The problem is that the JavaScript "prepareInputsForSubmission(this)", which is in the form's onsubmit attribute, is never called. This happens because the JavaScript from the form's onsubmit attribute is not called when the form is submitted via a commandLink, rather than a commandButton, and the DELETE button is a commandLink. So I figured I would try using the commandLink's onclick attribute to call prepareInputsForSubmission(), i.e. onclick="return prepareInputsForSubmission(this.form);". This almost works, but for some reason, the commandLink's child params are not included in the request that is submitted, and therefore the delete ends up being a no-op. Note, the reason commandLink was used rather than commandButton in the first place is because commandButton does not support child f:params, which were necessary for the DELETE button; a possible alternative is to use a commandButton and f:attributes, which are supported by both commandLinks and commandButtons. I spent several hours troubleshooting this and wasn't able to get it to work. I think it should be tabled until post-x.2. The workaround is to uncheck and then recheck the Unset checkboxes for any optional integer props before clicking SAVE to update the config. Otherwise, you will see validation errors for those props saying "" is not a valid value for integer property 'foobar'.
ian, what are the things we modeled as open map props today across all plugins? i'm trying to get a sense for how often end users will see this. if this is open map type is used for any common and/or important config props, i would want to get a proper fix into x.2 release. as for the problem itself, what happens if the commandLink makes the change, and then *redirects* back to the same page? if things are in the seam conversation (and they are) would this just naturally work instead of trying to get a javascript-based solution to work?
This bug was previously known as http://jira.rhq-project.org/browse/RHQ-1792
Temporarily adding the keyword "SubBug" so we can be sure we have accounted for all the bugs. keyword: new = Tracking + FutureFeature + SubBug
making sure we're not missing any bugs in rhq_triage
Closing per 19-Aug triage -- this bug will be made obsolete per UI redesign.