Dev meeting minutes 2010-11-15

Dev meeting minutes 2010-11-15

Minutes of the meetings between Spring developers are archived here.
Post Reply
zerver
Spring Developer
Posts: 1358
Joined: 16 Dec 2006, 20:59

Dev meeting minutes 2010-11-15

Post by zerver » 16 Nov 2010, 00:57

Present: <abma_irc> <[ARP]hoijui_g5> <[LCC]jK> <Kloot> <[RoX]Tobi> <zerver>

=== Release plan ===
<[ARP]hoijui_g5> aehh release...
<[ARP]hoijui_g5> Kloot, you still want something in .7?
<abma_irc> hey!
<[ARP]hoijui_g5> we just started abma, nothing said so far
<Kloot> yes I do
<[ARP]hoijui_g5> does it have to do wiht the thing that was too slow?
<Kloot> yes, that
<[ARP]hoijui_g5> ok
<Kloot> I've an alternate fix for it in the works
<[ARP]hoijui_g5> ok
<[ARP]hoijui_g5> did you also have a look at the last KAIK bug i reported?
<Kloot> only briefly, I assumed that assert triggered because a unit got blown off the map
<[ARP]hoijui_g5> mmm ok
<[ARP]hoijui_g5> shoudl we probably convert the asert to a warnign message then?
<Kloot> hmm
<Kloot> I'd rather not until I know if it's a more serious problem
<Kloot> s/if/that it is not
<[ARP]hoijui_g5> Tobi, stack-trace translator can not identify/parse the version number of builds of non-master branch, for example the infolog here: http://springrts.com/mantis/view.php?id=2219
<[ARP]hoijui_g5> [ 0] Spring 0.82+.6.1 ({0.82-branch}0.82.6.1-45-g79c95f5) has crashed.
<[ARP]hoijui_g5> ok Kloot
<[RoX]Tobi> hm
<zerver> .6 is working quite well i think, i don't have much to add
<[ARP]hoijui_g5> ok :-)
<zerver> pathing could be better ofc
<[ARP]hoijui_g5> yeah.. Kloot changed quite soem pathing stuff for .7 already
<zerver> ok, nice
<[ARP]hoijui_g5> is in the release branch
<[ARP]hoijui_g5> next?
<zerver> and this bug where commander stops building if attacked is still there
<zerver> i'll try to pinpoint it
<[ARP]hoijui_g5> nice :-)
<zerver> yeah, next

