ant-ivy-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Taylor <dt...@stanford.edu>
Subject Re: building only whats necessary with Ivy
Date Mon, 17 Sep 2007 20:11:11 GMT
The old IvyCruise page on the JayaSoft site points to IvyPublisher as 
the successor to IvyCruise. But on IvyPublisher's SourceForge site, I 
see no activity since its initial release a year ago.

Why no mention of IvyPublisher in this discussion? Did IvyPublisher 
follow IvyCruise to IvyHeaven?

dtayl

ivy-user-digest-help@incubator.apache.org wrote:
>
> Subject:
> Re: building only whats necessary with Ivy
> From:
> "Xavier Hanin" <xavier.hanin@gmail.com>
> Date:
> Thu, 13 Sep 2007 15:37:06 +0200
> To:
> ivy-user@incubator.apache.org
>
>
> This would be a good question for the xooctory user list :-) More comments
> below.
>
> On 9/13/07, Stephan Zeissler (KUTTIG) <stephan.zeissler@kuttig.com> wrote:
>   
>> I don't think thats a feature of Ivy, more a feature of xooctory. The CI
>> server may parse the ivy.xml files and build a dependency tree from
>> this. So when your projects depends on a library and you build a new
>> version of the lib, all dependent projects get rebuild, too.
>>     
>
>
> That's exactly what xooctory does (or is about to do, since this feature is
> still in developement). Basically some continuous integration servers do not
> implement project dependencies at all, and recommend to build your whole
> codebase at once whenever one module in your app change (this is what has
> been recommended with cruisecontrol for a long time). This does not scale
> very well, and you happen to rebuild and retest your whole application only
> when a very small subset is changed. Hence some CI servers implement project
> dependencies, so that you create one project in your server per module in
> your app, describe their dependencies, then the server takes care of
> rebuilding the right modules depending on your modifications. But then when
> you a dependency manager like Ivy, you still have to maintain your
> dependencies in your CI server by hand to keep them in sync with your
> metadata. So the idea is to make the CI server aware of metadata available
> in your modules. That's what the IvyCruise plugin tried to achieve a long
> time ago, but the CC architecture didn't make that easy. So that's what
> xooctory will achieve pretty soon.
>
> As i dont know xooctory, I dont know how it determines, what has
>   
>> changed, but from hudson I know that you can e.g. pull a scm repository
>> (svn,cvs,..) regularly or start a build via the invocation of an URL.
>> But all this is part of the CI server, not Ivy.
>>     
>
>
> Indeed. The singularity is that Xooctory considers module dependencies as
> first class citizens in the CI process, as the SCM is. Hudson has pretty
> good support of project dependencies, but you still need to maintain them by
> hand as far as I know (at least it was the case when I started the xooctory
> project some months ago).
>
> Xavier
>
> - Stephan
>   
>> bhatia schrieb:
>>     
>>> On the xooctory site, I read this:
>>>
>>> "leveraging Ivy to manage the dependencies between the modules, and
>>>       
>> rebuild
>>     
>>> exactly what is necessary, in the correct order, blocking builds when a
>>> dependency is currently building, and so on"
>>>
>>> Can someone provide more inputs/ideas on how to rebuild exactly what is
>>> necessary using Ivy ? What are the (best) practices you use to determine
>>> what has changed ?
>>>
>>> thanks
>>>
>>>       
>
>
>
>   


Mime
View raw message