http://gwt-code-reviews.appspot.com/1172801/diff/1/2 File dev/core/src/com/google/gwt/dev/shell/JavaDispatchImpl.java (right):
http://gwt-code-reviews.appspot.com/1172801/diff/1/2#newcode65 dev/core/src/com/google/gwt/dev/shell/JavaDispatchImpl.java:65: throw new RuntimeException("Field does not exist."); On 2010/12/01 16:28:32, jat wrote:
Is a plain RTE appropriate here? Would HostedModeException be better,
since it
seems to indicate either someone doing something they shouldn't or
something
gone seriously wrong?
Likewise the two places below.
These make sense as RuntimeExceptions with the corresponding rewrite changes in Jsni.java. If there is objection to the corresponding rewrite changes, then these should be transformed to HostedModeExceptions since then they would indicate potential user code problems. http://gwt-code-reviews.appspot.com/1172801/diff/1/3 File dev/core/src/com/google/gwt/dev/shell/Jsni.java (right): http://gwt-code-reviews.appspot.com/1172801/diff/1/3#newcode118 dev/core/src/com/google/gwt/dev/shell/Jsni.java:118: } On 2010/12/01 16:28:32, jat wrote:
This does mean that if you have bad JSNI that was never called, this
change will
stop your code from working.
I'm not objecting, just wanting to make sure this is a conscious
change. Yes, this is very much on purpose and is trying to keep in the spirit of not loading a class if it has errors. Given how hairy JSNI syntax can be, this is an especially useful change for the unfortunate users that don't use GPE. http://gwt-code-reviews.appspot.com/1172801/show -- http://groups.google.com/group/Google-Web-Toolkit-Contributors
