As a hack would setting the module component implementation to null
work? If this worked, I think it should only be temporary and we
should do the recursion solution since others will run into this.
Jim
On Feb 28, 2006, at 11:02 AM, Jeremy Boynes wrote:
Jeremy Boynes wrote:
There is a quick fix we can do for this if we are prepared to assume
that we only have a single level of composition. In that case, we can
register the module component first with no implementation, then
register and build it's implementation (the module element). However,
although this is a quick fix in the builder, we will need to make
more
extensive changes to the loader APIs so that they return Module
elements
rather than ModuleComponent elements. I think this is worth doing
as it
will make subsystem assembly easier.
I took a quick attempt at this and ran into a few issues that get me
concerned that this is not the right direction. In order to
register the
empty module component I had to do enough initialization that I don't
think registering a new implementation work.
If someone else can have a look or provide an alternative I would
appreciate it; if not I will look at the second option, getting the
recursion to work.
--
Jeremy
