`]test Package` errors, but `include(test_script)` succeeds

General Usage
1

When testing a package (FlashWeave.jl) via ]test FlashWeave, I’m getting errors related to internal matrix computations (only reliant on SparseArrays.jl), but if I instead manually include(<test_script>) the same test script and environment (default v1.9 env), it runs without errors, even though I’m using the same julia executable and package (in .julia/dev).

Curiously, if I make changes to FlashWeave and ]precompile, a subsequent include(<test_script>) runs as expected without precompilation, but ]test FlashWeave triggers a new round of precompilation (before running and then producing the error above).

Any idea what could be causing this? I have a hunch that this is related to the testing environment, even though package versions should be identical to my default v1.9 environment if I understand correctly.

Julia version:

julia> versioninfo()
Julia Version 1.9.1
Commit 147bdf428cd (2023-06-07 08:27 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin22.4.0)
  CPU: 16 × Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-14.0.6 (ORCJIT, skylake)
  Threads: 1 on 16 virtual cores
2

I think it would be useful if you included outputs and error messages in the description of the problem. Without this, it’s hard to tell what the issue is.

3

As far as I understand, ] test creates a new environment with possibly new versions.

If you do include("runtests.jl") it will use the packages and versions of the activated env.

To speed-up ]test I always use

1 Like
4

Thanks everyone for the responses, sorry for not providing more detail. I think I figured it out in the meantime: the code in question indeed had a bounds error, but is wrapped in @inbounds. It seems using the ]test workflow somehow activated boundschecks, resulting in the error, while include resulted in a silent out-of-bounds access that didn’t lead to any issues (in this case). Removing @inbounds and/or fixing the bug results in consistent behaviour.

For reference, this was the function in question:

function contingency_table_2d_optim!(X::Int, Y::Int, data::SparseArrays.AbstractSparseMatrixCSC{<:Integer}, test_obj::MiTest{<:Integer, Nz})
    # Initialize a 3x3 zero matrix to hold the contingency table
    fill!(test_obj.ctab, 0)
    rvs = rowvals(data)
    nzvs = nonzeros(data)

    # Get the pointers to the start and end of the non-zero elements in each column
    ptr_X, ptr_Y = data.colptr[X], data.colptr[Y]
    ptr_X_end, ptr_Y_end = data.colptr[X + 1], data.colptr[Y + 1]
    row_X, row_Y = rvs[ptr_X], rvs[ptr_Y]

    # While there are non-zero elements remaining in either column
    @inbounds while ptr_X < ptr_X_end && ptr_Y < ptr_Y_end        
        if row_X == row_Y
            val_X, val_Y = nzvs[ptr_X] + 1, nzvs[ptr_Y] + 1
            ptr_X += 1
            ptr_Y += 1
            row_X, row_Y = rvs[ptr_X], rvs[ptr_Y] # bounds error occurred here
            test_obj.ctab[val_X, val_Y] += 1
        elseif row_X < row_Y
            ptr_X += 1
            row_X = rvs[ptr_X]
        else
            ptr_Y += 1
            row_Y = rvs[ptr_Y]
        end
    end

    return nothing
end
1 Like