The source of the problem. :-)

unpossible · Post by **unpossible** » 02 Jul 2006, 22:58

i had no idea that there had been so many AIs, or that SAI and QAI had been abandoned

AF · Post by AF » 03 Jul 2006, 00:52

SAI was never patched to fix a few bugs nor was it under active development. When the version interface changed it wasnt recompiled. QAI was a one week wonder. Q-Man could have ahd a very powerful AI that would have put up a very good fight if he'd stayed a week longer.

Post by **Tim Blokdijk** » 07 Jul 2006, 01:01

I have been busy with some documentation concerning AI developer svn commit access.
The first draft has been posted on the mailinglist but you can find the document here: PDF
I'm interested in feedback from AI developers and other people that can give useful comments.

Please note that it's my intention to use this document as a basis for other types of developers to.

malric · Post by **malric** » 07 Jul 2006, 07:04

I think it would be great to have ai-s in svn. As I am running in linux, there are ocasional problems (like crashes) and sometimes I have the time to track them down... But if the sources are not the latest it would feel as a waste of time as maybe these things are already fixed.

Also it would make packaging for different distros easiear (I mean getting all the things from one place).

(I asked yesterday about ntai sources in ntai thread, but from this thread I understand AF doesn't commit directly into svn....)

submarine · Post by **submarine** » 07 Jul 2006, 15:24

Tim Blokdijk wrote: I'm interested in feedback from AI developers and other people that can give useful comments.

asuming its the same pdf file as the one attached to your mail i read it and it sounds very intersting to me.

perhaps i'll "reopen" source after aai 0.7 again. i didnt have the to read all that svn related stuff but i guess its not that hard to upload things

Post by **Tim Blokdijk** » 10 Jul 2006, 17:39

Well, I would fully support your open AI.

But other then that, do you have any other comments on it?
Krogothe, I understand you keep things closed but do you have any feedback?

Updated the document with AF's feedback PDF Version.

krogothe · Post by **krogothe** » 10 Jul 2006, 19:50

I dont really see how this would benefit my development. Sure, access to the interface would help, but wouldnt that break the other AIs?
I think its sad so many people start and stop developing their AIs, but thats because of motivation, It seems to me even sub has been losing motivation (not surprising since debugging is impossible atm). Anyways when im back home ill try and reply properly to this.

AF · Post by AF » 10 Jul 2006, 19:55

submarine got his debugging working again.

And no it isnt access to the itnerface just the option to commit source code to the AI folder and maitnain it in the berliOS svn.

As far as I am aware KAI already has a private SVN for internal development

Post by **jcnossen** » 10 Jul 2006, 19:57

It does save you from having to re-release everytime someone adds a bit in the UnitDef or anything else that breaks C++ ABI for AIs (which is almost every release).

Post by **Tim Blokdijk** » 11 Jul 2006, 00:47

As AF pointed out you can't commit to the interface as an AI developer. But if it would make sense to have commit access to it as an AI dev I would like to know. The procedure now (more or less) is to write a patch that has to be accepted by the (engine/AI) developers. (where AI developers is AF atm)
The benefit to development for you would be that discussion with other developers can be more effective. (as they see the source involved)

If your not working with the other developers in the Spring svn you more or less cut yourself from the overall development process. This "overall development process" (engine, mod, AI, lobby etc. developers working together) is not really here at this moment.
But I do hope to change that in the coming months.

Anyway, take your time as I presume the possibility to join in will be around for quite some time to.

krogothe · Post by **krogothe** » 12 Jul 2006, 19:23

jcnossen wrote:It does save you from having to re-release everytime someone adds a bit in the UnitDef or anything else that breaks C++ ABI for AIs (which is almost every release).

Id still have to release at every spring release AND any other releases i might want to make between releases.

And if AF is right and no changes can be made to the interface, then joining the spring SNV amounts to utter uselessness and waste of time, releasing after an interface changes takes minutes at worse, especially if youre prepared by simply checking the SVN now and then.
And yes, I have svn already, probably over 300 revisions by now, we work faster than spring itself

full interface write access rights are a must for even considering sticking KAI up on SVN and then theres a lot of thinking to be done...