Conclusions:
There are some remaining issues before 0.82.7 release



=== Color.h, vecX<type>, System/, Util/, Util/Math/ ===
<[ARP]hoijui_g5> related to this discussion
<[ARP]hoijui_g5> https://github.com/spring/spring/commit ... 8#comments
<[ARP]hoijui_g5> basically: cleanup math stuff
<[ARP]hoijui_g5> inclusing floatX, matrixX, vecX(new) and ColorABC
<[ARP]hoijui_g5> suggested was, introducing a new "package" (System, Sim, Rendering, ...)
<[ARP]hoijui_g5> called Util
<[LCC]jK> I don't think that anyone really wants to reimplement those vector types atm
<[LCC]jK> still there is left the thing with moving all data primitives into a separate
<[LCC]jK> +dir
<[ARP]hoijui_g5> i think, we should discuss.plan hwo to do it now, then implement the base stuff, and possibly convert the some of the little used stuff to use it
<[ARP]hoijui_g5> somewhen in the futture.. or never, maybe convert float3 and other often used stuff
<[ARP]hoijui_g5> it woudl be good to do now, cause of hte new Color thing
<[ARP]hoijui_g5> as so far, it is used only in one place
<[ARP]hoijui_g5> Kloot suggested making templeated vecX<type> classes
<[ARP]hoijui_g5> hmm.
<[ARP]hoijui_g5> there seems to be little enhusiasm on this topic :D
<[ARP]hoijui_g5> hehe
<zerver> lol
<[LCC]jK> lol
<[ARP]hoijui_g5> sucker!
<zerver> i was reading the comments...
<Kloot> SL crashed
<[ARP]hoijui_g5> ok ;-)
<[LCC]jK> btw there is the problem with unions when adding a template based vector class
<[ARP]hoijui_g5> mm
<zerver> yeah, unions only allow simple types right?
<[ARP]hoijui_g5> i think he means, you can not inherit form a class
<[ARP]hoijui_g5> oh.. is the same :D
<zerver> yeah
<[ARP]hoijui_g5> coudl we somehow get the same functionality liek wiht unions?
<[LCC]jK> btw I always wanted to include http://glm.g-truc.net/
<[LCC]jK> it's a glsl-like mathlib for c++
<zerver> union is a base language construct, hard to replace with something higher level
<[ARP]hoijui_g5> but it is executed on CPU, right?
<[LCC]jK> so you can do things like vec4(vec3(0.0),1.0).bgra
<[LCC]jK> yeah, it runs on the CPU
<zerver> good, otherwise it may desync :P
<[ARP]hoijui_g5> yeah... how woudl that work wiht sync/sync-checking?
<[ARP]hoijui_g5> would we but that on top of it, where needed?
<[LCC]jK> from the tutorials I sa, OpenCL might be sync safe
<[LCC]jK> *saw
<[LCC]jK> but I didn't tested that yet
<[ARP]hoijui_g5> ok so.. the question is:
<[ARP]hoijui_g5> * union
<[ARP]hoijui_g5> * vecx<type>
<[ARP]hoijui_g5> * GML (if tests show its sync-safe)
<[ARP]hoijui_g5> yeah?
<zerver> GLM
<[ARP]hoijui_g5> :D
<[ARP]hoijui_g5> can a class inherit from a union?
<zerver> no
<zerver> or?
<[ARP]hoijui_g5> can a union be templated? :D
<[ARP]hoijui_g5> mm i guess that is hte same again
<[LCC]jK> inheriting and templates aren't the same
<Kloot> it isn't that big of a problem in this case
<Kloot> a union would be needed to represent a color as a uint32 or a vec4<char>
<Kloot> which would have the same size
<[LCC]jK> the best would be if float4 would allow rgba & xyzw indexing
<Kloot> or possibly some more specialized type like colorRGBA
<zerver> or a function that returns the type, GetAsVec4()
<zerver> it may bloat the code
<Kloot> yeah swizzling would be nicest
<zerver> question is if vecx actually improves code readability
<[LCC]jK> problem is that vec2 & vec4 are nearly not implemented atm
<Kloot> it would be easy to typedef some common instances like vec3<float>
<Kloot> as vec3f
<[LCC]jK> and there is/was no global used colorprimitive
<zerver> yeah, the code would otherwise become bigger and have more <> in it
<zerver> which is not very readable
<zerver> if anyone enjoys doing it, i have no objections
<[ARP]hoijui_g5> the reason to do it, is to have a common base, where to put util/helper functions, like .mult(x)?
<Kloot> eh, those helper functions would just be part of vecX :)
<[ARP]hoijui_g5> yeah thats what i mean
<[ARP]hoijui_g5> woudl not have to duplicate in color4 and float4, for example
<Kloot> indeed
<zerver> if u gonna tamper with floatX better merge all branches first :)
<Kloot> plan was to replace int2/float2/float4 first since those are not used much at all
<[ARP]hoijui_g5> yeah.. as said, flaot3 wont be touched of course (now), except we can make the change happen only in float3.h/.cpp
<[ARP]hoijui_g5> mm
<zerver> ok
<[RoX]Tobi> [ARP]hoijui_g5: do you have a 0.82-branch stacktrace nearby to test?
<[RoX]Tobi> [ARP]hoijui_g5: you can try it yourself if you want, it *should* work now
<[ARP]hoijui_g5> yes
<[ARP]hoijui_g5> ok..
<[ARP]hoijui_g5> http://springrts.com/mantis/view.php?id=2219
<[ARP]hoijui_g5> there is one in there.. i will try
<[RoX]Tobi> thanks
<[ARP]hoijui_g5> yyep, works! :-)
<[ARP]hoijui_g5> nice one! :-)
<zerver> quick and dirty fixes are always the best
<[ARP]hoijui_g5> :-)
<zerver> if no one wants to comment, maybe proceed?
<[ARP]hoijui_g5> http://sl.pastebin.com/YFw3BcS9
<[ARP]hoijui_g5> tst.cpp
<[ARP]hoijui_g5> it compiles
<[ARP]hoijui_g5> is that of any use?
<[ARP]hoijui_g5> not nice to have rgba in vec4 of course :D
<[LCC]jK> problem is that it isn't modular anymore
<[ARP]hoijui_g5> mm :/
<[LCC]jK> you need to implement such a template for each single class
<[ARP]hoijui_g5> use GetR() GetG() ... functions in ColorRGBA : vec4<unsinged char> ?
<[ARP]hoijui_g5> hmm.. should we discuss this an other time?
<[ARP]hoijui_g5> looks like it woudl use too much time
<[RoX]Tobi> TBH I wouldn't do all too major rewrites since it's pretty bug free code at the moment afaik
<[ARP]hoijui_g5> and we still have other stuff
<[LCC]jK> I don't see that anyone wants to implement anything of these
<[ARP]hoijui_g5> yeah... rewrite is not hte main concern
<[LCC]jK> but moving the files in a separate dir is still valid
<[ARP]hoijui_g5> but to have the new Color thing be implemented as it should, before it is widely used
<[RoX]Tobi> yeah
<[ARP]hoijui_g5> ok.. in additiona to that, what shoudl we do wiht Color.h?
<[ARP]hoijui_g5> just not use anywhere?
<Kloot> +1
<[ARP]hoijui_g5> or make it a class, based ona a new vec4<type> class?
<[ARP]hoijui_g5> ok...
<[ARP]hoijui_g5> lets postphone it then
<[LCC]jK> there is no alternative for CColor atm :
<[LCC]jK> *:x
<[ARP]hoijui_g5> you mean SColor?
<[LCC]jK> erm SColor
<[ARP]hoijui_g5> still dont know why its S though its a union
<[ARP]hoijui_g5> casue it has a stucts? can be used as one?
<[LCC]jK> unions are structures afaik
<[ARP]hoijui_g5> ah
<[LCC]jK> but the diff between structure and class is just the default public/private thing ;)

