rhug failure on alpha

[Andrew Haley]

Tom Tromey writes:
 > 
 > Debugging shows that we are indeed generating the bit field operation
 > in fold_truthop().  I doubt I'll be fixing this for 3.1.
 > 
 > One fix might be to simply teach gcj how to undo the bit field
 > optimizations.  This would be a workaround.

This sounds rather frightening.

Andrew.