On Thu, 8 Mar 2012, Martin wrote:
Further more:
the program below, does not crash in delphi (turbo delphi)
I also noticed. I suspect the problem is the owner;
If it gets a free notification, it should pass it to all the children.
Michael.
On 08/03/2012 09:56, Martin wrote:
I found a behaviour, of which I am not sure, if it is intended. Maybe
someone can comment
My understanding is that FreeNotification are *always* set up
bi-directional ?
At least
C1.FreeNotification(C2);
will set up for C1 to inform C2, and for C2 to inform C1 whichever is
destroyed first to call the other
And
C1.RemoveFreeNotification(C2);
will remove both.
C2.Notification(C1, opRemove);
breaks that.
It will lead to C2 removing C1 from it's list of FreeNotifies, but C1 will
keep C2 on the list.
If then C2 is destroyed, it does not inform C1 about that, and C1 ends up
with a dangling pointer to the no longer existing C2.
Destroying C2 will then crash.
In the example below (compile with -gt) opRemove is caused by
Owner.RemoveComponent.
It crashes on the last line.
I was briefly wondering, if in case C1 did call "C1.FreeNotification(c2);",
then C1 should overwrite Notification and check for the opRemove and call
"RemoveFreeNotification" itself?
But if that was the case, then why does TComponent interfer with the
FreeNotify list at all?
program Project1;
{$mode objfpc}{$H+}
uses Classes;
var
Owner, C1, C2: TComponent;
begin
Owner := TComponent.Create(nil);
C1 := TComponent.Create(Owner);
C2 := TComponent.Create(Owner);
C1.FreeNotification(c2);
Owner.RemoveComponent(C1); // C2 Gets on opRemove
C2.Free;
C1.Free; // crash
ReadLn;
end.
_______________________________________________
fpc-devel maillist - fpc-devel@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-devel
_______________________________________________
fpc-devel maillist - fpc-devel@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-devel
_______________________________________________
fpc-devel maillist - fpc-devel@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-devel
