Use maximum capacity instead of a ratio of capacity for BufferedResourceLoader.

Use maximum capacity instead of a ratio of capacity for BufferedResourceLoader.

This change essentially moves internally buffered network data from WebURLLoader to BufferedResourceLoader as soon as there's capacity instead of waiting for an arbitrary threshold.

There's no fear of spamming progress events at a higher frequency since progress events are limited to firing at a minimum interval of 350ms. 

BUG=124719
TEST=none

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=146133

[scherkus (not reviewing), 2012-07-06 15:34:28]

Not 100% I want to land this. Need to test out how often we end up toggling
between IDLE and LOADING. I have a hunch this might spam a lot of events.

[Ami GONE FROM CHROMIUM, 2012-07-09 17:27:40]

To be clear, this CL is equivalent to setting kDisableDeferThreshold to 1.0
(except it'll be simpler code).
IWBN to have a rationale for making this change (specific scenario that'd be
improved / reasons why it's not harmful).

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
File webkit/media/buffered_resource_loader.cc (right):

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.cc:647: // TODO(scherkus): refer to
http://crbug.com/124719 for more discussion on
still needed?

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.cc:649: case kThresholdDefer: {
braces no longer necessary

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
File webkit/media/buffered_resource_loader.h (right):

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.h:44: kThresholdDefer,
Needs a rename.

[scherkus (not reviewing), 2012-07-11 00:49:37]

updated change is mostly driven by code simplification

doing some analysis it ends up being a no-op JS-wise due to the minimum interval
on firing progress events

network-wise it's mostly a no-op: the network stack has a series of internal
buffers. By siphoning off that data more frequently we free up room in their
internal buffers so we can pull more data off the wire (this is more or less an
extension of r135436)

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
File webkit/media/buffered_resource_loader.cc (right):

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.cc:647: // TODO(scherkus): refer to
http://crbug.com/124719 for more discussion on
On 2012/07/09 17:27:41, Ami Fischman wrote:
> still needed?

Done.

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.cc:649: case kThresholdDefer: {
On 2012/07/09 17:27:41, Ami Fischman wrote:
> braces no longer necessary

Done.

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
File webkit/media/buffered_resource_loader.h (right):

http://codereview.chromium.org/10694098/diff/1/webkit/media/buffered_resource...
webkit/media/buffered_resource_loader.h:44: kThresholdDefer,
On 2012/07/09 17:27:41, Ami Fischman wrote:
> Needs a rename.

Done.

[Ami GONE FROM CHROMIUM, 2012-07-11 05:05:58]

lgtm

http://codereview.chromium.org/10694098/diff/1005/webkit/media/buffered_resou...
File webkit/media/buffered_resource_loader.cc (right):

http://codereview.chromium.org/10694098/diff/1005/webkit/media/buffered_resou...
webkit/media/buffered_resource_loader.cc:646: bool
BufferedResourceLoader::ShouldDisableDefer() const {
Didn't notice until now, but it's hilarious that we have ShouldEnable and
ShouldDisable and they each return a bool (and have very similar but inverted
logic).  IWBN to collapse them into a single ShouldDefer().  Another day,
another CL.

[scherkus (not reviewing), 2012-07-11 18:15:38]

http://codereview.chromium.org/10694098/diff/1005/webkit/media/buffered_resou...
File webkit/media/buffered_resource_loader.cc (right):

http://codereview.chromium.org/10694098/diff/1005/webkit/media/buffered_resou...
webkit/media/buffered_resource_loader.cc:646: bool
BufferedResourceLoader::ShouldDisableDefer() const {
On 2012/07/11 05:05:58, Ami Fischman wrote:
> Didn't notice until now, but it's hilarious that we have ShouldEnable and
> ShouldDisable and they each return a bool (and have very similar but inverted
> logic).  IWBN to collapse them into a single ShouldDefer().  Another day,
> another CL.

lol.

see http://codereview.chromium.org/10694138/