fix #44705, fix tuple_tfunc on Union containing Type{...}

[aviatesk]

No description provided.

[Keno]

Why not Tuple{DataType}?

[vtjnash]

We would need a more precise idea of widenconst that does widenconcrete, and we don't have that right now

[aviatesk]

buildbot/package_xxx builds started to fail somehow, although I can't really reproduce it locally..

[aviatesk]

I encountered a weird behavior here -- without this !isvarargtype(x) check buildbot/xxx builds fail with the following error:

Assertion failed: (!jl_is_vararg(x) && !jl_is_vararg(y)), function subtype, file /Users/julia/buildbot/worker/package_macos64/build/src/subtype.c, line 1261.

(e.g. https://build.julialang.org/#/builders/34/builds/4697)
while the same commit successfully builds w/o any problem on my machine (after make cleanall or even a fresh clone).
Maybe buildbot uses a different method to build something, and in that case the bootstrap makes more use of interpreter or something?

[aviatesk]

Turned out that I needed to enable "assert build".

[aviatesk]

This causes SquadGames.jl test suite to fail.
Here is the reproducer:

function assertx(xs::Vector{T}) where T
    t = (xs[1], xs[2])
    t::Tuple{T,T}
end

@code_typed optimize=false assertx(Type{<:Number}[Int,Int])
@code_typed optimize=true assertx(Type{<:Number}[Int,Int])

diff --git a/_master.jl b/_pr.jl
index 03768efc773..54ccceaa2cc 100644
--- a/_master.jl
+++ b/_pr.jl
@@ -2,8 +2,8 @@ julia> @code_typed optimize=false assertx(Type{<:Number}[Int,Int])
 CodeInfo(
 1 ─ %1 = Base.getindex(xs, 1)::Type{<:Number}
 │   %2 = Base.getindex(xs, 2)::Type{<:Number}
-│        (t = Core.tuple(%1, %2))::Tuple{Type{<:Number}, Type{<:Number}}
-│   %4 = t::Tuple{Type{<:Number}, Type{<:Number}}
+│        (t = Core.tuple(%1, %2))::Tuple{Type, Type}
+│   %4 = t::Tuple{Type, Type}
 │   %5 = Core.apply_type(Main.Tuple, $(Expr(:static_parameter, 1)), $(Expr(:static_parameter, 1)))::Core.Const(Tuple{Type{<:Number}, Type{<:Number}})
 │   %6 = Core.typeassert(%4, %5)::Tuple{Type{<:Number}, Type{<:Number}}
 └──      return %6
@@ -13,6 +13,8 @@ julia> @code_typed optimize=true assertx(Type{<:Number}[Int,Int])
 CodeInfo(
 1 ─ %1 = Base.arrayref(true, xs, 1)::Type{<:Number}
 │   %2 = Base.arrayref(true, xs, 2)::Type{<:Number}
-│   %3 = Core.tuple(%1, %2)::Tuple{Type{<:Number}, Type{<:Number}}
-└──      return %3
+│   %3 = Core.tuple(%1, %2)::Tuple{Type, Type}
+│        Core.typeassert(%3, Tuple{Type{<:Number}, Type{<:Number}})::Tuple{Type{<:Number}, Type{<:Number}}
+│   %5 = π (%3, Tuple{Type{<:Number}, Type{<:Number}})
+└──      return %5
 ) => Tuple{Type{<:Number}, Type{<:Number}}

So it seems like this assertion error was wrongly erased before? I wonder if we want to disable constructing a container whose element type is Tuple{Type{...}}.

[aviatesk]

We discussed this behavior of Tuple type is expected and that typeassertion should not be erased. We may want to fix our isa-implementation in the future so that isa can work with Tuple whose parameter is Type though. For now, I'd like to just proceed this PR with fixing SquadGames separately.