gcc miscompiles gcj @ -O2

[Richard Henderson]

On Fri, Dec 15, 2000 at 02:03:26PM +1300, Bryce McKinlay wrote:
> Actually, it is not equivalent. The first call to exit_block() is
> delayed in the "broken" version of the method.

So what?  The resulting code is still correct.  I cropped out
that function and added enough of a test harness to verify that.

Anyway, I disbelieve that this is actually the problem, since
I see these same symptoms on alpha.



r~