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!