Present: BrainDamage, hoijui, Tobi, abma, Kloot
Agenda
- Welcome
conclusion: nothing new about pantheios (lib/headers for logging) - Release Plan
conclusion: http://springrts.com/mantis/view.php?id=2467 seems the only (test) release blocker currently, hoijui already removed the "n"'s from master, so: game-devs: please fix your lua-scripts! unit_morph for example is broken! afaik: to fix, replace .n with .# - how to split dev / gaming server infrastucture more? (springrts.com)
conclusion: nothing to be done for now - try to integrate/use http://en.wikipedia.org/wiki/OpenSUSE_Build_Service for daily builds?
conclusion: abma will try that - what about "divide financial powers"? see Meeting minutes 2010-07-18, 2010-07-25
conclusion: licho: please say something!
- Anything else? (WVTTK)
conclusion: you can get notifications about comments / pull requests when you in the team "spring", BrainDamage plans a new map format, -flto could increase fps of spring about ~30%, Kloot re-enabled USE_UNSYNCED_HEIGHTMAP
<abma>hey!
<hoijui>hey
<hoijui>i am kind of tired
<hoijui>should we add everythign still left ot do to mantis?
<hoijui>like.. new branchign model, using git tags as versions in CMake, ...
<Tobi>for the bugs/features things definitely yes
<Tobi>not sure it helps a lot with new branching model :)
<abma>imo the todo in mantis seems to work better than in etherpad
<Tobi>probably true yeah
<hoijui>we have a todo in etherpad?
<abma>not really
<abma>stuff like "notices"
<abma>[ARP]hoijui_g5: any news about http://www.pantheios.org/ cmakefile?
<abma>you wrote, that it will be added maybe in may 2011
<abma>hey! :)
<Kloot>hey
<hoijui>hello Kloot
<hoijui>still no news about pantheios
<abma>ok
<hoijui>the maintianer seems to be .. well
<hoijui> we were two, requesting .. bascially nothing
<abma>ok
<hoijui>and he was not able to respond wiht a short message even in the last 4 months or so
<abma>:D
<hoijui>the other guy and me wanted source code on github
<hoijui>adn the other guy was willing to put it there
<hoijui>just asked for permission
<BrainDamage>hi, mind if i spy a bit? i'm here for the drama ( your commit sucks, etc )
<hoijui>(its currently available as tarbals only)
<hoijui>ah yeah Kloot
<hoijui> wanted to say...
<hoijui>
<hoijui>kick BD!
<Kloot>sure, if he bends over
<hoijui>*bends BD over*
<abma>hmm
Release plan
<abma>do we have currently a major bug that would prevent a alpha/beta/test-release?
<Kloot>no
<Kloot>nothing critical on mantis roadmap
<Kloot>is 2482 reproducable?
<abma>can dedicaded-server run a replay?
<Kloot>nope, just scripts
<Tobi>the headless client should be able to, right?
<hoijui>yes
<hoijui>should be able
<abma>hm, spring can run it too...
<abma>or you mean, maybe it can be reproduced when serving the demo?
<hoijui>we'll.. we wanted to do the n's thigns before a test release
<abma>ok, thats a stopper
<hoijui>..what do yo umean wiht: spring can run it too?
<abma>i don't understand the difference between headless and non-headless in this case
<hoijui>there si no rendering and no soudn in headless
<hoijui>but there is the simulation
<hoijui>dedicated is jsut network re-reouting
<hoijui>rerouting*
<abma>yes, but the dedicaded server doesn't run sim
<hoijui>yes
<hoijui>headless does
<BrainDamage>it could read the replay data and forward it to clients, if that's what you meant, but currently it doesn't
<abma>and the network-routing has the memleak
<hoijui>ah memleak
<abma>i thought we are talking about #2484? :)
<Kloot>no, the /give all crash
<abma>ehm
<Kloot>but memleak in netcode sounds more serious
<abma>fuu
<abma>sorry
<abma>seems that i'm tired, too
<hoijui>yeah
<abma>Kloot: no, its not reproduceable
<hoijui>mm
<abma>or, maybe i didn't test enough
<hoijui>you have seen this 3 times while playing online?
<hoijui>what happend to the game? did it crash?
