JSON export/import in generate-bindings.pl

Source/bindings/scripts/generate-bindings.pl

Index: Source/bindings/scripts/generate-bindings.pl
diff --git a/Source/bindings/scripts/generate-bindings.pl b/Source/bindings/scripts/generate-bindings.pl
index b3f3d43298ac01f3e561873f5119d564763adf41..2abfb42985f16fe78092c1596d16bee5aca812f7 100755
--- a/Source/bindings/scripts/generate-bindings.pl
+++ b/Source/bindings/scripts/generate-bindings.pl
@@ -38,6 +38,7 @@ use Cwd;
 
 use IDLParser;
 use CodeGeneratorV8;
+use JSONSerializer;
 
 my @idlDirectories;
 my $outputDirectory;
@@ -184,6 +185,10 @@ foreach my $idlFile (@supplementedIdlFiles) {
     }
 }
 
+# FIXME: This code will be removed once IDLParser.pm and CodeGeneratorV8.pm

[dominicc (has gone to gerrit), 2013/06/12 04:39:33]

Good.

+# are connected via JSON files. See http://crbug.com/242795
+$targetDocument = deserializeJSON(serializeJSON($targetDocument));
+
 # Generate desired output for the target IDL file.

[dominicc (has gone to gerrit), 2013/06/12 04:39:33]

Is it possible to chop this up with functions or modules so that we know exactly
what data the following steps depends on? The above code sets up a bunch of
state and that is intertwined with parsing, which is not what we want.

I'm fine if that is in a follow-up changelist. But if so it would be nice to
have a separate refactoring patch that focuses specifically on that.

[Nils Barth (inactive), 2013/06/12 09:15:27]

Agreed.
I'd be happy to do a follow-up refactoring CL.
Alternatively, we could do it when we change the build,
as we'll be changing the flow then, and possibly breaking this script up into
two pieces (which would disentangle these pretty clearly).

Haraken?

 my @dependentIdlFiles = ($targetDocument->fileName(), @supplementedIdlFiles);
 my $codeGenerator = CodeGeneratorV8->new($targetDocument, \@idlDirectories, $preprocessor, $defines, $verbose, \@dependentIdlFiles, $writeFileOnlyIfChanged);