Question on -fbounds-check

[Tim Prince]

Am I correct in believing that -fbounds-check disables certain other
(possibly buggy) optimizations, including -funroll-loops?  I notice that
bugs associated with -funroll-loops (in gcc-2.96 snapshots) disappear
when -fbounds-check is invoked.

As a slight digression, should it be possible to step code and examine
loop counters under gdb, when code  has been compiled
with -funroll-loops?  Does the answer differ between gcc and g77?

Tim Prince