because different tests have different ideas about what it means to be
"close enough" to the right answer, depending on the properties of the
function being tested. In the process, I fixed some warnings and
added a few more 'volatile' hacks, which are sufficient to make all
the tests pass at -O2 with clang.
are workarounds for various symptoms of the problem described in clang
bugs 3929, 8100, 8241, 10409, and 12958.
The regression tests did their job: they failed, someone brought it
up on the mailing lists, and then the issue got ignored for 6 months.
Oops. There may still be some regressions for functions we don't have
test coverage for yet.
