Since #626 (probably revealed in that CI but overlooked)
the CI has been failing on 32 bit from one of tested added in that PR
== and hash: Test Failed at /home/runner/work/ChainRulesCore.jl/ChainRulesCore.jl/test/tangent_types/structural_tangent.jl:473
Expression: hash(MutableTangent{MDemo}(; x = 1.0f0)) == hash(MutableTangent{MDemo}(; x = 1.0))
Evaluated: 0x9d88e406 == 0x848801a3
There is no reasons obvious to me why these would disagree.
It is required that values that are == (like 1.0 and 1.0f0) also hash that same
Since #626 (probably revealed in that CI but overlooked)
the CI has been failing on 32 bit from one of tested added in that PR
There is no reasons obvious to me why these would disagree.
It is required that values that are
==(like 1.0 and1.0f0) alsohashthat same