Conclusions:
OK to make a template class for Color etc, but not to be used for refactoring, since the current code is fairly bug free



=== rename master -> develop, make master the release branch? ===
<[ARP]hoijui_g5> inspired by http://nvie.com/posts/a-successful-git-branching-model/
<[ARP]hoijui_g5> would mean. we always push to develop branch
<[ARP]hoijui_g5> and master branch is the release branch
<[ARP]hoijui_g5> they also use more branches there.. which i think woudl not make sense for us
<[ARP]hoijui_g5> it is not a big change really
<abma_irc> hm... what does it improve?
<[ARP]hoijui_g5> main outcomes are: a user doing a simple git clone woudl get the release sources instead of current dev
<[ARP]hoijui_g5> the release branch alwyas has the same name
<[ARP]hoijui_g5> (master)
<zerver> more noob friendly?
<[ARP]hoijui_g5> which simplifies some scripts, and makes wiki maintenence simpler
<[RoX]Tobi> version numbers as we have them now would be non-existent though in develop
<[ARP]hoijui_g5> yeah.. i woudl say so
<[RoX]Tobi> since master is never merged to develop as far as I see
<abma_irc> that sounds better then just a rename :-)
<[ARP]hoijui_g5> yes true
<[RoX]Tobi> other than that it sounds ok
<[RoX]Tobi> brb
<[ARP]hoijui_g5> woudl it have consequences to not have versions in develop?
<[ARP]hoijui_g5> looking at a git-describe in develop would not tell you much
<[ARP]hoijui_g5> though we seldomly use these for old commits
<[ARP]hoijui_g5> i came across this site by accident.. maybe there are ways that would fit us evne better... i did not search for them
<zerver> it should be fine i think
<zerver> as long as no one has to sit and determine what to merge into master
<[ARP]hoijui_g5> yeah.. someone has to do that, but that is the smae liek now
<zerver> not really
<[ARP]hoijui_g5> now its cherry picking "master -> 0.82-branch"
<zerver> the release branch we have now shall be deleted i assume
<[ARP]hoijui_g5> then it will be "develop -> master"
<zerver> once we move to 0.83
<[ARP]hoijui_g5> yes.. we woudl not have them anymore
<zerver> 0.83 = master
<[ARP]hoijui_g5> yep
<[ARP]hoijui_g5> all version tags would be on one branch, whihc is not the case now
<zerver> and there is no problem with merging the whole develop into master if you already cherry picked some stuff?
<[ARP]hoijui_g5> an alternative to the "develop & master" approach, is "master & release"
<[ARP]hoijui_g5> there coudl be merge conflicts
<[ARP]hoijui_g5> but thye are easy to solve
<[LCC]jK> using master for releases would be more noob safe
<[ARP]hoijui_g5> yeah
<zerver> i already pointed that out :)
<[ARP]hoijui_g5> :D
<[LCC]jK> we got a lot less-experienced linux users who try to compile spring themselves and always compile the development branch and wonder why things don't work
<[ARP]hoijui_g5> i never want to hear anybody say spring is not noob friendly! we so care for them
<[ARP]hoijui_g5> yeah true
<[ARP]hoijui_g5> though...
<[ARP]hoijui_g5> that would not entirely solve that problem
<Kloot> it is all documented
<Kloot> on the wiki
<[ARP]hoijui_g5> i mean.. the release branch (master) would now contian what we have in 0.82-branch, which does not sync wiht current release
<[LCC]jK> hmmm true
<zerver> that is a prob then
<[ARP]hoijui_g5> if we want the release branch to always sync with current release, we woudl need an extra branch
<zerver> so we have to make a temp branch...
<[ARP]hoijui_g5> like thier hotfix branch
<zerver> 82-branch is resurrected
<[ARP]hoijui_g5> no
<abma_irc> pre-release-branch ;-)
<[ARP]hoijui_g5> does not have to be temp
<[ARP]hoijui_g5> yeah... call it release, pre-release, fixes, ...
<[ARP]hoijui_g5> something
<zerver> pre-release, otherwise noobs may check that out instead
<zerver> so decide on this now or later?
<[ARP]hoijui_g5> i guess when Tobi is back
<[ARP]hoijui_g5> somewhen later in hte meeting
<[ARP]hoijui_g5> next?

