take you to run it. If its a area where I never used, some good matrix can be derived from that.
[14:31] Rob Linden: hrm...in other news, it looks like the Linux build broke coincident with the last checkin. however.....
[14:31] Rob Linden: I don't think Robin's checkin is to blame
[14:32] Digit Ditko: sighs
[14:32] Philip Linden: yes its OK if Snowglobe is somewhat less stable, but hopefully we'll have 10's of thousands of users.
[14:32] Techwolf Lupindo: I need to re-build all my svn ebuild to see how many got broken this week. :-)
[14:32] Robin Cornelius: it builds on GCC 4.3.3 but i've not had the build report yet
[14:32] Thickbrick Sleaford: last as of when?
[14:33] Robin Cornelius: is there a build log somewhere?
[14:33] Morgaine Dinova: Presumably some tests will require a dedicated sim with a predefined environment.
[14:33] Rob Linden: Robin: it went out to sldev-commit
[14:34] Rob Linden: I think the problem was the build machine switchover
[14:34] Techwolf Lupindo: Yea, expcially estate tools.
[14:34] Rob Linden: that seems a more likely culprit given the error
[14:34] Robin Cornelius: not see it yet in any case, laggy email
[14:34] Techwolf Lupindo: e-mail != IM.
[14:34] Rob Linden: here's the build log: http://s3.amazonaws.com/viewer-source-downloads/2009/http-texture/2297/failed-build.Linux
[14:34] Techwolf Lupindo: :-)
[14:35] Rob Linden: we just upgraded the Linux build machines to etch
[14:36] Techwolf Lupindo: How many cpu cores are use for the builds?
[14:36] Rob Linden: ...and it looks like they're trying to do 64-bit builds. we all know how that ends
[14:36] Robin Cornelius: very well for me
[14:36] CG Linden: as many as distcc will find :)
[14:36] Techwolf Lupindo: Rob, I have no problems with 64 bit builds. In fact, it fixes the gstreamer issue.
[14:36] CG Linden: oh, Hmmm....
[14:36] Digit Ditko: sometime works for me
[14:36] Dzonatas Sol: 64-bit builds work fine on lenny
[14:36] CG Linden: I need to updatye the distcc host list
[14:36] Robin Cornelius: it won't work non-standalone though
[14:36] Rob Linden: yup
[14:36] Robin Cornelius: and the build server is doing a non-standalone build in 64bit
[14:37] Robin Cornelius: now easybuild-2 would have told you off early for that ;-p
[14:37] Robin Cornelius: do i take it the build server is a AMD64 then?
[14:37] Robin Cornelius: (now)
[14:38] Rob Linden: doesn't know, but I suspect so
[14:38] Soft Linden: yep, it is
[14:38] Morgaine Dinova: Log suggests it
[14:39] Rob Linden: anyway....looks like the build succeeded on windows and mac
[14:39] Robin Cornelius: debootstrap into a chroot and fake a /usr/bin/uname and your are sorted
[14:39] Techwolf Lupindo: Looks like a X4. "-j 4"
[14:39] Robin Cornelius: hence my wanting to be able to build 32bit on a 64bit host
[14:40] Rob Linden: :)
[14:40] Robin Cornelius: must jira that whist i think about it
[14:40] Rob Linden: yeah, well, looks lie we'll be solving that problem :)
[14:40] Morgaine Dinova: This bombed out on paths though. No glib.h, etc
[14:41] Robin Cornelius: no gtk-ark-pango etc package for 64 bit, standard fail path
[14:41] Dzonatas Sol: I was able to compile 32 bit on 64 bit, it linked most items. There is something it needs for the final link the requires the host system to be setup a bit differently.
[14:41] CG Linden: we did solve that internally, I wonder why it didn't propagate...
[14:41] Rob Linden: so....here's what some of us will be working on:
[14:41] Robin Cornelius: CG solved which bit? the cross compile?
[14:42] Rob Linden: holds off on his update
[14:42] CG Linden: yeah
[14:43] Robin Cornelius: don't think its a biggie anyway, just two hard coded `uname` calls in develop.py and the cmake scripts and makeing sure the approprate flags are passed and paths are set up
[14:43] Rob Linden: Robin: worth filing the issue in JIRA then if it's still a problem
[14:44] Techwolf Lupindo: I ended up bypassing and calling cmake diretory so my ebuilds would work correctory.
[14:44] Techwolf Lupindo: cmake build that is.
[14:45] Rob Linden: anyway, the updates I was going to give:
[14:45] Rob Linden: we have three big known issues:
[14:45] Rob Linden: we have the windows freeze issue
[14:46] Rob Linden: http://jira.secondlife.com/browse/VWR-13437
[14:46] Rob Linden: (sorry....lost track of where I was cuting and pasting from)
[14:47] Rob Linden: OpenJPEG crashers: https://jira.secondlife.com/browse/VWR-13511
[14:47] Techwolf Lupindo: I had 13437 happen to me earler today on 1.23.2.RC2
[14:47] Rob Linden: ...and the curl crashers: "Curl crashers" https://jira.secondlife.com/browse/VWR-12952
[14:47] Techwolf Lupindo: Client crash and hung hard.
[14:48] Rob Linden: additionally, we've got the rebranding to get through
[14:48] Merov Linden: I'm working on 13511
[14:48] Rob Linden: the plan for 13511 is to revert to KDU for now
[14:48] Merov Linden: on 13437, I saw Mm Alder's comments on merge reverting fixes
[14:49] Merov Linden: some one with an access to the internal branches need to explore that
[14:49] Merov Linden: I
[14:49] Rob Linden: Merov: maybe I should take that on
[14:49] Merov Linden: I'm not convinced it's related to the described problem though
[14:50] Robin Cornelius: Merov, i believe the openjpeg fix is a one liner anyway, as you are shippng a version its probably worth fixing
