WIP: wayland: handle VK_SUBOPTIMAL_KHR returned by vkAcquireNextImageKHR()

[leandrohrb]

With the Vulkan WSI Wayland MR [1] that adds dma-buf hints support, vkcube may receive VK_SUBOPTIMAL_KHR when calling vkAcquireNextImageKHR().

In this case, we should recreate the swapchain and pass the old swapchain as argument. This results in an optimal modifier being selected to recreate the surface's buffers.

[1] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12226

cc: @fooishbar

[leandrohrb]

edit: this issue was related to a patch that I've added to the Vulkan WSI Wayland dma-buf feedback MR. This is fixed now.

Help required! From time to time I see a crash when recreate_swapchain() is called, this is the trace:

Thread 1 vkcube received signal SIGSEGV, Segmentation fault.
0x00007ffff7f5bdd8 in wl_proxy_marshal () from /usr/lib/libwayland-client.so.0
(gdb) backtrace
#0  0x00007ffff7f5bdd8 in wl_proxy_marshal () from /usr/lib/libwayland-client.so.0
#1  0x00007fffea17d8a0 in wl_buffer_destroy (wl_buffer=0x555555703270) at /usr/include/wayland-client-protocol.h:1990
#2  wsi_wl_swapchain_destroy (wsi_chain=0x5555556fdd20, pAllocator=0x5555556f7030) at ../src/vulkan/wsi/wsi_common_wayland.c:1245
#3  0x000055555555ab60 in recreate_swapchain (vc=0x7fffffffded0) at ../main.c:808
#4  0x000055555555bb5c in mainloop_wayland (vc=0x7fffffffded0) at ../main.c:1271
#5  0x000055555555d34b in mainloop (vc=0x7fffffffded0) at ../main.c:1726
#6  0x000055555555d49c in main (argc=1, argv=0x7fffffffe428) at ../main.c:1761

I'm not sure if I did something wrong in the swapchain destruction in vkcube, as I don't know much about Vulkan clients.

[rmader]

I think vkcube now lives in https://github.com/KhronosGroup/Vulkan-Tools. It also handles VK_SUBOPTIMAL_KHR but does not yet recreate the swapchain, see https://github.com/KhronosGroup/Vulkan-Tools/blob/master/cube/cube.c#L1053-L1056