Nicolas Cellier uploaded a new version of KernelTests to project The Trunk: http://source.squeak.org/trunk/KernelTests-nice.317.mcz
==================== Summary ====================
Name: KernelTests-nice.317 Author: nice Time: 27 January 2017, 10:04:37.372536 pm UUID: 42d2ef78-acd4-4909-aa38-c4a709f40482 Ancestors: KernelTests-eem.316
Change the expectation of -0.0 sign, it is now 0 rather than -1. Consequently adopt the new way to test a negativeZero result using signBit rather than sign.
=============== Diff against KernelTests-eem.316 ===============
Item was added: + ----- Method: FloatTest>>assertIsNegativeZero: (in category 'accessing') ----- + assertIsNegativeZero: aFloat + "Assert that aFloat is Float negativeZero" + self assert: aFloat = 0.0. + self assert: aFloat signBit = 1!
Item was added: + ----- Method: FloatTest>>assertIsPositiveZero: (in category 'accessing') ----- + assertIsPositiveZero: aFloat + "Assert that aFloat is Float zero (the positive one)" + self assert: aFloat = 0.0. + self assert: aFloat signBit = 0!
Item was changed: ----- Method: FloatTest>>testArSinh (in category 'tests - mathematical functions') ----- testArSinh + self assertIsPositiveZero: 0.0 arSinh. + self assertIsNegativeZero: Float negativeZero arSinh. - self assert: 0.0 arSinh = 0.0. - self assert: Float negativeZero arSinh = 0.0. - self assert: Float negativeZero arSinh sign = -1. self deny: Float infinity arSinh isFinite. self assert: (0.5 arSinh negated closeTo: 0.5 negated arSinh). self assert: (0.5 arSinh sinh closeTo: 0.5). self assert: (-2.5 arSinh sinh closeTo: -2.5).!
Item was changed: ----- Method: FloatTest>>testArTanh (in category 'tests - mathematical functions') ----- testArTanh + self assertIsPositiveZero: 0.0 arTanh. + self assertIsNegativeZero: Float negativeZero arTanh. - self assert: 0.0 arTanh = 0.0. - self assert: Float negativeZero arTanh = 0.0. - self assert: Float negativeZero arTanh sign = -1. self deny: 1 arTanh isFinite. self assert: (0.5 arTanh negated closeTo: 0.5 negated arTanh). self assert: (0.5 arTanh tanh closeTo: 0.5). self assert: (-0.5 arTanh tanh closeTo: -0.5).!
Item was changed: ----- Method: FloatTest>>testCharacterization (in category 'characterization') ----- testCharacterization
"Test the largest finite representable floating point value" self assert: Float fmax successor = Float infinity. self assert: Float infinity predecessor = Float fmax. self assert: Float fmax negated predecessor = Float negativeInfinity. self assert: Float negativeInfinity successor = Float fmax negated. "Test the smallest positive representable floating point value" self assert: Float fmin predecessor = 0.0. self assert: 0.0 successor = Float fmin. self assert: Float fmin negated successor = 0.0. self assert: 0.0 predecessor = Float fmin negated. "Test the relative precision" self assert: Float one + Float epsilon > Float one. self assert: Float one + Float epsilon = Float one successor. self assert: Float one + (Float epsilon / Float radix) = Float one. "Test maximum and minimum exponent" self assert: Float fmax exponent = Float emax. self assert: Float fminNormalized exponent = Float emin. Float denormalized ifTrue: [ self assert: Float fminDenormalized exponent = (Float emin + 1 - Float precision)]. "Alternative tests for maximum and minimum" self assert: (Float radix - Float epsilon) * (Float radix raisedTo: Float emax) = Float fmax. self assert: Float epsilon * (Float radix raisedTo: Float emin) = Float fmin. "Test sucessors and predecessors" self assert: Float one predecessor successor = Float one. self assert: Float one successor predecessor = Float one. self assert: Float one negated predecessor successor = Float one negated. self assert: Float one negated successor predecessor = Float one negated. self assert: Float infinity successor = Float infinity. self assert: Float negativeInfinity predecessor = Float negativeInfinity. + self assertIsNegativeZero: Float fmin negated successor. - self assert: Float fmin negated successor = 0.0. - self assert: Float fmin negated successor sign = -1. self assert: Float nan predecessor isNaN. self assert: Float nan successor isNaN. "SPECIFIC FOR IEEE 754 double precision - 64 bits" self assert: Float fmax hex = '7FEFFFFFFFFFFFFF'. self assert: Float fminDenormalized hex = '0000000000000001'. self assert: Float fminNormalized hex = '0010000000000000'. self assert: 0.0 hex = '0000000000000000'. self assert: Float negativeZero hex = '8000000000000000'. self assert: Float one hex = '3FF0000000000000'. self assert: Float infinity hex = '7FF0000000000000'. self assert: Float negativeInfinity hex = 'FFF0000000000000'.!
