Use sizeof(double) instead of hardcoding 8.
various cleanups; david says results are same
print out the bad values when the test fails
Fix pthread floatting point preemption test and enable it. Nextstep is to figure out why it fails on i386, yet works on sparc
convert everything to bsd.regress.mk
this shouldn't be commented out.
Regression tests for libc_r (pthreads) library.Thanks to pval@ for resolving the license stuff.