Here's what we're now doing on the renderer side:
- In the render widget, we now keep all invalidation as an independent
sub-rect in a vector as opposed to union them as we used to.
- Once, we are ready to paint:
- We create a bitmap that is the size of the whole view and wrap it in
a canvas.
- For each invalidation sub-rec
- We set a clip rect (on the canvas) to the invalidation sub-rect.
- And then we call WebWidget::Paint() with the canvas and the
sub-rect.
Anything smells fishy in there?
BYE
MAD
On Thu, May 14, 2009 at 3:14 PM, Avi Drissman <[email protected]> wrote:
> We haven't drawn one big rectangle for a long while. It took me a while to
> fiddle with the transforms to get it all right, but we correctly blit small
> rectangles for small updates (e.g. cursor blinks).
>
> OTOH, Mike has a point, that getting the coordinate transforms correct is
> important, tricky, and a pain in the rear. That still might be the issue.
> But if you're just calling into the RWHV repeatedly, it should handle it
> just fine.
>
> Avi
>
> On Thu, May 14, 2009 at 12:02 PM, Mike Pinkerton
> <[email protected]>wrote:
>
>>
>> Just off the top of my head, it could be a coordinate problem. Cocoa
>> uses a bottom-left coordinate system (the bottom left of the view is
>> 0,0 as opposed to the top left). Since before we were just drawing a
>> single big rectangle, it might not have mattered if we forgot to do
>> the coordinate flipping. Again, just a wild guess based on things I've
>> seen in the past and differences in the drawing paths.
>>
>> On Thu, May 14, 2009 at 2:33 PM, Marc-Andre Decoste <[email protected]>
>> wrote:
>> > Salut Adam,
>> > The problem is that I'm missing some refreshes... I see them more
>> clearly
>> > when interacting with Google maps, or a little HTML page I created that
>> has
>> > buttons that change the content of scattered table cells to force
>> multiple
>> > individual rect invalidations in one shot... And those don't display if
>> I
>> > try to paint individual rects in the bitmap...
>> > But it all works fine if I paint the whole bitmap, and then only draw
>> the
>> > sub-rects from the bitmap... This is why I think the new Mac specific
>> code
>> > is correct... Otherwise, I don't see how I would see the correct
>> refreshes
>> > in that latter case... Unless I missed something...
>> >
>> > Thanks!
>> > BYE
>> > MAD
>> > On Thu, May 14, 2009 at 2:27 PM, Adam Langley <[email protected]> wrote:
>> >>
>> >> On Thu, May 14, 2009 at 11:21 AM, Marc-Andre Decoste <[email protected]>
>> >> wrote:
>> >> > Since there is a change in the Windows specific code of the
>> renderer
>> >> > host
>> >> > backing store (to draw the sub-rects of the passed bitmap), I thought
>> my
>> >> > goof could have been in my adaptation of the Mac version of that code
>> >> > (http://codereview.chromium.org/108040/diff/2038/3040). But it seems
>> to
>> >> > be
>> >> > working fine if I disable the other part of that change (which is to
>> >> > paint
>> >> > only sub-rectangles in the bitmap). So if I paint a complete bitmap,
>> and
>> >> > pass it to the host with a list of sub-rectagles to draw... It
>> works...
>> >> > So I
>> >> > guess the goof is in the code that paint sub-rectangles in the
>> bitmap...
>> >> > But this code works fine on Windows and is not done in a Windows
>> >> > specific
>> >> > way... So I'm kind of lost and don't know where to look... Anybody
>> has a
>> >> > clue here?
>> >>
>> >> What's the problem that you see?
>> >>
>> >> Surely it could still be in the Mac specific code if you're painting
>> >> from the wrong regions in the bitmap?
>> >>
>> >>
>> >> AGL
>> >
>> >
>> > >
>> >
>>
>>
>>
>> --
>> Mike Pinkerton
>> Mac Weenie
>> [email protected]
>>
>> >>
>>
>
--~--~---------~--~----~------------~-------~--~----~
Chromium Developers mailing list: [email protected]
View archives, change email options, or unsubscribe:
http://groups.google.com/group/chromium-dev
-~----------~----~----~----~------~----~------~--~---
