Le 13/10/2014 18:15, Alfredo Braunstein a écrit :
The fix however seems bogus to me:
- cur.forwardPos();
+ cur.top().forwardPos();
Why bogus? Why do you want to revert it?
To me it seems wrong (for the reasons explained below). But maybe I
don't understand it, could you explain? Was your intention to make the
same change also to the other instance of cur.forwardPos() in that
function? Right now it doesn't achieve what the commit message says
and doesn't stop the crash...
If I remember correctly, I thought the change was obviously needed, but
I knew deep inside that it was not fixing the real bug. It may be that I
was wrong and that we want to go inside the inset if it can contains
changes.
I think that the functions should be rewritten from scratch :)
You forgot the attachment? ;-)
SeriousIy though, I totally agree. But for the moment let's fix it...
If nothing else for the sake of the stable branch.
Sometimes I think that it would be easier to drop all. But the fact that
I have never been brave enough to do it probably shows that it is not
that easy :)
That makes sense, but I do not have the time/energy to make sure that the
patch is correct. I cannot understand why this thing needs to be so
complicated.
IMO the exposed interface to the change tracking mechanism it's pretty
basic, and this is one reason why the code must be so complicated.
I suggest that you commit your changes.
JMarc
