Define Panning cursors only on Windows

Define Panning cursors only on Windows
Remove unused IDR_AURA_CURSOR_{BIG_}FLEUR

BUG=None

[Rick Byers, 2014-01-13 16:33:22]

Changing just chromium like this means that if blink changes in the future to
use these cursors (or I'm wrong and there is some other code path using them on
on-windows) then we'll hit the NOTREACHED - resulting in a DCHECK and using the
null cursor.  I think this would be pretty likely to go unnoticed in practice.

I was thinking it would be better to make the blink definitions (eg.
WebCursorInfo) conditional on OS_WIN as well - that way we can't possibly try to
use these special cursors on non-windows platforms.  But I don't see any other
example where the webkit API itself is platform specific.

Adam I see a few choices, which do you prefer?

1) Continue to require all platforms to support these mouse cursors, even though
they're only ever used on Windows (see the #ifdef in AutoscrollController.cpp).

2) Add #ifdef OS_WIN to blink (and the webkit API) so that these cursor types
are only available on Windows

3) Rename the blink enums to make it clear these are windows specific (eg.
'WindowsPanning...' instead of 'Panning...),  and add some ASSERTs in the
implementation that check that they're not used outside Windows, but leave them
exposed in the WebKit API.

Ideally we'd get an agreement to either remove this 'pan scrolling' feature from
Windows or add it to all platforms.  But I'm not sure anyone wants to invest
that sort of effort right now.

Rick

[abarth-chromium, 2014-01-13 17:18:43]

On 2014/01/13 16:33:22, Rick Byers wrote:
> 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these cursor types
> are only available on Windows

^^^ This option seems best.

[oshima, 2014-01-13 18:24:48]

On 2014/01/13 17:18:43, abarth wrote:
> On 2014/01/13 16:33:22, Rick Byers wrote:
> > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these cursor
types
> > are only available on Windows
> 
> ^^^ This option seems best.

I agree, but this CL still is still needed and has to be landed first right?

[Rick Byers, 2014-01-13 18:46:56]

On 2014/01/13 18:24:48, oshima wrote:
> On 2014/01/13 17:18:43, abarth wrote:
> > On 2014/01/13 16:33:22, Rick Byers wrote:
> > > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these cursor
> types
> > > are only available on Windows
> > 
> > ^^^ This option seems best.

Thanks Adam.

> I agree, but this CL still is still needed and has to be landed first right?

Yes.  And it looks like there are references on the other platforms (Mac, gtk)
that need to be removed from the chromium code first too.  If you prefer I'm
happy to land the blink side after you've cleaned up all the references in
chromium.

[oshima, 2014-01-13 18:53:43]

On 2014/01/13 18:46:56, Rick Byers wrote:
> On 2014/01/13 18:24:48, oshima wrote:
> > On 2014/01/13 17:18:43, abarth wrote:
> > > On 2014/01/13 16:33:22, Rick Byers wrote:
> > > > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these cursor
> > types
> > > > are only available on Windows
> > > 
> > > ^^^ This option seems best.
> 
> Thanks Adam.
> 
> > I agree, but this CL still is still needed and has to be landed first right?
> 
> Yes.  And it looks like there are references on the other platforms (Mac, gtk)
> that need to be removed from the chromium code first too.  If you prefer I'm
> happy to land the blink side after you've cleaned up all the references in
> chromium.

Oh, sorry I thought I did that but apparently the latest patch is missing them.
Uploaded new patch. PTAL.

[Rick Byers, 2014-01-13 19:16:16]

On 2014/01/13 18:53:43, oshima wrote:
> On 2014/01/13 18:46:56, Rick Byers wrote:
> > On 2014/01/13 18:24:48, oshima wrote:
> > > On 2014/01/13 17:18:43, abarth wrote:
> > > > On 2014/01/13 16:33:22, Rick Byers wrote:
> > > > > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these
cursor
> > > types
> > > > > are only available on Windows
> > > > 
> > > > ^^^ This option seems best.
> > 
> > Thanks Adam.
> > 
> > > I agree, but this CL still is still needed and has to be landed first
right?
> > 
> > Yes.  And it looks like there are references on the other platforms (Mac,
gtk)
> > that need to be removed from the chromium code first too.  If you prefer I'm
> > happy to land the blink side after you've cleaned up all the references in
> > chromium.
> 
> Oh, sorry I thought I did that but apparently the latest patch is missing
them.
> Uploaded new patch. PTAL.