Conclusions:
We will make master a release branch, so that noobs always check out a version that is valid for playing on the main server



== making shadow matrix linear (loss of quality in <2048 resolutions, but no distortions anymore, also faster) ==
<abma_irc> what means loss of quality?
<Kloot> aliasing
<Kloot> lots of i
<Kloot> t
<[ARP]hoijui_g5> can we see some screenshots?
<[ARP]hoijui_g5> maybe the game devs should have a say in this?
<[RoX]Tobi> uh I was back already, just didn't see anything I thought I needed to comment on :)
<[ARP]hoijui_g5> ok :-)
<[ARP]hoijui_g5> looks like jk is gone now :D
<[LCC]jK> re
<[LCC]jK> :)
<[LCC]jK> k then the matrix stuff now?
<[ARP]hoijui_g5> ok.. then lets stay with shadow
<[ARP]hoijui_g5> yep
<[LCC]jK> You all may known that the current shadow matrix is non-linear
<[LCC]jK> causing ugly misplaced shadows if the polygons are too large
<abma_irc> aaah, ok
<[LCC]jK> (visible on s44 bunkers etc.)
<[LCC]jK> http://home.arcor.de/jkei/s44_shaders_thumb.jpg
<[ARP]hoijui_g5> ahhhh
<[LCC]jK> so, this would get fixed with a linear shadow matrix and it is even possible w/o breaking backward compability (by changing p17 & p18)
<[LCC]jK> the only negative is that it reduce the quality of lowres shadowmaps
<[ARP]hoijui_g5> the edges of shadows would look pixelated?
<[LCC]jK> yup
<[ARP]hoijui_g5> ok.
<[LCC]jK> extremely blurred
<Kloot> and highres shadowmaps would be slower again
<[LCC]jK> not here
<[ARP]hoijui_g5> what defines the shadow map size? the unit model the map?
<Kloot> especially in shadows 1 mode
<[LCC]jK> for me there was never a difference in FPS for 1024 versus 2048 shadowmaps
<[ARP]hoijui_g5> ah.. its a user setting? spring settings?
<zerver> springsetting
<[ARP]hoijui_g5> ok
<abma_irc> hm... high-quality looks better (no misplaced shadows) and low quality looks more ugly but is faster?
<Kloot> also people would just start complaining about aliasing since it'd be a lot more visible than the nonlinear distortions
<[LCC]jK> also having a linear shadowmatrix would allow to use the normal terrain-renderer with LODing, and so reduce the CPU load
<[ARP]hoijui_g5> well. blurred does nto sound as bad as pixelated to me
<[LCC]jK> and the shaders would even be faster (no sqrt needed anymore)
<[LCC]jK> so in the end it would speedup shadow rendering
<[LCC]jK> it just makes the 1024 shadowmap size .. unusable
<[ARP]hoijui_g5> is it possible to have it as a setting?
<[ARP]hoijui_g5> to choose between the two methods
<[LCC]jK> it is, but then the optimization couldn't be done
<[ARP]hoijui_g5> ah :/
<Kloot> yes it could (#ifdef'ed glsl)
<[ARP]hoijui_g5> i mean.. springsetings
<zerver> Do like PathFinder sharing common interface :)
<[ARP]hoijui_g5> hehe :D
<[LCC]jK> shadow rendering isn't isolated as pathfinder
<[ARP]hoijui_g5> sounds like overkill
<[ARP]hoijui_g5> well.. i cant vote on that
<[ARP]hoijui_g5> maybe do some tests on different hardware?
<[ARP]hoijui_g5> coudl put it into the release branch right after .7
<[ARP]hoijui_g5> or even in a ,7.1 release
<[ARP]hoijui_g5> ti get some test coverage
<[ARP]hoijui_g5> it*
<[ARP]hoijui_g5> or is it too much work to be only possibly accepted?
<[ARP]hoijui_g5> should we just have a vote now?
<[LCC]jK> it's a 4 lines change
<[ARP]hoijui_g5> ah ok :-)
<[LCC]jK> later optimizations are much larger, but switching to a linear matrix is easy
<[ARP]hoijui_g5> so getting some testing before putting it in a require release woudl be good, yeah?
<[ARP]hoijui_g5> ah ok
<zerver> check end user acceptance before optimizing
<abma_irc> then put the 4 lines in the release and get some feedback?
<[ARP]hoijui_g5> anyone agasint that?
<zerver> i dont use shadows :P
<Kloot> I can see the shitstorm already
<[ARP]hoijui_g5> what is hte default shadow map size?
<Kloot> #define DEFAULT_SHADOWMAPSIZE 2048
<[ARP]hoijui_g5> ok...
<[ARP]hoijui_g5> cant see there beeing much problem then
<[ARP]hoijui_g5> this combined with most people not using minor releases...
<[ARP]hoijui_g5> patch-releases*
<Kloot> with a linear projection that number would have to be doubled imo
<Kloot> to satisfy people's "perceptions"
<[ARP]hoijui_g5> mmm
<[ARP]hoijui_g5> so what?
<[ARP]hoijui_g5> jk?
<[ARP]hoijui_g5> is it ok to put it in a patch release?
<[ARP]hoijui_g5> or what would you preffer?

