[richfaces-issues] [JBoss JIRA] Updated: (RF-11093) UIDataAdaptor context variable backup broken (reentrance problem)

UIDataAdaptor context variable backup broken (reentrance problem) ----------------------------------------------------------------- Key: RF-11093 URL: https://issues.jboss.org/browse/RF-11093 Project: RichFaces Issue Type: Bug Security Level: Public(Everyone can see) Components: component-tables Affects Versions: 4.0.0.Final Environment: Windows XP, Java 6, Tomcat 6.0.24, Mojarra 2.0.2. Reporter: Yannick Valot Attachments: UIDataAdaptor.java.patch I am encountering a reentrance problem with org.richfaces.component.UIDataAdaptor (using it as a dataTable). I need to perform a visit of the component tree while executing an action on a command link in one row of the same table. <rich:dataTable value="#{listOfItems}" var="myItem" > ... some stuff here ... <commandLink here> ... some stuff here ... </rich:dataTable> After the action on the command link has visited the component tree, myItem is no longer available in context. When the UIDataAdaptor sets or resets the "myItem" context variable, it backs up any previous value by calling captureOrigValue(FacesContext), and later restores it by calling restoreOrigValue(FacesContext), but it looks as if the visit backs the variable up twice (in a nested fashion), and it erases myItem between the two backups : - backup variable (old value is correctly backed up) (in visitTree(VisitContext visitContext, VisitCallback callback)) - erase variable (setRowKey(facesContext, null) in same function) - erase variable (doVisitChildren(VisitContext context, boolean visitRows)) - erase variable (visitTree(VisitContext visitContext, VisitCallback callback)) - backup variable (erases backed up value) (walk(FacesContext faces, DataVisitor visitor, Object argument)) - ... - restore value (restores null) (walk(FacesContext faces, DataVisitor visitor, Object argument)) - restore value (restores null) (in visitTree(VisitContext visitContext, VisitCallback callback)) etc. The problem can be solved by *stacking* backed up variables, instead of storing them in a simple attribute, to allow for any kind of reentrance. (I do not say that this is the best possible patch, but it has solved the problem for me).