Consider the following two IDL files: Base.idl: dictionary Base { unsigned long baseField; } Derived.idl dictionary Derived : Base { unsigned long derivedField; } When JSDerived.cpp is generated, its convertDictionary() implementation contains the code to convert both baseField and derivedField. JSBase.cpp's convertDictionary() has its own convertDictionary() to convert baseField, but JSDerived.cpp does not rely on it. If I add a second field to Base.idl and rebuild, JSBase.cpp is updated to convert the new field, but JSDerived.cpp is not updated, so it still only converts the original two fields. It seems like JSDerived.cpp's convertDictionary() should be calling JSBase.cpp's convertDictionary() to convert the base class fields. Either that or the bindings generator should be regenerating all dictionaries that inherit from Base.
We hit that issue in https://bugs.webkit.org/show_bug.cgi?id=215722
Created attachment 407094 [details] Patch
Created attachment 407097 [details] Patch
Comment on attachment 407097 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=407097&action=review > Source/WebCore/bindings/scripts/preprocess-idls.pl:244 > + my $makefileDeps = "#Supplemental dependencies\n"; Normally we’d have a space after the "#" in a comment. > Source/WebCore/bindings/scripts/preprocess-idls.pl:258 > + $makefileDeps .= "#Dictionaries dependencies\n"; Ditto. > Source/WebCore/bindings/scripts/preprocess-idls.pl:261 > + my $dependencies = $dictionaryDependencies{$derivedDictionary}; > + $makefileDeps .= "JS${derivedDictionary}.cpp: ${dependencies}\n"; Not sure we benefit from the local variable here.
Created attachment 407191 [details] Patch for landing
Committed r266120: <https://trac.webkit.org/changeset/266120> All reviewed patches have been landed. Closing bug and clearing flags on attachment 407191 [details].
<rdar://problem/67740737>