Thanks.  I'm building now with the blink side removed
(https://codereview.chromium.org/137183002) to make sure we didn't miss anything
else.  I'll let you know shortly.

[Rick Byers, 2014-01-13 20:01:10]

On 2014/01/13 19:16:16, Rick Byers wrote:
> On 2014/01/13 18:53:43, oshima wrote:
> > On 2014/01/13 18:46:56, Rick Byers wrote:
> > > On 2014/01/13 18:24:48, oshima wrote:
> > > > On 2014/01/13 17:18:43, abarth wrote:
> > > > > On 2014/01/13 16:33:22, Rick Byers wrote:
> > > > > > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these
> cursor
> > > > types
> > > > > > are only available on Windows
> > > > > 
> > > > > ^^^ This option seems best.
> > > 
> > > Thanks Adam.
> > > 
> > > > I agree, but this CL still is still needed and has to be landed first
> right?
> > > 
> > > Yes.  And it looks like there are references on the other platforms (Mac,
> gtk)
> > > that need to be removed from the chromium code first too.  If you prefer
I'm
> > > happy to land the blink side after you've cleaned up all the references in
> > > chromium.
> > 
> > Oh, sorry I thought I did that but apparently the latest patch is missing
> them.
> > Uploaded new patch. PTAL.
> 
> Thanks.  I'm building now with the blink side removed
> (https://codereview.chromium.org/137183002) to make sure we didn't miss
anything
> else.  I'll let you know shortly.

Shoot, looks like pepper exposes these - search for eg.
PP_MOUSECURSOR_TYPE_MIDDLEPANNING.  In theory a pepper app on ChromeOS could
presumably be using this cursor type, right?  I highly doubt that's happening in
practice.  But it's going to require a bit of digging on the pepper side to
determine if we can really remove these.  Sorry I didn't notice this sooner (I
got a compiler error here once I had removed the blink definitions).

[oshima, 2014-01-13 21:12:32]

On 2014/01/13 20:01:10, Rick Byers wrote:
> On 2014/01/13 19:16:16, Rick Byers wrote:
> > On 2014/01/13 18:53:43, oshima wrote:
> > > On 2014/01/13 18:46:56, Rick Byers wrote:
> > > > On 2014/01/13 18:24:48, oshima wrote:
> > > > > On 2014/01/13 17:18:43, abarth wrote:
> > > > > > On 2014/01/13 16:33:22, Rick Byers wrote:
> > > > > > > 2) Add #ifdef OS_WIN to blink (and the webkit API) so that these
> > cursor
> > > > > types
> > > > > > > are only available on Windows
> > > > > > 
> > > > > > ^^^ This option seems best.
> > > > 
> > > > Thanks Adam.
> > > > 
> > > > > I agree, but this CL still is still needed and has to be landed first
> > right?
> > > > 
> > > > Yes.  And it looks like there are references on the other platforms
(Mac,
> > gtk)
> > > > that need to be removed from the chromium code first too.  If you prefer
> I'm
> > > > happy to land the blink side after you've cleaned up all the references
in
> > > > chromium.
> > > 
> > > Oh, sorry I thought I did that but apparently the latest patch is missing
> > them.
> > > Uploaded new patch. PTAL.
> > 
> > Thanks.  I'm building now with the blink side removed
> > (https://codereview.chromium.org/137183002) to make sure we didn't miss
> anything
> > else.  I'll let you know shortly.
> 
> Shoot, looks like pepper exposes these - search for eg.
> PP_MOUSECURSOR_TYPE_MIDDLEPANNING.  In theory a pepper app on ChromeOS could
> presumably be using this cursor type, right?  I highly doubt that's happening
in
> practice.  But it's going to require a bit of digging on the pepper side to
> determine if we can really remove these.  Sorry I didn't notice this sooner (I
> got a compiler error here once I had removed the blink definitions).

ChromeOS does not have these cursors defined, so we're using X cursor if chrome
is really using it.
My guess is that ppapi is exposing them for the same reason that blink is
exposing them, but yeah we need to
confirm that.

I'll file a bug to track & resolve the issue. Thank you for testing.