Open Development User Group/Archive/2011-05-04

From Second Life Wiki
Jump to navigation Jump to search
13:30  Yuzuru  Hello Oz.
  oobscure  hi everybody
  Yuzuru  Hello
  Office Orb  mail_to set to
    Start transcription to
    Transcript will be sent to:
  oobscure  there were new web boxes in the login screen
13:31    new to me at least.
  Oz Linden  No new OPEN issues to look at today, so I have a topic:
13:32    Do you use DebugOS builds on Windows? (as opposed to RelWithDebInfoOS)
  caps router chum  I made it, weee
  teddy Zepp  hello all
13:33  caps router chum  stabs caps router
  Oz Linden  let me put it another way: suppose we were to remove Debug builds ... would you care?
13:34    no one?
  বরুনদাশ গুপ্তে  I use it on Linux
  Oz Linden  sleepy crowd
  teddy Zepp  ;_0
  Techwolf Lupindo  My OS build setup handles debugging builds. I run under gdb and portage does split-debug.
  caps router chum  You only asked about 3 seconds ago >.>
  Yuzuru  Sorry.. I am thinking about it..
  caps router chum  I've only used debug once, it was only slightly useful in that case
  Oz Linden  was longer than that for me.... must have been lagged
13:35  caps router chum  slightly more usefull* than RelWithDebInfo
  Yuzuru  If I can compile it, I don't need debug build.
  Techwolf Lupindo  I _think_ LL Debug setting only does a couple things to the code?
  Oz Linden  I ask because they don't work on the mesh branch, which I'm attempting to merge now
13:36  বরুনদাশ গুপ্তে  define "don't work"?
  caps router chum  Ah.
    Dependenacy or code problems?
  Techwolf Lupindo  Perhaps we can give you a better answer Oz if we knew just what DebInfo does.
13:37  বরুনদাশ গুপ্তে  Main difference between RelWithDebInfo and Debug is that the former is optimized, I think.
  Oz Linden  They are using tcmalloc to deal with some alignment issues for SSE support, but it breaks the MS heap debugger
13:38    I actually don't know what the difference is specifically
  oobscure  personally I don't need it. I only build to provide feedback to LL, so if LL doesn't need user Debug builds it's fine for me.
  caps router chum  I think RelWithDebInfo uses the release libs but without optomizations and generates the debug symbols
  Oz Linden  personally I am not in habit of ever using debuggers (life as a real time networking dev)
    I use external observation and logging
13:39  Techwolf Lupindo  There is a couple #if defines that on debug, crash while Release will warning log and continue.
13:40  বরুনদাশ গুপ্তে  does RelWithDebInfo execute llassert()s or only llalwaysassert()? I know Release doesn't execute llasert()s.
  caps router chum  I think it does
  Oz Linden  don't know, but that should be easy to figure out
  Techwolf Lupindo  On linux, all debugging symbles are gen at compile with a gcc flag and those symbles can be stripped at the end and end up wiht a bin just like the gcc debug wasn't there. Hence why split-debug works. The bin is stripped and debugging is in a seperate file.
13:41  caps router chum  There's bugs I'm hit were you'd only generate an error on RelWithDebInfo
  বরুনদাশ গুপ্তে  ok
13:42    Then Debug probably isn't too important. There should just be some way to test whether non-realease assertions are hit, and with RelWithDebInfo that's available, then.
    But if you want more feedback, better ask on the mailing list. I suspect there aren't many windows based devs here, today.
13:43  Oz Linden  I'll do that
13:44    very likely, I'll check in the mesh merge, which will break it, and then we'll figure out what to do about it after
  Techwolf Lupindo  I would also help to clearify just what is the differences are to begine with, so that would be a good secondary question for the list.
13:45    After that mesh merge, going to play with it and see if one huge patch can be generated of just the mesh stuff.
  বরুনদাশ গুপ্তে  what for, Techwolf?
  Oz Linden  well, it's around 2100 changesets....
13:46  বরুনদাশ গুপ্তে  I doubt that hube patch would apply to anything but exactly what it was made for.
  Techwolf Lupindo  Looking at single changeset does'nt give you the idea of what all was changed.
  Oz Linden  you certainly have an interesting way to entertain yourself....
13:47  caps router chum  Is it possible to build the mesh viewer under OS config? I'm assuming that the missing convex lib isn't a blocker to build.
  Techwolf Lupindo  hehehe
  Oz Linden  there is a stub version of the lib available on bitbucket
13:48    you'll have to build that locally and substitute the package into the autobuild configuration file
    which is something we need to make easier....
13:49  বরুনদাশ গুপ্তে  Different topic: Can EXP-749 be made publicly visible, please? VWR-25595 was closed as a duplicate of it.
  Office Orb