Item was changed: ----- Method: FloatTest>>testCopySign (in category 'zero behavior') ----- testCopySign self assert: (0.0 copySignTo: 1) = 1. self assert: (Float negativeZero copySignTo: 1) = -1. + self assertIsNegativeZero: (-1 copySignTo: 0.0). + self assertIsPositiveZero: (1 copySignTo: Float negativeZero).! - self assert: (-1 copySignTo: 0.0) = 0.0. - self assert: (-1 copySignTo: 0.0) sign = -1. - self assert: (1 copySignTo: Float negativeZero) sign = 0.!
Item was changed: ----- Method: FloatTest>>testFractionAsFloatWithUnderflow (in category 'testing - conversion') ----- testFractionAsFloatWithUnderflow "test rounding to nearest even"
| underflowPower | underflowPower := Float emin - Float precision. self assert: (2 raisedTo: underflowPower) asFloat = 0.0. self assert: (2 raisedTo: underflowPower) negated asFloat = 0.0. + self assert: (2 raisedTo: underflowPower) negated asFloat signBit = 1 description: 'a negative underflow should return a negative zero'.! - self assert: (2 raisedTo: underflowPower) negated asFloat sign = -1 description: 'a negative underflow should return a negative zero'.!
Item was changed: ----- Method: FloatTest>>testLiteralEqualityOfZeroAndNegativeZero (in category 'testing compare') ----- testLiteralEqualityOfZeroAndNegativeZero + self assert: 1 + equals: (Compiler evaluate: '1>2 ifTrue: [0.0] ifFalse: [-0.0]') signBit - self assert: -1 - equals: (Compiler evaluate: '1>2 ifTrue: [0.0] ifFalse: [-0.0]') sign description: 'Float zero and negativeZero are not literally substituable'.!
Item was changed: ----- Method: FloatTest>>testNegativeZeroAbs (in category 'zero behavior') ----- testNegativeZeroAbs + self assert: Float negativeZero abs signBit = 0 description: 'the absolute value of a negative zero is zero'! - self assert: Float negativeZero abs sign positive description: 'the absolute value of a negative zero is zero'!
Item was changed: ----- Method: FloatTest>>testNegativeZeroSign (in category 'zero behavior') ----- testNegativeZeroSign + self assert: Float negativeZero sign = 0. + self assert: Float negativeZero signBit = 1 "That's how we can distinguish from positive zero"! - self assert: Float negativeZero sign = -1!
Item was changed: ----- Method: FloatTest>>testSign (in category 'testing') ----- testSign
"Set up" | negatives negz positives strictNegatives strictPositives zero | strictPositives := {2. 2.5. Float infinity}. strictNegatives := {-3. -3.25. Float negativeInfinity}. zero := 0.0. negz := Float negativeZero. positives := strictPositives copyWith: zero. negatives := strictNegatives copyWith: negz. "The sign of non zeros" strictPositives do: [:aPositive | self assert: aPositive sign = 1]. strictNegatives do: [:aNegative | self assert: aNegative sign = -1]. "The sign of zeros" self assert: zero sign = 0. + self assert: negz sign = 0. "remark that negz can't be distinguished from zero and is thus considered positive..." + self assert: negz signBit = 1. "but we can differentiate" - self assert: negz sign = -1. "remark though that negz >= 0.0, and is thus considered positive... Weird" "Test the copy sign functions" positives do: [:aPositiveSign | positives do: [:aPositive | self assert: (aPositive sign: aPositiveSign) = aPositive]. negatives do: [:aNegative | self assert: (aNegative sign: aPositiveSign) = aNegative negated]. + (zero sign: aPositiveSign) signBit = 0. + (negz sign: aPositiveSign) signBit = 0]. - (zero sign: aPositiveSign) sign = 0. - (negz sign: aPositiveSign) sign = 0]. negatives do: [:aNegativeSign | positives do: [:aPositive | self assert: (aPositive sign: aNegativeSign) = aPositive negated]. negatives do: [:aNegative | self assert: (aNegative sign: aNegativeSign) = aNegative]. + (zero sign: aNegativeSign) signBit = 1. + (negz sign: aNegativeSign) signBit = 1].! - (zero sign: aNegativeSign) sign = -1. - (negz sign: aNegativeSign) sign = -1].!
Item was changed: ----- Method: FloatTest>>testSinh (in category 'tests - mathematical functions') ----- testSinh + self assertIsPositiveZero: 0.0 sinh. + self assertIsNegativeZero: Float negativeZero sinh. - self assert: 0.0 sinh = 0.0. - self assert: Float negativeZero sinh = 0.0. - self assert: Float negativeZero sinh sign = -1. self deny: Float infinity sinh isFinite. self assert: (2.0 cosh squared - 2.0 sinh squared closeTo: 1). self assert: (2.0 sinh negated closeTo: 2.0 negated sinh).!
Item was changed: ----- Method: FloatTest>>testTanh (in category 'tests - mathematical functions') ----- testTanh + self assertIsPositiveZero: 0.0 tanh. + self assertIsNegativeZero: Float negativeZero tanh. - self assert: 0.0 tanh = 0.0. - self assert: Float negativeZero tanh = 0.0. - self assert: Float negativeZero tanh sign = -1. self assert: (Float infinity tanh closeTo: 1). self assert: (2.0 cosh squared - 2.0 sinh squared closeTo: 1). self assert: (2.0 tanh negated closeTo: 2.0 negated tanh).!