Failure: Test: scalbln (-0X1.5555555555556P-2, 4294966274) == -0X0.5555555555556P-1022 Result: is: -inf -inf should be: -7.41691286169067457074e-309 -0x0.55555555555560000000p-1022
Suspended until somebody comes up with a patch.
scalbln takes a "long" second argument, which on x86_64 is a signed 64-bit value. When that is 4294966274, a large positive value, a result of -inf as reported here is the correct result. If you pass (int)4294966274, a negative value, then you do get -0X0.5555555555556P-1022 as expected. So there is no bug here.