2.95 compiler bug with optimization level > 0...

[Martin v. Loewis]

> Any help would be greatly appreciated!!!!!!!!!!!!!!!!

Without further analysis, I recommend to compile this code with
-fno-strict-aliasing, and see whether this improves anything.

If it does, there are two possibilities:
a) there is a bug in the alias analysis, or
b) the bug is not strictly conforming with standard C

Please note that option b) is more likely, since gcc now exposes a
common misunderstanding about C.

Regards,
Martin