hiebeler@turing.cs.rpi.edu (Dave Hiebeler) (04/14/89)
Well, after asking about backing_store, and playing around with
it on various machines, I have finally found this out:
in the "fix2" patches, the following 2 lines are added
to server/dix/window.c:
if (!pWin->firstChild)
return;
This is supposed to fix "core dump on no-op UnmapSubwindows";
these 2 lines are located at the beginning of UnmapSubwindows().
Now, I have tested my program which uses backing_store on a
Sun-3/50, using a version of the X11R3 server with those 2
lines, and without them. With them, backing_store does NOT
work. Without them, it DOES work fine.
2 questions:
- Under what circumstances does this core dump occurr? I don't
know what a "no-op UnmapSubwindows" is. I.e. is it
"dangerous" to run the server without that patch?
I've never gotten a core-dump while running a server
without the patch.
- Is this patch really supposed to disable backing_store? I
am not extremely familiar with X, but those 2 lines seem
to say "if you're not the first child, don't do anything",
not giving the server a chance to do backing_store. The
window I'm trying to use backing_store on is a box widget,
which is the child of a viewport widget, which is the
child of a top-level shell widget. This seems to be
hitting that return.
Again, much thanks for any help...
--
Dave Hiebeler Internet: hiebeler@cs.rpi.edu (preferred address)
Computer Science Dept. hiebeler@itsgw.rpi.edu
Amos Eaton Bldg. Bitnet: userfrzk@rpitsmts.bitnet
Rensselaer Polytechnic Institute / Troy, NY 12180-3590rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (04/14/89)
Now, I have tested my program which uses backing_store on a
Sun-3/50, using a version of the X11R3 server with those 2
lines, and without them. With them, backing_store does NOT
work. Without them, it DOES work fine.
I suppose stranger things are true, but I find this hard to believe.
- Under what circumstances does this core dump occurr?
When calling UnmapSubwindows on a window with no children.
I.e. is it "dangerous" to run the server without that patch?
Not unless you have something that calls it.
- Is this patch really supposed to disable backing_store?
No.
those 2 lines seem
to say "if you're not the first child, don't do anything",
No, they say "if the window doesn't have any children, don't do anything".
This seems to be hitting that return.
Have you actually witnessed this with a debugger?hiebeler@turing.cs.rpi.edu (Dave Hiebeler) (04/14/89)
In article <8904132030.AA00612@EXPIRE.LCS.MIT.EDU> rws@EXPO.LCS.MIT.EDU (Bob Scheifler) writes: > >> those 2 lines seem >> to say "if you're not the first child, don't do anything", > > No, they say "if the window doesn't have any children, don't do anything". > >> This seems to be hitting that return. > > Have you actually witnessed this with a debugger? I haven't witnessed this with a debugger, but I think my test was fairly conclusive: I compiled the same exact source code, both with and without those 2 lines. When the 2 lines are there, backing_store does not work (in my particular case). When the lines are not there, backing_store works fine. No other changes were made between compiles. I am enabling backing_store on the window of a box widget with no children, so it would seem logical that this return is getting executed. -- Dave Hiebeler Internet: hiebeler@cs.rpi.edu (preferred address) Computer Science Dept. hiebeler@itsgw.rpi.edu Amos Eaton Bldg. Bitnet: userfrzk@rpitsmts.bitnet Rensselaer Polytechnic Institute / Troy, NY 12180-3590
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (04/19/89)
I am enabling backing_store on the window of a box widget with
no children, so it would seem logical that this return is getting
executed.
Logical only if something calls UnmapSubwindows. I am not aware of any
such call in the Xt or Xaw code, but perhaps you call it directly in your
application. Perhaps you could send me a (small) complete application
which you believe demonstrates the problem, so that we can test it against
our server.