Currently, apps using WebRTC for window capture only get the benefits of
using CroppingWindowCapturer on Windows (described below) after changing
calls to DesktopCapturer::CreateWindowCapturer to instead call
CroppingWindowCapturer::CreateCapturer. This change adds a new flag to
DesktopCaptureOptions to allow opting in to the faster capture-screen-
and-crop path via the older & more discoverable API.
Benefits of using CroppingWindowCapturer's capture-screen-and-crop path
when possible:
1) It's significantly faster, up to ~36ms/frame (~160x) faster than the
capture-window-contents path in my testing (more details are in the
bug). This difference increased with the recent fix for
https://crbug.com/webrtc/10734 .
2) It allows capture of menus & tooltips (plus dialogs if
https://crbug.com/webrtc/10767 is fixed), partially mitigating
https://crbug.com/980864 .
Downsides of using it:
1) It may inadvertently capture occluding windows that aren't detected
properly, e.g. some system UI: https://crbug.com/webrtc/10835 .
2) It may capture some neighboring regions when moving/resizing the
captured window.
The new flag is not enabled by default, so the default behavior is
unchanged. This could perhaps be revisited after addressing
https://crbug.com/webrtc/10835 .
Bug: webrtc:10825
Change-Id: Ib77e5facc7240c5df311fe1fe204d0d8ea22a96a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/146823
Commit-Queue: Jamie Walch <jamiewalch@chromium.org>
Reviewed-by: Jamie Walch <jamiewalch@chromium.org>
Cr-Commit-Position: refs/heads/master@{#28695}
366ac4d84e