Error during perform isposdef on own Matrix type at Julia 1.6

[AnzhiZhang]

Version

The julia is installed via winget, and 1.6 installed by juliaup.

julia> versioninfo()
Julia Version 1.6.7
Commit 3b76b25b64 (2022-07-19 15:11 UTC)
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: 11th Gen Intel(R) Core(TM) i5-11320H @ 3.20GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.1 (ORCJIT, tigerlake)

MRE

using LinearAlgebra
import Base: getindex, size

struct TestMatrix{T} <: AbstractMatrix{T}
    n::Int
end

size(A::TestMatrix) = (A.n, A.n)

getindex(A::TestMatrix{T}, i::Int, j::Int) where {T} = T(i + j)

m = TestMatrix{Int}(5)

isposdef(m)

Output

julia> isposdef(m)
ERROR: MethodError: no method matching cholesky(::Hermitian{Int64, TestMatrix{Int64}}; check=false)
Closest candidates are:
  cholesky(::Number) at C:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.6\LinearAlgebra\src\cholesky.jl:406 got unsupported keyword argument "check"
  cholesky(::Number, ::Symbol) at C:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.6\LinearAlgebra\src\cholesky.jl:406 got unsupported keyword argument "check"    
  cholesky(::Diagonal) at C:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.6\LinearAlgebra\src\diagonal.jl:725 got unsupported keyword argument "check"
  ...
Stacktrace:
 [1] isposdef(A::TestMatrix{Int64})
   @ LinearAlgebra C:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.6\LinearAlgebra\src\dense.jl:91
 [2] top-level scope
   @ REPL[19]:1

Related Code

1.10

julia> @which cholesky(Hermitian(TestMatrix{Int}(5)))
cholesky(A::AbstractMatrix; ...)
     @ LinearAlgebra ~\.julia\juliaup\julia-1.10.4+0.x64.w64.mingw32\share\julia\stdlib\v1.10\LinearAlgebra\src\cholesky.jl:401

cholesky(A::AbstractMatrix, ::RowMaximum; tol = 0.0, check::Bool = true) =
    cholesky!(cholcopy(A), RowMaximum(); tol, check)
@deprecate cholesky(A::Union{StridedMatrix,RealHermSymComplexHerm{<:Real,<:StridedMatrix}}, ::Val{true}; tol = 0.0, check::Bool = true) cholesky(A, RowMaximum(); tol, check) false

1.6

julia> @which cholesky(Hermitian(TestMatrix{Int}(5)))
ERROR: no unique matching method found for the specified argument types
Stacktrace:
 [1] error(s::String)
   @ Base .\error.jl:33
 [2] which(tt::Type)
   @ Base .\reflection.jl:1249
 [3] which(f::Any, t::Any)
   @ Base .\reflection.jl:1238
 [4] top-level scope
   @ REPL[21]:1

julia> @which cholesky(Matrix(Hermitian(TestMatrix{Int}(5))))
cholesky(A::Union{Union{Hermitian{Complex{var"#s809"}, var"#s808"}, Hermitian{var"#s809", var"#s808"}, Symmetric{var"#s809", var"#s808"}} where {var"#s809"<:Real, var"#s808"<:(StridedMatrix{T} where T)}, StridedMatrix{T} where T}) in LinearAlgebra at C:\Users\andya\.julia\juliaup\julia-1.6.7+0.x64.w64.mingw32\share\julia\stdlib\v1.6\LinearAlgebra\src\cholesky.jl:378

cholesky(A::Union{StridedMatrix,RealHermSymComplexHerm{<:Real,<:StridedMatrix}},
    ::Val{false}=Val(false); check::Bool = true) = cholesky!(cholcopy(A); check = check)

There is no method accepts AbstractMatrix in Julia 1.6, so this will raise the above error.

[dkarrasch]

Yes, relaxing many LinearAlgebra methods to AbstractMatrix was a big step somewhere between v1.6 and v1.10. The necessary changes were non-trivial, and very likely built on progressive changes. So I'd say that this is definitely not going to be fixed (backported to) in v1.6.x. Recommendation: upgrade to v1.10+.