Re: bug in dominance algorithm

[Jan Hubicka <jh at suse dot cz>]

> Hi,
> 
> On Fri, 15 Jun 2001, Jan Hubicka wrote:
> 
> > > 2) B is the entry to a noreturn section (i.e. B is postdominated by a
> > >    noreturn block, without being one itself), for this dominance info in
> > >    the above way can be used.
> > This is what I am shooting for!
> > Thanks a ton!
> 
> Hmm, below is a patch (only the essential parts) to do something like
> this.  It essentially links all successorless nodes to EXIT.  Tell me, if
> it works like you want (it bootstraps without regressions on x86).
As Richard mentioned, this way we will get each successorless node postdominated
by EXIT, that is also not nice.
But wait... do we have postdominance of EXIT in datastructure at all?
Hmm, not, so it is OK.
I see, rest of your email has similar ideas to mines when I was thinking
about the issue. I will try to summarize my ideas later.

Other problem can be with infinite loops.  It can be nice to have "sane
postdominance" there too.

But for my problem, this change should work perfectly.  I will give it a try
after exam. Thanks!

Honza