Regression in 1.8.0-rc3 re collection of tuples of types

[mchitre]

On 1.8.0-rc3, I get the following error:

julia> versioninfo()
Julia Version 1.8.0-rc3
Commit 33f19bcbd25 (2022-07-13 19:10 UTC)
Platform Info:
  OS: macOS (arm64-apple-darwin21.3.0)
  CPU: 8 × Apple M1
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, apple-m1)
  Threads: 1 on 4 virtual cores

julia> abstract type X end

julia> struct Y <: X end

julia>  a = Tuple{Type{<:X}}[]
Tuple{Type{<:X}}[]

julia> push!(a, (Y,))
ERROR: TypeError: in typeassert, expected Tuple{Type{<:X}}, got a value of type Tuple{DataType}
Stacktrace:
 [1] setindex!(A::Vector{Tuple{Type{<:X}}}, x::Tuple{DataType}, i1::Int64)
   @ Base ./array.jl:966
 [2] push!(a::Vector{Tuple{Type{<:X}}}, item::Tuple{DataType})
   @ Base ./array.jl:1059
 [3] top-level scope
   @ REPL[4]:1

for code that used to work fine in 1.7.x (and even in the 1.8.x-beta versions):

julia> versioninfo()
Julia Version 1.7.2
Commit bf53498635 (2022-02-06 15:21 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin19.5.0)
  CPU: Apple M1
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-12.0.1 (ORCJIT, westmere)

julia> abstract type X end

julia> struct Y <: X end

julia> a = Tuple{Type{<:X}}[]
Tuple{Type{<:X}}[]

julia> push!(a, (Y,))
1-element Vector{Tuple{Type{<:X}}}:
 (Y,)

[jakobnissen]

Is it maybe a bugfix? On Julia 1.7.3 and 1.8-rc2, while it works, we also have

julia> first(a) isa eltype(a)
false

which is certainly confusing, and I'd argue, a bug.

[jakobnissen]

This is a bug that has been fixed. This issue explains it: #44705 . Briefly:

julia> isconcretetype(Type{Y})
false

julia> #= concrete type is... =# typeof((Int,))
Tuple{DataType}

julia> #= therefore =# (Y,) isa Tuple{Type{Y}}
false

julia> #= therefore =# (Y,) isa eltype(a)
false