Dev Meeting Minutes (2011-01-24)

[jK]

Date: 2011-01-24
Present: abma, hoijui, jK, Kloot, Tobi, zerver



_Agenda_____________________________________________________________________

• Welcome
• Release plan
• revert separate main thread
• improve hangcheck during loadingstage

_Main Conclusions___________________________________________________________

• Boost'ification of the Mainthread is reverted
• Watchdog is improved during LoadingStage & new ClearWatchDogTimer call-out for Lua was added

_The meeting________________________________________________________________
>> jK joined!
>> Tobi joined!
>> zerver joined!
>> hoijui joined!
>> Kloot joined!

Welcome
hoijui: hello everyone
Kloot: lo
zerver: hi
Tobi: hey

Release plan
hoijui: .. i have not done any testing
zerver: are we talking .83 now?
hoijui: no, still 0.82.8
zerver: :)
zerver: i may have some commits to fix that water thing with MT
zerver: should i merge them into release branch?
hoijui: maybe commit to your own repo on current release branch...
hoijui: and tell me when done, is that ok?
zerver: yeah, if i can figure out how to do it
hoijui: (i mena, your github repo)
hoijui: i'll explain:
hoijui: https://github.com/spring/spring
hoijui: go there, and click the "Fork" button
hoijui: somewhere on top of the page
hoijui: in your local spring repo:
hoijui: git add remote zerver git@github.com:zerver/spring.git
hoijui: and hwen yo uhave the commits on the release branch ready in your local repo:
hoijui: git push zerver 0.82-branch
zerver: ok thx
hoijui:
zerver: i think i knew it already except that last git push syntax
hoijui: ah :D
hoijui: yeah.. i guess no more ot say there then
hoijui: next?
zerver: y

revert separate main thread
hoijui: aehm...
zerver: not a good idea, it's like wiping the problem under the rug imo
jK: this is now >1month on the meeting lists
hoijui: yeah... we found out .. about 1.5 weeks ago, that we do not need it really (apart from the problems it creates)
jK: and you didn't really tried to fix it nor is it easy to fix it at all
hoijui: as the sole reason was, for handling exceptions (where we do not need it)
zerver: if exception handling or stack trace fails with a separate main thread, then same thing will happen with the threaded game loading as well
zerver: so we have to revert that also
jK: ???
hoijui: if i remember right, findeings were as follows:
jK: threaded loading was still using the main thread and all exception were reached to it
hoijui: it is better to let exceptions fall through (as in.. abvoe main, to the OS .. or.. dont know how ot explain)
hoijui: or catch them in local places, where it makes sense
zerver: no, game loading is not in main thread
zerver: gameLoadThread = new COffscreenGLThread( boost::bind(&CGame::LoadGame, game, mapName) );
zerver: i think the problem is we are trying to exit cleanly after a crash
zerver: this causes multiple subsequent crashes
hoijui: the problem is, that this additional thread creates problems, and solves none
zerver: maybe so, but same problem would occur if there is a crash during game loading
hoijui: hmm.. yeah kind of,... i agree
hoijui: (i mena. with the trying to exit cleanly)
zerver: so better fix the underlying issue, if we can
jK: not true at all
jK: you are twisting things
hoijui: if you call catching all exceptions before they reach the OS clean
jK: also you had now >2months time to fix and didn't tried to
hoijui: that is kind of obsolte anyway... with what we foudn out 1.5 weeks ago
zerver: i have no personal interest in it
jK: then revert it
zerver: so we can revert it
hoijui: YEAH!! :D conclusion!
hoijui: who will do it?
zerver: but game loading crashes still have the issue
jK: trust me I test my code ...
zerver: i can make a switch so it is enabled / disabled with LoadingMT flag
jK: I debugged it for a very long time
hoijui: :/
hoijui: a switch?
zerver: so how is your thread better than mine
hoijui: hehe
jK: mine doesn't create SDL
hoijui: thing is, we will not try to catch the exceptions, so it cant happen the same way
zerver: this is not about SDL, ok?
jK: mine reaches exceptions to the mainthread and not a boost-thread
zerver: ?
>> abma joined!
abma: hey!
zerver: offscreenglthread is a boost thread
hoijui: hey
jK: but it isn't handling the exception (with exiting the libraries, showing the messagebox etc.)
zerver: true
hoijui: someone should revert it, and when that is done, we can put a throw new exception() into jks thread
hoijui: and see what happens
hoijui: and if it woudl fail bad, it coudl still be sovled then
hoijui: it does not interfere with the matter at stacke here
jK: as said I already tested that, but you can redo that when you don't trust me
hoijui: yeah, i wanted to say about that :D
zerver: is multiple crashes the problem we are facing, or something more?
hoijui: the SDL problem, and the destruction of crash information
zerver: SDL problem, what was that again?
hoijui: initialization/destruction in non-main thread causing problems
abma: problems at init / destruct
hoijui: anyway, there is not a single arguemnt pro this thread
zerver: how does the problem show?
abma: hang on exit, some strange x-server errors
jK: http://stackoverflow.com/questions/4372 ... break-down
abma: but it can't be really reproduced, it only happens sometimes
zerver: k, different behavior on linux obviously
zerver: so we revert it, i dont have linux
zerver: next?
zerver: (i will do it)
hoijui: who will revert it?
hoijui: ok

