| Commit-Queue | +1 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
+msisov@ could you help on reviewing this? thanks
// TODO remove workaround after libdrm in sysroot is updatedI assume this is blocking this CL from landing?
Is there any CL under review addressing it? If not, should we add a TODO+crbug to track it?
DrmIoctlWrapper* drm() const { return drm_.get(); }Lets avoid using const for functions that return pointers to non-const.
static bool fail_on_syncobj_handle_to_fd;please use methods, unless the style guide mentions exceptions for mock/test classes?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
+msisov@ could you help on reviewing this? thanks
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
lgtm given Nick's comments are addressed
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
static bool fail_on_syncobj_handle_to_fd;please use methods, unless the style guide mentions exceptions for mock/test classes?
Per offline discussion and after looking at how these are used in the subsequent CL, I believe it complies with the coding style. So, please, just add a comment above them advice about their usage/reset requirements or typical usage. thanks
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
// TODO remove workaround after libdrm in sysroot is updatedI assume this is blocking this CL from landing?
Is there any CL under review addressing it? If not, should we add a TODO+crbug to track it?
I think it's not a blocker as this just adds the API here if the `DRM_IOCTL_SYNCOBJ_EVENTFD` macro is not defined in libdrm version being used.
I added a crbug here.
Lets avoid using const for functions that return pointers to non-const.
Done.
static bool fail_on_syncobj_handle_to_fd;Nick Yamaneplease use methods, unless the style guide mentions exceptions for mock/test classes?
Per offline discussion and after looking at how these are used in the subsequent CL, I believe it complies with the coding style. So, please, just add a comment above them advice about their usage/reset requirements or typical usage. thanks
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
// TODO remove workaround after libdrm in sysroot is updatedOrko GaraiI assume this is blocking this CL from landing?
Is there any CL under review addressing it? If not, should we add a TODO+crbug to track it?
I think it's not a blocker as this just adds the API here if the `DRM_IOCTL_SYNCOBJ_EVENTFD` macro is not defined in libdrm version being used.
I added a crbug here.
Ack.
+thomasanderson@ who might help with the sysroot update I think.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
class DrmIoctlWrapper {Can this be named to specify for syncobj ioctl? I do not foresee that we do other drm ioctl's here.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Structurally drm syncobj related stuff is a gpu process responsibility, I think it's possible that we can keep drm node in the gpu process and transport the syncobj for the buffer from ozone/wayland/gpu to ozone/wayland/host, in a future possibility that shared_image maybe exporting timeline to syncobj.
But I understand it's an over-optimization so we don't want to do it now, at least we right now there's no syncobj equivalent of GpuFenceHandle.
class DrmIoctlWrapper {Can this be named to specify for syncobj ioctl? I do not foresee that we do other drm ioctl's here.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Structurally drm syncobj related stuff is a gpu process responsibility, I think it's possible that we can keep drm node in the gpu process and transport the syncobj for the buffer from ozone/wayland/gpu to ozone/wayland/host, in a future possibility that shared_image maybe exporting timeline to syncobj.
But I understand it's an over-optimization so we don't want to do it now, at least we right now there's no syncobj equivalent of GpuFenceHandle.
That sounds good. Yeah I agree in future syncobj support could be added to chromium's GPU fence (or equivalent), and we could do the wait using that in the GPU process.
class DrmIoctlWrapper {Orko GaraiCan this be named to specify for syncobj ioctl? I do not foresee that we do other drm ioctl's here.
Acknowledged
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
@fang...@chromium.org The +1 got removed due to the renaming. Please see if all is good now.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Commit-Queue | +2 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
linux_drm_syncobj_v1: Part 4: Add DrmIoctlWrapper
This is a change in preparation for the linux_drm_syncobj_v1
implementation.
Summary of changes:
- Add DrmIoctlWrapper class which wraps libDrm ioctl APIs.
- Set the drm wrapper in WaylandBufferManagerHost from
OzonePlatformWayland::InitializeUI().
- Add mock for testing.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Good job thanks for all the hard work
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Thanks again
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |