In overlay_CreateSwapchainKHR, if IsPresentModeSupported returns false (which it will when FIFO_RELAXED is not supported) it sets HUDElements.cur_present_mode to VK_PRESENT_MODE_FIFO_KHR but it does not update createInfo.presentMode. This can cause swapchain creation to fail since it's trying to use an unsupported present mode.
This updates the createInfo's presentMode as well to prevent this.
In GLIBC, this behavior is controlled by per-architecture definition
DL_RO_DYN_SECTION, which is only set to 1 on MIPS (according to the
comments, it's because of requirements of MIPS ABI) and RISC-V (which
seems to be a left-over of copying MIPS code, and kept because of
compatibility to older GLIBC versions).
Signed-off-by: Icenowy Zheng <uwu@icenowy.me>
Previously the summary file and the benchmark both had their own code
to calculate the avg/percentiles, so instead we unified so all cases
use the fps_metrics class
Inside steam runtime we should use `steam-runtime-launch-client --alongside-steam --host`
This should resolve issues where the user tries to use commands
that don't exist inside the runtime
this workflow will update the meson version in meson.build
when a new tag is created. This should cause less headache
for the developer who keeps forgetting to change the version manually
ASan complains about allocated size mismatch in ctor/dtor.
Scott Meyers says: "Polymorphic base classes should declare virtual destructors. If a class has any virtual functions, it should have a virtual destructor."