Just days after ndarray introduced no_std, approx did the same. As a result, users can use both no_std and approx together via --no-default-features --features approx. Currently, our CI/CD runs one test batch for --no-default-features and one test batch for --features docs, which includes approx; as a result, it misses a number of tests which are configured with feature = approx but not configured with feature = std. Examples include:
var_axis, std_axis, etc in numeric.rs
The bottommost test of serial_many_dim_serde in serialization-tests, which uses linspace (requires std) but isn't marked with feature = serde
These should be pretty simple fixes: add the appropriate cfgs and alter the CI/CD to include a test run that has --no-default-features with approx.
Just days after
ndarrayintroduced no_std,approxdid the same. As a result, users can use bothno_stdandapproxtogether via--no-default-features --features approx. Currently, our CI/CD runs one test batch for--no-default-featuresand one test batch for--features docs, which includesapprox; as a result, it misses a number of tests which are configured withfeature = approxbut not configured withfeature = std. Examples include:var_axis,std_axis, etc innumeric.rsserial_many_dim_serdeinserialization-tests, which useslinspace(requiresstd) but isn't marked withfeature = serdeThese should be pretty simple fixes: add the appropriate
cfgs and alter the CI/CD to include a test run that has--no-default-featureswithapprox.