improve hangcheck during loadingstage
zerver: just need a ClearDrawWDT() in pathing code
zerver: and maybe export the functions to lua
jK: or in SetLoadMessage
zerver: y, that should do it
jK: (would need 2 diff commits for 0.82 & master, because of new loadingscreen class)
zerver: mmm
jK: k seems we are finished, any WVTTK?
hoijui: who will do that stuff?
hoijui: improvign hang-check
jK: those are one-liners
hoijui: ok
abma: then... jk, you'll do?
jK: when zerver hasn't done it, before my next commit day, then ya
abma: ok
abma: then... next point?
hoijui: that was the last one already
abma: hmm
abma: maybe we should remove the "Next Meeting" point?
hoijui: btw... about the discussion we had about the projectiles refactor
hoijui: i still think hte same way
hoijui: would anyone know where i coudl find Betalords email?
hoijui: i'd like to add it to TASServer
Tobi: I have one, but don't know if he'd want that listed publicly
<< Tobi left!
<< Kloot left!

[Licho]

Btw atm test builds of engine are still screwed..

* many gadgets are broken (terraforming, unitAI)
* terrain rendering suffers from jitter
* units are rendered incorrectly (too black or too white etc)
* pathing speed still not fixed (take 40 gators and order them to crater on comet catcher redux -> 4% cpu when away from crater, 23% cpu when they are trying to get into crater and can only do it 1by1. After a while most gators give up, though they can actually get there one by one)

That happens in tge most played open source ota-free game for this engine.

[Licho]

Also, google_frog made this funny video:

[smoth]

Btw atm test builds of engine are still screwed

What build numbers did you test? So people know.

bugs

wow!

That happens in tge most played open source ota-free game for this engine.

what game is that?

Also, google_frog made this funny video:

if there was a lua unit toggle for this, it could be a feature

[Google_Frog]

That happens in tge most played open source ota-free game for this engine.

what game is that?

Why don't you ask the pie chart? http://tracker.caspring.org/stats/view.php

I would just like an engine release that fixes the things that were broken in the past few releases and doesn't break anything new. The pathfinding has improved with this current version but it is prone to slowing the game to a crawl with only 20 units. There are many little bugs like turn in place (which is critical for zk), area attack and map icons. I checked a build recently and it was good to see turn in place was fixed.

[smoth]

That happens in tge most played open source ota-free game for this engine.

what game is that?

Why don't you ask the pie chart? http://tracker.caspring.org/stats/view.php

I'll be more blunt and less playful. What relevance does it being open source, most played etc? He could have just said ZK. He didn't, ZK is the only completely open source project outside of that very unimportant point, what does it matter that is is more popular.

that would like me saying "this happens in that game with the giant cartoon robots"

[AF]

This is how I see it:


Licho decided that merely mentioning ZK would not be enough, and seen as insignificant as a result. He mentioned the statistic to lend weight to his argument in hope that it would carry more weight against the 'we only bother with BA' crowd

Smoth:

I don't think Smoth sees the BA crowd problem as an issue at all, and thus failed to understand Lichos attempt at all, and instead saw it as arrogance and decided to call him on his bullshit.

If either party wishes to clarify, please feel free, my thoughts are momentary and in no way verified, but that's what it looks like from here.

[smoth]

Hardly calling licho on anything, I feel like I missed his point or am missing some crucial part of the context of his statement. Nothing to do with me "calling him" on anything. If anything ZK is one of the projects I see as an ally in spring content development.