Do not record CPU profile samples when stack collection is failed.

src/profiler/cpu-profiler.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/profiler/cpu-profiler.h"
 
 #include "src/debug/debug.h"
 #include "src/deoptimizer.h"
 #include "src/frames-inl.h"
 #include "src/locked-queue-inl.h"
@@ skipping 96 matching lines @@
   }
 
   const TickSampleEventRecord* record = ticks_buffer_.Peek();
   if (record == NULL) {
     if (ticks_from_vm_buffer_.IsEmpty()) return NoSamplesInQueue;
     return FoundSampleForNextCodeEvent;
   }
   if (record->order != last_processed_code_event_id_) {
     return FoundSampleForNextCodeEvent;
   }
-  generator_->RecordTickSample(record->sample);
+  if (record->sample.pc) {

[titzer, 2016/02/19 09:28:37]

Will this tick just be discarded? Is there a bucket for discarded ticks to avoid
showing skewed results?

[alph, 2016/02/19 16:24:29]

It is just discarded, so effectively it adds to all other samples a bit.
There's no such a bucket. If I add one I'm not sure how that would be presented
to the user.
Actually currently the results are skewed as well because if we fail to collect
a stack the sample is not attributed to the function currently being executed,
but to a program node instead. So the function node becomes smaller than it is
in the reality.

[titzer, 2016/02/19 16:36:33]

As long as the missing samples go somewhere, so as not to skew the existing
samples, then this sgtm.

[alph, 2016/02/19 17:54:22]

Agree.

+    generator_->RecordTickSample(record->sample);
+  }
   ticks_buffer_.Remove();
   return OneSampleProcessed;
 }
 
 
 void ProfilerEventsProcessor::Run() {
   while (!!base::NoBarrier_Load(&running_)) {
     base::TimeTicks nextSampleTime =
         base::TimeTicks::HighResolutionNow() + period_;
     base::TimeTicks now;
@@ skipping 398 matching lines @@
     Builtins::Name id = static_cast<Builtins::Name>(i);
     rec->start = builtins->builtin(id)->address();
     rec->builtin_id = id;
     processor_->Enqueue(evt_rec);
   }
 }
 
 
 }  // namespace internal
 }  // namespace v8