> > What I dont get is why we would need a FB/Eclipse plugin if we manage to > > make these changes to the compiler. As you example points out, a valid > > class would be created in the src path. This would take care of the > > auto-complete requirements, right? All IDEs should be able to work with > > it. > IMO, something like an "R" class should not be put into the compiler or the > IDE. The compiler should be as framework agnostic as possible. It is > probably possible to create a set of classes that do what R does. > > Alex, I think I lost you there. What Dirk is talking about is to have the compiler auto-generate the class R as part of pre-compilation. Except that this class would actually be added to the source path. What do you mean by 'framework agnostic'?
> My proposal, is to at least show compiler warnings if other locales dont > > have matching strings for each entry in the 'master' bundle. > > having compiler warnings for missing keys sounds good - but this is > maybe a bit inconsisten with the runtime behaviour: > > IIRC, if you compiled the SWF with the locale chain ["en_US", "fr_FR"] > and at runtime try to get a key from the fr_FR bundle and that key > does not exist, Flex will try to get the key from the en_US bundle > (and failing that I think you get an RTE, not sure though). So no > warning at runtime but a "soft" fallback to the previous locale in the > locale chain. > Dirk, why would this cause inconsistent behavior? All the compiler warning would say is that "hey, you dont have the key in your locale: fr_FR whereas the master locale: en_US has it". If the developer ignores the warning and the scenario you mention occurs, it will cause the same exact runtime behavior. What the compiler warnings help with is to avoid having to run into a runtime error to actually realize the error.
