Fix tests on Julia 1.8-beta1

[rikhuijzer]

I've tested these changes on Julia 1.6, 1.7 and 1.8-beta1.

I've also changed

on:
  - push
  - pull_request

to

on:
  push:
    branches:
      - master
  pull_request:

to avoid triggering the jobs for pushes and pull requests at the same time. See, for example, https://github.com/JuliaData/CSV.jl/blob/2c02394b4ce11ee673ed1b6d0c811a98316bc2f9/.github/workflows/ci.yml#L3-L5

[rikhuijzer]

Can someone approve the workflow?

[codecov]

Codecov Report

Merging #205 (a6f6d9d) into master (de595e0) will increase coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #205      +/-   ##
==========================================
+ Coverage   93.79%   93.81%   +0.02%     
==========================================
  Files          13       13              
  Lines        2208     2265      +57     
==========================================
+ Hits         2071     2125      +54     
- Misses        137      140       +3     

Impacted Files	Coverage Δ
src/LazyArrays.jl	100.00% <ø> (ø)
src/lazybroadcasting.jl	96.20% <0.00%> (-0.90%)	⬇️
src/lazyapplying.jl	84.89% <0.00%> (-0.37%)	⬇️
src/cache.jl	97.21% <0.00%> (+0.02%)	⬆️
src/linalg/mul.jl	84.12% <0.00%> (+0.08%)	⬆️
src/lazyoperations.jl	96.71% <0.00%> (+0.19%)	⬆️
src/lazyconcat.jl	95.79% <0.00%> (+0.20%)	⬆️
src/linalg/add.jl	86.27% <0.00%> (+0.27%)	⬆️
src/lazysetoperations.jl	85.71% <0.00%> (+1.50%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update de595e0...a6f6d9d. Read the comment docs.

[dlfivefifty]

The fixes look like they are working around regressions so I don't think this should be merged as is. E.g. replacing ≡ with == in

@test apply(*, Eye(10), Ones(10)) ≡ Ones(10)

is an issue: we want it to be identically a Ones, not just have the same entries.

[rikhuijzer]

is an issue: we want it to be identically a Ones, not just have the same entries.

Do you have a suggestion on how to fix it? I think that the failures are due to the new representations for sparse matrices

EDIT: Specifically, JuliaLang/julia#42577

[dlfivefifty]

Just need to step through the debugger and see what's going wrong

[rikhuijzer]

I've added this one because currently this doesn't even hold on Julia 1.8-beta1 (due to JuliaArrays/FillArrays.jl#172). Testing this may save a lot of time in the future when trying to figure out why n .+ v !== v .+ n.

[rikhuijzer]

Tests are still broken. Patch for FillArrays is at JuliaArrays/FillArrays.jl#172.

[rikhuijzer]

Awesome! Thanks for fixing this and MatrixFactorizations, @dlfivefifty