Conclusions:
Implement linear shadow map, wait for end-user input, if OK, proceed with further optimization



=== streflop::sqrt ===
<zerver> yeah
<[LCC]jK> I will commit and Kloot can see
<[ARP]hoijui_g5> ok
<zerver> i remember back long ago someone did some profiling and streflop::sqrt was 25% of cpu
<zerver> now with SSE and stuff, has that changed?
<[ARP]hoijui_g5> i am on the way to get profling going, wiht BDs help
<zerver> in other words, is sqrt still emulated using integer math? does anyone know?
<Kloot> no, that is the fastmath sqrt
<zerver> eh? emulation is not fast
<Kloot> sqrt_nosse to be precise
<zerver> ok, so the performance problem disappeared when SSE was introduced?
<[LCC]jK> yup
<zerver> good
<zerver> next
<[LCC]jK> IIRC SSE is 20x faster than streflop
<zerver> oh sure

Conclusions:
zerver now knows that streflop::sqrt is fast with SSE



== switch wiki to a default style? ==
<[ARP]hoijui_g5> yeah.. i have listed my arguments in the etherpad
<[ARP]hoijui_g5> *
<[ARP]hoijui_g5> * less glitches
<[ARP]hoijui_g5> * ugly colors
<[ARP]hoijui_g5> * background similar to foreground
<[ARP]hoijui_g5> * less manual color-coding required
<[ARP]hoijui_g5> * people would probably realise that it is a wiki
<[ARP]hoijui_g5> * as it is an engine project site (not game), we are no frontend to the user/player -> no need to have a cool/nice style
<[LCC]jK> what do you mean with default style?
<[ARP]hoijui_g5> black on white
<[LCC]jK> the wiki is still implemented on the spring website
<[ARP]hoijui_g5> like wikipedia
<zerver> ugh... my eyes hurt
<zerver> im a white on black person, to 100%
<[ARP]hoijui_g5> i hate how .. for example code stuff works now
<[LCC]jK> green/yellow on blck here :D
<[ARP]hoijui_g5> i preffer light green on dark green
<[ARP]hoijui_g5> but i cant see why we should care for style
<[LCC]jK> the current code-syntax is because of flexibility ...
<[ARP]hoijui_g5> i jsut though.. the defautl style must have everythign ready
<[ARP]hoijui_g5> hm?
<[ARP]hoijui_g5> we have no syntax coloring
<zerver> prefer is spelled with one f
<[ARP]hoijui_g5> nothign automatic at least
<zerver> fyi
<[ARP]hoijui_g5> :D :P
<zerver> ich preffer pfeffer und michelle pfeiffer
<[LCC]jK> syntax coloring ...
<[ARP]hoijui_g5> and when you use <code> or <code><pre> it has different colors
<[ARP]hoijui_g5> it looks different on different browsers
<[LCC]jK> code w/o pre is used for inline code declaration
<[LCC]jK> ppl just don't get it ...
<[LCC]jK> e.g. "open <code>rts/System/float3.h</code> and edit line 30 ..."
<[ARP]hoijui_g5> .. .. so?
<[ARP]hoijui_g5> why shoudl it have a differtn color then the other code?
<[ARP]hoijui_g5> with pre, the code has hte smae color as normal text, just a slightly differnt font (hardly noticeable on my browser
<[ARP]hoijui_g5> but well.. looks like nobody else woudl liek to change, so next...
<[ARP]hoijui_g5> .. lets do the wvote for git branch changing now
<[ARP]hoijui_g5> 1. keep as is
<[LCC]jK> your browser is broken
<zerver> :P
<[ARP]hoijui_g5> 2. master = release branch, develop = what is now master
<[ARP]hoijui_g5> 3. like 2. but wiht an additional pre-release branch (so master woudl always be in sync with current release)
<abma_irc> hmm.. IMHO 2. looks like no real improvement
<zerver> yeah, anything but 2
<zerver> 3 will mean some initial confusion, as always
<[ARP]hoijui_g5> 2 is still an improvement, cause the release branch is always the same, and all version tags are on a single branch, and current release candaidate is alwasy there
<abma_irc> 3. means also to get rid of 0.8X-branches?
<[ARP]hoijui_g5> les maintenence, eg on wiki
<[LCC]jK> http://home.arcor.de/jk3064/wikicodepre.png
<[ARP]hoijui_g5> yes, no temp release branches anymore, neihter wiht 2 nor 3
<[ARP]hoijui_g5> hmm ok jk... thats how it looks in konqueror here too, but not in firefox
<[LCC]jK> firefox is broken then
