Menu
▾
▴
#232 runway lights not rendered on ATI graphics
Originally created by: g...@av8n.com
Originally owned by: g...@av8n.com
Just now I visited three different airports at night.
I saw no runway lights at any of them.
This is wildly unrealistic.
This is a major regression relative to previous releases.
This was observed on branch "next" commit 92be000dbdfdffc0
Date: Wed Dec 29 23:58:18 2010 +0100
Related
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: aeitsch...@yahoo.de
Wow- what a super helpfull bugreport....*rolleyes*
How about adding:
-airport name
-your system specs like GPU, RAM, processor...
-your OS
Otherwise this report can't be accepted
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: g...@av8n.com
Wow, that's not exactly a super-helpful response/comment.
You could try setting a good example. In particular, have you tried it?
Do *you* observe runway-lights anywhere? If so, please say so.
Until I hear otherwise, I will assume this bug is 100% reproducible,
at all airports, with all GPUs, all RAMs, all processors, all OSs,
et cetera.
See attached file.
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: g...@av8n.com
Also FWIW: The absence of runway lights is observed no matter what
settings are applied (via menu --> view --> rendering options) for
the following items:
-- use point sprites for runway lights
-- enhanced runway lighting
-- runway light distance attenuation
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: aeitsch...@yahoo.de
I have runway lights, with GIT from yesterday.
It is known that some GPU's especially ATI has problem to render them.
As this is more something for Tim Moore and he needs more information, I recommend you again:
-airport name
-your system specs like GPU, RAM, processor...
-your OS
Otherwise this report can't be accepted.
This are btw. basic rules, when reporting bugs!
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: aeitsch...@yahoo.de
Btw. If you have updated your data to match the latest GIT you can't change the settings via menu -->view-->rendering options anymore!
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: tommi... (code.google.com)@gmail.com
I have the same problem. I noticed there are no runway lights in every airport I tried. For example it was RJTT.
I have ATI Mobility Radeon HD 3470, 3 GB RAM, Intel Core 2 Duo P8700 2,53 GHz
Win 7 Home Premium 32-bit
I use GIT build #106 (2010-12-31 05:02:20)
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: aeitsch...@yahoo.de
timmik26:
did you had before? Like in 2.0.0 or earlier builds?
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: tommi... (code.google.com)@gmail.com
Yes, I have still that problem since 2.0.0.
Sincerly, I don't remember whether was it ok in 1.9.x, but probably it worked well.
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: aeitsch...@yahoo.de
FYI Tim Moore
Labels: OSG Rendering ATi
Status: Accepted
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
(No comment was entered for this change.)
Summary: runway lights not rendered on ATI graphics
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: robertw... (code.google.com)@gmail.com
I can reproduce this with ATI Radeon x1300 in Windows XP.
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
Apparently caused by an issue with the ATI driver not supporting point sprites. There is a hack in the forum (from Jester) removing the use of point sprites completely, which is said to revive runway lights for ATI graphics.
http://www.flightgear.org/forums/viewtopic.php?f=37&t=7138
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: dogg... (code.google.com)@gmail.com
Same issue here.
I tested this on my radeon 6850 with git version from 19. Feb.
It seems like the lights depend on the view angle. Tested at EDDL.
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: gijsrooy
(No comment was entered for this change.)
Labels: Priority-High
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
By popular demand: apparently runway lights aren't working on (some/many) ATI graphics cards. The hack (see comment #13) removes the use of point sprites, which seems to help with ATI graphics. It's trivial to convert that patch to use a command-line parameter (or maybe even run-time property) to enable/disable the use of point sprites in pt_lights.cxx. Maybe we could even detect unsupported graphics cards/drivers and disable the sprites automatically...
I'd vote for adding that patch along with an fgfs command-line switch to the 2.4.0 branch - so ATI users can also fly at night time...
@Torsten: agreed?
@Tim/Mathias: any idea why point sprites aren't working - or what exactly is affected?
Labels: -Priority-High Priority-Immediate
Cc: timoor...@gmail.com cumuluni...@gmail.com Mathias....@googlemail.com
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
See attachment: suggested (dirty) hack to optionally disable sprites (--disable-point-sprites).
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: ad... (code.google.com)@truthsolo.net
Great idea, beats having two sets of binaries! I patched and rebuilt; I also patched FGRun to include this option, but fgfs.exe returns "Unknown option --disable-point-sprites".
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: cumuluni... (code.google.com)@gmail.com
When we discussed into which branch to put a patch, we agreed on putting ugly hacks into the release branch and not backport them but create a better solution for the next release. Now, that's a candidate for that rule ;-)
For 2.6.0, I'd prefer one of these (or both) options:
- Have a property (and checkbox in rendering dialog) for use-point-sprites
- Check if GL_ARB_point_sprite and GL_ARB_point_parameters are supported extensions and let's hope, ATI tells the truth about it's abilities.
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
A fix is pushed. It resurrects the "point sprites for runway lights" switch:
"/sim/rendering/point-sprites" or use switch in View / Rendering dialog. Requires restart when changed.
Feature has already been there, but was probably disabled when ported to OSG. The actual graphics driver capability check is still disabled though - since it requires a graphics "context". Mathias may know more how to enable this (and how do we check capabilities when multiple graphics cards/drivers are used?).
And I've pushed this to all branches after all - since the feature had been there before, so it's not _that_ much of a hack. ;-)
http://www.gitorious.org/fg/fgdata/commit/bb191bc70f24f95445142ce97a9186126154f0e8
http://www.gitorious.org/fg/flightgear/commit/b0a4299d466005c8182e805ed6e9c5e57ddb4f3c
http://www.gitorious.org/fg/simgear/commit/b53c53d9d81cf8a8a5115e52c09086ceeb9a720f
Status: Testing
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: ad... (code.google.com)@truthsolo.net
Latest patches work on ATI HD5830. Cheers!
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
(No comment was entered for this change.)
Labels: Milestone-2.4.0
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: Mathias....@googlemail.com
Hi,
As I remember right, the SGSceneFeatures class was written in the transition to osg.
The problem with this was, in the first implementation that it asks for extensions in the main thread. This works in the single threaded case but does not work well in the multithreaded case. I do not remember to remove this feature myself from git. But I expect that point sprites - which are available on every hardware, even that old so that you can buy that old thing only on ebay - were just enabled unconditionally to get rid or one of these single threaded extension checks.
The truth of or point sprites for the lights is that we render triangles in point mode to get directional lights by the front and back faces of these triangles. On to of that we enable point sprites. In this combination, I guess we are the only application on the earth that tries to do this. Any I believe tha OpenGL standard does not guarantee this to work.
Even if this does not work with fglrx, I can see that working with the open source driver stack. So the hardware as such must be able to do that. But I can well understand that the driver is not tested on this codepath ...
Greetings
Mathias
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
Ok, thanks for clearing that up, Mathias!
I guess our work-around (providing a purely manual switch) could be the permanent solution then. Detecting hardware/driver capabilities probably wouldn't even help. So we could close this issue for good...
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: bre... (code.google.com)@gmail.com
Heard from several ATI users that it's working well for them. Since auto-detection of the sprite feature isn't an option here (see comment #24) => issue closed.
Status: Fixed
View and moderate all "codetickets Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Tickets"
Originally posted by: gijsrooy
For some reason the userarchiving of the point-sprites property doesn't seem to work.? I get this lighting issue on my new laptop (NVidia GT 540M), and had to use a --prop in order to disable the point sprites (I took me several times unchecking the Rendering Options checkbox and restarting several times, before I found out why it didn't work :P)
I'm clueless though, why FG doesn't save my setting...