apparent compilation hang in 1.9.0+rc2

[lmiq]

Maybe related to: #49321

Apparently compilation is hanging when I'm trying to use the JSON3 package in Julia 1.9.0-rc2.

The minimal script to reproduce the bug is:

import Pkg
Pkg.activate(temp=true)
Pkg.add("ComplexMixtures")
using ComplexMixtures
ComplexMixtures.load("./test.txt")

where the input file is: test.txt

I have installed 1.9.0-rc1 and 1.9.0-rc2 using juliaup, and I run:

julia +1.9.0-rc1 script.jl
#or
julia +1.9.0-rc2 script.jl

in rc2 the execution hangs (it works in 1.6, 1.8.5 and 1.9.0-rc1).

If I kill the execution I get the following stack trace: stacktrace.txt

The function in case is this one: https://github.com/m3g/ComplexMixtures.jl/blob/ea0d48880c08b6178034fa7c603ec03eefa186eb/src/results.jl#L679

but I don't think it has anything to do with the bug.

I suspect that the problem is compilation because if I add @show statements in that function, even in the first line, they don't get immediately printed.

[lmiq]

Here is a "more minimal" example.

The script hangs, as above.

it does not hang if I replace @with_kw from Paramters.jl by the base @kwdef.

Thus, this bug may be related to Parameters.jl, in association with some change in rc1.

Run the example with:

julia +1.9.0-rc2 minimal1.jl

with test.txt in the directory. The files are available in the zip: minimal1.zip

I will try to make the example shorter.

edit: tentatives to simplify the structure tend to make the test pass. I noted that for a relatively complex data structure the test passes but may take a long time (while being fast in other Julia versions than 1.9.0-rc2).

Maybe the code is not completely stalled, but just taking way too long.

[maleadt]

2a2068d2677eae3fe2101c89f9ed6c98e51f3a62 is the first bad commit
commit 2a2068d2677eae3fe2101c89f9ed6c98e51f3a62
Author: N5N3 <2642243996@qq.com>
Date:   Thu Feb 2 11:41:26 2023 +0800

    Subtype: avoid false alarm caused by eager `forall_exists_subtype`. (#48441)

ref #48441, cc @N5N3

Doesn't happen on master, so let's see if I can bisect a fix.

[maleadt]

3d0278b7b2e4cb62323ef4d5b2222737522d2690 is the first fixed commit
commit 3d0278b7b2e4cb62323ef4d5b2222737522d2690
Author: N5N3 <2642243996@qq.com>
Date:   Thu Feb 2 14:13:49 2023 +0800

    Subtype: skip more identical check for `X::Tuple <: Y::Tuple`

So back-port #48534?

[N5N3]

MWE

S = Tuple{Type{Result{Vector{Float64}}}, Union{Nothing, Int64}, Union{Nothing, Float64}, Union{Nothing, Float64}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Bool}, Union{Nothing, ComplexMixtures.SolSummary}, Union{Nothing, ComplexMixtures.SolSummary}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, Vector{Float64}}, Union{Nothing, ComplexMixtures.Density}, Union{Nothing, ComplexMixtures.Volume}, Union{Nothing, Options}, Union{Nothing, Int64}, Union{Nothing, Int64}, Union{Nothing, Int64}, Union{Nothing, Vector{String}}, Union{Nothing, Vector{Float64}}}
T = Tuple{Type{Result{T1}}, Vararg{Any, 29}} where T1
S <: T

Looks like the fake union split on T1 blocks the separatable subtyping.