[Paraview] Strange X11 error (BadDrawable) in strange circumstances

[William Sherman]

Hello ParaViewers,

I just encountered a very strange error on a new system that
I was building ParaView on.

I decided to build the just released version 4.4.0 of ParaView,
and used the version of QT that I've been using for all of the 4.x
versions of ParaView (QT 4.8.2).  Also, I'm using Python 2.6 (with
Python enabled), and MPICH2 3.1.4.  All of this is running on
RedHat Enterprise Linux release 6.7 (Santiago).

And here's some GLX info:
         % glxinfo | grep -i opengl
         OpenGL vendor string: NVIDIA Corporation
         OpenGL renderer string: Quadro K6000/PCIe/SSE2
         OpenGL version string: 4.5.0 NVIDIA 352.30
         OpenGL shading language version string: 4.50 NVIDIA
         OpenGL extensions:


So here's what I get when I run ParaView from the shell:
	% Build/bin/paraview
         X Error: BadDrawable (invalid Pixmap or Window parameter) 9
           Major opcode: 62 (X_CopyArea)
           Resource id:  0x0

I googled on that and searched recent ParaView mailing list
discussions, but nothing relevant turned up.

I then went back and compiled ParaView 4.3.1, and accidentally
discovered something bizarre: when I run ParaView from a shell
that is ssh'd into the machine (but with the DISPLAY environment
set to the local display (:0.0), ParaView Works!  But even version
4.3.1 if I run it from a local shell it gives me the "BadDrawable"
error.

And I then confirmed that ParaView 4.4.0 is giving me the same
behavior!  It will run if I do this:
	% ssh localhost
	% export DISPLAY=:0.0
	% paraview

That works!


I then started comparing environment variables and "set" variables,
and started to unset some of the environment variables in the local
shell, but none of that seemed to be working.  But it did cause me
to notice something -- the area where the splash screen would normally
appear retained the buffer of whatever was on the screen at that location
when these failed paraview executions were active -- ie. if I move
a window across that area, only the original screen content remains
visible.

So recalling there was a "--no-splash" option for ParaView, I figured
I'd try that and see what happens -- because when I run ParaView from
the remote shells, there is no splash screen, so maybe that's the issue.
Well, either my memory is wrong, or the "--no-splash" option was removed
from ParaView at some point becuase "--help" doesn't report that option.

But I did notice another option that although it says it's only for
MS-Windows I tried it out of desperation.  And so, I got the console
window appearing to tell me that the "--print-monitors" option is only
available for MS-Windows.  But!!!! somehow, because the Console window
is open, the BadDrawable error also appeared on the Console window.  And
somehow by doing that, ParaView ran!  So now ParaView is running from
having been executed on the local shell -- but only because I caused an
error message to appear before the splash screen was created (which
caused the console window to appear, and then the BadDrawable error
appears on the console window, and things seem to be working fine.

...  Until I close the console window!  So it turns out whenever I
first access one of the pulldown menus ("File", "Edit", "Sources", etc.)
that will also generate a BadDrawable error -- but only the first time,
once that menu has been seen, it works fine the rest of the time.  The
same seems to be true for pulling panels off the ParaView client (e.g
the "Animation" panel).  If the console is closed when a BadDrawable
error occurs, the ParaView hangs, and I have to kill it.


So, this points to QT as being the issue.  I don't have any other QT
applications to test against to see whether there's something in how
ParaView uses QT that is the problem.

So I wonder if anyone else has seen this behavior, and if so, has a
better solution than to run ParaView through a shell via "ssh".

I find it all rather bizarre.

	Thanks!
	Bill

--
Bill Sherman
Sr. Technology Advisor
Advanced Visualization Lab
Pervasive Technology Inst
Indiana University
shermanw at indiana.edu