[solved] GLSL Link fails with no error on intel graphics

OpenGL Development
#1

Hi, I am using similar code to http://lwjgl.org/wiki/index.php?title=GLSL_Shaders_with_LWJGL and when I run my program on my intel HD 4000 graphics card, the link stage fails:

ARBShaderObjects.glGetObjectParameteriARB(shaderID, ARBShaderObjects.GL_OBJECT_COMPILE_STATUS_ARB)

returns GL_FALSE but when using

ARBShaderObjects.glGetInfoLogARB(obj, ARBShaderObjects.glGetObjectParameteriARB(obj, ARBShaderObjects.GL_OBJECT_INFO_LOG_LENGTH_ARB));

an empty string is returned.

This only happens on my intel card, my ATI one runs fine.
Also a couple of simple shaders did link fine on the intel card but just drew a black screen (They worked fine on the ATI card too)
Is there anything I can do? Or the intel card just sucks and can’t handle the shaders?

Thanks,
roland

#2

sometimes shader compilers fail so hard, they don’t even tell you why. depends on the code.

try using https://www.opengl.org/registry/specs/ARB/debug_output.txt for more info.

#3

What is “similar”? One thing that I’ve found fails on Intel but not other cards is using integers where you mean floats (eg 1 rather than 1.0). I think the Intel behaviour is actually correct unless you specify #version but it’s the one that always seems to bite me. :persecutioncomplex:

#4

Thanks :slight_smile: You’re right about the shader compilers fail so hard, they don’t even tell you why. I commented out everything and added back one thing at a time, this introduced some errors. I finally got it linking but it still crashed when I ran it.

Turns out using an array of uniform sampler2D was the issue (As I said, worked fine on ATI). So I just replaced it with 4 uniform sampler2Ds instead.


#define NUM_MAP_TEXTURES 4
uniform sampler2D mapTextures[NUM_MAP_TEXTURES]; //fails on intel

Thanks, I got some of those errors too before, however they printed out fine in the compile stage. Thinking about why the similar shaders linked fine and this one didn’t helped though. :slight_smile:

#5

Could even be the same reason - should have been more generic in my answer! :wink: Looking at something like this ( http://stackoverflow.com/a/12031821 ) arrays of samplers are not valid below 1.30. Are you using a #version pragma?

According to the GLSL spec ( https://www.opengl.org/wiki/Core_Language_(GLSL)#Version )

[quote]If a #version​ directive does not appear at the top, then it assumes 1.10, which is almost certainly not what you want.
[/quote]
What I’ve found is that AMD & nVidia cards seem to allow features above version to work, whereas Intel cards seem to follow that spec and fail. The int <> float thing is not the only time I’ve fallen foul of this, but just the one I keep repeating! ::slight_smile:

#6

Thanks, I wasn’t using a #version directive. However, I don’t want to go right up to GLSL 4.0 just so that I can loop over the sampler array :-\

[quote]In GLSL 1.30 to 3.30, you can have sampler arrays, but with severe restrictions on the index. The index must be an integral constant expression. Thus, while you can declare a sampler array, you can’t loop over it.
[/quote]
Oh well, it seems to be fine how I currently have it, apart from a bit of duplicated code. I’ll get over it :slight_smile:

Thanks again,
roland