VWR-25595 Profile Icons (group, user) and other art missing or defaulting to generic

13:51  Oz Linden  it should be visible
  caps router chum  Ah.
  Oz Linden  it's set to be visible, anyway
  caps router chum  Not for me
  oobscure  VWR-25477 - same problem
  Techwolf Lupindo  Permission Violation
  বরুনদাশ গুপ্তে  I get "Permission Violation" on that, (like on most exp issues)
  Oz Linden  are you logged in to jira?
13:52  caps router chum  Yes.
  বরুনদাশ গুপ্তে  yes
  Oz Linden  hmmm....
  Techwolf Lupindo  I'me logged in
  Yuzuru  I am login but I can't see EXP-749
  Ima Mechanique  same here for "Permission Violation"
13:53  Yuzuru  Permission Violation, same
  বরুনদাশ গুপ্তে  VWR-25477 is actually a even worser case, as that was moved to EXP-752, so even the originally public issue isn't visible anymore.
  Oz Linden  I'll see what I can do... I'll have to file an issue against jira
13:54  বরুনদাশ গুপ্তে  Thanks.
  caps router chum  Lol.
  Oz Linden  if it's any comfort, both those issues are marked as fixed
  caps router chum  Is there an issue on the openID problems with jira killing open tabs?
  Oz Linden  but may not have been merged yet, I don't know
13:55    I don't know what you're referring to, caps
  বরুনদাশ গুপ্তে  I think the fixes are both in viewer-development, now
13:56  Office Orb  End transcription
  caps router chum  When resuming a session in firefox and the openID auth redirects all open jira tabs to the dashboard.
    I found the caps issue jira, SVC-6917
  Office Orb

SVC-6917 Apache2-worker/caps router on sim hosts is relaying 499 HTTP status codes back to clients - not a valid wire status code

  Oz Linden  oh... I don't know if there's an issue for that or not
    if there is, it would be in WEB
  বরুনদাশ গুপ্তে  ah, yeah, jira doesn't seem to be very RESTful :-P
  caps router chum  I couldnt find one
13:57    And it would be nice if the caps problems were fixed, I havent been able to do anything on SL for 2 weeks
14:00  Oz Linden  I don't know anything about that, but it doesn't look good....
  caps router chum  Its not, I can't do anything on v2 till its fixed ;~;
14:01  বরুনদাশ গুপ্তে  re openID issues: maybe WEB-2771 should be re-opened?
  Office Orb

WEB-2771 Following a direct link to a JIRA issues shows the system dashboard instead on the first try

14:02  Oz Linden  it's not that simple
  caps router chum  !!
  Techwolf Lupindo  That one breaks my "jira <search>" trick in firefox.
  caps router chum  THat looks like the one
  Oz Linden  that works fine - firefox is doing something other than just repeating the request when it restores the session
14:03  caps router chum  When i log in, it always redirects me to the dashboard, everytime
  Oz Linden  I follow jira links all day, and often (too often) have to log in, and I get there fine
  caps router chum  Unless I'm logged in already, which is most of the time
14:04  বরুনদাশ গুপ্তে  I think the problem is when jira thinks you are not (anymore) logged in, but the central login page thinks you are logged in and redirects you back to jira without asking for credentials.
  Oz Linden  I don't think I've seen that, but I may be accessing things differently
14:05  বরুনদাশ গুপ্তে  If I have to log in manually, I get redirected correctly, but if the login page redirects me back without asking me, I end up on the dashboard.
14:08    Why did the transcription stop at [13:56]? I thought this UG is now till 14:30, isn't it?
  Oz Linden  it does... I set the timer wrong and stopped it
  বরুনদাশ গুপ্তে  ah
14:09  Oz Linden  I'll have to fix the timestamp by hand when I create my transcript (laziness fail)
    I'll have to enhance that script... the current Orb assumes meetings end on the hour....
14:10  teddy Zepp  when writing a complier is they anyway of using varibles which will pick up on programs bugs so as to shorten the testing methodolgy?
  বরুনদাশ গুপ্তে  didn't you want to make your script public some time ago, so that we could help you tweak it?
14:11  teddy Zepp  seems a lot of time is spent in this processes of translating the software program
14:12  বরুনদাশ গুপ্তে  "complier"? do you mean "compiler"?
    we don't write our compilers ourselfs ...
14:13    but maybe you're looking for something like assertions or unit tests?
    Though, while those help catch mistakes, they make compiletime longer, not shorter.
14:14  Oz Linden  well... there are languages that are closer to "if it compiles then it will run" than C++
  বরুনদাশ গুপ্তে  indeed
  Oz Linden  but there's no way to make any language foolproof - we fools are way to inventive
14:15  বরুনদাশ গুপ্তে  :-P
  teddy Zepp  ahhh
    thank you Boroondas
  বরুনদাশ গুপ্তে  you're welcome :-)
  Cron Stardust  C++ isn't a pretty language, but it's what the industry is stuck with until something replaces it (D is trying...) :P
14:16  teddy Zepp  using C++ how much leak on resourse is there
  Oz Linden  There are lots of ways to write C++ to facilitate the ability to write good small unit tests, and doing that can greatly improve code correctness.
  teddy Zepp  ?
  Oz Linden  Alas, there are few examples of those class designs in the viewer code
14:17    C++ is notoriously easy to leak things in
14:18  বরুনদাশ গুপ্তে  One could say, memory leaking is the default modus operandi of C++ :-P
  teddy Zepp  yes
    happens to me many times
14:19  Cron Stardust  If the existing client was converted into another language, how open would LL be to switching to it? Or is such an idea too soon?
  Oz Linden  I created an issue just recently in hopes that someone might refactor some code to make it testable: STORM-1178
  Office Orb

