[PATCH v2] winex11.drv: Handle X error from vkGetRandROutputDisplayEXT().

[Paul Gofman]

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=49407

Nvidia implementation of vkGetRandROutputDisplayEXT() generates X exception
when rrOutput is from different provider (or if rrOutput is just
invalid). That can happen on certain multiple GPU configurations, on which
Wine is currently unable to initialize display driver and thus create
any window.

According to Vulkan spec, vkGetRandROutputDisplayEXT is supposed to just
return VK_NULL_HANDLE if there is no corresponding VkDisplayKHR. But it is
probably better to workaround the problem to avoid long standing regression.

Signed-off-by: Paul Gofman <pgofman(a)codeweavers.com>
---
v2:
    - add XSync();
    - add comment for the patch.

 dlls/winex11.drv/xrandr.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/dlls/winex11.drv/xrandr.c b/dlls/winex11.drv/xrandr.c
index f61a5b190e9..b1a43e48eab 100644
--- a/dlls/winex11.drv/xrandr.c
+++ b/dlls/winex11.drv/xrandr.c
@@ -756,9 +756,11 @@ static void get_vulkan_device_uuid( GUID *uuid, const XRRProviderInfo *provider_
     {
         for (output_idx = 0; output_idx < provider_info->noutputs; ++output_idx)
         {
+            X11DRV_expect_error( gdi_display, XRandRErrorHandler, NULL );
             vr = pvkGetRandROutputDisplayEXT( vk_physical_devices[device_idx], gdi_display,
                                               provider_info->outputs[output_idx], &vk_display );
-            if (vr != VK_SUCCESS || vk_display == VK_NULL_HANDLE)
+            XSync( gdi_display, FALSE );
+            if (X11DRV_check_error() || vr != VK_SUCCESS || vk_display == VK_NULL_HANDLE)
                 continue;
 
             memset( &id, 0, sizeof(id) );
-- 
2.26.2

[Liam Middlebrook]

Agreed that this patch makes sense to avoid long-standing breakage for 
WINE users with this configuration. Thanks for filing a bug directly 
with us.

Signed-off-by: Liam Middlebrook <lmiddlebrook(a)nvidia.com>


Thanks,

Liam Middlebrook

On 6/18/20 7:28 AM, Paul Gofman wrote:
> Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=49407
>
> Nvidia implementation of vkGetRandROutputDisplayEXT() generates X exception
> when rrOutput is from different provider (or if rrOutput is just
> invalid). That can happen on certain multiple GPU configurations, on which
> Wine is currently unable to initialize display driver and thus create
> any window.
>
> According to Vulkan spec, vkGetRandROutputDisplayEXT is supposed to just
> return VK_NULL_HANDLE if there is no corresponding VkDisplayKHR. But it is
> probably better to workaround the problem to avoid long standing regression.
>
> Signed-off-by: Paul Gofman <pgofman(a)codeweavers.com>
> ---
> v2:
>      - add XSync();
>      - add comment for the patch.
>
> dlls/winex11.drv/xrandr.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/dlls/winex11.drv/xrandr.c b/dlls/winex11.drv/xrandr.c
> index f61a5b190e9..b1a43e48eab 100644
> --- a/dlls/winex11.drv/xrandr.c
> +++ b/dlls/winex11.drv/xrandr.c
> @@ -756,9 +756,11 @@ static void get_vulkan_device_uuid( GUID *uuid, const XRRProviderInfo *provider_
>       {
>           for (output_idx = 0; output_idx < provider_info->noutputs; ++output_idx)
>           {
> +            X11DRV_expect_error( gdi_display, XRandRErrorHandler, NULL );
>               vr = pvkGetRandROutputDisplayEXT( vk_physical_devices[device_idx], gdi_display,
>                                                 provider_info->outputs[output_idx], &vk_display );
> -            if (vr != VK_SUCCESS || vk_display == VK_NULL_HANDLE)
> +            XSync( gdi_display, FALSE );
> +            if (X11DRV_check_error() || vr != VK_SUCCESS || vk_display == VK_NULL_HANDLE)
>                   continue;
>
> memset( &id, 0, sizeof(id) );