VS 2017 complains about the same thing that clang does: casting S32 to GLvoid*
can't possibly produce a valid pointer value because S32 can't fit a whole
64-bit pointer. To appease it, not only must we use reinterpret_cast, but we
must first cast S32 to intptr_t and then reinterpret_cast THAT.

The previous implementation went to some effort to crash if anyone attempted
to create or destroy an LLInstanceTracker subclass instance during traversal.
That restriction is manageable within a single thread, but becomes unworkable
if it's possible that a given subclass might be used on more than one thread.

Remove LLInstanceTracker::instance_iter, beginInstances(), endInstances(),
also key_iter, beginKeys() and endKeys(). Instead, introduce key_snapshot()
and instance_snapshot(), the only means of iterating over LLInstanceTracker
instances. (These are intended to resemble functions, but in fact the current
implementation simply presents the classes.) Iterating over a captured
snapshot defends against container modifications during traversal. The term
'snapshot' reminds the coder that a new instance created during traversal will
not be considered. To defend against instance deletion during traversal, a
snapshot stores std::weak_ptrs which it lazily dereferences, skipping on the
fly any that have expired.

Dereferencing instance_snapshot::iterator gets you a reference rather than a
pointer. Because some use cases want to delete all existing instances, add an
instance_snapshot::deleteAll() method that extracts the pointer. Those cases
used to require explicitly copying instance pointers into a separate
container; instance_snapshot() now takes care of that. It remains the caller's
responsibility to ensure that all instances of that LLInstanceTracker subclass
were allocated on the heap.

Replace unkeyed static LLInstanceTracker::getInstance(T*) -- which returned
nullptr if that instance had been destroyed -- with new getWeak() method
returning std::weak_ptr<T>. Caller must detect expiration of that weak_ptr.

Adjust tests accordingly.

Use of std::weak_ptr to detect expired instances requires engaging
std::shared_ptr in the constructor. We now store shared_ptrs in the static
containers (std::map for keyed, std::set for unkeyed).

Make LLInstanceTrackerBase a template parameterized on the type of the static
data it manages. For that reason, hoist static data class declarations out of
the class definitions to an LLInstanceTrackerStuff namespace.

Remove the static atomic sIterationNestDepth and its methods incrementDepth(),
decrementDepth() and getDepth(), since they were used only to forbid creation
and destruction during traversal.

Add a std::mutex to static data. Introduce an internal LockStatic class that
locks the mutex while providing a pointer to static data, making that the only
way to access the static data.

The LLINSTANCETRACKER_DTOR_NOEXCEPT macro goes away because we no longer
expect ~LLInstanceTracker() to throw an exception in test programs.
That affects LLTrace::StatBase as well as LLInstanceTracker itself.

Adapt consumers to the new LLInstanceTracker API.

Fix for bad fullbright shiny shininess values and inconsistency between materialF and fullbrightShinyF

Simplify LLSingleton::SingletonLifetimeManager to SingletonInitializer: that
struct has not been responsible for deletion ever since LLSingletonBase
acquired dependency-ordered deleteAll().

Move SingletonData::mInitState changes from SingletonLifetimeManager to
constructSingleton() method. Similarly, constructSingleton() now sets
SingletonData::mInstance instead of making its caller store the pointer.

Add variadic arguments to LLSingleton::constructSingleton() so we can reuse it
for LLParamSingleton.

Add finishInitializing() method to encapsulate logic reused for
getInstance()'s INITIALIZING and DELETED cases.

Make LLParamSingleton a subclass of LLSingleton, just as LLLockedSingleton is
a subclass of LLParamSingleton. Make LLParamSingleton a friend of LLSingleton,
so it can access private members of LLSingleton without also granting access
to any DERIVED_CLASS subclass. This eliminates the need for protected
getInitState().

LLParamSingleton::initParamSingleton() reuses LLSingleton::constructSingleton()
and finishInitializing(). Its getInstance() method completely replaces
LLSingleton::getInstance(): in most EInitStates, LLParamSingleton::getInstance()
is an error.

Use a std::mutex to serialize calls to LLParamSingleton::initParamSingleton()
and getInstance(). While LLSingleton::getInstance() relies on the "initialized
exactly once" guarantee for block-scope static declarations, LLParamSingleton
cannot rely on the same mechanism.

LLLockedSingleton is now a very succinct subclass of LLParamSingleton -- they
have very similar functionality.

Giving the LLSINGLETON() macro variadic arguments eliminates the need for a
separate LLPARAMSINGLETON() macro, while continuing to support existing usage.

Provide method of storing joint indices sep from weight data for faster runtime processing.

Remove all references to unused shader var global_gamma.

Remove many unused decls for gamma from shaders.

Make post-deferred gamma correction use display_gamma.

Make setting display_gamma use the correct RenderDeferredDisplayGamma setting.

Remove binding and use of glIsShader for now (only useful for avoiding warnings about deleting shader objects re-used across programs).

Make vograss objects set their face vert/index counts to 0 when setting
the number of blades to 0 to disable rendering so we don't allocate space
and then skip setting it up with actual index data in getGeometry.

Fix broken throttling of sky/envmap updates.

Add code to force a usage hint when VBOs are disabled and make useBVOs reflect that as well.

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).

Convince Geenzo code to get a little DeMorgan in it.

Trophy Unlocked: make all alpha textures partcipate in picking.

Fix light atten and transmittance calcs breaking ambient color contrib to fog.

Needs to be tested on DRTSIM-415 or better to get legacy viewer settings that match EEP.

Fix ambient light inputs to the renderer.

Fix 3rd sky shader w/ mistaken density mod conversion.

Make ambient clamp apply to all modes.

Tune ALM ambient clamp to match non-ALM.

Make OpenGL 3.3 cards not false alarm on version 140 shader reqs.
Makes deferred rendering work on low-end GPUs more often.

Make haze look more like WL.

Make lack of sun and moon mean black and fixed dir to prevent spooky underlighting.

Fix direct light matching across alpha/blended-material/deferred.

Get diffuse lighting to match from Low to Ultra.

Given that mShaderObjects only ever expects one copy of a given shader object (leading to lots of "shaderV" and "shaderF"), it made having utility shaders a bit more difficult to incorporate.
This change makes sure the two are separated.  Using this over a multimap largely for readibility.

Also, actually, you know, use this functionality.

--HG--
branch : OPEN-340

Still some work to go here.  Largely related to how certain values (i.e., SSAO mix values) are supplied.  But so far, so good.

--HG--
branch : OPEN-340