fix the creation of BigDecimal from a String

follow what MRI is doing and ignore the precision

fixes #2650

Sponsored by Lookout Inc.

Author: mkristian

core/src/main/java/org/jruby/ext/bigdecimal/RubyBigDecimal.java

@@ -526,7 +526,9 @@ public static RubyBigDecimal newInstance(IRubyObject recv, IRubyObject[] args) {
             MathContext context = MathContext.UNLIMITED;
 
             if (args.length == 2) {
+		System.err.println( "margs" + args[1]);
                 int digits = (int)args[1].convertToInteger().getLongValue();
+		System.err.println( "margs" + digits);
                 if (digits < 0) {
                     throw runtime.newArgumentError("argument must be positive");
                 }
@@ -562,6 +564,8 @@ public static RubyBigDecimal newInstance(IRubyObject recv, IRubyObject[] args) {
                     return new RubyBigDecimal(runtime, (RubyClass)recv, new BigDecimal(((RubyFixnum)args[0]).getLongValue(), context));
                 } else if (args[0] instanceof RubyBignum) {
                     return new RubyBigDecimal(runtime, (RubyClass)recv, new BigDecimal(((RubyBignum)args[0]).getBigIntegerValue(), context));
+                } else {
+                    context = MathContext.UNLIMITED;
                 }
                 // fall through to String coercion below
             }
@@ -592,7 +596,10 @@ public static RubyBigDecimal newInstance(IRubyObject recv, IRubyObject[] args) {
             strValue = NUMBER_PATTERN.matcher(strValue).replaceFirst("$1");
 
             try {
+		System.err.println("newInstance " + strValue );
+		System.err.println("newInstance " + context );
                 decimal = new BigDecimal(strValue, context);
+		System.err.println("newInstance " + decimal );
             } catch(NumberFormatException e) {
                 if (isOverflowExceptionMode(runtime)) {
                     throw runtime.newFloatDomainError("exponent overflow");
@@ -1769,6 +1776,7 @@ public IRubyObject sqrt(IRubyObject arg) {
 
     @JRubyMethod(name = "to_f")
     public IRubyObject to_f() {
+	System.err.println("to_f " + value );
         if (isNaN()) {
             return RubyFloat.newFloat(getRuntime(), Double.NaN);
         }

test/test_bignum.rb

@@ -1,4 +1,5 @@
 require 'test/unit'
+require 'bigdecimal'
 
 class TestBignum < Test::Unit::TestCase
   # others tested in FixnumBignumAutoconversion test
@@ -41,6 +42,11 @@ def test_bignum_should_respond_to_array_operator
   def test_bignum_aref_with_bignum_arg_no_exception
     assert_equal(0, (2**64)[2**32])
   end
+
+  def test_GH_2650
+    assert_equal(BigDecimal.new("10.91231", 1).to_f, 10.91231)
+    assert_equal(BigDecimal.new("10.9", 2).to_f, 10.9)
+  end
   # more to come
 
-end
+end