[test] ofTexture more straightforward way to see if it support GL_RGB16F

[dimitre]

No description provided.

[dimitre]

@ofTheo I think this one is good to go. it fixes things for RPI and completes compilation on aarch64.
it reveals apothecary or system libraries issues (GLEW) on previous RPI versions

[ofTheo]

Feels like this would maybe be a safer change as it keeps the original logic, but also checks for the define

		if(ofGLESVersionFromGL() >= 300) {
				#if defined(GL_RGB16F)
						glInternalFormat = GL_RGB16F;
				#else
						glInternalFormat = GL_RGB;
				#endif
			textureTarget = GL_TEXTURE_2D;
		} else {
			glInternalFormat = GL_RGB;
			textureTarget = GL_TEXTURE_2D;
		}

[dimitre]

My main idea was depart from ofGLESVersionFromGL which is very imprecise.
if you read the function it tries to parse from a string (that greatly varies from platform, video card) and I am suspecting this function returns both false positives and negatives.

a more deterministic way of define would be get this info from glGetString(GL_SHADING_LANGUAGE_VERSION);

but the simpler way would be just checking the presence of the define. if it exists it can be set.

OF uses this approach a lot in ofCubeMap

			#if defined(GL_RGBA16F)

cc: @NickHardeman

[ofTheo]

yeah - I think I was just imagining possible regressions where it might now default to GL_RGB16F instead of GL_RGB - one is a floating point texture and the other is unsigned char.

So that could mean some shaders might behave a little differently - curious if that would be very noticeable or not really an issue.

I think in general bug fixes should correct breaking bugs or existing issues without changing behavior, hence my caution.

[ofTheo]

Thanks @dimitre - unrelated curious why we still have the armv7l CI checks - thought we dropped armv7 builds?

[dimitre]

@ofTheo I've put back working runners and fixed cross compiling for armv6l and armv7l about 7 months ago.
They were running great. now they are missing GLEW so I'm not sure if we were using apothecary ones or via apt.
we talked about dropping but there were other ideas, like 32 bits and rpi zero (or something like this)

[ofTheo]

@dimitre can you double check / fix the apothecary changes that are pulled into this PR and if it all looks good feel free to merge?

[dimitre]

@ofTheo I've cleaned up this PR to reflect the needed changes, so it is good to go in my opinion.
it can potentially help with emscripten issues

[ofTheo]

Amazing - thanks @dimitre !