review of per and post bake protocols showed that the CAPTURING shader macro was not being properly recompiled in. as opengl was not playing nice with a simple batch shader recompilation for all effected shaders, a full lightmanager restart is at time of writing required. once we have a proper globally cached scene structure stored off GPU side, we'll want to change GFXShader::addGlobalMacro("CAPTURING", String("1")); on over to dirtying that value in the cached buffer via setting a shader global uniform
review of prefilter examples shows a fixed sample count of 1024 across multiple implementations, so we'll use the standard barring further research into where that number is comming from for a scalar approach
review of gl shaders shows a doubleup in compiled state testing, so slimmed that down and added additional debugging reports
The fix adds a sanity check to the D3D and GL device enumeration function that exits early if there's no registered display from the OS(implying it's running in commandline).
With newer versions of SDL, attempting to enumerate the GL device causes a segfault because the OS will return back a valid context, but the context isn't actually valid. So when tested against, it crashes. Avoiding enumerating the device when you're not in a position to render works around the issue.
TODO: If the machine is running in terminal-only mode, and thus has no valid render context, but still has a monitor plugged into the machine, it can sidestep the sanity check and still result in a segfault. Need a more robust check for that circumstance in a future fix.
Workaround: MacOS does not have access to glClipControl.
Because the lower range of the 32F buffer is not available using this workaround, the improvements to depth behavior from depth reversal will be limited on MacOS. Behavior is equivalent to reversed depth on a 24UINT buffer; Better than stock, but not dramatically.
-Adds reversed depth projection model, dramatically increasing depth buffer effective resolution.
-Adds 32F depth 8U stencil format GFXFormatD32FS8X24 (following DX naming conventions). Note this is a 64-bit format, and likely not suitable for mobile platforms. Revert to GFXFormatD24S8 in renderManager.tscript for mobile & "ancient" platforms.
-Corrects alignment of texture type details array.
refactored detection of texture sizes for cubemaps
removed the assumption that if we give U32 GFXTextureManager::getTextureDownscalePower( GFXTextureProfile *profile ) no profile, it should go right ahead and downscale anyway
sniped the downscaling strings in the resulting U32 getProbeTexSize(); and void setCubeTexSize(const U32 cubemapFaceSize); until sucj time as we can properly follow up all possible combinations of shiping in one scale, and a customer choosing to use lower resolution textures as the current result is a hard shutoff entirely
Swapped the water's disableTrueReflections variable to be enableTrueReflections for simplicity and consistency(also fixed a persistent typo)
Swapped disableVerticalSync to be enableVerticalSync for simplicity and consistency
Swapped disableParallaxMapping to be enableParallaxMapping for simplicity and consistency
Fix click detection on slider mode entries for guiGameSettingsCtrl so the click position correctly matches the percentage
Fixed problem where postFX initialization would always exec default.postfxpreset.tscript, even if a level's got it's own preset, which can cause problems
Fixed range field type behavior so that editing the values applies in real time, and also consistently applies between slider and text field
clean up mismatched PROFILE_START+PROFILE_END() that was in/out of the mip packign for loop on the gl side, leaving just the self-cleaning PROFILE_SCOPE(GFXGLTextureObject_copyToBmp); similar to the dx11 side
void GuiGameListMenuCtrl::removeRow(const S32& row); (the GuiGameListMenuCtrl:: bit in the class iteslf is redundant and making linux unhappy)
GL layer was missing _captureBackBuffer() implementation which caused screenShot() function to fail. Also caused ScreenShot::capture() to be called every frame afterward which caused significant performance issues.
largely cleanups, though does include glCopyImageSubData for gl revs that support it
TODOS:
1) non glCopyImageSubData for `void GFXGLCubemapArray::copyTo(GFXCubemapArray *pDstCubemap)`
2) while we don't get corruption showing in >0 mips for irradiance maps using the newer code, it nevertheless renders wrong.
not 100% sure if renderdoc is showing blank mips because it expects it due to size, or if we are in fact generating em even though we don't use em in the end for irradiance references.
Fixed multiple case sensitivity issues
Adds GL ARB check for cubemap arrays for certain cases where the driver only supports the extension, not as core
Fixes undeeded namespace declaration in an inline operator function
Cleaned up/reordered some modules default in the data dir
WIP of Das Boot test map
Begun refactoring visualizer tools
Added default cloud and water textures to core/rendering so water and cloud objects work correctly
Added default Fog_Cube mesh so volumetric fog works correctly