[14:50] Robin Cornelius: seperate the the shipping kdu issue *
[14:50] Rob Linden: Robin: where's the fix at?
[14:51] Rob Linden: we keep getting pointed to whole new repositories and ways of getting code
[14:51] Robin Cornelius: I probably should post it some where convien and for that fix you probably want the minimum delta
[14:51] Rob Linden: that would be wonderful
[14:51] Merov Linden: Robin: can't you post that to the JIRA?
[14:51] Dzonatas Sol: The calloc patch?
[14:51] Robin Cornelius: yes
[14:52] Robin Cornelius: i can post it and it is the calloc patch image.c s/opj_malloc/opj_calloc/
[14:53] Rob Linden: worth a shot. the other factor at work in the KDU switch, though, is performance
[14:54] Robin Cornelius: sure, but even if KDU is shipped by default thats not a good reason to leave know crashers in that have fixes ;-)
[14:54] Dzonatas Sol: Yes, I don't know what issues Merov had with trying the patches from my repository (which includes that patch). It would have been good for you to give them a whirl.
[14:55] Merov Linden: it's hard to argue with Robin on that one :)
[14:55] Rob Linden: Robin: true
[14:56] Soft Linden: So, one other quickie here. Right now, we've got a single viewer_1-23 branch that we use for ongoing development, RCs, and final releases...
[14:56] Rob Linden: Dz: we'd like to make sure we have an official upstream provider, or short of theat, understand every patch we apply
[14:57] Soft Linden: Oops - too soon
[14:57] Rob Linden: anyway....you were saying Soft
[14:57] Soft Linden: Sure.
[14:58] Soft Linden: So, that mostly works, except that we have to do this weird thing when we release an RC, where we roll back some changes to match the RC, then roll forward again. Makes for confusing diffs
[14:58] Robin Cornelius: Rob, do it the debian way, apply serious bug/security fixes, keep these as seperate patches and a clean up stream then version a 1.3-2 1.3-3 etc for internal changes
[14:58] Soft Linden: Are those back-and-forth diffs problematic for people to read through? Yesterday we were wondering if we should have separate branches for the RC and/or final viewer
[14:59] Rob Linden: internally, we branch off of viewer-1.23 before making a specific rc, right?
[14:59] CG Linden: yep
[14:59] Techwolf Lupindo: counts the number to svn ebuilds he has...
[14:59] Aimee Trescothick: it the very least they could do with being clearly marked as such, quite a few times they've made me go "Huh, what?" in the past before realising what was going on
[14:59] Soft Linden: Yeah. So the RC may be a week older than the current work in the viewer branch.
[14:59] Aimee Trescothick: it=at
[15:00] Techwolf Lupindo: I have 9 ebuilds for 9 different brances of LL code.
[15:00] Dzonatas Sol: Rob, to followup my concern. (sorry wasn't finished): it you know how to reproduce those crashes and there is a fix posted, then the upstream provider (which partically includes me), needs to know if the patch fixes the know crash
[15:00] Dzonatas Sol: partiially*
[15:01] Robin Cornelius: and like wise i will be pushing similar patches into debian's openjpeg
[15:02] Dzonatas Sol: steps away due to disability... sorry
[15:02] Rob Linden: we can keep talking about the openjpeg stuff on IRC
[15:02] Rob Linden: in the meantime, we're about out of time here
[15:03] Soft Linden: No strong feelings for/against separate branches for ongoing work and RCs, on the previous question?
[15:03] Rob Linden: on the RC front.....I think we should probably bite the bullet and find a place for those separate branches
[15:03] Philip Linden: see you guys later.
[15:03] Rob Linden: see ya
[15:03] Morgaine Dinova: Cyu Philip :-)
[15:04] Soft Linden: My big worry was huuuuuge numbers of branchs. But not having to refer to wiki to know which revision is the RC would be nice
[15:04] Thomas Shikami: how about using tags?
[15:04] Rob Linden: maybe we should take the RC discussion to IRC as well
[15:04] CG Linden: tags are branchs for all intents :)
[15:04] Soft Linden: yar. In svn, tag = branch
[15:04] Thomas Shikami: yep, except they aren't updated
[15:04] Thomas Shikami: you can say, a patch is the diff between tag_a and tag_b
[15:05] CG Linden: but yeah, getting rid of the wiki page would be cool, since some person has to manually update it.
[15:05] CG Linden: even if the macros are quite spiffy :)
[15:05] Rob Linden: anyway....we're pretty much out of time
[15:05] Rob Linden: let's follow up on both of the open things on IRC
[15:05] Soft Linden: kk
[15:05] Rob Linden: irc.efnet.org #opensl
[15:06] Morgaine Dinova: You could keep a generic helpful wiki page which points to a mechanically generated page of links.
[15:06] Rob Linden: thanks everyone for coming
[15:06] Thomas Shikami: as a tip, you don't need to use trunk, branch, tags. But you can add additional folders, like patch
[15:07] Morgaine Dinova: Cyu Rob :-)
[15:07] Robin Cornelius: later all
[15:08] Digit Ditko: bye all
[15:08] Morgaine Dinova: waves to all
[15:08] CG Linden: byes
[15:08] Office: Going: to office
[15:08] Aimee Trescothick: bye
[15:08] Rob Linden: bye all
[15:08] Merov Linden: ok, let's kill VWR-13511...

Open Source Meeting/2009-05-21

Navigation menu

Search