Please review: JBS: https://bugs.openjdk.java.net/browse/JDK-8177280 Webrev: http://cr.openjdk.java.net/~hannesw/8177280/webrev.00/
As I said previously there are some things in this patch I’m unsure or not quite happy about. Some notes: - While the implementation of the new DocTrees method is quite simple, I’m not sure if I should install a new Log.DeferredDiagnosticHandler for the runtime of the method, like the attributeDocReference method in the same class does. - In HtmlDocletWriter.seeTagToContent I changed the handling of the tag text to escape HTML characters if no label is specified. This causes „<„ and „>“ to be displayed in the browser instead of being interpreted as HTML tag if a generic link target can’t be resolved. This is potentially problematic since @see and @link can contain plain HTML content, but I think it’s ok since those cases are handled further up in HtmlDocletWriter.seeTagToContent. - That same method in HtmlDocletWriter contained some never-used code (dependent on the BaseConfiguration.backwardCompatibility flag which is always true) to use the fully qualified class name for links in some cases. I removed that code and the field in BaseConfiguration since it adds to that already long-winding method. - Setting the label on a LinkInfoImpl basically disables rendering of type arguments. While I understand the rationale behind this, it might still be useful to set the label for the main link of a generic type. I’ve tried to remove the restriction, but ran into a lot of problems (i.e. failing tests) in other places. Since the current behaviour does what we need I decided to not change this. - There was a potential infinite recursion in LinkFactoryImpl.getTypeParameterLinks caused by following the type parameters of linkInfo.typeElement when linkInfo.type is set. The problem was that linkInfo.typeElement may be set to the owner of linkInfo.type, and the solution is to only follow that path if linkInfo.type is null. - I removed two unused LinkInfo Kind enum values. - I CommentHelper there were previously two and now three Visitors to extract ReferenceTrees, so I added a common base class called ReferenceDocTreeVisitor. - As an completely unrelated cleanup I removed the unused javafx field in IndexBuilder. Thanks, Hannes
