FDEs for virtual thunks

[Matthew Austern]

On Monday, June 17, 2002, at 05:18 PM, Richard Henderson wrote:

> On Mon, Jun 17, 2002 at 03:29:11PM -0700, Matthew Austern wrote:
>> Is there a reason that g++ generates FDEs for virtual thunks?
>
> Because they're generated by a mechanism that does not
> guarantee that they'll be implemented as tail-calls.
> And if that doesn't happen, we have to unwind through them.

Does that imply that it would be safe to eliminate those FDEs if
we change the mechanism so as to guarantee that we will
always get a tail call?

			--Matt