http://gwt-code-reviews.appspot.com/1074801/diff/7001/8004 File dev/core/src/com/google/gwt/dev/CompilePerms.java (right):
http://gwt-code-reviews.appspot.com/1074801/diff/7001/8004#newcode228 dev/core/src/com/google/gwt/dev/CompilePerms.java:228: PrecompileOptions precompilationOptions = On 2010/11/04 20:58:24, scottb wrote:
I don't understand why this is changing. Shouldn't Precompile consume AnalyzeModuleOptions and produce PrecompilationResult unconditionally?
Why does
this phase sometimes want to consume AnalyzeModuleOptions?
things didn't really change, I just intercepted the 'new' sharding mode early in order to short circuit the old logic. The problem stems from the fact that in the previous sharing solutions the file used to save 'options' was overloaded to be either serialized options or serialized AST. In this new mode, we will always have an options file and it will always have a serialized options object. And we will have 1 or more serialized permutation file. I agree it would be clearer to just make a new entry point for Precompile and CompilePerms. I could do the same for the Link step - what do you think? http://gwt-code-reviews.appspot.com/1074801/diff/7001/8005 File dev/core/src/com/google/gwt/dev/Link.java (right): http://gwt-code-reviews.appspot.com/1074801/diff/7001/8005#newcode506 dev/core/src/com/google/gwt/dev/Link.java:506: } On 2010/11/04 20:58:24, scottb wrote:
Unrelated to patch?
Its related in that you can't debug this thing when you specify the wrong input. The default error is "error in file" and it exits. What file? What kind of error? http://gwt-code-reviews.appspot.com/1074801/show -- http://groups.google.com/group/Google-Web-Toolkit-Contributors
