The whole methode is questionable in doing nothing and it seems there was a deliberate change
that made sure it does nothing. I am not sure what this method really is supposed to do, thus I did opt
for the most basic version of "do nothing" like it was before.
(cherry picked from commit 1c1a93e42e209e959f09ec46ca414a8936f4ac20)
Make shadowAlphaMask shader ignore vertex alpha before discards iff we're a fullbright object.
Keep llappearance from stuffing dangling pointers into the alpha cache (crash iff you're using nSight debugging and disable readbacks).
Fix glow calcs in sky shaders (not just the not shared enough atmospherics funcs).
Revert 10625 attempted fix as it breaks baggy clothes as much as skipping readbacks does.
Keep texture baking from pointlessly mercy flushing GL like a kid that just discovered the toilet handle.
Make glReadPixels call use GL_ALPHA8 (a texture format) instead of GL_ALPHA (an enum to indicate the alpha channel).
Move dynamic texture update (which does local back readback work) to just before swap
which may keep the Intel driver from tripping on its own shoelaces and is also subjectively faster for bakes.
Fix emissives without normal maps clobbering the alpha output in materialF.
Modify glow size calc to get close to pre-EEP sun glow behavior (may require sky vert settings >= mid).
Make bake ignore alpha readback for Intel until we can determine why their drivers now go boom.
Try aligning data used for glReadPixels to see if we can coax the Intel driver stack into being a less souciant pile of detritus.
Replace face color based test replaced with other logic in error.
These changes are only enabled if RenderNsightDebugSupport is true and eliminate use of
some OpenGL legacy functionality which is incompatible with nSight capture
(mostly glReadPixels and other fixed-function pipe rendering calls).
When wearing multiple layers of pants, the morph masks for all layers would be composited,
resulting in the smallest intersection of all layers. Instead we should only be applying the
top layer. This change will affect all consumers of the llappearance library.
Looks like we were caching bad morph masks that were generated on login. Adjusting
algorithms to always replace the cache when we have invalidated the other channels
of the baked texture.
consolidated most indra-specific constants in llcommon under indra_constants.h
fixed issues with operations on mixed unit types (implicit and explicit)
made LL_INFOS() style macros variadic in order to subsume other logging methods
such as ll_infos
added optional tag output to error recorders
Despite incrementing the COF version, beard edits were not updating the head hash.
We were not applying layers/parameters to the baked texture hashes for
layers that did not contain a user-defined image. However, many layers have standard
morph masks, etc that are affected by wearable parameters.
Edited the LLTexLayerInterface::getWearable() method to return the proper wearable type
if there is no associated local texture index, and all parameters in the wearable
refer to a particular wearable type. Should be safe even if some (but not all) params
have no wearable type defined.
Andrey Lihatskiy