0.13RC1r1/OpenGL fails on Geforce 8200/WinXP.

Bug #1341861 reported by gzotti
20
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Stellarium
Won't Fix
Medium
Unassigned

Bug Description

Tried 0.13RC1r1 on 2008 PC with WinXP and GeForce 8200 integrated GPU.

Mesa version works. The OpenGL version starts, but crashes after a few seconds.

Interestingly, GeForce 8200 should support OpenGL3.3 (GPU caps viewer), but startup diagnostics (logfile) only report 2.1 support.

BTW: The Atom netbook (Win7) runs MESA version at 1-2fps, the hardware-OpenGL version crashes on startup. Logfile should show an entry "Insufficient OpenGL support. (<maxOGLVersion> < 2.1). Please try to run MESA version." and the program might terminate more gracefully (?) with a similar message in a box.

Tags: crash opengl

Related branches

Revision history for this message
gzotti (georg-zotti) wrote :
Changed in stellarium:
importance: Undecided → High
Revision history for this message
Alexander Wolf (alexwolf) wrote :

I can make solution for avoiding crash of application (exit from application) if system has insufficient OpenGL support. Now question: which minimal version of OpenGL support required now? Georg, can you attach log from MESA version?

Revision history for this message
gzotti (georg-zotti) wrote :

This older PC is at home, I can add this in the evening. But Mesa support always provides the required OpenGL2.1. (From what I have read, Mesa may also be the recommended version for Virtual PCs.)

So please add in the startup checks (preudocode)

if (OpenGL_max_version < 2.1) {
   display_message("Insufficient OpenGL version. Please update drivers, graphics hardware, or use MESA version.");
   exit(1);
}

This will at least provide a graceful "expected" exit for netbooks etc. with OpenGL1.* hardware.

What I don't understand is the problem with hardware that should be OpenGL3.3 capable. It was reported in the logfile to support 2.1 (which should be enough), and then fails with some texture trouble. I have not updated drivers to the very latest version, but this should not be the problem. Wil try this however tonight.

Kind regards,
Georg

Revision history for this message
Alexander Wolf (alexwolf) wrote :

Patch to avoiding crash of application if system has insufficient OpenGL
support: http://bazaar.launchpad.net/~stellarium/stellarium/trunk/revision/6924, improvements: http://bazaar.launchpad.net/~stellarium/stellarium/trunk/revision/6925 (check version of OpenGL ES and platform).

Georg, can you check it?

Changed in stellarium:
milestone: none → 0.13.0
assignee: nobody → Alexander Wolf (alexwolf)
status: New → In Progress
Revision history for this message
gzotti (georg-zotti) wrote :

Code looks good. I hope I can check the graceful exit on my old netbook tonight. (depends on current build environment there ...)
Thanks!

Revision history for this message
gzotti (georg-zotti) wrote :

Dear Alex, I have already MSVC/Angle on my old Atom netbook. ... Trunk compiles and runs with about 8fps, although with problems (catalog3). I will check this functionality with some installable RC1r2 or RC2.

After driver update on the 2008 NVidia 8200, during startup it properly mentions openGL3.3 support. Still, the mentioned crash persists. I still have to re-install mingw build environment here.

Revision history for this message
gzotti (georg-zotti) wrote :

I have found a related bug report which hints at better diagnostics:

https://bugreports.qt-project.org/browse/QTBUG-33454?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

I set the environment variable: QT_QPA_VERBOSE=opengl:1

The next log is from an Intel Core-i3/HD3000 notebook, representing last year's 400€ notebook class. It works, but reports issues in the logfile

BTW, there is another error message not yet discussed. Any idea?

QSslSocket: cannot resolve TLSv1_1_client_method

Revision history for this message
gzotti (georg-zotti) wrote :

Again a log file, this time again from the NVidia 8200, with that environment variable set. There are differences in OpenGL diagnostics, but I don't see the crucial one. The NVidia 8200 should still be better than the IntelHD3000, but crashes :-(

I tested it also on my new Win8.1 AMD APU-1200 netbook (OpenGL4.2) All works here at least!

Revision history for this message
gzotti (georg-zotti) wrote :

And another for reference: Again the Core i3 notebook, but with the MESA version.

All works, of course slower than with hardware-OpenGL.

Revision history for this message
gzotti (georg-zotti) wrote :

another: NVidia GeForce 210 in a 2007 PC running Win7. Seems to work flawlessly. I have 2 more PCs to test (Win7/some Intel GMA and WinXP/Geforce 9600), hopefully tomorrow.

