Bug 167656

Summary: [Mac] Update CARingBuffer class
Product: WebKit Reporter: Eric Carlson <eric.carlson>
Component: MediaAssignee: Eric Carlson <eric.carlson>
Status: RESOLVED FIXED    
Severity: Normal CC: commit-queue, dino, jer.noble, thiago.lacerda, webkit-bug-importer
Priority: P2 Keywords: InRadar
Version: Other   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Proposed patch
none
Updated patch, with API test. none

Description Eric Carlson 2017-01-31 11:47:20 PST
Update class to be able to copy or mix input, and to allocate storage from the information in a stream description.
Comment 1 Eric Carlson 2017-01-31 12:48:45 PST
Created attachment 300248 [details]
Proposed patch
Comment 2 WebKit Commit Bot 2017-01-31 12:51:16 PST
Attachment 300248 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/audio/mac/CARingBuffer.h:58:  The parameter name "mode" adds no information, so it should be removed.  [readability/parameter_name] [5]
Total errors found: 1 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.
Comment 3 Jer Noble 2017-01-31 14:17:22 PST
Comment on attachment 300248 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=300248&action=review

> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:140
> +        case CARingBuffer::Mix:
> +            float* destination = static_cast<float*>(dest->mData);
> +            vadd(destination, 1, reinterpret_cast<float*>(*buffers + srcOffset), 1, destination, 1, std::min<size_t>(nbytes, dest->mDataByteSize - destOffset));
> +            break;

This doesn't feel right. A float is not the same size as a byte, so I feel like this is going to overrun outside the bounds of the passed in list. (Well, the std::min will keep that from happening, but it'll still not behave correctly.)

> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:163
> +    ZeroRange(static_cast<Byte**>(m_buffers->data()), m_channelCount, 0, m_capacityBytes);

You shouldn't have to zero anything here. Rather, just update the bounds below.

> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:184
> +    uint64_t start, end;
> +    getCurrentFrameBounds(start, end);
> +    if (start == end)
> +        setCurrentFrameBounds(startFrame, startFrame);
> +

What's this for?
Comment 4 Eric Carlson 2017-02-01 15:12:40 PST
Comment on attachment 300248 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=300248&action=review

>> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:140
>> +            break;
> 
> This doesn't feel right. A float is not the same size as a byte, so I feel like this is going to overrun outside the bounds of the passed in list. (Well, the std::min will keep that from happening, but it'll still not behave correctly.)

Good point. While we the upcoming patches only use it with floats, we might as well support other the formats.

>> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:163
>> +    ZeroRange(static_cast<Byte**>(m_buffers->data()), m_channelCount, 0, m_capacityBytes);
> 
> You shouldn't have to zero anything here. Rather, just update the bounds below.

Fixed.

>> Source/WebCore/platform/audio/mac/CARingBuffer.cpp:184
>> +
> 
> What's this for?

Oops, leftover testing.
Comment 5 Eric Carlson 2017-02-01 15:13:29 PST
Created attachment 300359 [details]
Updated patch, with API test.
Comment 6 WebKit Commit Bot 2017-02-01 15:16:36 PST
Attachment 300359 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/audio/mac/CARingBuffer.h:59:  The parameter name "mode" adds no information, so it should be removed.  [readability/parameter_name] [5]
Total errors found: 1 in 14 files


If any of these errors are false positives, please file a bug against check-webkit-style.
Comment 7 Jer Noble 2017-02-01 15:20:08 PST
Comment on attachment 300359 [details]
Updated patch, with API test.

View in context: https://bugs.webkit.org/attachment.cgi?id=300359&action=review

r=me, with nit

> LayoutTests/webaudio/mediastreamaudiosourcenode.html:32
> -                stream.getAudioTracks()[0].stop();
> +//                stream.getAudioTracks()[0].stop();

We probably still need this. :)
Comment 8 Eric Carlson 2017-02-01 15:51:59 PST
Committed r211539 and r211540: https://trac.webkit.org/r211539, https://trac.webkit.org/r211540
Comment 9 Radar WebKit Bug Importer 2017-02-01 15:53:25 PST
<rdar://problem/30318329>
Comment 10 Eric Carlson 2017-02-01 17:50:39 PST
Plus https://trac.webkit.org/r211545 to fix the 32-bit build.