Closed daviehh closed 1 year ago
Base: 81.28% // Head: 0.00% // Decreases project coverage by -81.29%
:warning:
Coverage data is based on head (
3140bb0
) compared to base (d46c7fa
). Patch coverage: 0.00% of modified lines in pull request are covered.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Not sure why codecov is unable to pick up the changes, looks like the test runs ok https://github.com/JuliaMath/AppleAccelerate.jl/actions/runs/4048059274/jobs/6962842738#step:6:190
but the upload to codecov failed https://github.com/JuliaMath/AppleAccelerate.jl/actions/runs/4048059274/jobs/6962842738#step:8:46
Locally coverage seems fine
using Pkg
using Coverage
Pkg.test(; coverage = true)
coverage = process_folder()
clean_folder("src")
clean_folder("test")
covered_lines, total_lines = get_summary(coverage)
gives
(121, 151)
I believe I saw some discussion a few days ago around codecov fixes. Trying again.
@daviehh I will also give you commit access soon.
Reverts the part of the commits made in https://github.com/JuliaMath/AppleAccelerate.jl/commit/e843a3aca63b0b76e03e0fcd66322f8242e171ad where the first argument to
ccall
is changed to be a function pointer. Now it reverts back to the tuple form(function_name, library)
Should fix #51
Due to changes in how macos ships system dylibs, the original check using
isfile
would result in errors, and e843a3a was first made to change the existence check of the accelerate library usingdlopen
. In that commit, since I though it would be good to pass handles to theccall
since we already have the handles from thedlopen
check, the filessrc/Array.jl
andsrc/DSP.jl
is also changed. Now I see this might not be necessary and can actually be fragile, so the changes toccall
is reverted in this PR but we keep the change to check whether the dylib exists.