Mailing-List: contact users-help@buildr.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@buildr.apache.org
Received-SPF: pass (athena.apache.org: domain of rteabeault@rallydev.com
 designates 74.125.149.155 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAHuk3y54wzqNYXYW8abwxdmxhBSRybe6+otsjbYA+F=uJBY3vw@mail.gmail.com>
References: 
 <CALcALS+MtauD=yahVw2OR=9uc-m7etb4W5B6EDzyuuQVFFZpFA@mail.gmail.com>
	<CAHuk3y54wzqNYXYW8abwxdmxhBSRybe6+otsjbYA+F=uJBY3vw@mail.gmail.com>
Date: Fri, 11 Nov 2011 09:51:04 -0700
Message-ID: 
 <CALcALSLfiHjq8HKVZ9TPpAVo+Vxw7vjEtabtwoX17wrRmTbkvQ@mail.gmail.com>
Subject: Re: Problem using build.yaml to specify groups of dependencies
From: Russ Teabeault <rteabeault@rallydev.com>
To: users@buildr.apache.org
Content-Type: multipart/alternative; boundary=90e6ba6e8ebca3dede04b1785329

--90e6ba6e8ebca3dede04b1785329
Content-Type: text/plain; charset=ISO-8859-1

Thanks for taking a look at this Alex.  I tried adding your suggestion and
it almost works but I think there might be another bug here.  The problem
is with specs that have a classifier.  For example,
consider net.sf.json-lib:json-lib:jar:jdk15:2.3 which has the 'jdk15'
classifier.  When this is listed under my compile dependencies it ends up
raising the error "RuntimeError : Expecting a String, Hash or object that
responds to to_spec"  After looking at the code I see the following:

When the build.yaml is loaded the ArtifactNamespace.use method determines
that compile is a group and sets is_group to true.  It then adds the each
artifact in the group to the registry using the unversioned spec string
"net.sf.json-lib:json-lib:jar:jdk15".  It then calls the group method
passing in 'compile' and all the spec names.  The group method creates a
lambda that when called will do a values_at using the spec names.  The
values_at method then generates a key
using ArtifactRequirement.unversioned_spec(name).  But name here is
"net.sf.json-lib:json-lib:jar:jdk15" and so the key comes out
as "net.sf.json-lib:json-lib:jar".  So when it calls get on the registry
using this key it returns false since the registry is using
""net.sf.json-lib:json-lib:jar:jdk15" key.  And 'false' is passed to the
satisfied_by method which ends up throwing the error.

So in short it seems that unversioned_spec is called twice on the spec
name.  The first call leaves the spec name with the classifier dangling off
the end.  This is used as the key to the registry. The second call, used to
find the artifact in the registry ends up chopping the classifier off
because it thinks it is the version.

Thoughts?

Thanks,
-russ

On Fri, Nov 11, 2011 at 4:56 AM, Alex Boisvert <alex.boisvert@gmail.com>wrote:

> On Thu, Nov 10, 2011 at 7:26 PM, Russ Teabeault <rteabeault@rallydev.com
> >wrote:
>
> > I am trying to use buildr.yaml to specify my dependencies
> >
> > artifacts:
> >  compile:
> >    - ant:ant:jar:1.6.5
> >    - ant-contrib:ant-contrib:jar:1.0b2
> >    - ...
> >
> > Based on everything I have read I should then be able to do
> >
> > compile.with artifact_ns[:compile]
> >
> > However, this fails in the artifacts method of artifact.rb since each
> spec
> > in the array is an ArtifactRequirement and there is no case handling that
> > type.
> >
> > Is this a bug?  Is there a better way to do this?  What are other people
> > doing?
> >
>
> Yes, that is a bug.  artifacts() doesn't correctly handle objects that
> ActsLikeArtifact. Will be fixed in trunk in a few minutes.
>
> In the mean time or for previous versions, you can write,
>
> compile.with artifact_ns.compile.map(&:artifact)
>
> instead.
>
> alex
>

--90e6ba6e8ebca3dede04b1785329--