LibC: Casting a large enough long double to double and printing it results in an infinite loop.

[hiperbolt]

When casting a long double that overflows the double-precision floating point exponent range (10^308) to double and trying to print it, it results in an infinite print loop:

Looking at the binary representation, it seems the casting is going very wrong:

With an exponent smaller than 308 this doesn't happen, however the lack of precision is extremely high, which might be a bug on its own:

For comparison, with glibc, casting a long double to overflowing double results in "inf" and the precision is much higher:

Maybe we should go with the glibc approach and print "inf" in these cases?