ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Bodewig <bode...@apache.org>
Subject Re: RFC - rewrite of the delegating classloader FAQ
Date Tue, 10 May 2005 14:26:04 GMT
On Wed, 04 May 2005, Peter Reilly <peterreilly@apache.org> wrote:
> Stefan Bodewig wrote:

>>I also shortly thought about explaining the dangerous solutions
>>(which probably work), but I'm not sure we want to talk about
>><taskdef reverseloader="true">
>>
> I tried this, it does not work too good (well ....  I could not get
> it to work at all) - it seems to get defeated because some of the
> classes are already paritially loaded already.

In my test, DefBase#createLoader() is the problem, in particular

            // need to load Task via system classloader or the new
            // task we want to define will never be a Task but always
            // be wrapped into a TaskAdapter.
            ((AntClassLoader) createdLoader)
                .addSystemPackageRoot("org.apache.tools.ant");

which means that reverseloader is ignored for all classes that are in
Ant's package, including the JUnit task itself.

>>The major thing I'm unsure about is that I called the classloader
>>loading Ant in 1.6.x the "coreloader", something that may give false
>>ideas to people reading Ant's source.
>>
> Indeed, we should try to clean this up -- maybe deprecate the
> Project#setCoreloader() method.

Are we sure they are really dead code?  If so, yes, let's deprecate
them - and document that they are not used for what one could think by
looking at the name.

Stefan

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@ant.apache.org
For additional commands, e-mail: dev-help@ant.apache.org


Mime
View raw message