Carsten Ziegeler wrote:
Nicola Ken Barozzi wrote:
I have backported the pass-through stuff from the trunk (where it works, and is being used by Forrest) to the 2.1 branch (where it does not work right).
In particular, it seems that the source resolving is not happening correctly, because part of the processing start happening if I change from cocoon:/ to cocoon:// and if I insert full pathnames for files.
The same processing part work ok if not called by a passthrough mount.
It seems like things are not being set back correctly after the child sitemap ends and the parent continues processing.
The files to compare is MountNode.java (and maybe PipelineNode.java).
Any ideas?
As far as I remember, the code in 2.1.x has never been designed to reset the state when a child sitemap is left. The rewriting of the internal environment handling for 2.2 introduces this "new" behaviour. So I think, we either have to introduce this into 2.1.x as well or introduce the pass-through stuff for 2.2 and do not add it to 2.1.x
Are you sure? How about <handle-errors/> - it does exactly the same, it *has* to properly exit child sitemap(s), and it was working since 2.0.
Vadim