<abma>no, it just hanged
<hoijui>ok
<hoijui>i have never sen that
<hoijui>seen
<hoijui>coudl it be that it was all on hte same server?
<hoijui>maybe they use a patched binary or so
<hoijui>and i never was on that server
<hoijui>just saying cause.. i spoectate wuite a few games
<abma>hu?
<hoijui>quite*
<abma>the hang happened when testing stuff with braindamage
<hoijui>ahh..
<abma>#2482
<hoijui>ah.. i was talking about 84 now :D
<hoijui>dedicated server
<abma>about 84: i think it didn't crash, it only caused a very laggy game
<hoijui>mm ok
<hoijui>ah yeah .. can be i saw that too
<hoijui>then again... no. it was high CPU usage there
<hoijui>well.. when we have a test release, we coudl run dedicated server in valgrind :D
<abma>hm.. then maybe its better to test that in the release...
<abma>maybe its already fixed there
<hoijui>yeah
<hoijui>aeh..
<hoijui>you mean in the test-release?
<abma>yes...
<abma>#82
<abma>hm
<abma>confusing :D
<BrainDamage>http://blogs.msdn.com/cfs-file.ashx/__k ... C6E986.png
<hoijui>ahh k :D doh
<hoijui>lotso mixup
<hoijui>yeah.. in release woudl be bad :D
<hoijui>in test release is ok
<hoijui>before would be better
<hoijui>wait..
<Kloot>feature freeze at this point might also not be unwise
<hoijui>#82 is not reproducible?
<hoijui>but we are all so active these days
<hoijui>do not let the release kill it!
<hoijui>ah well.. no i am fine with it
<hoijui>abma, why woudl you do #82 in the test release, nto before?
<hoijui>not reproducible?
<abma>yes, hard to reproduce
<hoijui>ok, we dont consider it blockign for the test release then
<hoijui>what about the n's? :D
<hoijui>i guess nobody except abma and me woudl even consider doing this, but we are noobs there
<hoijui>i woudl have to guess and talk about eahc case wiht mod devs or soemthing
<Kloot>mod devs will want them to stay ofc :)
<hoijui>no
<hoijui>surely not all
<hoijui>at least FLOZi and.. i think smoth would want them gone
<hoijui>or was it someoen else.. well at least FLZOi and someone else
<hoijui>and nobody was agasint it
<hoijui>remvove only for array-like tables
<hoijui>but i think i foudn tables that are filled liek this: { key1, value1, key2, value2, key3, value3, ...}
<Kloot>yeah, that's most of where they are used
<hoijui>with n=3
<hoijui>mm
<hoijui>so they shoudl stay there?
<hoijui>or shoudl they use #/2, and hte n removed?
<Kloot>for key=value tables it's very convenient to have the .n
<Kloot>because #n doesn't work on those
<hoijui>yeah... we woudl not want to remove them there for sure
<hoijui>my question is: shoudl i only remvoe it wehre n == #, or also where n == #/2 ?
<Kloot>eh, n should always equal number of keys
<Kloot>keys as in key=value pairs
<Kloot>mixed tables can use the # operator for the rest
<hoijui>ummm
<hoijui>yeah i know
<hoijui>i am not talkign abotu the map liek tables
<hoijui>only about array liek ones
<hoijui>but there are array like ones that are kind of maps/property lists
<hoijui>and there, n == #/2
<Kloot>oic
<Kloot>I would rather convert those to normal array format
<hoijui>.. it is also possible i am reading the code wrong
<hoijui>i guess i am
<Kloot>can you give an example of a function where you think this is happening?
<hoijui> for (it = groupUnits.begin(); it != groupUnits.end(); ++it) {
<hoijui> count++;
<hoijui> lua_pushnumber(L, count);
<hoijui> lua_pushnumber(L, (*it)->id);
<hoijui> lua_rawset(L, -3);
<hoijui> }
<Kloot>that's a plain array :)
<hoijui>(theres a lua_newtable(L); before that)
<hoijui>ok :D
<hoijui>good then
<hoijui>sorry
<hoijui>fugly shit that :D
<Kloot>yeah lua API takes some mental exercise
<hoijui>i guess i will have a try then, to remove the ns
<hoijui>in the array forms
<Kloot>glhf
<hoijui>thanks
<hoijui>:D
<hoijui>next?