<[ARP]hoijui_g5> is syntax coloring possible in wiki?
<[ARP]hoijui_g5> mm ok
<[LCC]jK> the css isn't even very special for this
<[LCC]jK> so it is a clear bug
<[LCC]jK> and not a missing ccs3 thingy or something like that
<[ARP]hoijui_g5> mm .. especially that it does not even get the color right
<abma_irc> hmm... the springrts page has many html-syntax errors...
<[ARP]hoijui_g5> ok so.. cna we vote on the git change?
<[ARP]hoijui_g5> +1 on 3
<[ARP]hoijui_g5> or are there more questions?
<abma_irc> [RoX]Tobi: does this break many stuff on the buildbot/translate stuff?
<[RoX]Tobi> good point
<zerver> will there be more merging work?
<[ARP]hoijui_g5> no, not more
<zerver> im ok with 3 then
<[ARP]hoijui_g5> at every mayor release, you will merge develop into master
<zerver> just make sure all commiters know to reduce the confusion
<[RoX]Tobi> I don't really know right now, I guess it will break everything but I think most things should be easy to fix
<[ARP]hoijui_g5> if there are conflicts, you simply replace the whole repoof master with what is in develop, and commit that (can be automated with one or two commands)
<[ARP]hoijui_g5> so what?
<[ARP]hoijui_g5> you want to decide it next time?>
<[ARP]hoijui_g5> jk, kloot?
<[ARP]hoijui_g5> i guess we will have the other change ready before next mayor release too
<[LCC]jK> +1 for a 3 branched system
<[ARP]hoijui_g5> versions beeing only tracked in git tags
<abma_irc> +1
<Kloot> I vote blank on this, don't see clear advantages or disadvantages yet
<[ARP]hoijui_g5> i guess i can do the buildbot related changes too, now that i have access there
<[ARP]hoijui_g5> ok
<[ARP]hoijui_g5> so.. i note it down as accepted.. but will not wokr on it until the AI interface binary version check is done, and will probably also try the version-as-git-tag-only thing before that
<[ARP]hoijui_g5> or say.. it can not be done before the mayor release anyway
<[ARP]hoijui_g5> next thing..
<[ARP]hoijui_g5> Licho asked for a change in the installer
<[ARP]hoijui_g5> as it is now, the exe installed for ZeroK lobby is only an installer/downloader
<[LCC]jK> btw here a comparision between linear & non-linear http://home.arcor.de/jk3064/shadowmatrix.png http://home.arcor.de/jk3064/shadowmatrix2048.png
<[ARP]hoijui_g5> it has a generic installer icon, and when executed for the first time, will show a message like: "this exe is unverified, woudl you really like to proceed?"
<[ARP]hoijui_g5> if oyu click on yes, it will downlaod latest zeroK lobby, and the icon on the desktop changes to the ZeroK one
<[ARP]hoijui_g5> now...
<[ARP]hoijui_g5> Licho wants the installer to execute the exe right when install is finnished/it was copied ot the install dir
<[ARP]hoijui_g5> so the user will never see the ugly instalelr icon on his desktop
<[LCC]jK> add a "run XYZ" checkbox at the end of the installer?
<[ARP]hoijui_g5> what i kind of disklike is, that we execute .. kind of a 3rd party exe, we downloaded from a URL behind the scenes
<[ARP]hoijui_g5> yeah... sounds better
<[ARP]hoijui_g5> is it defautl on or off? :D
<[ARP]hoijui_g5> on i guess
<[ARP]hoijui_g5> Licho siad, he is not 3rd party ;-)
<[ARP]hoijui_g5> and the exe will never change.. we could md5 check it before inclusion
<[ARP]hoijui_g5> so.. checkbox at last screen, default on?
<[ARP]hoijui_g5> any objections?
<[LCC]jK> yup
<[LCC]jK> no objections
<zerver> lol

Conclusions:
Keep current wiki style



== Next meeting ==
<[RoX]Tobi> side note: actually the site is completely broken (as abma said already). until the site validates fine you can't judge browsers based on it
<[RoX]Tobi> same date/time/place is fine for me
<abma_irc> for me it is fine, too
<[LCC]jK> fine
<zerver> same here
<Kloot> same
<zerver> plz test luasplit and tell me what the problems are until next meeting
<[ARP]hoijui_g5> here too
<zerver> you can change the splitting mode via MultiThreadLua in springsettings

