Fold WebDataSource into BufferedDataSource.

Fold WebDataSource into BufferedDataSource.

With the removal of SimpleDataSource in r127309 we no longer require multiple implementations of WebDataSource.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=129192

[Ami GONE FROM CHROMIUM, 2012-03-27 05:16:19]

LGTM % nit (and a couple of q's that can be taken off reviewlog)

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
File webkit/media/buffered_data_source.cc (right):

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.cc:113: return loader_.get() ?
loader_->HasSingleOrigin() : true;
(predates your CL, but)
Is it right to default open like this, or should this be a DCHECK(loader_) to
prevent being asked this question before Initialize() is called, and thus before
we can pass a judgement on the single-origin-ness of this object?
(I worry about someone caching the answer and then Initializing with a
not-single-origin URL/loader).

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
File webkit/media/buffered_data_source.h (right):

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.h:27: // in-memory sliding window.
doco ctor/dtor thread restrictions?  (I think this object must be created &
destroyed on the render thread, but I'm not sure)

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.h:28: class BufferedDataSource : public
media::DataSource {
How long till DataSource can become a pure interface?  (or die?)

[scherkus (not reviewing), 2012-03-27 14:10:06]

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
File webkit/media/buffered_data_source.cc (right):

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.cc:113: return loader_.get() ?
loader_->HasSingleOrigin() : true;
On 2012/03/27 05:16:19, Ami Fischman wrote:
> (predates your CL, but)
> Is it right to default open like this, or should this be a DCHECK(loader_) to
> prevent being asked this question before Initialize() is called, and thus
before
> we can pass a judgement on the single-origin-ness of this object?
> (I worry about someone caching the answer and then Initializing with a
> not-single-origin URL/loader).

Good catch!

Technically we can get a false value even if loader_ exists (i.e., BRL hasn't
received an HTTP response yet so it returns default value).

DCHECK() added to make sure initialization has completed.

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
File webkit/media/buffered_data_source.h (right):

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.h:27: // in-memory sliding window.
On 2012/03/27 05:16:19, Ami Fischman wrote:
> doco ctor/dtor thread restrictions?  (I think this object must be created &
> destroyed on the render thread, but I'm not sure)

Done.

https://chromiumcodereview.appspot.com/9836102/diff/1/webkit/media/buffered_d...
webkit/media/buffered_data_source.h:28: class BufferedDataSource : public
media::DataSource {
On 2012/03/27 05:16:19, Ami Fischman wrote:
> How long till DataSource can become a pure interface?  (or die?)

set_host() is the only non-pure-virtual method and the interface also includes
ref-counting -- it'd be nice to get rid of both

the only other DataSource implementations are:
  * FileDataSource (tests, tools)
  * MockDataSource (tests)
  * DataSourceLogger (tools)

...so mostly test/tool support code, however DataSource is relatively low on the
list of code that needs to be nuked other than BDS/BRL itself.