Open Source Meeting/2008-04-10

From Second Life Wiki
Jump to navigation Jump to search
  • [14:00] Carjay McGinnis: hello Liana
  • [14:00] Michelle2 Zenovka: Hi Liana
  • [14:01] Liana Linden: no agenda items today
  • [14:01] Liana Linden: Do you folks have topics you want to hit today?
  • [14:02] Opensource Obscure: this is about copyright, CC licenses and SL DRM
  • [14:02] Opensource Obscure: [1]
  • [14:02] Opensource Obscure: [2]
  • [14:02] Opensource Obscure: and [3]
  • [14:02] Opensource Obscure: kind of 'new feature request' thing.
  • [14:03] AlexanderThe Benelli: Hallo...entering a lot of chatranges but nobody rezzing...^^
  • [14:04] Squirrel Wood: patience ^^
  • [14:04] AlexanderThe Benelli: I'm very patient...i didn't crash nothing today because of Secondlife..
  • [14:05] AlexanderThe Benelli: perhaps because im online for a few minutes only :P
  • [14:05] Squirrel Wood: the current rc0 client need a LOT of work
  • [14:05] Liana Linden: Opensource: I think we've covered this ground before. Let me check the list.
  • [14:05] AlexanderThe Benelli: starts to see people
  • [14:05] Squirrel Wood: textures flash every once in a while and removing the friends list dropdown from the map window was a bad decision
  • [14:06] Michelle2 Zenovka: I think opensource was away when that was on the ajenda so rob posponed it
  • [14:06] Opensource Obscure: checks past meeting transcripts too.
  • [14:06] Liana Linden: Ah, and now Rob's away.
  • [14:06] Opensource Obscure: oh
  • [14:06] Opensource Obscure: ok : )
  • [14:06] Liana Linden: I recomend that you put your specific question on the agenda for next week. Can you make it then, Opensource?
  • [14:06] Opensource Obscure: fine, liana
  • [14:07] Liana Linden: Thanks.
  • [14:07] Aimee Trescothick: prepare for possible invasion, someone just mentioned this meeting at the Windlight meeting in Pooley which is stacked with peopl
  • [14:07] Aimee Trescothick: e
  • [14:07] Liana Linden: For those of you just joining us, Rob is away today.
  • [14:08] Liana Linden: We don't have any specific agenda items, which means we resort to default agenda, which is Updates and jira triage.
  • [14:08] Michelle2 Zenovka: Liana, can we bring up cmake very quickly at some point (again)
  • [14:08] Liana Linden: Notice: chat transcript of this meeting will be posted on the wiki.
  • [14:09] Liana Linden: Michelle2: Sure, what's your question?
  • [14:09] Opensource Obscure: /
  • [14:09] Michelle2 Zenovka: Weve been having a bit of a session on #opensl testing cmake-8 and its more of a case we are filing a number of minor issues that cause build fails on all 3 platforms
  • [14:10] Michelle2 Zenovka: Just silly little thinks like missing libraries in libs package
  • [14:10] Anders Falworth: is building xmlrpc-epi now
  • [14:11] Liana Linden: Michelle, sounds frustrating. What's your question exactly? Just looking for a project update?
  • [14:12] Michelle2 Zenovka: An update would be good but more of a case just wanted to note these issues and we are attaching them all via VWR-2871 to keep them together
  • [14:12] Liana Linden: FYI for late commers: though we don't have much of one today, the agenda is at https://wiki.secondlife.com/wiki/Open_Source_Meeting/Agenda
  • [14:12] Soft Linden: Thanks, I'll check VWR-2871, see if I can knock those off after I get 1.20-RC0 building cleanly
  • [14:13] Michelle2 Zenovka: Anders don't forget to file your xmlrpc issue there too
  • [14:13] Liana Linden: Perfect place to do that, thanks Michelle. I'll ping Sardonyx Linden and ask him to send an update to the sldev list.
  • [14:13] Anders Falworth: Michelle2, will do.
  • [14:13] Michelle2 Zenovka: Yea i sent sardonyx a mail with a list and fixes for many issues, but i think there may be a differentce between copying from internal to public svn than files are being dropped
  • [14:14] Soft Linden: I don't see missing libs on VWR-2871. But I'd bet 3Dconnexion at the least needs updating, it did in 1.20
  • [14:14] Tara Walworth: Q may I disturb you please?
  • [14:14] Q Linden: go for it, tara
  • [14:14] Soft Linden: If you can make sure the missing libs are listed there or on sldev, I'll be sure to get 'em, Michelle2.
  • [14:15] Liana Linden: Does anyone have agenda items to add? If not, let's head into the default agenda.
