Re: bug in dominance algorithm

[Jan Hubicka <jh at suse dot cz>]

> As said, it currently works in exactly this way for the postdom case
> (although I need to check to be really sure, it's some time since I wrote
> the code), except that to support the old behaviour everything is screwed
> up afterwards.  On the weekend I'll try to add another function providing
> this information without emulating the old behaviour.  But note, that you
Do we need the old behaviour somewhere?  I think it should be safe to change
it by default.

Only DCE uses the postdominance tree and it is broken on the moment.
Also I beleive that the new behaviour is what DCE want.
> still have two cases:
> The interesting edge: A--e-->B
> 1) B itself is already the noreturn block, for this case no postdominance
>    information can be used.
It can, B is postdominator of B, but not A, in case edge e comes from conditional
jump, so I can handle it in same way as I do 2).
> 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!

Honza