It's a little distressing how often we have historically coded S32 or U32 to
pass a length or index.
There are more such assumptions in other viewer subdirectories, but this is a
start.
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.
removed complicated and unnecessary fast timer gapless handoff logic (it should be gapless anyway)
improved MemTrackable API, better separation of shadow and footprint
added memory usage stats to floater_stats.xml
made getPrimaryAccumulator return a reference since it was an
always non-null pointer
changed unit conversion to perform lazy division in order to avoid truncation
of timer values
various fixes to lltrace
start() on started recording no longer resets
fixed various instances of unit forgetfullness in lltrace
recording split now has gapless timing
scene monitor now guarantees min sample time
renamed a bunch of stats
added names to debug thread view on windows
removed LLThreadLocalSingleton
collapsed all thread recorder classes to single type, LLTrace::ThreadRecorder
moved fasttimer stack head to llthreadlocalsingletonpointer via ThreadRecorder
fixed some lltrace logic errors
more consistent syncing of timestamps of sample values in recording stack
selection of primary buffers was completely incorrect
assignment of recordings got wrong play state due to implicit
operator = defined in base class
fixed asset stats only working up to the first send
removed implicit flushes on reads from recorders for better performance
made sure stack timers were updated on recorder deactivate
faster rendering and better ui for fast timer view
removed unnecessary templates from accumulator types...now always
track data in double precision floating point, using templated accessors to
convert to and from arbitrary types
fixed multithreading lltrace causing values to be interpolated towards 0
added Radians unit
improved sceneloadmonitor restart heuristic to use accumulated camera motion
reverted SlaveThreadRecorder update gating
moved processTimes() outside of Recording, so it is called only once per frame
refined sample merge logic so that multi-threaded samples do not stomp on linear history
of a stat
fixed LLTrace::ExtendablePeriodicRecording::extend() to include *all* frame extensions
gated SlaveThreadRecorder pushing to master based on master update rate
reverted changes to LLThreadLocalSingletonPointer to not use offset-from-default trick
fixed copy construction behavior of Recordings to not zero out data
split measurement into event and sample, with sample representing
a continuous function
removed extra dereference for copy on write pointer
moved copyonwrite mechanism to RecordingBuffers from individual buffer
fixed logic that was leaving scene unfrozen when camera moved during metrics gathering
fix for unit test failures...cleanup apr without destroying
pools, allowing LLProxy to clean itself up as a singleton (and
avoiding spurious dependencies associated with manually destorying
singletons that rely on apr pools)
made fast timer stack thread local
added LLThreadLocalSingleton
made LLThreadLocalPointer obey pointer rules for const
added LLThreadLocalSingletonPointer for fast thread local pointers
improvements on lifetime of lltrace core data structures
tweaks to thread local pointer handling so that static constructors/destructors
can safely call functions that use lltrace
Alexander Gavriliuk