<Kloot>y
<abma>hmm
how to split dev / gaming server infrastucture more? (springrts.com)
<hoijui>abma...
<abma>yes...
<abma>was because of the recent server problems i thought about that
<abma>but the move of the buildbots to koski's server made this a bit obsolete
<abma>only big point that could help a bit would be, to use a foreign etherpad server
<abma>piratenpad.de for example has some auth functions integrated
<abma>i know this is a bit boring...
<abma>mhm
<abma>just a suggestion
<abma>anything pro/con?
<abma>con is: links will change
<hoijui>hm?
<abma>pro is: (hopefully) fewer time to maintain
<hoijui>you mean a pads link is not fixed there?
<abma>con: unknown backup
<abma>ehm, no
<hoijui>Tobi, i reemmber there were problems wiht etherpad using lots of ram, and crashing often..
<abma>the currently known links will change
<hoijui>but it did not crash since soem time now..
<hoijui>is it all good and clean now?
<hoijui>ah ok
<abma>ping [RoX]Tobi
<abma>if it doesn't crash/eat mem, i would keep it
<abma>if it still does... i would suggest to try a foreign etherpad
<abma>ehm.. next
try to integrate/use http://en.wikipedia.org/wiki/OpenSUSE_Build_Service for daily builds?
<abma>opensuses build-service can build packages for most linux-distributions
<abma>i think we should try to use that...
<abma>this is a "free" build-farm
<abma>with that we can create packages for : http://en.opensuse.org/openSUSE:Build_S ... ld_targets
<hoijui>sounds good.. what are the obstacles?
<abma>setting it up
<abma>imo thats all
<abma>but someone is using it already for springrts
<hoijui>ah
<hoijui>...
<hoijui>can you find out who it is? is it successfull there?
<abma>yep, will try
<abma>was only some informational topic
<abma>...next
what about "divide financial powers"? see Meeting minutes 2010-07-18, 2010-07-25
<abma>hoijui tried to talk about that, but there was no interest at all
<abma>so.. keep it like it is?
<abma>what i currently dislike is: its not clear, where the donation goes
<hoijui>i think it was though of as a godo idea, but nobody wants to do somethign for it :D
<hoijui>especially not be one of the group that has shares "the power"
<abma>on springrts is: website, some game repos, zero-k windows stuff, ...
<abma>can we make that clear and promote that more?
<hoijui>would have to invite Licho for this
<abma>(he still can answer to the minutes)
<abma>so.. ehm... next
<hoijui>but if you bring this up here, it measn it shoudl be discussed (and it makes sense for that topic) and discussing here makes more sense then in forum
<hoijui>its too cumbersome there
<abma>hmm
<hoijui>but well.. as siad.. i am tired.. so i dont care if its skipped
<hoijui>whats next then?
<abma>np
<abma>thats all
<hoijui>ok
Anything else? (WVTTK)
<hoijui>i will do the n's stuff, and maybe try to do git/cmake/new-brnaching stuff
<hoijui>abma.. will check out that build service
<hoijui>and.. Kloot will fix all the bugs
<hoijui>right?
<Kloot>and I get all the $$$
<abma>on github: you can get notifications about comments / pull requests when you in the team "spring"
<hoijui>:D
<abma>(how to mark that fat)
<abma>?
<hoijui>you get the hatred for eveyr single bug thats left
<BrainDamage>me & koshi had a dreadful experience with buildservice, when trying to build something non-suse
<hoijui>huh?
<BrainDamage>basically, the buildservice is supposed to build for many distros
<hoijui>abma.. what fat?
<abma><strong><fat>**___on github: you can get notifications about comments / pull requests when you in the team "spring"___**</fat></strong>
<BrainDamage>but in reality, any distro that isn't suse has obsolte and poorly maintained packages
<hoijui>ah
<Tobi>[ARP]hoijui_g5, abma_irc, etherpad is ok-ish, it's cramped in a little bit of ram yeah but seems to cope with the occassional kick in the butt from monit
<BrainDamage>so you end up struggling regularry, or having to resort to use other people's packages
<abma>[RoX]Tobi: ok
<BrainDamage>which aren't exactly always top notch
<abma>BrainDamage: that sounds..awfull
<hoijui>mm :/
<abma>when did you test it?
