GPU: recreate swapchain on window pixel size change event

[thatcosmonaut]

This avoids querying window size in a relatively hot function, SDL_AcquireGPUSwapchainTexture.

Also removes a pointless Vulkan surface extent check, the limits are guaranteed by the spec.

[slouken]

This seems good to me. We might need some locking in case something is recreating the swap chain while this event comes in on the main thread.

I assume this PR obsoletes #10984?

[thatcosmonaut]

This seems good to me. We might need some locking in case something is recreating the swap chain while this event comes in on the main thread.

I assume this PR obsoletes #10984?

It actually complements that PR - they should both go in. You're right, we should lock on the window data.

[slouken]

This seems good to me. We might need some locking in case something is recreating the swap chain while this event comes in on the main thread.
I assume this PR obsoletes #10984?

It actually complements that PR - they should both go in. You're right, we should lock on the window data.

Okay, then @Akaricchi's comment on that PR is valid, and that shouldn't be merged until it's resolved.

[thatcosmonaut]

@slouken do the mutexes in the latest commit look good to you? If so this can go in before the other PR.

[thatcosmonaut]

This is good to go once CI passes

[flibitijibibo]

Latest looks good - we could maybe put the watcher in gpu.c and make a driver function to be sure all drivers are always aware of resizes but that could be done later if needed.

[thatcosmonaut]

Yeah a NotifyResize function on sysgpu.h could be nice, but we can do that later.