Remoting client rectangle decoder and VP8 decoder fixes.

Remoting client rectangle decoder and VP8 decoder fixes.
These are required to avoid artefacts when linear scaling is implemented.

BUG=93451
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=115088

[Wez, 2011-11-30 07:24:26]

Some changes to the rectangle decoder to avoid artefacts when we switch to
linear interpolated scaling.

[Wez, 2011-12-01 18:31:47]

sergeyu: PTAL :)

[Sergey Ulanov, 2011-12-01 20:02:32]

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc
File remoting/base/util.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode156
remoting/base/util.cc:156: int left = floor(rect.left() * horizontal_ratio);
Why do we need floor() here? just converting to int should do what we need here.

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode158
remoting/base/util.cc:158: int right = ceil(rect.right() * horizontal_ratio);
int right = rect.right() * horizontal_ratio + 1.0; ?

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
File remoting/client/rectangle_update_decoder.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:144:
refresh_rects_.push_back(frame_rect);
Would it make sense to call refresh_rects_.clear() before this line? We want to
update the whole screen so we don't care about previous content refresh_rects_

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:154: DoRefresh();
Looks like there should be public Refresh() method so that the plugin can call
it when screen needs refreshing.

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:174: #if 0
Can this be changed to if (false) {...} or just commented?

[Wez, 2011-12-01 21:20:06]

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc
File remoting/base/util.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode156
remoting/base/util.cc:156: int left = floor(rect.left() * horizontal_ratio);
On 2011/12/01 20:02:32, sergeyu wrote:
> Why do we need floor() here? just converting to int should do what we need
here.

We don't, but it makes the floor() / ceil() clamping more explicit.

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode158
remoting/base/util.cc:158: int right = ceil(rect.right() * horizontal_ratio);
On 2011/12/01 20:02:32, sergeyu wrote:
> int right = rect.right() * horizontal_ratio + 1.0; ?

Nope.  That's out-by-one if the multiplication results in an integer.

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
File remoting/client/rectangle_update_decoder.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:144:
refresh_rects_.push_back(frame_rect);
On 2011/12/01 20:02:32, sergeyu wrote:
> Would it make sense to call refresh_rects_.clear() before this line? We want
to
> update the whole screen so we don't care about previous content refresh_rects_

We could do, but really |refresh_rects_| should be replaced with an SkRegion,
which I planned to do in a follow-up CL.

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:154: DoRefresh();
On 2011/12/01 20:02:32, sergeyu wrote:
> Looks like there should be public Refresh() method so that the plugin can call
> it when screen needs refreshing.

I'm not sure how your comment relates to this change?

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:174: #if 0
On 2011/12/01 20:02:32, sergeyu wrote:
> Can this be changed to if (false) {...} or just commented?

It's just as easy to remove the code entirely for now.

[Sergey Ulanov, 2011-12-01 21:49:41]

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc
File remoting/base/util.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode156
remoting/base/util.cc:156: int left = floor(rect.left() * horizontal_ratio);
On 2011/12/01 21:20:06, Wez wrote:
> On 2011/12/01 20:02:32, sergeyu wrote:
> > Why do we need floor() here? just converting to int should do what we need
> here.
> 
> We don't, but it makes the floor() / ceil() clamping more explicit.

Can we use SkFloatToIntFloor() and SkFloatToIntCeil() here? Or maybe define
IntFloor() and IntCeil() locally?

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
File remoting/client/rectangle_update_decoder.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:154: DoRefresh();
On 2011/12/01 21:20:06, Wez wrote:
> On 2011/12/01 20:02:32, sergeyu wrote:
> > Looks like there should be public Refresh() method so that the plugin can
call
> > it when screen needs refreshing.
> 
> I'm not sure how your comment relates to this change?

I'm just suggesting that we should let the upper layer call Refresh(), and never
trigger refresh internally in this class -  but it's for a separate CL.

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:174: #if 0
On 2011/12/01 21:20:06, Wez wrote:
> On 2011/12/01 20:02:32, sergeyu wrote:
> > Can this be changed to if (false) {...} or just commented?
> 
> It's just as easy to remove the code entirely for now.

Yes, removing it would be even better.

[Wez, 2011-12-15 02:16:24]

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc
File remoting/base/util.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/base/util.cc#newcode156
remoting/base/util.cc:156: int left = floor(rect.left() * horizontal_ratio);
On 2011/12/01 21:49:41, sergeyu wrote:
> On 2011/12/01 21:20:06, Wez wrote:
> > On 2011/12/01 20:02:32, sergeyu wrote:
> > > Why do we need floor() here? just converting to int should do what we need
> > here.
> > 
> > We don't, but it makes the floor() / ceil() clamping more explicit.
> 
> Can we use SkFloatToIntFloor() and SkFloatToIntCeil() here? Or maybe define
> IntFloor() and IntCeil() locally?

We could do, but that would add a dependency on Skia to remoting_base, which
would pull in <math.h> and use floor() and ceil() to implement those methods...

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
File remoting/client/rectangle_update_decoder.cc (right):

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:154: DoRefresh();
On 2011/12/01 21:49:41, sergeyu wrote:
> On 2011/12/01 21:20:06, Wez wrote:
> > On 2011/12/01 20:02:32, sergeyu wrote:
> > > Looks like there should be public Refresh() method so that the plugin can
> call
> > > it when screen needs refreshing.
> > 
> > I'm not sure how your comment relates to this change?
> 
> I'm just suggesting that we should let the upper layer call Refresh(), and
never
> trigger refresh internally in this class -  but it's for a separate CL.

The Decoder is responsible for indicating what parts of the frame have changed,
either because the peer sent an update, or because the peer's dimensions
changed.  Handling refresh in response to changes in the plugin's scaling
factors or size in here keeps it all in one place.

http://codereview.chromium.org/8745008/diff/1/remoting/client/rectangle_updat...
remoting/client/rectangle_update_decoder.cc:174: #if 0
On 2011/12/01 21:49:41, sergeyu wrote:
> On 2011/12/01 21:20:06, Wez wrote:
> > On 2011/12/01 20:02:32, sergeyu wrote:
> > > Can this be changed to if (false) {...} or just commented?
> > 
> > It's just as easy to remove the code entirely for now.
> 
> Yes, removing it would be even better.

Done.

[Sergey Ulanov, 2011-12-20 01:07:16]

lgtm

[commit-bot: I haz the power, 2011-12-20 01:15:22]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/8745008/6001

[commit-bot: I haz the power, 2011-12-20 03:19:34]

Try job failure for 8745008-6001 (retry) on win_rel for steps
"safe_browsing_tests, browser_tests" (clobber build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...

[commit-bot: I haz the power, 2011-12-20 04:07:30]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/8745008/6001

[commit-bot: I haz the power, 2011-12-20 07:09:03]

Change committed as 115088