geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jason Dillon <ja...@planet57.com>
Subject Re: Local module repos and SNAPSHOT artifacts
Date Fri, 23 Feb 2007 10:46:03 GMT
Oh ya... insert :-) every few lines... :-P

--jason


On Feb 23, 2007, at 2:27 AM, Jason Dillon wrote:

> Folks, the local module repository thing is a massive hack, not  
> meant to be used as much as we are using it... and really it should  
> *not* have SNAPSHOT artifacts in it.
>
> When using snaps in these local artifacts, strange artifact  
> resolution failures are bound to occur when Maven decides its time  
> to go update snaps (daily for us).  This causes problems when  
> building a module which depends on another module which has a  
> module local repo that contains snapshots, since the current module  
> does not have access to the dependency's repo it will cause an  
> artifact resolution exception.
>
> I did not check _all_ of the module local repos that we have in  
> server/trunk, but I know that at least in configs/jasper/repository  
> there are myfaces 1.2.0-SNAPSHOT artifacts.  This is not really  
> acceptable.  First, its bad enough that we have to have this repo  
> here in the first place (the myfaces team should just publish snaps  
> like all other projects) and second, these artifacts are SNAPSHOT  
> which causes build problems as noted above.
>
> *If you must* use a module local repo, then *do not* put SNAPSHOT  
> artifacts in there... *if you must* then include the _timestamp- 
> build_ artifacts and configure your pom to use them, so that  
> dependent projects don't freak-out when mvn tries to update snaps.
>
>  * * *
>
> I'm still working out a simply/elegant way to solve this problem of  
> remote repos and local repos, etc... but until I get that we need  
> to becareful about how we use these local repos and which remote  
> repos we include (I notice we still have a few legacy repos, which  
> are a big no-no).
>
> David Jencks mentioned in IRC today that it might be better if we  
> just had one location where all of these module-local repos  
> artifacts are kept.  IMO, the module-local repos suck... but having  
> one sucks less than having more than one, so I'm inclined to agree  
> this is a good idea for the short-term.  To make this work, we  
> basically create another top-level module (peer to modules and  
> configs, etc) say named "repository" (or whatever).  This module  
> contains the single local-module repository in m2 format, which is  
> only configured in that modules pom.  Then the pom lists all of the  
> artifacts as dependencies which are in the repo to force them to  
> get installed into Maven's local cache.
>
> This module is added first, before all other children in the  
> reactor, which will *hopefully* always get that module executed  
> first, those deps installed and then other modules which depend on  
> those custom bits will already have them resolved in the local  
> maven cache.  It sucks... but it sucks-less IMO than handful of  
> repos that we have now scattered through the project tree.
>
> I'm still working on getting a better solution to the entire remote  
> repo/repeatable build/blah, blah stuff... but until then I think  
> this is a decent step to simplify things a tiny bit more and help  
> reduce strange problems from popping up.
>
>  * * *
>
> Comments?  Unless there are any objections, I'll implement this in  
> the next day or so... BUT, still need someone to deal with the  
> myfaces 1.2.0-SNAPSHOT bits which really need to be fixed even if  
> we leave the repos asis, or move to this one single short-term  
> local repo.
>
> --jason


Mime
View raw message