Nicolas Cellier uploaded a new version of KernelTests to project The Trunk: http://source.squeak.org/trunk/KernelTests-nice.285.mcz
==================== Summary ====================
Name: KernelTests-nice.285 Author: nice Time: 26 December 2014, 3:27:41.92 pm UUID: 49e9fec1-53df-4f38-9b76-455d6196bd5f Ancestors: KernelTests-bf.284
Test that the conversion 32bits float -> 64 bits -> 32bits will preserve the original bit pattern for any single precision float but nan(s)
=============== Diff against KernelTests-bf.284 ===============
Item was added: + ----- Method: FloatTest>>test32bitConversion (in category 'tests - IEEE 754') ----- + test32bitConversion + "Except for NaN, we can convert a 32bits float to a 64bits float exactly. + Thus we can convert the 64bits float to the original 32bits float pattern." + + #(16r0 "zero" + 16r80000000 "negative zero" + 16r1 "min unormalized" + 16r12345 "a unnormalized" + 16r801FEDCB "a negative unnormalized" + 16r7FFFFF "largest unnormalized" + 16r800000 "smallest normalized" + 16r468ACDEF "a normalized float" + 16rCABD1234 "a negative normalized float" + 16r7F7FFFFF "largest finite float" + 16r7F800000 "positive infinity" + 16rFF800000 "negative infinity" + ) + do: [:originalWord | self assert: (Float fromIEEE32Bit: originalWord) asIEEE32BitWord = originalWord]!