Nicolas Cellier uploaded a new version of Kernel to project The Trunk:
http://source.squeak.org/trunk/Kernel-nice.1260.mcz==================== Summary ====================
Name: Kernel-nice.1260
Author: nice
Time: 28 August 2019, 6:50:57.499373 pm
UUID: f05b8d6f-4211-c84b-89b5-fec01139f4f8
Ancestors: Kernel-nice.1259
Provision for the case when we disallow mixed SmallInteger/Float arithmetic/comparison in primitives.
Since (rcvr positive) is implemented as self >= 0 (in Number), it would trigger another Float>=SmallInteger comparison, that will primitiveFail, trigger another adaptToFloat:andCompare: etc... (infinite loop)
There are other possible solutions like defining Float>>positive, but it sounds more reasonable to be very careful in this method which specifically handle primitive failure, rather than delegating such care.
=============== Diff against Kernel-nice.1259 ===============
Item was changed:
----- Method: Number>>adaptToFloat:andCompare: (in category 'converting') -----
adaptToFloat: rcvr andCompare: selector
"If I am involved in comparison with a Float, convert rcvr to a
Fraction. This way, no bit is lost and comparison is exact."
rcvr isFinite
ifFalse: [
selector == #= ifTrue: [^false].
selector == #~= ifTrue: [^true].
rcvr isNaN ifTrue: [^ false].
(selector = #< or: [selector = #'<='])
ifTrue: [^ rcvr positive not].
(selector = #> or: [selector = #'>='])
+ ifTrue: [^ rcvr >= 0.0].
- ifTrue: [^ rcvr positive].
^self error: 'unknow comparison selector'].
"Try to avoid asTrueFraction because it can cost"
self isAnExactFloat ifTrue: [^rcvr perform: selector with: self asExactFloat].
selector == #= ifTrue: [^false].
selector == #~= ifTrue: [^true].
^ rcvr asTrueFraction perform: selector with: self!