Fix profile updating in tailcall

[Richard Guenther]

Does this fix profiledbootstrap now?

Thanks,
Richard.

On 7/27/05, Jan Hubicka <jh@suse.cz> wrote:
> Hi,
> tailcall reduces number of invocations of function but forgets to update
> profile.  Interestingly this causes code estimating number of iterations
> to always return 1 iteration that results in newly formed loop beeing
> peeled instead of unrolled and cause us to lose relative to ICC on very
> important ackerman function benchmark ;).  Fixed thus.  I've also took
> oppurtunity to fix cleanup of dump files as pointed out by Janis ages
> ago.