> From: Stefan Bodewig [mailto:[EMAIL PROTECTED]
>
> Craeg K. Strong <[EMAIL PROTECTED]> wrote:
>
> > This is persuant to the patch I sent recently for the <style> task
> > (adding a "force" attribute).
>
> which I hope to get to these days, thanks.
>
> > Note that these dependencies may not actually be _computable_ --
> > rather they are simply asserted.
>
> But this doesn't exclusively apply to <style>. Say I have Java code
> that uses a different class via reflection - neither <javac> nor
> <depend> will catch this.
>
> For this situation I've been playing with the idea of
>
> <explicit-depend src="foo" dest="bar">
> <dependency target="targetfile">
> <source name="source1" />
> <source name="source2" />
> </dependency>
> </explicit-depend>
>
> which would - similar to depend - delete bar/targetfile if foo/source1
> or foo/source2 is newer than that.
>
> This would help in more than just these situations but could be used
> to support the more generic tasks we have in several other fields.
>
It would be nice if we could define this in a more general way as a sort of
dependency pattern. The idea being that you can define them and then add it
as references to <depend>-like tasks that are specific to the particular
compiler.
<depend> for java code;
<cc> which does its own dependency analysis; etc.
> This task would handle several dependencies at once (I'd translate all
> source tags to includes in a fileset with dir "src" and create a
> single mapper from all dependency rules).
>
> Stefan
>
> who doesn't really like the name explicit-depend.
>
If it where some kind of pattern, then it may be better called <dependset>
or something like that.
Jose Alberto