Revision history for this message
gzotti (georg-zotti) wrote :

Test on small office desktop PC: Intel CoreDuo with Intel integrated graphics GMA4500 (chipset Q43/Q45). It has OpenGL2.1 and works, but again cannot resolve some functions. In night mode, there is still the semitransparent box overlaid in the bottom left corner (image). I add log files with and without the enhanced diagnostics env.variable.

Revision history for this message
Alexander Wolf (alexwolf) wrote :

I uploaded RC2 on launchpad.net

Revision history for this message
gzotti (georg-zotti) wrote :

Thanks! Quick test on that Intel CoreDuo/GMA4500: no change. The semitransparent box in Night Mode is limited exactly by the bottom and left toolbars.

Revision history for this message
gzotti (georg-zotti) wrote :

OK, a quick test of RC2 now on my last winXPsp3 with a GeForce9600. Similar to the other XP with Geforce 8200, it crashes :-(

BTW, these crashes are not immediately after launch, but only when the moon (or maybe also other textured objects?) enters the viewport.

Attached are logs with and without that env.var. setting.

Of course, MESA runs also here, at about 3fps.

Revision history for this message
Alexander Wolf (alexwolf) wrote :

Hm. Can you use old textures and check it again?

Changed in stellarium:
assignee: Alexander Wolf (alexwolf) → nobody
Revision history for this message
gzotti (georg-zotti) wrote :

I will try at home tonight. Both failing machines are XP with NVidia. Maybe there is something missing already in the Qt/XP support?

Not the same problem, but also others have XP problems: http://qt-project.org/forums/viewthread/28578

If all else fails, we must concede XP is as dead as M$ says it is, and declare MESA the only working solution here. A pity, given otherwise sufficient hardware.

Revision history for this message
Alexander Wolf (alexwolf) wrote :

OK. Are we solve the initial issue? Because looks like we have troubles with GL context switch now.

Revision history for this message
gzotti (georg-zotti) wrote :

Hmm, the initial issue is the crash on WinXP with OpenGL capable hardware. This is not solved. The graceful exit is the small side-issue, which I can test again only on my Atom tonight. The other logfiles were for reference how well it works on currently used hardware in my reach. I expect many Intel-GMA-based complaints about the strange red box, and of course many unhappy (NVidia only or all?) users on XP (still about 25% market share?!?). I have no non-NVidia GPU's on WinXP to test this.

Revision history for this message
Alexander Wolf (alexwolf) wrote :

I have one WinXP on the work (hardware) but it contains Intel GPU with OpenGL 1.4 support only and I can use only MESA version on it.

Revision history for this message
gzotti (georg-zotti) wrote :

I tried with the old small textures on the Geforce 8200. Same crash.
The graceful exit on detected insufficient (OpenGL1.4) hardware (old netbook etc.) works, and is nicer indeed, thanks.

Changed in stellarium:
milestone: 0.13.0 → 0.13.1
Slava (iqal)
tags: added: crash
Revision history for this message
Alexander Wolf (alexwolf) wrote :

Well, looks like we have issue in shaders :(

Revision history for this message
gzotti (georg-zotti) wrote :

Possible, but not sure. All (?) failing systems were XP/Nvidias. I still have no report from any ATI/AMD Radeon.

Revision history for this message
Alexander Wolf (alexwolf) wrote :

We have first report from ATI/AMD.

Revision history for this message
gzotti (georg-zotti) wrote :

+/-?

If crash, with the same issues in the logfile?
(The other bug report with the ~2003 Radeon/OpenGL1.3 does not count here. )

Revision history for this message
Alexander Wolf (alexwolf) wrote :
gzotti (georg-zotti)
tags: added: opengl
Changed in stellarium:
milestone: 0.13.1 → 0.13.2
Changed in stellarium:
milestone: 0.13.2 → none
Revision history for this message
gzotti (georg-zotti) wrote :

I think we can declare "won't fix" on this. With Win7 on the same PC it works. Must have been an NVidia XP driver issue. XP is no longer supported.

Changed in stellarium:
importance: High → Medium
status: In Progress → Won't Fix
milestone: none → 0.14.0
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Duplicates of this bug

Other bug subscribers

Related questions

Remote bug watches

Bug watches keep track of this bug in other bug trackers.