This is more robust but introduces a small change in behavior:
For .pcf.gz fonts, the new code calculates the hash of the uncompressed
font data whereas the original code was calculating the hash of the
compressed data.

No big deal IMO.