Fixes bug in handling of nil fields

Fixes a bug that was causing the encoder to skip fields after other fields were set to nil.

lib/protocol_buffers/runtime/field.rb

@@ -142,7 +142,7 @@ def has_#{name}?; true; end
           def #{name}=(value)
             field = fields[#{tag}]
             if value.nil?
-              @set_fields.delete_at(#{tag})
+              @set_fields[#{tag}] = false
               @#{name} = field.default_value
             else
               field.check_valid(value)

spec/nil_fix_spec.rb

@@ -0,0 +1,53 @@
+# Test by Mark Herman, II <mherman@iseatz.com>
+#
+# This is a test to demonstrate a significant bug with the handling of nil
+# values.  I will be contributing a fix that will cause these tests to pass.
+
+# Here is an explanation of the bug...
+
+# There is an array called @set_fields in our protobuf Message class.
+# The Field class (field.rb) adds field accessors to the Message class that incorrect
+# call delete_at on @set_fields rather than setting the value to false when a nil
+# value is passed.  So, @set_fields shrinks from the end rather than getting false
+# values in the proper place.  As more fields are set to nil, the Message class
+# thinks fewer and fewer fields are set.  The value_for_tag? method depends on the
+# value of @set_fields.  The encoder depends on value_for_tag?.  So, as more nils
+# are set, the encoded protobuffers shrink.  The data isn't actually lost.  It's
+# never encoded.
+
+require File.expand_path(File.dirname(__FILE__) + '/spec_helper')
+require 'protocol_buffers'
+require 'base64'
+
+# This is a compiled Ruby protobuffer class that will demonstrate the bug
+# with the handling of nil classes.
+
+class TestMessage < ::ProtocolBuffers::Message
+  optional :string, :field1, 1
+  optional :string, :field2, 2
+
+  gen_methods! # new fields ignored after this point
+end
+
+describe "Error in nil handling" do
+  it "will flag field2 as not having a value when setting field1 to nil" do
+    tp = TestMessage.new  # a test protobuffer
+    tp.field1 = "field1_value"
+    tp.field2 = "field2_value"
+
+    tp.field1 = nil
+
+    tp.value_for_tag?(2).should be_true
+  end
+
+  it "will still encode field2 after setting field1 to nil twice" do
+    tp = TestMessage.new  # a test protobuffer
+    tp.field1 = "field1_value"
+    tp.field2 = "field2_value"
+
+    tp.field1 = nil
+    tp.field1 = nil
+
+    Base64.encode64(tp.to_s).length.should be > 0
+  end
+end