Bug 9539

Summary: Must-be-hidden surfaces are shown
Product: Mesa Reporter: Kent Eschenberg <eschenbe>
Component: Drivers/DRI/i810Assignee: haihao <haihao.xiang>
Status: RESOLVED DUPLICATE QA Contact:
Severity: major    
Priority: medium CC: facorread, sitsofe
Version: 6.5Keywords: NEEDINFO
Hardware: Other   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Bug Depends on:    
Bug Blocks: 16029    
Attachments: Snapshots/notes

Description Kent Eschenberg 2007-01-04 10:10:37 UTC
Some triangles disappear at some viewing angles. This problem shows up with
perspective projection only, never orthogonal.

Although it looks like some triangles disappear what might really be happening
is that those triangles are covered by others that should have been behind them,
i.e., the Z buffer test is wrong. For example, triangles only seem to disappear
when the viewing angle is right AND another part of the surface is behind the
part that seems to have disappeared.

In addition I've found that the problem depends on the angle of view. For
example, the test case below is a roughly rectangular patch that has a slight
arc. The problem only shows up when the patch is viewed from near the edge and
from one particular size of the rectangle.

The application, paraview 2.4.4, when used in client/server mode, can easily
switch between real OpenGL (on my workstation) or Mesa (on the server). The
server does not have a graphics card. The image is always correct using real OpenGL.

The problem is seen with both Mesa 6.5.2 and 6.4.2. "Make" was run twice for
each, with configuratons "linux" and "osmesa".

The system has 16 dual-core Itanium2 processors. The application, however, is
serial and uses only 1 processor. The OS is Suse SLES8.

Example files related to this bug can be downloaded via anonymous ftp from
ftp.psc.edu under pub/outgoing/eschenbe/Mesa. Here's a summary:

mesaBug.mpg - a short video showing the triangles coming and going as the
surface is rotated

mesaBug.vtk - The smallest subset of the surface that I could make that exhibits
the problem; in legacy VTK binary format. All other images use this data.

mesaBugAxes.png, mesaBugWire.png - The small surface with axes turned on to show
the size of the surface. Real OpenGL rendering was used because Mesa wouldn't
draw the axes (may be a ParaView problem). Also shown in wireframe mode to
better see the triangles.

mesaBugTop.png - A view from the top just to help you get oriented. No problems
from this view angle.

mesaBugEdge.png - Some triangles have disappeared.
Comment 1 Kent Eschenberg 2007-01-04 10:18:03 UTC
Typo, paragraph 3, should end with "... one particular side of the rectangle."
Comment 2 Kent Eschenberg 2007-01-04 10:33:46 UTC
The problem has been duplicated on another system that is the same as for the
original report except for the following:

Mesa 6.4.2 only
Red Hat Enterprise Linux WS release 4
Intel Xeon 64 bit processor
Comment 3 Fabio Correa 2007-01-23 16:01:46 UTC
Hello,

After watching the video, I can say that I had exactly the same problem while
running XScreenSavers like stairs and tangram. Somehow inexplicably, after
rebuilding mesa, the problem got fixed. No patches are involved in the
compilation process.

Gentoo Linux 2006.2
mesa-6.5.2.ebuild
Graphics: Intel 945 GM
Processor: Intel Pentium M 1.4GHz
Comment 4 Fabio Correa 2007-01-25 15:11:37 UTC
The problem got back.

I am afraid this problem is hard to reproduce; Some runs of any GL XScreenSaver will present it, some will not. It has to do with initialization, because I haven't seen the situation change in the middle of the program.

I do not have enough time to install paraview, so I am not of further help here.
Comment 5 Fabio Correa 2007-01-26 14:15:52 UTC
Created attachment 8510 [details]
Snapshots/notes

I found _two_ ways to reproduce the bug. Here goes the first:

1. Execute an OpenGL XScreenSaver with an invalid visual. Examples:
cube21 -visual 0 -cubesize 2
gears -visual 0 -planetary

Please see the attachment for further comment.

OS: Gentoo Linux 2006.1, Updated to Jan 26, 2007
Graphic Unit: Intel 945GM
OpenGL renderer string: Mesa DRI Intel(R) 945GM 20061017 x86/MMX/SSE2
Comment 6 Gordon Jin 2007-03-14 19:44:34 UTC
The bug priority was upgraded (P2->high) with the bugzilla configuration change.
I'm Changing the priority back to the normal one.
Sorry for the spam.
Comment 7 Sitsofe Wheeler 2007-08-04 00:56:34 UTC
Might be related to https://bugzilla.novell.com/show_bug.cgi?id=239188 .
Comment 8 Michael Fu 2008-05-03 20:02:21 UTC
(In reply to comment #5)
> Created an attachment (id=8510) [details]
> Snapshots/notes
> 
> I found _two_ ways to reproduce the bug. Here goes the first:
> 
> 1. Execute an OpenGL XScreenSaver with an invalid visual. Examples:
> cube21 -visual 0 -cubesize 2
> gears -visual 0 -planetary
> 
> Please see the attachment for further comment.
> 
> OS: Gentoo Linux 2006.1, Updated to Jan 26, 2007
> Graphic Unit: Intel 945GM
> OpenGL renderer string: Mesa DRI Intel(R) 945GM 20061017 x86/MMX/SSE2
> 

Fabio, are you still able to reproduce this on a newer environment? e.g. newer Gentoo build, latest stable Ubuntu or lastest stable Fedora.
Comment 9 Sitsofe Wheeler 2008-05-04 02:24:11 UTC
If you look at https://bugzilla.novell.com/show_bug.cgi?id=239188 you will see a report that this is still here with openSUSE 11 with Mesa 7.0.3 . I can also confirm that this bug is most certainly still here in Ubuntu 8.04 with libgl1-mesa-glx 7.0.3~rc2-1ubuntu3 on a 945 card in a Thinkpad T60. It is reproducible every time.
Comment 10 haihao 2008-06-23 22:20:27 UTC
This bug should be a dup of 16072, although 16072 is reported against 965.


*** This bug has been marked as a duplicate of bug 16072 ***

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.