Fix CI #424
Fix CI #424
Conversation
The package doesn't have any binary dependendencies so there is not much value in testing on all platforms except sometimes detecting issues in test dependencies.
|
Hm. Do we really want to allow the GR brittleness to cause test failures for this package? |
|
Thanks, I didn't know about |
|
I tried that but I'm getting julia> @testset "plot recipes" begin
x = categorical(["B", "A", "C", "A"], levels=["C", "A", "B"])
y = categorical([10, 1, missing, 2], levels=[10, 2, 1])
res = RecipesBase.apply_recipe(Dict{Symbol, Any}(:plot_object => nothing), x, y)[1]
@test res.args[1] isa Formatted
@test res.args[1].data == [3, 2, 1, 2]
@test res.args[2] isa Formatted
@test res.args[2].data == [1, 3, 4, 2]
end
plot recipes: Error During Test at REPL[19]:1
Got exception outside of a @test
MethodError: no method matching apply_recipe(::Dict{Symbol, Any}, ::CategoricalVector{String, UInt32, String, CategoricalValue{String, UInt32}, Union{}}, ::CategoricalVector{Union{Missing, Int64}, UInt32, Int64, CategoricalValue{Int64, UInt32}, Missing})
The function `apply_recipe` exists, but no method is defined for this combination of argument types.
Closest candidates are:
apply_recipe(::AbstractDict{Symbol, Any}, ::AbstractVector{<:TimeZones.ZonedDateTime}, ::Any)
@ TimeZonesRecipesBaseExt ~/.julia/packages/RecipesBase/BRe07/src/RecipesBase.jl:296
apply_recipe(::AbstractDict{Symbol, Any}, ::Type{T}, ::T) where T<:CategoricalValue
@ CategoricalArraysRecipesBaseExt ~/.julia/packages/RecipesBase/BRe07/src/RecipesBase.jl:296
apply_recipe(::AbstractDict{Symbol, Any}, ::RecipesBase.__RecipesBasePrecompileType)
@ RecipesBase ~/.julia/packages/RecipesBase/BRe07/src/RecipesBase.jl:296if I don't add and load |
| os: | ||
| - ubuntu-latest | ||
| - macOS-latest | ||
| - windows-latest |
There was a problem hiding this comment.
I think we need the tests on 32-bit machines. Does ubuntu-latest include them?
There was a problem hiding this comment.
OK - I see it does, so I think it is OK.
|
Are you sure you loaded RecipesPipeline? I had the same error but then I found I could use that package. I've pushed a commit which works locally at least. |
|
I did not. I read it as |
|
OK. We still have a failure on 1.6.0 due to a method ambiguity in Arrow. Let's stick to target |
|
Yes. Let's just stick with 1.6 for now. Then we can switch to min once we bump the required version. |
ambiguity in Arrow on 1.6.0.
There are two separate commits here. One that adjust the versions to
min,lts,1, andpre. The first one is just a convenient way to keep CI i sync with the requirement in the projects file. The latter is in my opinion a better choice for most packages as it avoids tests failures due to unintended breakage on nightly. Instead, we'll detect the regressions only during the testing period of new releases.The second commit removes the Mac and Windows testing. I generally think there is little value to testing on these targets when the package doesn't have any direct binary dependencies. It will provide faster CI and it will also avoid an unrelated issue with Plots triggering a segfault on Windows.