Closed r-potter closed 1 week ago
Author r-potter not on autobuild list. Waiting for curator authorization before starting CI build.
Author r-potter not on autobuild list. Waiting for curator authorization before starting CI build.
Author r-potter not on autobuild list. Waiting for curator authorization before starting CI build.
Author r-potter not on autobuild list.
I added you to the list and manually kicked off CI as well
CI Vulkan-ValidationLayers build queued with queue ID 206605.
CI Vulkan-ValidationLayers build # 16929 running.
The following tests failed:
[ FAILED ] NegativeParent.Instance_PhysicalDeviceAndSurface [ FAILED ] NegativeParent.Instance_DeviceAndSurface [ FAILED ] NegativeParent.Instance_Surface_2 [ FAILED ] NegativeWsi.LeakASwapchain
However, they are all expecting surface creation to have succeeded in a headless environment (i.e. they only worked due to silently swallowing errors when XCB fails to open a valid connection). If we want this to work, then one approach would be to make use VK_EXT_headless_surface (https://gitlab.freedesktop.org/mesa/vulkan-wsi-layer/-/blob/master/README.md)
CI Vulkan-ValidationLayers build # 16929 passed.
I think a better fix (for sure simpler) is going
- ASSERT_EQ(VK_SUCCESS, CreateSurface(surface_context, instance2_surface.handle, instance2));
+ if (CreateSurface(surface_context, instance2_surface.handle, instance2) != VK_SUCCESS) {
+ GTEST_SKIP() << "Cannot create 2nd surface";
+ }
on them all
CI Vulkan-ValidationLayers build queued with queue ID 206654.
CI Vulkan-ValidationLayers build # 16930 running.
CI Vulkan-ValidationLayers build # 16930 passed.
CI Vulkan-ValidationLayers build queued with queue ID 206684.
CI Vulkan-ValidationLayers build # 16931 running.
CI Vulkan-ValidationLayers build # 16931 passed.
Without this change, running tests headlessly via SSH will fail to create a valid surface, but report success and subsequently crash in the ICD.