<hoijui>(thansk Tobi)
<abma>maybe they improved stuff ...
<BrainDamage>1-2 years ago
<abma>ok
<abma>time to recheck :)
<BrainDamage>it should still be running, but build targets other than suse should be off
<hoijui>BD, what do you mena iwht using other peoples packages? for SL? or for dependencies?
<BrainDamage>deps
<hoijui>mm ok
<BrainDamage>they settle for old distro releases or lts
<BrainDamage>so you end up having outdated packages
<hoijui>but the list includes the newest ubunut at least
<hoijui>which is not too old
<hoijui>ah...
<BrainDamage>check how much it takes to update
<abma>at least they promote with "User friendliness."
<hoijui>you mean deps from when the OS version was released
<hoijui>:D
<BrainDamage>in the past, it used to lag several months behind releases
<hoijui>mm ok
<BrainDamage>even 1 year at some point
<hoijui>latest ubuntu is like.. 2 months or so, right?
<hoijui>well... if it is not too hard ot setup, it might be worth a try
<hoijui>at worst, we will alos only use OpenSUSE packages
<BrainDamage>oh, one thing, kinda unrelated
<BrainDamage>if i get some spare time after exams, i might make few map format changes/extensions
<BrainDamage>if none has a prob with the
<BrainDamage>i'll put them in a branch first, and make them retrocompatible
<BrainDamage>i'll make smf be able to load files directly, instead of that silly binary packing
<BrainDamage>so you can keep heightmap, texture tiles raw, etc
<BrainDamage>should make the maps more maintainable, and skip 1 compiling step which imo is kinda useless since spring de-serializes the on load
<BrainDamage>second thing, is to make start position è
<BrainDamage>profiles
<BrainDamage>as in, you can have more than 1 start position set, and spring can expose them over unitsync for lobbies to get the list
<BrainDamage>so the same map could get a ffa position profile, 4v4, 2v2v2, etc
<hoijui>mmm nice
<BrainDamage>oh, and let pos be able to be tagged by teams, so you could make random be team-sensitive, so when shuffling positions you would not jump map side
<BrainDamage>ofc latter being optional
<hoijui>mmm
<BrainDamage>any toughts?
<hoijui>(zerver is on etherpad :D )
<Kloot>map profiles sounds like a job for mapoptions to me
<hoijui>map-info maybe, even?
<BrainDamage>it needs a standard infrastructure tought
<BrainDamage>so you don't get lobbies etc seeing a different thing that the map does
<hoijui>yeah true.. but it sounds godo to put it there
<hoijui>good
<Kloot>for mapinfo engine would need a bit of extra code, mapoptions allow us to be lazier :]
<hoijui>eventually, there will be some in-game lua tool
<hoijui>i have to do support arbitrary map-infos anyway, if that is what you mean
<hoijui>it seems to fit better there in my eyes, logically
<hoijui>well..
<hoijui>i dont know
<hoijui>in a way yes, in a way not
<Kloot>heh
<hoijui>but one fo the two
<BrainDamage>this would be not just passively storing the info, it would have to use it
<hoijui>ah..
<hoijui>ok yeah.. then i guess options fit better
<hoijui>i though abotu the lobby doing the work..
<hoijui>but yeah.. makes no sense
<BrainDamage>the lobby would basically get a profile list, with their characteristics, and pick it
<hoijui>mmm
<BrainDamage>it could be used for chose before game start pos mode, or when using fixed/random
<BrainDamage>actually, with this idea, fixed start pos start to become useful again
<hoijui>
<BrainDamage>because you can make it react to amount of players, chosing best profile
<hoijui>to ensure mappers stick to the standard format...
<BrainDamage>yes, you basically have to enforce it :p
<hoijui>once your other change is done, it shoudl be possible ot have maps as sdd folders, right?
<hoijui>then they coudl run it wiht a lua tool in the engine
<BrainDamage>it's already possible, but it should be making them much easier
<Kloot>that already is, just the smt is compressed
<hoijui>and clcik the start pos thigns together
<hoijui>ah ok
<abma>BrainDamage: +1 for that :)
<hoijui>then the tool writes the info entries in the standard format
<hoijui>yeah.. also like it, especially this wya with using options
(and the other thing anyway)<abma>BrainDamage: you want to make it possible that spring directly can load three .png's as heightmap/map/metalmap?
