Here's a patch that fixes the test that Collin mentioned to reflect what's actually been 
happening for the last nearly 5 years, and changes _PyFloat_Pack4 and _PyFloat_Pack8, as 
follows.

When packing a float that's too large for the destination format (e.g. pack(">f", 1e39)):

 - before the patch, _PyFloat_Pack* gives an OverflowError on non-IEEE-754 platforms and
   an IEEE infinity on IEEE-754 platforms.

 - after the patch, _PyFloat_Pack* gives an IEEE infinity on all platforms.

This patch doesn't fix the problem that the cast from double to float on IEEE machines 
involves potentially undefined behaviour;  I think that should be considered a separate 
issue.