Dev meeting minutes Monday, 12. September, 2011

Dev meeting minutes Monday, 12. September, 2011

Minutes of the meetings between Spring developers are archived here.
Post Reply
Tobi
Spring Developer
Posts: 4598
Joined: 01 Jun 2005, 11:36

Dev meeting minutes Monday, 12. September, 2011

Post by Tobi »

Date: 12-9-2011
Present: _koshi_, hoijui, abma, zerver, jK, Tobi

Misc.
<hoijui> so...
<hoijui> :D
<zerver> yawn
<zerver> i think we need a release to get some action in here
<hoijui> :D
<hoijui> Tobi, can you probably suggest how to best what abma suggests in his comment here?
<hoijui> http://springrts.com/mantis/view.php?id=2642
<hoijui> the problem with AIs and bandwidth limit has ot be done before a release, definitely
<zerver> I don't understand the design choice to make AI:s different from real players
<zerver> I would never have taken that route if I implemented it
<hoijui> i guess it was just easier that way
<zerver> maybe
<hoijui> woooot!
<hoijui> jk is in hte us
<jK> ?
<zerver> some proxy maybe
<jK> need to use ircbridge (not at home)
<hoijui> ah k
<hoijui> ddint know we have such one in the US
<jK> possibly it detects the ip of the irc webclient
<hoijui> what else..
<hoijui> ahhh
<hoijui> we coudl now do the change to the develop branch
<hoijui> as Tobi is back
<hoijui> we said it may make sense to wait for him, in case something in the buildbot needs a root present
<hoijui> means, we would fix a date (at best, right now i guess), where we start to use develop, isntead of master
<hoijui> and only one of us will touch master after that
<hoijui> hmm..
<hoijui> buildbot should probably build both master and devellop from now on
<hoijui> when we do the release, we will set master HEAD to develop HEAD a last tiem, and form then on, they will be separate
<hoijui> so, we add a buildbot stuff for branch develop (equal to what we have for master)
<hoijui> when that is done, we can all switch to develop
<jK> [RoX]Tobi
<Tobi> hmm ok
<hoijui> changeing name locally, when you need to do it (you shoudl be on master), dont do yet: git checkout -b develop master
<Tobi> so when is the idea we will switch?
<hoijui> though you may have to set the tracking branch then
<hoijui> when the buildbot is setup to build branch develop
<Tobi> ok
<hoijui> and .. that could be doen now.. or.. tomorow i guess
<hoijui> can it all be done in the spring repo?
<hoijui> or cna some stuff onyl be done by you?
<Tobi> master would need a restart
<hoijui> ok
<Tobi> abma and you can do that too iirc
<Tobi> but I guess I know most about buildbot
<hoijui> ahh k :-)
<hoijui> yeah
<Tobi> I can do it on wednesday, am occupied tomorrow and now its a bit late to start on something imho
<hoijui> ook, thanks :-)
<Tobi> I'll probably push some test crap to develop then at first, to test buildbot setup, before finishing with force push that makes it equal to master again
<hoijui> ook :-)
<hoijui> so nobody use develop until tobi says so
<Tobi> anyway I can make a post on the forum when buildbot builds it
<hoijui> perfect
<Tobi> kk
<zerver> ok

conclusions:
  • Tobi will make buildbot build develop branch
  • when this is done we switch to it
include http://code.google.com/p/bitwise-enum/
<jK> he couldn't relicense it (duno why), still he allowed us to use his version as base for our own implementation
<hoijui> mmm ok...
<hoijui> but is that .. possible?
<hoijui> i mena, if he .. disapears
<hoijui> and someone says spring uses that code
<hoijui> which it does..
<hoijui> then what?
<jK> sure, you are always allowed to write your own implementation, as long as you don't copy paste (or violate any IP)
<hoijui> mm ok
<hoijui> so yo will not copy paste anything?
<hoijui> change all var names.. stuff like that?
<hoijui> is it very small?
<jK> he said I should take his code and rename the variables, but I will rewrite it totally (also needed to fix a bug in it)
<hoijui> ok
<hoijui> :-)
<jK> it's just a very nice source to see how to do it and I will still credit him
<hoijui> ok. good :-)

conclusions:
  • he can't relicense, but he allows us to use his version as a reference for our own implementation
extend coding standards? http://springrts.com/wiki/Coding_Standards
<hoijui> thats stil in from last time?
<hoijui> guess does not need to be here
<hoijui> or did we agree to reevaluate it now?
<hoijui> i didnt do anything
<jK> no it was moved to this meeting

