Add 4 VP9 profiles to the media::VideoCodecProfile enum.

media/ffmpeg/ffmpeg_common.cc

 // Copyright (c) 2012 The Chromium 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 "media/ffmpeg/ffmpeg_common.h"
 
 #include "base/logging.h"
 #include "base/memory/scoped_ptr.h"
 #include "base/sha1.h"
 #include "base/strings/string_number_conversions.h"
@@ skipping 443 matching lines @@
     aspect_ratio = stream->sample_aspect_ratio;
   else if (stream->codec->sample_aspect_ratio.num)
     aspect_ratio = stream->codec->sample_aspect_ratio;
 
   VideoCodec codec = CodecIDToVideoCodec(stream->codec->codec_id);
 
   VideoCodecProfile profile = VIDEO_CODEC_PROFILE_UNKNOWN;
   if (codec == kCodecVP8)
     profile = VP8PROFILE_ANY;
   else if (codec == kCodecVP9)
-    profile = VP9PROFILE_ANY;
+    // TODO(servolk): Find a way to obtain VP9 profile from FFmpeg.
+    profile = VP9PROFILE_PROFILE3;

[fgalligan1, 2016/03/23 18:14:09]

I read the comments on why this should be profile 3 in another file. So this
will only be used to setup a decoder that says it can handle profile 3 and
below? If so then this should be fine.

But if you use this to say the stream is a profile 3 and then use the config to
see if it will play in a decoder, then I think you should make this profile 0.

[servolk, 2016/03/24 18:05:43]

I'm not sure I understand the distinction, this is going to be used  both for
deciding which decoder to use and for initializing the decoder later.
I just did a bit more research and unfortunately it looks like current code is
somewhat fragile here, since currently we just send this VP9PROFILE_ANY to all
decoders and leave it up to them to interpret what it means. This is problematic
because for example DXVAVideoDecodeAccelerator doesn't check which profiles are
actually supported
(https://code.google.com/p/chromium/codesearch#chromium/src/content/common/gpu...)
and VAAPI simply maps VP9PROFILE_ANY to VAProfileVP9Profile0
(https://code.google.com/p/chromium/codesearch#chromium/src/content/common/gpu...).
I guess this means that we might get unexpected VP9 playback failures on
DXVA/VAAPI platforms when trying to play VP9 streams with profiles higher than
0.
I've opened crbug.com/597693 for this issue.

Having said this, I now agree that profile0 is probably a better choice here,
since software libvpx decoder ignores profile value completely in
InitializeVpxContext
(https://code.google.com/p/chromium/codesearch#chromium/src/media/filters/vpx_...),
and for accelerated decoders using profile0 is going to be most similar to the
current situation with VP9PROFILE_ANY and thus should minimize the chances of
worsening the current situation. Once this CL lands and we are able to
distinguish various VP9 profiles in Chromium media stack, we can start making
progress on fixing crbug.com/597693.

   else
     profile = ProfileIDToVideoCodecProfile(stream->codec->profile);
 
   // Without the FFmpeg h264 decoder, AVFormat is unable to get the profile, so
   // default to baseline and let the VDA fail later if it doesn't support the
   // real profile. This is alright because if the FFmpeg h264 decoder isn't
   // enabled, there is no fallback if the VDA fails.
 #if defined(DISABLE_FFMPEG_VIDEO_DECODERS)
   if (codec == kCodecH264)
     profile = H264PROFILE_BASELINE;
@@ skipping 284 matching lines @@
 }
 
 int32_t HashCodecName(const char* codec_name) {
   // Use the first 32-bits from the SHA1 hash as the identifier.
   int32_t hash;
   memcpy(&hash, base::SHA1HashString(codec_name).substr(0, 4).c_str(), 4);
   return hash;
 }
 
 }  // namespace media