  • [14:15] Michelle2 Zenovka: Soft: will do i can upate windows untill tomorrow now
  • [14:16] Liana Linden: OK, so default agenda of Updates and jira triage it is. :-)
  • [14:16] Soft Linden: Actually
  • [14:16] Liana Linden: Lindens want to start with updates?
  • [14:16] Soft Linden: There were three agenda items posted to sldev while the wiki was down. I'm digging those up
  • [14:16] Liana Linden: Thanks, soft.
  • [14:17] Soft Linden: Okay - the first was "What's coming in 1.20" - I posted release notes to sldev. Also, there's a text file in a non-obvious place (indra/newview, in the source tree) that should list those by the first RC release of a branch going forward.
  • [14:17] Soft Linden: Because of the non obvious location, I dropped a "releasenotes-where.txt" doc in doc/ that points to it.
  • [14:17] Liana Linden: Who were those from, Soft?
  • [14:18] Soft Linden: Mike Monkowski
  • [14:18] Carjay McGinnis: yes, thanks for that, Soft
  • [14:18] Soft Linden: One sec actually, I'll get the list archive link to his email
  • [14:19] Soft Linden: https://lists.secondlife.com/pipermail/sldev/2008-April/009011.html
  • [14:19] Mm Alder: Soft: That would be me. The questions were answered on-list except for the voice moderation.
  • [14:20] Soft Linden: The second item, he'd wanted to know about skinning for dazzle, after some strong reactions to the default skin. A Linden posted a link to a wiki guide on editing the XML in response to that.
  • [14:20] Soft Linden: Ah right, I thought it was you but wasn't 100%. I don't like getting the list to in-world names wrong. :)
  • [14:21] Mm Alder: I did put my name in the secret decoder list on the wiki. :-)
  • [14:21] Soft Linden: So for the last, on the question of group voice moderation - I followed up with the voice team. That's not deployed, but it hasn't been dropped and is still a priority. Can't give a firm date just yet.
  • [14:21] Mm Alder: OK
  • [14:21] Soft Linden: Looks like that's it for stuff from when the wiki was read-only.
  • [14:22] Liana Linden: Cheers for the follow up on that.
  • [14:23] Liana Linden: Launching into updates for the week -- Mine is pretty much same as last week. Helping with CMake project, working on licensing issues, lots of other inhouse program management stuff.
  • [14:24] Q Linden: is working on Google summer of code stuff. Anyone here involved in that?
  • [14:25] Mm Alder: Since you brought up CMake: Are theMS project files in teh RC source built with CMAke or are they still made manually?
  • [14:25] Liana Linden: SoC: [4]
  • [14:26] Soft Linden: Still manual right now. cmake hasn't been merged back to release yet, so all three are done by hand.
  • [14:26] Mm Alder: Thanks
  • [14:26] Soft Linden: We're really hoping our official switchover to VS2005 or newer internally will coincide with making cmake the one true way. At that point, that's what will be exported for the RC as well.
  • [14:27] Soft Linden: Nobody on Q's question?
  • [14:28] Soft Linden: For me, I'm getting the 1.20 sources into shape on all three platforms. I'm hoping to drop this evening or early tomorrow. Currently banging out some issues introduced by havok that need to be stripped out of the master project, 3Dconnexion and something unhappy with GL under Windows.
  • [14:29] Squirrel Wood: Ta! Second Life 1.20.0 (84432) Apr 9 2008 03:59:09 (Second Life Release Candidate) STILL does the camera freezes :(
  • [14:29] Liana Linden: Community updates?
  • [14:29] Aimee Trescothick: Loving the 3Dconnexion support ^^
  • [14:30] Opensource Obscure: 3Dconnexion -> is SpaceNavigator supposed to work on Linux too?
  • [14:30] Michelle2 Zenovka: My 1.20 build just broke
  • [14:30] Soft Linden: I saw on 3Dconnexion/Logitech's site that they do have Linux support. But as far as I know, we only have Windows and Mac so far.
  • [14:31] Opensource Obscure: thanks.
  • [14:32] Soft Linden: Sadly, the library isn't all open. So we'd not want that in a pure open source build anyway. Are there many of you building Linux viewers from source who are concerned about the closed components?
  • [14:32] Michelle2 Zenovka: yep
  • [14:32] Opensource Obscure: I do.
  • [14:32] Carjay McGinnis: since they break 64 bit builds, yes
  • [14:32] Ron Overdrive: well with exception at work, I use linux almost exclusively
  • [14:33] Soft Linden: I don't know where the priority on Linux support sits then. But the Mac version did come to us as an open source patch, following the Windows version done in-house.