STORM-1178 Develop tests for GPU table changes

14:20  Oz Linden  Converted how?
  Cron Stardust  those parts that would be roughly the same transferrred over and other parts re-written, all using unit tests, etc.
  Oz Linden  That would be a monumental undertaking
14:21  Techwolf Lupindo  The Linux kernel is in english for a good reson. There is no good way to translate error messages and the auto-transator can have bugs in it also.
  Cron Stardust  The language would have to be cross-platform and support existing (or replacement) dependencies of course.
  Oz Linden  Realistically, if someone wants to incrementally improve the code (as in that issue), that's more than welcome
14:22  teddy Zepp  wouldn't it be an ideal world for certian parts of the code to be hands off
    or is it pen for all develpoment and testing ?
  Oz Linden  meaning what? that we don't improve it?
  teddy Zepp  open
14:23  Oz Linden  I don't think that we consider anything "hands off" in general.... but proving compatibility for some things might be pretty challenging
  Ima Mechanique  teddy, no need for hands off. If LL don't like a patch, they just decline to accept it
14:24  teddy Zepp  ahhh
    good point
    I like to start from the ground u just didn't know how much would be allowed
    thank you
14:25    up
  বরুনদাশ গুপ্তে  We only change code if we have a reason to do so. But there might be some parts of the code that don't get changed (maybe although they /should/ get changed), because there's no one left who understands that code enough.
  teddy Zepp  yes, this appens a lot in the programmers world
  Oz Linden  happens on all successful software projects, in my experience
14:26  teddy Zepp  yes
  বরুনদাশ গুপ্তে  What do you want to start from ground up? Do you want to write a new second life client from scratch?
  Cron Stardust  AS an aside, I don't remember: is C++ object-code compatible with C? (such that you could compile a section of C++ code and then link it into a C program)
  teddy Zepp  so at this point what specific areas are we looking on developing or improving ?
14:27  Oz Linden  yes, but you have to declare the C symbols differently
14:28  বরুনদাশ গুপ্তে  yes and no. C has a flat namespace, so C++ names get mangled. I don't remember whether the mangling rules are implementation dependent or whether you might rely on them.
14:29  Cron Stardust  ok: the reason I was asking is that D is object-code compatible with C, and I was wondering if we could do a slow language transition of the source code... But many assumptions would have to be tested before that could even start.
  Oz Linden  /me's mind boggles
14:30  Cron Stardust  hehe, pipe dream, I know...
  Phoenix-FirestormViewer Rocks  why would you want to? :P
  Oz Linden  some of my best dreams.... never mind
  Techwolf Lupindo  D hasn't reached production level yet.
  Cron Stardust  D has a host of feature (not the least of which is option garbage collection) that could make the viewer much more stable.
14:31  Techwolf Lupindo  In fact, it still a changeing .
  Oz Linden  I've got to go meet with someone about getting a build farm up for contributors....
  বরুনদাশ গুপ্তে  \o/
  Oz Linden  so thank you all for coming
  teddy Zepp  thank you
  caps router chum  Thanks Oz
  teddy Zepp  Oz
  Techwolf Lupindo  garbage collection impact performace.
  oobscure  thank you Oz
    have a nice day everybody
  Phoenix-FirestormViewer Rocks  tc Oz