DO NOT MERGE Do not require signed zero support in ldexp tests()

This relaxes ldexp() tests to allow -0/+0 to be interchanged.

Bug: 19145770
Change-Id: I46c6a8f224606a15be0f974af36c281612f2398b
(cherry picked from commit 1e44f197ded674677381f65c840820728cf36efb)
diff --git a/modules/gles31/functional/es31fShaderCommonFunctionTests.cpp b/modules/gles31/functional/es31fShaderCommonFunctionTests.cpp
index 1bac102..00ad997 100644
--- a/modules/gles31/functional/es31fShaderCommonFunctionTests.cpp
+++ b/modules/gles31/functional/es31fShaderCommonFunctionTests.cpp
@@ -1879,7 +1879,6 @@
 		const glu::DataType		type			= m_spec.inputs[0].varType.getBasicType();
 		const glu::Precision	precision		= m_spec.inputs[0].varType.getPrecision();
 		const int				scalarSize		= glu::getDataTypeScalarSize(type);
-		const bool				signedZero		= supportsSignedZero(precision);
 
 		const int				mantissaBits	= getMinMantissaBits(precision);
 		const deUint32			maxUlpDiff		= getMaxUlpDiffFromBits(mantissaBits);
@@ -1890,7 +1889,7 @@
 			const int		in1			= ((const int*)inputs[1])[compNdx];
 			const float		out0		= ((const float*)outputs[0])[compNdx];
 			const float		refOut0		= ldexp(in0, in1);
-			const deUint32	ulpDiff		= signedZero ? getUlpDiff(out0, refOut0) : getUlpDiffIgnoreZeroSign(out0, refOut0);
+			const deUint32	ulpDiff		= getUlpDiffIgnoreZeroSign(out0, refOut0);
 
 			const int		inExp		= tcu::Float32(in0).exponent();