Forboding Angel · Post by **Forboding Angel** » 12 Jul 2006, 19:32

Umm may I interject here?

Maybe AI makers could benefit from having a seperate forum (Just like the mappers forum), to where only AI devs (and maybe spring devs) could post.

THat way you could have discussions, fights, resolutions, etc without the rest of the community giving their 2 cents.

Regardless of what any of you may think, the mappers only forum has helped a TON. It has a way of uniting people with the same cause.

Anyway, sorry for the break in discussion.

AF · Post by AF » 12 Jul 2006, 19:42

And I would pull NTai out the svn if you started messing with the interface at will. You can already change it yourself by submitting patches, but that sort of access would let you change the interface with no notice or discussion with other AI devs.

And the idea of new NTai releases co-inciding with Spring releases is extremely bad as Jelmer is notorious for giving Zero delay, which is a huge annoyance.

I did think SVN access would be alright as it'd allow linux users to have the most upto date NTai possible, but I'm not happy with windows users getting an NTai copy I'm not willing to give support to. Afterall I can just fix the issue and tell a linux user to recompile from svn, I cant do that with a windows user without relesing hundreds of tiny updates, people just wouldnt be bothered and I'd be forever telling them to upgrade.

Post by **Tim Blokdijk** » 12 Jul 2006, 20:50

AF wrote:...
And the idea of new NTai releases co-inciding with Spring releases is extremely bad as Jelmer is notorious for giving Zero delay, which is a huge annoyance.

I did think SVN access would be alright as it'd allow linux users to have the most upto date NTai possible, but I'm not happy with windows users getting an NTai copy I'm not willing to give support to. Afterall I can just fix the issue and tell a linux user to recompile from svn, I cant do that with a windows user without relesing hundreds of tiny updates, people just wouldnt be bothered and I'd be forever telling them to upgrade.

The problem you describe can also be fixed with some update system for Spring components.
But as long as it's not availeble you would like releases without NTai?

And thanks for supporting us Linux users

Post by **Tim Blokdijk** » 12 Jul 2006, 21:56

krogothe wrote:...

And if AF is right and no changes can be made to the interface, then joining the spring SNV amounts to utter uselessness and waste of time, releasing after an interface changes takes minutes at worse, especially if youre prepared by simply checking the SVN now and then.
And yes, I have svn already, probably over 300 revisions by now, we work faster than spring itself

full interface write access rights are a must for even considering sticking KAI up on SVN and then theres a lot of thinking to be done...

It's optional to work with and within the development enviorment the Spring project has build. (and continues building)
And if you work better in a seperate envirement,.. alright do that.

Still standing is the problem that you are making a propertary extension for an open source engine.
Now I understand there is competition in AI development and this ain't bad in itself, competition drives people to do beter.
But at the end of the day this is still an open source project.
And I value that aspect of the project more then the competition.

AF · Post by AF » 12 Jul 2006, 22:58

No I'd rather the latest upto date windows version was released with spring rather than the latest svn code.

Post by **hollowsoul** » 12 Jul 2006, 23:27

Allow AI devs to have 2 folders for AI i.e release AI & svn AI
And then by default scons builds the release AI unless u pass an argument i.e NTAI-svn to compile the release AI instead ?

( just a random though

)

edit @AF i though of that straight away, but had just hit submit

and couldnt edit post so soon had to wait :/

AF · Post by AF » 12 Jul 2006, 23:31

scons should build the svn AI, but the installer or specified manually should do the release AI

submarine · Post by **submarine** » 18 Jul 2006, 20:11

i didnt read all posts above, i just wanted to ask what i have to do to make aai public source on svn?

AF · Post by AF » 18 Jul 2006, 20:29

Sign up on berliOS, and tell tim your username and that you want access. He gives you svn access to the AI folder and you can commit an AAI folder to the svn.

Spring RTS Engine

The source of the problem. :-)