Interpolating variables does not work anymore in v1.2.1

[ranocha]

Here is an example:

julia> using Pkg; Pkg.activate(temp=true); Pkg.add("BandedMatrices"); Pkg.add(name="BenchmarkTools", version=v"1.2.1")
[...]
  [6e4b80f9] + BenchmarkTools v1.2.1
[...]

julia> using BandedMatrices, BenchmarkTools, LinearAlgebra

julia> u = randn(3); du = zero(u); A = brand(3, 3, 1, 1);

julia> @benchmark mul!(du, A, u)
BenchmarkTools.Trial: 10000 samples with 932 evaluations.
 Range (min … max):  106.950 ns … 367.095 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     108.638 ns               ┊ GC (median):    0.00%
 Time  (mean ± σ):   113.436 ns ±  17.081 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  █▅▆▆▁              ▁                                          ▂
  █████▇▆▃▄██▇▇▇▇▅▅▆▆██▇█▇▆▇▆▆▅▇█▇▇▇▇▅▅▅▆▆▅▆▆▄▄▄▄▅▅▆▅▆▅▄▁▄▅▄▅▅▅ █
  107 ns        Histogram: log(frequency) by time        206 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark mul!($du, $A, $u)
ERROR: MethodError: convert(::Type{AbstractArray{Float64}}, ::BandedMatrix{Float64, Matrix{Float64}, Base.OneTo{Int64}}) is ambiguous. Candidates:
  convert(::Type{T}, a::T) where T<:AbstractArray in Base at abstractarray.jl:16
  convert(::Type{AbstractArray{T}}, M::BandedMatrix) where T in BandedMatrices at /home/hendrik/.julia/packages/BandedMatrices/2lafE/src/banded/BandedMatrix.jl:99
Possible fix, define
  convert(::Type{AbstractArray{T}}, ::BandedMatrix{T}) where T
Stacktrace:
 [1] setindex!(A::Vector{AbstractArray{Float64}}, x::BandedMatrix{Float64, Matrix{Float64}, Base.OneTo{Int64}}, i1::Int64)
   @ Base ./array.jl:903
 [2] copyto!(dest::Vector{AbstractArray{Float64}}, src::Tuple{Vector{Float64}, BandedMatrix{Float64, Matrix{Float64}, Base.OneTo{Int64}}, Vector{Float64}})
   @ Base ./abstractarray.jl:893
 [3] vect(::Vector{Float64}, ::Vararg{Any})
   @ Base ./array.jl:147
 [4] top-level scope
   @ ~/.julia/packages/BenchmarkTools/3UOOD/src/execution.jl:442

The expected behavior is obtained on v1.2.0:

julia> using Pkg; Pkg.activate(temp=true); Pkg.add("BandedMatrices"); Pkg.add(name="BenchmarkTools", version=v"1.2.0")
  Activating new project at `/tmp/jl_PVnuoc`
    Updating registry at `~/.julia/registries/General.toml`
   Resolving package versions...
    Updating `/tmp/jl_PVnuoc/Project.toml`
  [aae01518] + BandedMatrices v0.16.11
    Updating `/tmp/jl_PVnuoc/Manifest.toml`
  [4c555306] + ArrayLayouts v0.7.8
  [aae01518] + BandedMatrices v0.16.11
  [1a297f60] + FillArrays v0.12.7
  [56f22d72] + Artifacts
  [8f399da3] + Libdl
  [37e2e46d] + LinearAlgebra
  [9a3f8284] + Random
  [ea8e919c] + SHA
  [9e88b42a] + Serialization
  [2f01184e] + SparseArrays
  [10745b16] + Statistics
  [e66e0078] + CompilerSupportLibraries_jll
  [4536629a] + OpenBLAS_jll
  [8e850b90] + libblastrampoline_jll
   Resolving package versions...
    Updating `/tmp/jl_PVnuoc/Project.toml`
  [6e4b80f9] + BenchmarkTools v1.2.0
    Updating `/tmp/jl_PVnuoc/Manifest.toml`
  [6e4b80f9] + BenchmarkTools v1.2.0
  [682c06a0] + JSON v0.21.2
  [69de0a69] + Parsers v2.1.2
  [ade2ca70] + Dates
  [56ddb016] + Logging
  [a63ad114] + Mmap
  [de0858da] + Printf
  [9abbd945] + Profile
  [cf7118a7] + UUIDs
  [4ec0a83e] + Unicode

julia> using BandedMatrices, BenchmarkTools, LinearAlgebra

julia> u = randn(3); du = zero(u); A = brand(3, 3, 1, 1);

julia> @benchmark mul!(du, A, u)
BenchmarkTools.Trial: 10000 samples with 930 evaluations.
 Range (min … max):  116.087 ns … 232.405 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     119.637 ns               ┊ GC (median):    0.00%
 Time  (mean ± σ):   120.453 ns ±   6.769 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

   ▂▆▅▄▇█▆▃▁                                                    ▂
  ▆█████████▇▄▄▃▃▄▃▁▁▁▁▃▁▁▁▃▄▆▄▅▆▇▆▆▇▆▄▄▃▄▁▁▃▁▁▁▃▄▁▃▁▁▅▃▅▅▇▇██▇ █
  116 ns        Histogram: log(frequency) by time        151 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark mul!($du, $A, $u)
BenchmarkTools.Trial: 10000 samples with 937 evaluations.
 Range (min … max):  104.303 ns … 255.039 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     107.078 ns               ┊ GC (median):    0.00%
 Time  (mean ± σ):   108.293 ns ±   9.135 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▇▇▄█▅▂▄             ▁                   ▁                     ▂
  ███████▇▇▇▆▅▄▄▄▃▁▅▅▆█▇▇█▇▇▆▅▄▃▁▁▁▃▁▄▁▇████▆▆▃▄▁▁▃▃▃▁▁▃▁▃▅▆▅▆▅ █
  104 ns        Histogram: log(frequency) by time        156 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

[vchuravy]

@ettersi could you take a look? this is probably related to #265

[vchuravy]

@ranocha do you have an idea for a test-case, that would use only Base?

[ranocha]

Sorry, that's the minimal version of types I have found so far...