On Thu, Jan 15, 2009 at 12:01 PM, Scott Blum <[email protected]> wrote: > This sounds great. My only comment involves this statement: " > > Casts and instanceof checks are accomplished by adding a flag > to JCastOperation and JInstanceOf to allow a non-null, non-Java-derived > (i.e.o.typeMarker$ != nullMethod) object to pass the type checks. Additional > methods are added to the Cast utility class and CastNormalizerupdated." I > believe based on reading this, but wanted to verify, that the 'flag' exists > only in the compiler (not the compiled code) and different exact variants of > dynamicCast/instanceOf are called at runtime.
That's the flag I asked about. In general it makes perfect sense to update the AST to hold any information that is needed. In this case, though, if I understand correctly, the information is soley determined by the target type of the cast. Mainly I worry that any code constructing a cast operation will need to be sure and set allowJso correctly, won't it? How does the code know what to supply? As a concrete example, there is an upcoming patch to add bridge methods in GenerateJavaAST. The bridge methods will need cast operations. How does the bridge-adder know what to set allowJso to, and should the bridge-adder have to worry about JSO's at all, anyway? -Lex --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/Google-Web-Toolkit-Contributors -~----------~----~----~----~------~----~------~--~---
