Speculative fix for division by zero in Vp8EncoderImpl.

webrtc/modules/video_coding/media_optimization.cc

Index: webrtc/modules/video_coding/media_optimization.cc
diff --git a/webrtc/modules/video_coding/media_optimization.cc b/webrtc/modules/video_coding/media_optimization.cc
index a30ac9521abe3371567f427c661d7777e5ee86a7..fec5c4c33b6e7163598ebba8bac344f7bc80f4fa 100644
--- a/webrtc/modules/video_coding/media_optimization.cc
+++ b/webrtc/modules/video_coding/media_optimization.cc
@@ -10,6 +10,8 @@
 
 #include "webrtc/modules/video_coding/media_optimization.h"
 
+#include <limits>
+
 #include "webrtc/base/logging.h"
 #include "webrtc/modules/video_coding/utility/frame_dropper.h"
 #include "webrtc/system_wrappers/include/clock.h"
@@ -129,7 +131,9 @@ uint32_t MediaOptimization::InputFrameRate() {
 
 uint32_t MediaOptimization::InputFrameRateInternal() {
   ProcessIncomingFrameRate(clock_->TimeInMilliseconds());
-  return uint32_t(incoming_frame_rate_ + 0.5f);
+  uint32_t framerate = static_cast<uint32_t>(std::min<float>(
+      std::numeric_limits<uint32_t>::max(), incoming_frame_rate_ + 0.5f));
+  return framerate;

[tommi, 2017/03/10 12:55:45]

should we have a check here?  At least RTC_DCHECK(framerate)?

[stefan-webrtc, 2017/03/10 13:06:55]

framerate == 0 is handled outside this class, which I think may be reasonable
since it should be OK to compute a framerate of zero.

The issue I'm fixing here is that incoming_frame_rate_ can overflow when being
assigned to maxFramerate in VideoCodec in common_types.h, since that type was
uint8_t that wasn't too unlikely. With some bad luck that ends up being 0. :)

 }
 
 uint32_t MediaOptimization::SentFrameRate() {