Separate implementation of sinpi / cospi that ensure that e.g. radius…
…(sinpi(0.5..0.5)) == 0. Fixes #412.
Now all tests pass, but running for instance cospi(@biginterval(0.5)) returns a StackOverflowError on my machine (Ubuntu 20.04, Julia 1.5.0).
julia> cospi(@biginterval(0.5))
ERROR: StackOverflowError:
Stacktrace:
[1] cospi(::BigFloat, ::RoundingMode{:Down}) at /home/sad/.julia/packages/CRlibm/NFCH5/src/CRlibm.jl:137 (repeats 79984 times)
Coverage decreased (-0.1%) to 90.974% when pulling 92f3d6278e2e0ffa4264c8b0fcb57f2ef5bfb7d1 on fph:sinpi into f80b6835814cfb05f3b73173d25a55b50914ff10 on JuliaIntervals:master.
This seems very weird -- the tests pass on Travis but not on Appveyor. Any idea why? Maybe it is because one of the two environments is using MPFR fallback mode and the other isn't?
I'm not sure what happened but there are now lots of other commits included in this PR. Could you please rebase against master?
Yes, sorry, I was trying to rebase today but I must have done something wrong with git. I will get the PR in order and then get back to you soon.