ch-math seems broken on big-endian systems?

[RAOF]

Or at least the tests? ch-self-test.c has a bunch of tests of the form ch_double_to_packed_float(3.1415, &pf) followed by pulling out the low 16 bits and high 16 bits of the ChPackedFloat and then expecting them to have certain values. I don't think this works on big-endian systems.

Empirically, I'm now running these tests in the Debian packaging and they're failing on big-endian systems.