Forum OpenACS Development: Problem with content_item__new() after TIP #42 ...

This commit (part of TIP 42, 43, 53 by Timo) has introduced a bug on HEAD.

For example, you will run into trouble if you have an application based on Workflow and your workflow cases are not associated with a cr_item but with a plain acs_object.

workflow_case_log_entry__new() gets called when a new case is created and this function in turn calls content_item__new() (the one with 16 params which then directly calls the now modified 17 param version). The part that causes the error is this, at line 359:

	   if new__package_id is null then
  	     v_package_id := acs_object__package_id(content_item__get_root_folder(v_parent_id));
  	   else
  	     v_package_id := new__package_id;
  	   end if;
If you try "SELECT content_item__get_root_folder(719);" with a plain object id you will get this:
ERROR:   -20000: Could not find a root folder for item ID 719. Either the item does not exist or its parent value is corrupted.
Any ideas how this should be solved?

TIA.

Collapse
Posted by Jeff Davis on
It probably should check if the parent_id is a cr_item before it tries to deduce the package_id from the root folder, and the workflow_case_log_entry__new should probably propigate a package id down so it does not have to do any of this either way.
Collapse
Posted by Ola Hansson on
I have committed a few changes to Workflow which will allow case objects to be plain acs objects.

I can't test the upgrade script on oracle so if someone wants to at least visually inspect it that would be nice.

I figured that workflow_case_log_entry__new() needed to be augmented with the package_id param since by fetching that id inside the function by a call to acs_object__package_id() 'null' would be returned (and the above error would reoccur) if the app at hand stored the package_id in a table of its own as opposed to in acs_objects ...

Should package owners be suggested to move 'package_id' to acs_objects and provide upgrade scripts that will move the existing entries from the package-specific tables?