conclusions:
  • no progress
use a git-hook for checking code-style? for example https://github.com/abma/pr-downloader/b ... pre-commit
<Tobi> can that even be done on github?
<hoijui> yeah exactly Tobi, that is also what i odnt know..
<hoijui> an other project i work in has such hooks, but they setup thier own git server
<hoijui> i tried to find out once... but i ma still not sure
<hoijui> github does not mention this kind of hooks
<hoijui> i guess they could be abused
<Tobi> I don't think it is possible, I've never seen such a thing in the admin interface
<Tobi> yeah
<hoijui> mmm yeah i guess so too
<Tobi> oh, ppl could install it locally in their repo ofc
<Tobi> if it is a pre-commit hook that is where it should be
<Tobi> was confusing it with pre-receive or so
<hoijui> ahhh..
<Tobi> nothing that can be 100% enforced though
<hoijui> yeah
<hoijui> that shure could be done, and is a good idea i think
<Tobi> better first have good coding standards and only then when enforcing using wooden stick doesn't work think about this :P
<hoijui> :D
<hoijui> one sentence got into the other?
<Tobi> too much subsentences and lack of commas, I guess
<hoijui> :D ok
<hoijui> yeah well..
<hoijui> in the other project, the hook-enforced standards are pretty low
<hoijui> i dont even remember...
<hoijui> something like.. very basic white-space ugliness
<hoijui> and it makes sure you don't submit files of certian types (eg microsoft word)
<hoijui> or certain binaries
<hoijui> the thing is, if you make strict rules with hooks, then the whole repo has to conform to these rules, or else you can not commit anymore
<hoijui> thats with the hook on hte receiving end, at least
<hoijui> so if you do a clean commit on an unclean repo, and want to push, it fails
<Tobi> sure?
<Tobi> might depend on the code in the hook though
<hoijui> yeah possible
<hoijui> but it is either the stadnard, or not preventable even
<hoijui> i forgot why it is so
<hoijui> it also makes sense kind of...
<hoijui> if oyu think about it..
<hoijui> eve nif all uf us used the hooks locally
<hoijui> contributors woudl not
<Tobi> in any case I don't really believe in technically enforcing such things, I'd rather see good standard + comments on github when stuff doesn't adhere to it + quick improvement of ppl so those comments are seldomly needed after a few weeks/months
<Tobi> yeah true
<hoijui> aif we used githubs pull request stuff
<hoijui> yeah, that woudl be better
<Tobi> if we actually have a way to check whether code adheres to 'our standard', then we could run it on buildbot too
<hoijui> btw zerver, you still do "type *name" sometimes
<Tobi> i.e., build fails if code doesn't adhere to standard
<hoijui> ahh yeah
<hoijui> true..
<Tobi> or continues but is still marked red
<hoijui> good idea
<hoijui> next?
<Tobi> sure, if there is anything left

conclusions:
  • not possible on github
  • pre-commit hook for every dev could work, but isn't strictly enforcing (in particular w.r.t. merging from outside contributors...)
  • possibly best option: run a style check on buildbot
try etherpad lite? http://etherpad.org/2011/08/22/major-re ... d-lite-v1/
<jK> added by abma
<Tobi> he told me that in private too, I've looked at it quickly and it seems nice
<jK> I am happy with current etherpad, except memleaks all few months
<Tobi> seems much more lightweight than normal etherpad (not surprising since it's nodejs instead of rhino on jvm), and has approx. same feature set
<Tobi> IMHO we migrate once I figured out how to migrate the database
<Tobi> or rather, once I figured that out I can make a test instance with our content so we can try
<hoijui> nice! :-)
<jK> don't forget that we aren't the only ones that use the current etherpad
<Tobi> kk, will do public test then
  • etherpad lite is an etherpad clone fully in javascript, with nodejs on the server
  • Tobi tried this, looks promising: much less resources needed than etherpad, and identical (or at least similar) feature set
  • Tobi will continue with a migration to this as time allows
[teh]Beherith[PiRO] has some questions to kloot related to roam + unsynced heightmap (and maybe others)
[00:12] ** abma_irc joined the channel.
<abma> hi! :-/
<jK> hi
<abma> this was added by me, Beherith asked me when is meeting and if he maybe can take part
<abma> it looks like he is offline, so skip this
<jK> then we are finished
<abma> kloot: maybe you can help in this thread: http://springrts.com/phpbb/viewtopic.php?f=12&t=26776
<jK> help? the thread contains nothing
<abma> Beherith asks for a summary how the unsynced heigtmap is implemented in .83
<hoijui> koshi wanted to ask us something
<jK> that's like asking how sun works
<jK> he needs to be more precious

