Rendering of buildings broken under r300g/Mobility FireGL V5250 (poor driver/hardware support?) about openage

[TheJJ]

Probably the shader compilation fails. Can you try to get some error messages? Mesa supports debug environment variables: http://www.mesa3d.org/envvars.html

[oldlaptop]

I tried setting LIBGL_DEBUG and MESA_DEBUG to "verbose", nothing of real relevance:

libGL: screen 0 does not appear to be DRI3 capable
libGL: pci id for fd 8: 1002:71d4, driver r300
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/tls/r300_dri.so
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/r300_dri.so
libGL: Can't open configuration file ~/.drirc: No such file or directory.
libGL: Can't open configuration file ~/.drirc: No such file or directory.
<lots of openage's spew about loading textures>
<much further down, the same message I reported earlier>

I do see a few messages from openage about '[asset file] not present, ignoring...'; it looks to me like most/all buildings are being loaded correctly though. If you are interested in the full 4 kiloline log, I can post it somewhere.

[TheJJ]

Are you able to get any more info about the r300 FP: Compiler Error you posted first? If we can figure out why the compilation fails, we can probably fix this for radeon cards.

[oldlaptop]

After some discussion on IRC (thanks especially to mareko in #radeon), the conclusion seems to be that the shader in question (https://github.com/SFTtech/openage/blob/master/assets/shaders/teamcolors.frag.glsl) is more than this particular hardware can handle (although in the process it was discovered that the shader compiler was unrolling loops where it shouldn't). Specifically the hardware doesn't support indexing arrays with non-constant values, and the compiler has to work around that (which gets too wordy with big arrays). mareko did have a suggestion for optimizing the shader for this hardware, though:

[16:22:58] <mareko> oldlaptop: it's definitely indirect addressing, sorry, that shader cannot run on R500, the solution is to set the player_color array as a 1D texture and use texture1D() to read it, that should work
[16:24:08] <mareko> oldlaptop: that's the only way to do indexing to a huge read-only array with a non-constant expression in a fragment shder

[LeviSchuck]

Is the solution to make it a compatible shader (and not taking advantage of current hardware?) or to have shader overrides for certain chipsets that fulfill the basic function?

[jamesmintram]

Gonna look into this as part of #34

[franciscod]

are the shaders only being used for painting buildings with the player's color? see #112

[TheJJ]

is this still an issue? mesa should have advanced a lot since back then.

[TheJJ]

stil broken, but at some other place now:

    INFO Loading time  [data]: 476.962 s
    Traceback (most recent call last):
      File "run.py", line 15, in init run (/home/peter/projects/openage/run.cpp:810)
        main()
      File "/home/peter/projects/openage/openage/__main__.py", line 87, in main
        return args.entrypoint(args, cli.error)
      File "/home/peter/projects/openage/openage/game/main.py", line 39, in main
        return run_game(args, assets)
      File "openage/game/main_cpp.pyx", line 20, in openage.game.main_cpp.run_game (/home/peter/projects/openage/openage/game/main_cpp.cpp:845)
        result = run_game_cpp(args_cpp)
      File "libopenage/util/opengl.cpp", line 66, in void openage::util::gl_check_error()
        throw Error(MSG(err) <<
    openage.cppinterface.exctranslate.CPPException: OpenGL error state after running draw method: 1281
            invalid value passed to opengl call

[TheJJ]

Now that we have multiple FBOs and render a Qt GUI into one of them, you may wanna test again :P

[Linkaan]

I get a similar error as above running on a 64bit kubuntu 16.04.1 with a GeForce GTX 460 graphics card. I get invalid operation performed during some state instead of invalid value passed to opengl call. The full error looks like:

INFO [T1] Loading meta file: ./assets/converted/terrain/15001.slp.docx
Traceback (most recent call last):
  File "run.py", line 15, in init run (/home/linkan/openage/run.cpp:891)
    main()
  File "/home/linkan/openage/openage/__main__.py", line 94, in main
    return args.entrypoint(args, cli.error)
  File "/home/linkan/openage/openage/game/main.py", line 47, in main
    return run_game(args, assets)
  File "openage/game/main_cpp.pyx", line 26, in openage.game.main_cpp.run_game (/home/linkan/openage/openage/game/main_cpp.cpp:989)
    result = run_game_cpp(args_cpp)
  File "", line 0, in 
  File "", line 0, in openage::run_game(openage::main_arguments const&)+0x55f
  File "", line 0, in openage::Engine::run()+0x1f
  File "", line 0, in openage::Engine::loop()+0x430
  File "", line 0, in openage::util::gl_check_error()+0x165
  File "", line 0, in openage::error::Error::Error(openage::log::message const&, bool, bool)+0x185
  File "", line 0, in openage::error::StackAnalyzer::analyze()+0x9f
  File "libopenage/util/opengl.cpp", line 66, in void openage::util::gl_check_error()
    throw Error(MSG(err) <<
openage.cppinterface.exctranslate.CPPException: OpenGL error state after running draw method: 1282
        invalid operation performed during some state
Segmenteringsfel (minnesutskrift skapad)

Any work around for this problem?

[VelorumS]

Try to run it like that to get more info:

./run game --gl-debug

[zuntrax]

We'll get rid of our current renderer soon, closing.