Conclusions:
Next meeting same time and day



== Anything else? (WVTTK) ==
<abma_irc> what about an mod, that (nearly) always work with the develop-branch?
<zerver> NullMod
<abma_irc> imho, currently every developer has an .sdd to test stuff
<abma_irc> it should be an full-featured mod, like ba/ca/1944...
<abma_irc> for example, current master + ba doesn't work, because the spawning of units changed
<abma_irc> make an mutator that fixes that problems on github?
<abma_irc> this would also allow to generate an real .zip from this directory for other users, to test current master for example
<abma_irc> ehm, develop-branch :)
<abma_irc> just an idea...
<[ARP]hoijui_g5> yeah.. i am sure nobody here has somethign agasint that
<Kloot> will you be maintaining this mutator on every engine change? :
<[ARP]hoijui_g5> just have to ask the author of the mod you want to use for it
<Kloot> engine/lua*
<[ARP]hoijui_g5> woudl be good if the dev that makes the change in the engine, that requires a change in the mods also makes that in the mutator
<abma_irc> best would be, if the author of the mod would maintain it
<[ARP]hoijui_g5> we would also have a perfect reference then
<abma_irc> ...or so, that would be better
<abma_irc> what mod do you prefer?
<abma_irc> for such stuff
<abma_irc> imho ca is the mod that uses most features
<[RoX]Tobi> it seems to me what would work best is a kind of repo (not necessarily in the VCS kind of way) where we can easily share any fixed mod
<[RoX]Tobi> (as opposed to picking a single mod)
<[RoX]Tobi> also, not maintained, but only updated when the need arises
<[RoX]Tobi> for example: I try to fix bug X in Spring that occurs with game Y, but first I need to patch game Y to run with master. Then I share somehow this patched game Y with other devs so that if they want to fix a bug in game Y too they don't need to patch it themselves.
<abma_irc> is a mutator enough to fix all problems in a mod?
<[RoX]Tobi> pretty much yeah
<[RoX]Tobi> only not if a file may not exist for a game to work
<[RoX]Tobi> as you can't "delete" files using a mutator
<abma_irc> ok...
<[RoX]Tobi> but don't listen too much to my opinion as I probably won't have any significant time to work on Spring in the near future..
<[RoX]Tobi> three words that come up to me when thinking about a solution to this problem are: shared dropbox account ;-)
<abma_irc> my personal problem in this case is, that i've an very slow upstream...
<abma_irc> [LCC]jK: can't say if that would be an problem for other users
<abma_irc> for me the screenshots look fine, but i play always with low settings...
<[RoX]Tobi> from what I've heard dropbox is pretty good at transferring only what is really needed, but I have not tried it myself
<[RoX]Tobi> I'm off, laters
<abma_irc> good night! :)
<[ARP]hoijui_g5> with Tobis method.. that woudl mean...
<[ARP]hoijui_g5> if you change somethign that needs changes in mods
<[ARP]hoijui_g5> all mod devs come yelling at you to fix it for thier mod
<[ARP]hoijui_g5> and ... well i woudl not want that ;-)
<abma_irc> i don't want to do that, too
<abma_irc> its for testing, not for fixing
<[ARP]hoijui_g5> yeah..
<[ARP]hoijui_g5> i think it woudl be good to stick to a single mod
0 x

User avatar
FLOZi
MC: Legacy & Spring 1944 Developer
Posts: 6109
Joined: 29 Apr 2005, 01:14

Re: Dev meeting minutes 2010-11-15

Post by FLOZi » 16 Nov 2010, 13:36

Minor point wrt using CA for testing; lua that works correctly with CA gadgetHandler does not always play nicely with the base content gadgerHandler.

I look forward to testing out the shadowmap changes.
0 x

User avatar
Licho
Zero-K Developer
Posts: 3803
Joined: 19 May 2006, 19:13

Re: Dev meeting minutes 2010-11-15

Post by Licho » 16 Nov 2010, 13:39

Note that normal game installers install other stuff often without asking, usually directx, openal or whatever they need.
0 x

User avatar
hoijui
Former Engine Dev
Posts: 4342
Joined: 22 Sep 2007, 09:51

Re: Dev meeting minutes 2010-11-15

Post by hoijui » 16 Nov 2010, 14:10

What is Normal? (from a guy that was in a mental health institution)
What "Normal People" Want from TV
What's Normal? (from Psychologists)

for the lulz

Is It Normal To Be Late On Your Period?