python wrapper for unitsync
<hoijui> koshi made(makes) a python wrapper for unitsync
<hoijui> and wants to know if/how to integrate it into the spring repo
<hoijui> most importantly, he needs to use a wrapper tool (forgot the name), whihc he said is not packaged by the linuxes
<abma> jk: yes, true. i think he doesn't what to ask for... so its more a incremental thread... more specific question more specific answers
<hoijui> so it probably would have to be in rts/lib or in a subdir of his wrapper (whihc would make sense if it would be a submodule
<abma> pybindgen?
<hoijui> yep, that coudl be it
<hoijui> ill invite koshi
[00:20] ** _koshi_ joined the channel.
<_koshi_> hi
<hoijui> hey
<hoijui> i just bascially explained that you did a unitsync python wrapper
<hoijui> and that it requries pybindgen
<hoijui> which is not packaged in linuxes
<_koshi_> debian
<_koshi_> dunno about redhat
<_koshi_> or source distros
<jK> hoij, why should such a wrapper be in rts/lib/ ?
<hoijui> hmm yeha i guess it shoudl not
<_koshi_> sidenote: wrapper is somewhat of a misnomer. this actually compiles a shared object file that is directly loadable as a python module
<_koshi_> and that object file does not depend on libunitsync shared object
<Tobi> what is the advantage compared to a python module with ctypes?
<_koshi_> aside from the automation, this could be versioned for diff unitsync versions in the future
<_koshi_> with ctype you'd have to handle loading diff unitsync so yourself
<jK> how does that matter
<jK> when we have multiple engine version support we need a diff interface no matter what
<jK> and even, with your module you still need to load different module for different version
<jK> so you just replace lib with module
<_koshi_> no
<_koshi_> you'd be needing multiple unitsync versions and wrappers with ctypes
<_koshi_> with the self contained lib/module you only need that

conclusions:
  • koshi made python bindings for unitsync
  • con: requires pybindgen, which isn't packaged in any major Linux distros
  • advantages are a bit unclear
  • no decisions made
0 x

User avatar
koshi
Lobby Developer
Posts: 1058
Joined: 14 Aug 2007, 16:15

Re: Dev meeting minutes Monday, 12. September, 2011

Post by koshi »

a couple points re the python unitsync bindings:
  • Atm no wrapper is shipped with spring
  • The ctypes wrapper tobi alluded to can be found in rapid, iirc it's originally by Hugh: A generator that parses unitsync_api.h and outputs the actual ctypes based wrapper. This you can import and need point to a fitting unitsync lib to load at runtime.
  • The pypindgen based generator generates C code (runs relevant usync files thru gccxml) that gets compiled into a standalone shared object file that is directly loadable in python.
  • I wrote tests for that.
0 x

Tobi
Spring Developer
Posts: 4598
Joined: 01 Jun 2005, 11:36

Re: Dev meeting minutes Monday, 12. September, 2011

Post by Tobi »

Buildbot builds the develop branch now.

The schedulers that run after every commit build both master and develop; the nightlies build develop only.
0 x

zerver
Spring Developer
Posts: 1358
Joined: 16 Dec 2006, 20:59

Re: Dev meeting minutes Monday, 12. September, 2011

Post by zerver »

<hoijui> btw zerver, you still do "type *name" sometimes
Yup, I did learn this style at university. And tbh I tend to prefer it, but only very slightly. It takes time to change.
0 x

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

Re: Dev meeting minutes Monday, 12. September, 2011

Post by hoijui »

zerver:
What about the ouh so almighty VisualStudio?
you want to tell me that the best IDE ever created by mankind can't help you there?
VS, the only reason for any dev to use windows at all?
or do they worsen C++ features, so devs are more likely to use C#?
(btw, i am not making fun of you here)


thanks Tobi, nice! :-)

so that means we should start using develop now...
ahh am too tired to figure out perfect commands right now.
0 x

User avatar
aegis
Posts: 2456
Joined: 11 Jul 2007, 17:47

Re: Dev meeting minutes Monday, 12. September, 2011

Post by aegis »

koshi wrote:ctypes wrapper tobi alluded to can be found in rapid, iirc it's originally by Hugh
not by Hugh.

possibly first incarnation of the code: http://springrts.com/phpbb/viewtopic.ph ... 21#p319821
0 x

Post Reply

Return to “Meeting Minutes”