Flux.stack causes StackOverflowError on macOS ARM

[gabrevaya]

Stacking vectors of length > 1352 causes a StackOverflowError:

julia> a = [[1] for i in 1:1352];
julia> Flux.stack(a, 2); # works

julia> a = [[1] for i in 1:1353];
julia> Flux.stack(a, 2);
ERROR: StackOverflowError:
Stacktrace:
 [1] __cat_offset!(::Matrix{Int64}, ::Tuple{Int64, Int64}, ::Tuple{Bool, Bool}, ::Tuple{Int64, Int64}, ::Matrix{Int64}, ::Matrix{Int64}, ::Vararg{Matrix{Int64}})
   @ Base ./abstractarray.jl:1746
 [2] __cat_offset!(::Matrix{Int64}, ::Tuple{Int64, Int64}, ::Tuple{Bool, Bool}, ::Tuple{Int64, Int64}, ::Matrix{Int64}, ::Matrix{Int64}, ::Vararg{Matrix{Int64}}) (repeats 1351 times)
   @ Base ./abstractarray.jl:1749
 [3] __cat(::Matrix{Int64}, ::Tuple{Int64, Int64}, ::Tuple{Bool, Bool}, ::Matrix{Int64}, ::Vararg{Matrix{Int64}})
   @ Base ./abstractarray.jl:1744
 [4] _cat_t(::Int64, ::Type{Int64}, ::Matrix{Int64}, ::Vararg{Matrix{Int64}})
   @ Base ./abstractarray.jl:1737
 [5] _cat(::Int64, ::Matrix{Int64}, ::Vararg{Matrix{Int64}})
   @ LinearAlgebra /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/LinearAlgebra/src/special.jl:424
 [6] cat(::Matrix{Int64}, ::Vararg{Matrix{Int64}}; dims::Int64)
   @ Base ./abstractarray.jl:1916
 [7] stack(xs::Vector{Vector{Int64}}; dims::Int64)
   @ MLUtils ~/.julia/packages/MLUtils/OojOS/src/utils.jl:101
 [8] stack(x::Vector{Vector{Int64}}, dims::Int64)
   @ MLUtils ./deprecated.jl:72

Moreover, if the length is sufficiently large, there is no Stacktrace included:

julia> a = [[1] for i in 1:2000];
julia> Flux.stack(a, 2);
ERROR: StackOverflowError:

(Flux_stack) pkg> st
Status `~/Documents/issues/Flux_stack/Project.toml`
  [587475ba] Flux v0.13.5

julia> versioninfo()
Julia Version 1.8.0
Commit 5544a0fab76 (2022-08-17 13:38 UTC)
Platform Info:
  OS: macOS (arm64-apple-darwin21.3.0)
  CPU: 8 × Apple M1 Pro
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, apple-m1)
  Threads: 1 on 6 virtual cores

[ToucheSir]

stack lives in https://github.com/JuliaML/MLUtils.jl/, but unfortunately I can't transfer the issue cross-org so you'll have to re-open it there (or comment on JuliaML/MLUtils.jl#119). As for a solution that works now, you could try using Base.stack (which shouldn't cause stack overflows) through Compat.jl >= 4.2.