Town of Normal, Illinois - "Committed to service Excellence"

Michael Jackson was a normal guy
facebook - Being Normal wrote:Being Normal is now available from http://www.amazon.co.uk at £5.99 - that's a 25% saving. Alternatively, it can be purchased from ...
http://www.facebook.com/group.php?gid=18507937175
Image
0 x

User avatar
Licho
Zero-K Developer
Posts: 3803
Joined: 19 May 2006, 19:13

Re: Dev meeting minutes 2010-11-15

Post by Licho » 16 Nov 2010, 15:59

We made lobby that is simple to use and targeted to nubs. With purpose to increase spring players. As such it should install itself easily and without unclear questions.

Now what is purpose of adding extra checkbox step?
0 x

User avatar
zwzsg
Kernel Panic Co-Developer
Posts: 7017
Joined: 16 Nov 2004, 13:08

Re: Dev meeting minutes 2010-11-15

Post by zwzsg » 16 Nov 2010, 16:44

It would let met install Spring without installing Zero-K lobby! :-)

I don't like installing Zero-K lobby because I feels it puts so much mess on my computer, dropping files everywhere, in so many places it feels dirty, the kind of app that when you try to remove it you'll never be sure it didn't left files lingering in some place you'll never check.

Before uninstalling Spring was a matter of running the installer, then checking the Spring folder had disappeared and deleting whatever remained in it if done.
Plus some regkeys I seldom cared much about. Spring regkeys got removed and replaced with a .cfg files, TASClient regkeys remains to this day.
Then some people thought it would be great to separate program and data, and from then on I also had to check and delete some springsettings in my documents and settings. But it was just one extra folder, with one extra file (or two, depending if springlobby had run).
But now with Zero-K, there's the Net framework, its installer, Zero-k, its installer, Zero-k autoupdater and autodownlader and their half completed files, so it ends up with settings and data and installer and temp files stored all over the places!

I wish Zero-K lobby would be a simpler self contained application, a couple files all in the same folder, instead of a tentacular windozy takeover. But I suppose now is too late, you won't reprogram in the raw now that you got used to your comfy .NET and so.
0 x

User avatar
Licho
Zero-K Developer
Posts: 3803
Joined: 19 May 2006, 19:13

Re: Dev meeting minutes 2010-11-15

Post by Licho » 16 Nov 2010, 17:23

What a BS zwzsg - there is already checkbox in installer!

It wont install if its unchecked!

Talk is about extra second checkbox.

(And if you uninstall it there wont be program left, it deletes self and temp stuff except stuff shared by other programs - maps and mods. Whole point of clickonce system is to have it self contained and not pollute your system)
0 x

User avatar
SinbadEV
Posts: 6475
Joined: 02 May 2005, 03:56

Re: Dev meeting minutes 2010-11-15

Post by SinbadEV » 16 Nov 2010, 17:28

Licho wrote:Waht a BS zwzsg - there is already checkbox in installer!

It wont install if its unchecked!

Talk is about extra second checkbox.
You are referring to an extra second check-box for the .Net framework or whatever right? Yeah, I don't hink ZWZSG got that and so his comments are slightly out of place...

Personal Opinion... maybe you could make the Z-Lobby checkbox description include "requires .Net X.X Framework, installation of this package will install or update your computer to the required version of the framework"

In general though, I've seen many installers that make required components optional in the installer... it's annoying when this happens.
0 x

User avatar
zwzsg
Kernel Panic Co-Developer
Posts: 7017
Joined: 16 Nov 2004, 13:08

Re: Dev meeting minutes 2010-11-15

Post by zwzsg » 16 Nov 2010, 17:30

there is already checkbox in installer!
It wont install if its unchecked!
Talk is about extra second checkbox.
Ah good! Sorry I misunderstood then.

And if you uninstall it there wont be program left, it deletes self and temp stuff except stuff shared by other programs - maps and mods. Whole point of clickonce system is to have it self contained and not pollute your system)
U shure? wasn't so last time I tried (maybe a month or two ago).
0 x

User avatar
Licho
Zero-K Developer
Posts: 3803
Joined: 19 May 2006, 19:13

Re: Dev meeting minutes 2010-11-15

Post by Licho » 16 Nov 2010, 18:09

Not even about second Framework checkbox, I'm talking about THIRD checkbox that Hoi wants to add :)

It's not there yet..
0 x

User avatar
Beherith
Moderator
Posts: 4934
Joined: 26 Oct 2007, 16:21

Re: Dev meeting minutes 2010-11-15

Post by Beherith » 16 Nov 2010, 20:57

linear transform for shadowmaps makes me happy :)
0 x

luckywaldo7
Posts: 1397
Joined: 17 Sep 2008, 04:36

Re: Dev meeting minutes 2010-11-15

Post by luckywaldo7 » 17 Nov 2010, 01:02

Ditto, I'm so glad work is being done on shadows.

Not to sound like the other work being done isn't important. Good work all around :)
0 x

Post Reply

Return to “Meeting Minutes”

cron