UUM-147204 - Unity release-note mentions

Version	Section	Impact	Note
6000.7.0b8	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b8	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b7	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b7	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b6	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b6	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b5	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b5	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b4	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b4	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b3	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b3	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b2	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b2	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0b1	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0b1	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0a6	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.7.0a6	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0a6	Fixes Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behavior from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable.
6000.6.0f1	Fixes Performance Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases. (UUM-147204) First seen in 6000.6.0a8.
6000.6.0f1	Fixes Performance Test Framework	Fix	Measure.Method().GC() again forces a full garbage collection immediately before each timed region, restoring the behaviour from before the switch to the ProfilerRecorder API. Without it, a collection triggered by the measured code's own allocations lands inside the timed region and shows up as periodic spikes across the samples. As before: this only happens when GC() is specified; warmup and probing iterations are not affected; and because SetUp/CleanUp cause each of the IterationsPerMeasurement inner executions to be timed separately, that combination gets one collection per inner execution rather than one per sample, which makes it considerably slower. Reported times for allocating methods measured with GC() will change, generally becoming lower and more stable, so baselines recorded since 3.2.1 are not comparable. (UUM-147204) First seen in 6000.6.0a8.
6000.7.0a5	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0a4	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.
6000.7.0a4	Fixes Test Framework	Fix	Profiler marker recorders used by Measure.Method now start when the measured iterations begin instead of when ProfilerMarkers() is called, so warmup and probing iterations no longer contribute to the reported marker values or add sample collection overhead to those phases.