squeeze_notes vs. g++.other/loop2.C (resend)

[Richard Henderson]

On Thu, Feb 08, 2001 at 05:28:31PM +0000, Jason Merrill wrote:
> But it also seems to me that squeeze_notes shouldn't be messing with EH
> notes at all; how can it possibly be right to remove EH region notes from
> around a block of instructions?  Won't that screw up EH semantics?

Yes it does screw up EH semantics.  So the squeeze_notes change
is clearly wrong.

> What did you intend when you changed squeeze_notes to move EH notes?

It's hard to tell, but my guess is that I was attempting to prevent
one EH region from being moved out from inside a surrounding region.
I've been unable to construct a testcase for which this happens, but
there's nothing in loop that appears to check for this...

No testcase fails when reverting the change, so if the reason for the
change ever got added to the testsuite, it succeeds for different
reasons now.  I guess I'll commit the reversion and we'll wait for
something bad to happen...


r~