http://build.webkit.org/builders/Apple%20MountainLion%20Debug%20WK1%20%28Tests%29/builds/316/steps/bindings-generation-tests/logs/stdio One example failure: FAIL: (ObjC) DOMTestActiveDOMObject.h --- WebCore/bindings/scripts/test/ObjC/DOMTestActiveDOMObject.h 2012-08-19 00:12:14.000000000 -0700 +++ /var/folders/rg/3xbzr5_x083c8kzwn20v95nm0000gn/T/tmpQeyWTd/DOMTestActiveDOMObject.h 2012-08-29 08:24:34.000000000 -0700 @@ -32,7 +32,8 @@ @class NSString; @interface DOMTestActiveDOMObject : DOMObject -- (int)excitingAttr; +@property(readonly) int excitingAttr; + - (void)excitingFunction:(DOMNode *)nextChild; - (void)postMessage:(NSString *)message; @end The binding tests are expecting incorrect results on Mac. The versions using the @property syntax are correct for Leopard and above. CodeGeneratorObjC.pm determines which syntax to use based on the value of MACOSX_DEPLOYMENT_TARGET. That particular bot has MACOSX_DEPLOYMENT_TARGET=10.8 set in the environment (not sure why), while other invocations of the bindings generation tests must have it unset. The actual bindings generation that is performed for our use as API does have MACOSX_DEPLOYMENT_TARGET set correctly as it's invoked from within the context of xcodebuild. We should remove the code in CodeGeneratorObjC.pm that deals with the old pre-@property syntax, since it is only used on Tiger and earlier, and fix the expected results. <rdar://problem/12147126>
Sounds like an excellent plan.
Created attachment 161280 [details] Patch
Comment on attachment 161280 [details] Patch LGTM. You might want tim or someone else from apple to review though.
(In reply to comment #3) > (From update of attachment 161280 [details]) > LGTM. You might want tim or someone else from apple to review though. Thanks! Tim said it looks good.
Comment on attachment 161280 [details] Patch Committed in http://trac.webkit.org/changeset/127032