Fixes #3326. JRuby 9.0.1.0 vs. MRI 2.2.2 define_method inconsistency

Problem was when we switched from arity to signature we added a new call and we passed the block arg passed to the symbol proc rather than the symbol proc itself. Getting this to trigger did not really fit into any standard ruby behavior so I made a regression spec for it. My guess is we only internally call the broken Java method if the block is set in an uncommon way (although I am not entirely sure what that is...but define_method will cause it to happen).
jruby · Nov 2, 2016 · 643d442 · 643d442
1 parent 07ad439
commit 643d442
Show file tree

Hide file tree

Showing 2 changed files with 16 additions and 1 deletion.
diff --git a/core/src/main/java/org/jruby/RubySymbol.java b/core/src/main/java/org/jruby/RubySymbol.java
@@ -1043,7 +1043,7 @@ private IRubyObject yieldInner(ThreadContext context, RubyArray array, Block blo
 
         @Override
         public IRubyObject yield(ThreadContext context, Block block, IRubyObject[] args, IRubyObject self, Block blockArg) {
-            RubyProc.prepareArgs(context, block.type, blockArg.getBody(), args);
+            RubyProc.prepareArgs(context, block.type, this, args);
             return yieldInner(context, RubyArray.newArrayMayCopy(context.runtime, args), blockArg);
         }
 

diff --git a/spec/regression/GH-3326_symbol_proc_should_be_callable.rb b/spec/regression/GH-3326_symbol_proc_should_be_callable.rb
@@ -0,0 +1,15 @@
+require 'rspec'
+
+describe "sym.to_proc in define_method" do
+  it "should not raise an ArgumentError" do
+    expect do
+      # Weirdly using the proc generated will end up changing
+      # something about the block which would in turn call a
+      # java version of yield in SymbolProcBody which was passing
+      # in its block param instead of itself for signature arity check.
+      s = :to_s.to_proc
+      define_method :gh3326, s
+      s[1]
+    end.not_to raise_error
+  end
+end