<BrainDamage>correct
<BrainDamage>texture should still be stored as a bunch of tile files
<BrainDamage>or it'd ruin efficiency
<BrainDamage>btw, after this change it should be also much easier to extend the map file format without doing crazy thing to the binary format
<hoijui>what kind of extensions? partial loading?
<Kloot>not really, you still can't put anything new in the header (but you could look for presence of extra files of specific names)
<BrainDamage>yes, but it's much more kludged than it needs to be
<Tobi>map format has extensible header
<Kloot>oh lol, ExtraHeader sections
<Tobi>it's actually used for the vegetation map
<Kloot>needs mapconv changes to fill them though
<Tobi>yeah
<Tobi>adding stuff to a compiled map also requires shifting the rest and adapting offsets in other headers (of which you may not know the type)
<Tobi>extraheaders should have been a linked list
<Tobi>then it would be easy to add/remove stuff to the binary file without needing to know anything about other things stored in it
<Tobi>but well, zip files have that property too and are supported by common tools already
<Kloot>sometimes I want to go back in time
<Kloot>and "steer" the SY's toward different design choices
<Tobi>lol yeah :)
<Kloot>btw, should I re-enable USE_UNSYNCED_HEIGHTMAP?
<Kloot>it looks fast enough with your optimisation, but haven't tested extensively (400 hawks on patrol seem to cause ~10% bigger load than in 82.7)
<Tobi>suppose it would be good to enable it for a test release to have it tested
<hoijui>+1
<hoijui>just wanted to write the same
<Tobi>we can see if anyone complains about lower LOS performance out of the blue
<hoijui>night!
<Kloot>yeah, so long as people do not read this and remain unbiased :]
<Kloot>nn
<BrainDamage>btw, kinda related
<BrainDamage>tizbac pointed to me that gcc now has an higher level of optimization options
<BrainDamage>that work on both compiler & linker level
<BrainDamage>apparently it makes quite some difference for other games
<BrainDamage>and he compiled spring with it, and synced fine
<BrainDamage>option is -flto
<BrainDamage>http://gcc.gnu.org/projects/lto/lto.pdf
<BrainDamage>might be worth having some more tests?
<BrainDamage>he said his fps went 70 -> 100 on the same replay, on the standard version, but i guess some cpu time profiles should be better
<BrainDamage>the resulting binary was ~1MB smaller too
<Kloot>cool
<abma>25% performance gain, 10% loss through LOS changes... :D
<BrainDamage>as long as net is positive, players shouldn't complain :p
<Kloot>you'd think
<abma>so.. just enable it and see what happens?
<BrainDamage>pretty much
<BrainDamage>needs some sync testing, i'll try it also myself with current version
<BrainDamage>so far it shown no problems
<abma>i open a mantis, so we can't forget that :D
<Tobi>nice
<Tobi>I'm off too, laters
<abma>good night! :)
<BrainDamage>gn
<abma>btw, BrainDamage, did you found the problem with the lua-handler?
<BrainDamage>no, i had plenty of fun with microwave dielectric resonators tought today :p
<BrainDamage>i'm studying for exams, so free time changes wildly
<Kloot>did you cook any rats?
<Kloot>(poooor DNF ref)
<abma>aah, ok
<BrainDamage>sadly nothing ------- related :/
<BrainDamage>btw, behe's new roam iteration looks promising, maybe assist him more closely?
<abma>roam iteration?
<abma>aaah
<abma>terrain rendering?
<Kloot>his work is not public yet afaik
<BrainDamage>yes, ROAM is an algorithm that changes ground detail according to distance & frustum
<abma>hm, so ask him for pushing it to github?
<abma>at least he has an outdated repo there..
<BrainDamage>i think he dislikes git heavily, so you might have a problem there
<Kloot>yeah, plus I get the impression he likes to figure things out by himself
<abma>i asked him about pushing to the repo, let's see what happens...
<Kloot>I'm off for tonight, ciao :)
<abma>ciao! :)
Would BrainDamage system offer a clean solution to such need, and how?