  • [14:36] Liana Linden: If there aren't any more updates, perhaps we can move into jira triage.
  • [14:37] Liana Linden: [5]
  • [14:38] Soft Linden: I usually take a second to preload a few of these since our JIRA likes to get a bit pokey...
  • [14:38] Soft Linden: [6] Failure to compile v1.19.1(4) from current release snapshot under Linux with gcc4.1 (extra qualifier in lllineeditor.h
  • [14:39] Squirrel Wood: Whoever uses it - [14:37
  • [14:39] Rookiie Roux: smiles
  • [14:39] Soft Linden: Ah, this looks like about the safest patch in the history of patches.
  • [14:39] Soft Linden: I'll snag this one.
  • [14:40] Soft Linden: [7] Decoded audio WAV files have 8 bytes missing at the end (llaudiodecodemgr.cpp)
  • [14:40] Gigs Taggart: This has been reported before, like a year ago, IIRC
  • [14:40] Soft Linden: So the problem on this - to clarify after a bit of back and forth - isn't that the header size is wrong. We're actually writing 8 bytes too few.
  • [14:40] Carjay McGinnis: exactly, before the header size was wrong, too
  • [14:40] Carjay McGinnis: that's why OpenAL worked before
  • [14:41] Carjay McGinnis: then somebody tried to fix it but not 100%
  • [14:41] Soft Linden: I don't remember the discussion. But this time we've got a JIRA with a patch. It looks pretty straightforward - has anyone tried this?
  • [14:41] Soft Linden: I'm wondering if this might be the source of our OpenAL clicking, even.
  • [14:41] Carjay McGinnis: I use OpenAL currently, that's why I noticed, without that patch you don't get any in-world sounds
  • [14:42] Carjay McGinnis: but I don't notice any clicking
  • [14:42] Carjay McGinnis: but could be my ears of course
  • [14:42] Gigs Taggart: so with this patch, no clicking?
  • [14:42] Soft Linden: (Squirrel is a stealthy camera ninja)
  • [14:42] Squirrel Wood: hehe
  • [14:42] Squirrel Wood: I wasn't even there :p
  • [14:43] Soft Linden: This looks straightforward. I'm going to import this and assign it to internal triage. It wants more examination than the last patch.
  • [14:43] Carjay McGinnis: sure
  • [14:43] Soft Linden: [8] Do not error on gcc warning converting string to const char *
  • [14:43] Gigs Taggart: That warning is not in -Wall
  • [14:43] Gigs Taggart: why is it turned on in the first place?
  • [14:44] Michelle2 Zenovka: String conversion warnings are on in gcc 4.x
  • [14:44] Michelle2 Zenovka: wiht -Wall
  • [14:44] Soft Linden: So the warning is on writing to strings - does this mean treating a hard-coded string as a non-constant?
  • [14:44] Gigs Taggart: The reason it wasn't in Wall is because your entire program has to be const correct or you get spurious warnings
  • [14:44] Soft Linden: It sure looks like the correct solution is to make those gchars into const gchars, not turning off the warning.
  • [14:44] Gigs Taggart: Why did they change that in 4.x?
  • [14:45] Soft Linden: Are we otherwise const correct with strings at least?
  • [14:46] Q Linden: gigs, gcc wasn't pedantic enough yet.
  • [14:46] Soft Linden: Does gcc put strings in a read-only segment by default, as with Visual Studio?
  • [14:46] Gigs Taggart: -Wwrite-strings When compiling C, give string constants the type "const char[length]" so that copying the address of one into a non-"const" "char *" pointer will get a warning; when compiling C++, warn about the deprecated conversion from string constants to "char *".
  • [14:46] Gigs Taggart: -Wstallman
  • [14:47] Soft Linden: Well, that's not just pedantic. It allows the compiler to safeguard against intent to do something that will cause an access violation, under VS.Net on forward, at least.
  • [14:47] Gigs Taggart: soft there is the warning documentation
  • [14:47] Gigs Taggart: from 3.x at least
  • [14:48] Soft Linden: I'd like to put a note on this one, asking how many times that warning comes up, and whether it makes more sense to correct the misuse of string pointers. Anyone disagree?
  • [14:48] Gigs Taggart: I agree -Werror should not be turned off
  • [14:48] Carjay McGinnis: sure, I like Werror
  • [14:48] Soft Linden: okie
  • [14:49] Soft Linden: [9] Mozilla libraries interfer with memory types tracking (Client => Consoles => Memory)
  • [14:49] Soft Linden: Taking a sec just to parse this...
