Re: [PATCH] ui/dbus: Implement damage regions for GL

["Philippe Mathieu-Daudé" <philmd@linaro.org>]

Hi,

On 9/5/23 13:59, Bilal Elmoussaoui wrote:
> From: Christian Hergert <chergert@redhat.com>
> 
> Currently, when using `-display dbus,gl=on` all updates to the client
> become "full scanout" updates, meaning there is no way for the client to
> limit damage regions to the display server.
> 
> Instead of using an "update count", this patch tracks the damage region
> and propagates it to the client.
> 
> This was less of an issue when clients were using GtkGLArea for
> rendering,
> as you'd be doing full-surface redraw. To be efficient, the client needs
> both a DMA-BUF and the damage region to be updated.
> 
> In the future, when additional methods are allowed on the D-Bus
> interface,
> this should likely be updated to send damage regions as a single RPC to
> avoid additional message processing.
> 
> Currently, Linux does not propagate damage rectangles when using the
> virtio-gpu drm driver. That means compositors such as Mutter which
> utilize
> drmModePageFlip() will be sending full or near-full surface damages.
> 
> https://lists.freedesktop.org/archives/dri-devel/2023-March/395164.html
> contains a patch to fix that too.
> 
> Signed-off-by: Bilal Elmoussaoui <belmouss@redhat.com>
> ---
>   meson.build        |  8 ++++++++
>   ui/dbus-listener.c | 25 +++++++++++++++++++------
>   ui/meson.build     |  2 +-
>   3 files changed, 28 insertions(+), 7 deletions(-)
> 
> diff --git a/meson.build b/meson.build
> index 229eb585f7..72678ef78e 100644
> --- a/meson.build
> +++ b/meson.build
> @@ -1761,6 +1761,14 @@ dbus_display = get_option('dbus_display') \
>              error_message: '-display dbus is not available on Windows') \
>     .allowed()

^ dbus strictly required deps, ...

> +cairo = not_found
> +if dbus_display
> +  cairo = dependency('cairo',
> +                     kwargs: static_kwargs,
> +                     method: 'pkg-config',
> +                    )

cairo declared as optional dep, ...

> +endif
> +
>   have_virtfs = get_option('virtfs') \
>       .require(targetos == 'linux' or targetos == 'darwin',
>                error_message: 'virtio-9p (virtfs) requires Linux or macOS') \
> diff --git a/ui/dbus-listener.c b/ui/dbus-listener.c
> index 911acdc529..047be5cb3a 100644
> --- a/ui/dbus-listener.c
> +++ b/ui/dbus-listener.c
> @@ -25,6 +25,7 @@
>   #include "qemu/error-report.h"
>   #include "sysemu/sysemu.h"
>   #include "dbus.h"
> +#include <cairo.h>

cairo used unconditionally.

Shouldn't we now declared it as a strict dependency in meson?

>   #include <gio/gunixfdlist.h>
>   
>   #ifdef CONFIG_OPENGL