Tests for powm error

[kshyatt]

This little throw and the pass-through for LowerTriangular are not tested.

[dkarrasch]

Wow, this is an amazing example of what testing is good for. The tests currently fail because powm[!] is not exported. BUT, adding a LinearAlgebra specifier still keeps both tests failing, because

1. julia/stdlib/LinearAlgebra/src/triangular.jl

   Line 2172 in 3935491

   ArgumentError("p must be a real number in (-1,1), got $p")

   requires a throw before the ArgumentError, otherwise it simply doesn't throw at all;

2. julia/stdlib/LinearAlgebra/src/triangular.jl

   Line 2231 in 3935491

   powm(A::LowerTriangular, p::Real) = copy(transpose(powm(copy(transpose(A)), p::Real)))

   should be

powm(A::LowerTriangular, p::Real) = copy(transpose(powm!(copy(transpose(A)), p::Real)))

because otherwise it throws saying there's no method matching powm(::UpperTriangular{Complex{Float64},Array{Complex{Float64},2}}, ::Float64). Sure enough, we don't have such a method, only one with !.

@kshyatt Would you be willing to expand this PR to a "bugfix" one, or do you prefer to to do that elsewhere?

[kshyatt]

I'll expand it. Shows what you get when you let CI run the tests for you.

…

On Mar 13, 2020, at 1:52 PM, Daniel Karrasch ***@***.***> wrote:  Wow, this is an amazing example of what testing is good for. The tests currently fail because powm[!] is not exported. BUT, adding a LinearAlgebra specifier still keeps both tests failing, because https://github.com/JuliaLang/julia/blob/3935491dc5028f23587ca8df6d4e0de75cf50e53/stdlib/LinearAlgebra/src/triangular.jl#L2172 requires a throw before the ArgumentError, otherwise it simply doesn't throw at all; https://github.com/JuliaLang/julia/blob/3935491dc5028f23587ca8df6d4e0de75cf50e53/stdlib/LinearAlgebra/src/triangular.jl#L2231 should be powm(A::LowerTriangular, p::Real) = copy(transpose(powm!(copy(transpose(A)), p::Real))) because otherwise it throws saying there's no method matching powm(::UpperTriangular{Complex{Float64},Array{Complex{Float64},2}}, ::Float64). Sure enough, we don't have such a method, only one with !. @kshyatt Would you be willing to expand this PR to a "bugfix" one, or do you prefer to to do that elsewhere? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.