  • [14:49] Gigs Taggart: Soft it can be a major job to retrofit const correctness though.
  • [14:49] Soft Linden: Yeah. I'd like to see how far off we are on that, though.
  • [14:49] Soft Linden: If we're close, I'd rather fix it than break that window.
  • [14:50] Q Linden: spent a month once convincing gcc of const correctness, hence the arrant pedantry. ;-)
  • [14:51] Gigs Taggart: well is there a way to say -Wall-except ?
  • [14:51] Gigs Taggart: :P
  • [14:51] Michelle2 Zenovka: Yes thats what the -Wno-write-strings does
  • [14:51] Soft Linden: Yes. That's what that patch does.
  • [14:51] Gigs Taggart: oh ok
  • [14:51] Q Linden: so there are two options
  • [14:51] Gigs Taggart: Sorry I haven't looked at the patch
  • [14:51] Soft Linden: But I'd rather keep that specific warning if possible.
  • [14:51] Gigs Taggart: I assumed it just removed -Werror
  • [14:51] Soft Linden: Oh! No. That would be hella bad :)
  • [14:51] Michelle2 Zenovka: no
  • [14:52] Soft Linden: I understand your previous comment now.
  • [14:52] Soft Linden: 2765 here - does anyone have input on that?
  • [14:53] Harleen Gretzky: Memory console is gone now in 1.19.1.4 :)
  • [14:53] Mm Alder: Just a comment: I started looking at the viewer in Purify and the embedded browser looks nasty. Memory corruption.
  • [14:53] Soft Linden: So the problem seems to be that new and delete are overloaded in our code, but it's something in the preprocessor so it's affected differently in llxul headers when compiling the viewer vs the mozlib component?
  • [14:53] Gigs Taggart: Mm there has been other evidence of that, it would be good to track down.
  • [14:53] Carjay McGinnis: I only enabled the memory tracking once but haven't tried since then
  • [14:53] Mm Alder: I can send you the Purify output if you like.
  • [14:53] Carjay McGinnis: memory corruption where, Mm?
  • [14:53] Q Linden: if memory crosses lib boundaries, the allocators need to agree with each other
  • [14:54] Carjay McGinnis: did you investigate?
  • [14:54] Soft Linden: Right. I'm worried about two things here. First, we shouldn't have any preprocessor overloads for these intrinsics while parsing third party library headers. That's scary...
  • [14:54] Mm Alder: In order to do it right, I need to compile Mozlib with Vs2005 and I haven't figured out how to do that yet
  • [14:54] Soft Linden: But the other thing is, it bothers me that this means there's some data we're allocating and freeing inside a lib, or vice-versa?
  • [14:55] Michelle2 Zenovka: I don't think it helps that you staticly link a large chunk of mozilla
  • [14:55] Carjay McGinnis: ah, ok, right, you were on Windows
  • [14:55] Soft Linden: Er - that was ambiguous - we're allocating in the client and freeing in the lib or vice-versa.
  • [14:56] Carjay McGinnis: could happen inside a header, Soft
  • [14:56] Soft Linden: Ah, that's fair. So there may be some inline function/method in the header that's getting polluted.
  • [14:56] Carjay McGinnis: an inlined new or something
  • [14:56] Soft Linden: Right.
  • [14:57] Soft Linden: What's the right fix here then - making sure we disable new/delete overload in the preprocessor before including this header?
  • [14:57] Carjay McGinnis: I had similar fun with the Qt lib once, took me a while, I ended up selectively excluding certain headers
  • [14:58] Gigs Taggart: reimplements the client using only preprocessor metacode.
  • [14:58] Soft Linden: Maybe we should highlight this on sldev for a discussion. We're about out of time, not hearing any confident answers.
  • [14:59] Carjay McGinnis: memory tracking is always delicate I think
  • [14:59] Carjay McGinnis: I doubt there's a one fit all solution
  • [14:59] Soft Linden: Well, spending a little time actually looking at the code to verify what's happening here should get us somewhere.
  • [15:00] Soft Linden: The next issue's a flex quirk. Probably don't want to take that on in the last remaining minute. :)
  • [15:00] Liana Linden: Thanks for leading the triage, Soft. And thanks everyone for participating.
  • [15:01] Harleen Gretzky: tc all
  • [15:01] Soft Linden: Thanks for filling in for Rob, Liana!
  • [15:01] Liana Linden:  :-)
  • [15:01] Liana Linden: I'll post the minutes ... in a minute.
  • [15:01] Q Linden: time compression!
  • [15:02] Soft Linden: Awesome :>
  • [15:02] Carjay McGinnis: indeed