Hi Frank,

On Jul 17, 2014, at 7:48 AM, Frank Shearar <frank.shearar at gmail.com> wrote:

> On 16 July 2014 23:46,  <commits at source.squeak.org> wrote:
>> Eliot Miranda uploaded a new version of KernelTests to project The Trunk:
>> http://source.squeak.org/trunk/KernelTests-eem.274.mcz
>> 
>> ==================== Summary ====================
>> 
>> Name: KernelTests-eem.274
>> Author: eem
>> Time: 16 July 2014, 3:46:21.204 pm
>> UUID: bdd6cda6-6011-4885-b368-877c23cb305e
>> Ancestors: KernelTests-dtl.273
>> 
>> Fix compact class index tests for Spur.
>> Fix TimeTest>>testHash for hashing (i.e. Spur).
>> Add a test of the ranges of SmallInteger>>maxVal/minVal that
>> will serve to catch forgetting to change thse methods when
>> Spur 64-bits arrives.
>> 
>> =============== Diff against KernelTests-dtl.273 ===============
>> 
>> Item was added:
>> + ----- Method: IntegerTest>>testRange (in category 'tests - basic') -----
>> + testRange
>> +       self assert: SmallInteger maxVal class equals: SmallInteger.
>> +       self assert: (SmallInteger maxVal + 1) class equals: LargePositiveInteger.
>> +       self assert: SmallInteger minVal class equals: SmallInteger.
>> +       self assert: (SmallInteger minVal - 1) class equals: LargeNegativeInteger!
>> 
>> Item was changed:
>>  ----- Method: LargeNegativeIntegerTest>>testCompactClassIndex (in category 'tests') -----
>>  testCompactClassIndex
>> +       Character instSize > 0 ifTrue: "No compact classes on Spur..."
>> +               [self assert: 4 equals: LargeNegativeInteger indexIfCompact]
>> -       self assert: 4 equals: LargeNegativeInteger indexIfCompact.
>>  !
>> 
>> Item was changed:
>>  ----- Method: LargePositiveIntegerTest>>testCompactClassIndex (in category 'tests') -----
>>  testCompactClassIndex
>> +       Character instSize > 0 ifTrue: "No compact classes on Spur..."
>> +               [self assert: 5 equals: LargePositiveInteger indexIfCompact]!
>> -       self assert: 5 equals: LargePositiveInteger indexIfCompact.!
>> 
>> Item was changed:
>>  ----- Method: TimeTest>>testHash (in category 'testing') -----
>>  testHash
>> +       self assert: aTime hash = (Character instSize > 0 ifTrue: [607004] ifFalse: [101409191])!
>> -       self assert: aTime hash =  607004.!
> 
> I wouldn't mind seeing a "No compact classes on Spur..." comment here too.


Ah, but the cause is quite different. The source is the seed to String>>#hash, ByteString class>>#identityHash, which in turn affects the hash of Array species, which is based on Array name (i.e. #Array), which produces a different hash.  Just step through in the debugger and you'll see the source.  So that comment would be misleading; this has nothing to do with compact class indices.

Personally I'd like to see a constant seed to stringHash:initialHash: (if that's the method; I'm on the phone right now).

> frank


Eliot (phone)