skip known math failures mark known failures with X and let the test pass if the error is acceptable and not expected to be fixed soon. this way the excessive math failures are easier to grep out.
more consistent feature test macros in math tests
add _DEFAULT_SOURCE wherever _BSD_SOURCE was used
math: emit less errors for known broken functions
add feature test macros to non-c99 math tests
new test system
math: fix some invalid checks (remquo, lrint, lround, y0, y1, yn)
math: add special case test vectors
math: add fenv pragma, relax the error check, simplify util.h
math: print file:line in the error msg, fix nearbyint tests
fix math test gen script
add generated math tests