[kepler-dev] ksw builds

Kevin Ruland kruland at ku.edu
Fri Sep 16 13:38:05 PDT 2005


Chad,

Does it still create files in the magical mystery directory
build/classes?  That directory is intentionally left off my eclipse
setup in order to allow me to switch between the ant build system and
the eclipse build easily.

Kevin

Chad Berkley wrote:

>Hey,
>
>I'm ready to change kepler over to use the new Object Manager/KSW 
>architecture.  I wanted to see what people think should be done with the 
>build process because this is going to affect the way we build kepler on 
>a daily basis.  Here's the scenario of how kepler is now going to work:
>
>When you first checkout a clean copy of kepler, the ant task 
>'buildkswlib' needs to be run.  This builds all of the ksw files from 
>the actorList.moml and directorList.moml files.  The ksw files contain 
>metadata and the implementation class of each actor and director in the 
>system.  The ksw files are placed in the directory kepler/ksw.  When you 
>run kepler for the first time after building the ksw library, the ksw 
>files are read from this directory and cached using the object manager. 
>  This takes about 45 secs to a minute (on my P4 2.8GHz machine).  Once 
>the ksw files are cached, any subsequent runs of kepler takes the normal 
>8-12 seconds to start up.
>
>What I would like feedback on, is how would you like the build system 
>setup and how would you like to distribute this to users?  I'm perfectly 
>happy to do 'ant buildkswlib' as a first step to building kepler (just 
>like I do 'ant ptolemy' as a first step) but I know other people use 
>different build systems and may like to see this streamlined (note that 
>you should only have to do an 'ant buildkswlib' once unless you want a 
>fully clean build).  One way around this is to have not just the ksw 
>files, but also the cache pre-built by ant.  There could be an ant 
>'setup' target or the like that would build the ksw library and run the 
>object manager to cache all of the ksw files.  This setup stage could 
>also include building ptII which would make one setup step for a clean 
>checkout of kepler.
>
>On the distribution side, I think we should distribute a pre-built cache 
>and install it with the installAnywhere installer.  This will keep users 
>from having to wait the initial 45 secs - 1 min. for the caching to take 
>place.  I'm not sure how hard this would be to do with InstallAnywhere 
>but it seems pretty easy.
>
>Anyway, suggestions are appreciated.  I am going to switch the build 
>process to use the new ksw system this afternoon, unless I hear any 
>major objections within the next few hours.  When I do, i'll send out 
>another email detailing what you have to do to get kepler to run.  I 
>think we can work the kinks in the build system as time goes on and we 
>get used to the new system.  Please let me know if you have any 
>questions/comments.
>
>Thanks,
>chad
>_______________________________________________
>Kepler-dev mailing list
>Kepler-dev at ecoinformatics.org
>http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/kepler-dev
>  
>


More information about the Kepler-dev mailing list