Program panics when waitAny is used in C++

[nahahahah]

Hello, I'm having trouble with buffer mapping. I'm trying to understand buffer mapping by testing the different options available but when I try to use waitAny to wait for the Future to get a result, the program panics and it tells me that it's not implemented.

I'm using Elie Michel's C++ wrapper, I generate a webgpu.hpp header from their generate.py script and I'm compiling with MSVC. I am using wgpu-native v25.0.2.1.

Here's the whole program output :

<wgpu::Instance 00000200A7FAA0D0>
<wgpu::Adapter 00000200A837CFD0>
<wgpu::Device 00000200A82EBB00>
<wgpu::Queue 00000200A83813B0>
<wgpu::Buffer 00000200A82ED600>
<wgpu::Buffer 00000200A82ECD00>
<wgpu::CommandEncoder 00000200A82ECD40>

thread '<unnamed>' panicked at src\unimplemented.rs:230:5:
not implemented
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

thread '<unnamed>' panicked at library\core\src\panicking.rs:218:5:
panic in a function that cannot unwind
stack backtrace:
   0:     0x7ff71727fbf2 - <unknown>
   1:     0x7ff717258a0b - <unknown>
   2:     0x7ff71727df97 - <unknown>
   3:     0x7ff71727fa35 - <unknown>
   4:     0x7ff7172813ac - <unknown>
   5:     0x7ff717281142 - <unknown>
   6:     0x7ff717281f8f - <unknown>
   7:     0x7ff717281cf2 - <unknown>
   8:     0x7ff7172802ff - <unknown>
   9:     0x7ff71728192e - <unknown>
  10:     0x7ff717606ad5 - <unknown>
  11:     0x7ff717606b83 - <unknown>
  12:     0x7ff717606c01 - <unknown>
  13:     0x7ff717606933 - <unknown>
  14:     0x7fff1a410430 - _CxxFrameHandler3
  15:     0x7fff1a40342d - is_exception_typeof
  16:     0x7fff3e2c6006 - RtlCaptureContext2
  17:     0x7ff71760691a - <unknown>
  18:     0x7ff7176053ba - <unknown>
  19:     0x7ff7176057e0 - <unknown>
  20:     0x7fff3d6ae8d7 - BaseThreadInitThunk
  21:     0x7fff3e1dc34c - RtlUserThreadStart
thread caused non-unwinding panic. aborting.

Here's the thing I was trying :

struct OnBufferMappedContext {
    bool operationEnded = false;
    bool mappingSucceeded = false;
};

auto OnBufferMapped = [](WGPUMapAsyncStatus status, WGPUStringView message) {
    //OnBufferMappedContext& context = *reinterpret_cast<OnBufferMappedContext*>(userdata1);
    //context.operationEnded = true;
    if (status == WGPUMapAsyncStatus_Success) {
        std::cout << "Buffer mapped" << std::endl;
        //context.mappingSucceeded = true;
    }

    else {
        std::cerr << "Could not map buffer: " << status << std::endl;
        std::cerr << message.data << std::endl;
    }
};

OnBufferMappedContext mapContext {};

wgpu::Future bufferMapFuture = buffer2.mapAsync(wgpu::MapMode::Read, 0, 16 * sizeof(float), wgpu::CallbackMode::AllowProcessEvents, OnBufferMapped);

wgpu::FutureWaitInfo bufferMapFutureInfo {};
bufferMapFutureInfo.future = bufferMapFuture;

wgpu::WaitStatus status = instance.waitAny(1, &bufferMapFutureInfo, 200 * 1000);

while (!bufferMapFutureInfo.completed) {
    //std::cout << "Waiting..." << std::endl;
}

buffer2.unmap();

Did I do something wrong ? Thanks you kindly for your help.

ajvp

[nahahahah]

Ok, I am an idiot. It is indeed not implemented and it was right in front of my eyes in src. Sorry for that. Is there any workaround to check that a my buffer was mapped successfully (and by extension, use Future) ?

With my webgpu.hpp that was generated by generate.py in the given repo from Elie Michel (plus some editing of the file), I have this for mapAsync :

Future mapAsync(MapMode mode, size_t offset, size_t size, CallbackMode callbackMode, const Lambda& callback) const {
	auto* lambda = new Lambda(callback);
	auto cCallback = [](WGPUMapAsyncStatus status, WGPUStringView message, void* userdata1, void*) -> void {
		std::unique_ptr<Lambda> lambda(reinterpret_cast<Lambda*>(userdata1));
		(*lambda)(static_cast<MapAsyncStatus>(status), message);
	};
	BufferMapCallbackInfo callbackInfo;
	callbackInfo.nextInChain = nullptr;
	callbackInfo.mode = callbackMode;
	callbackInfo.callback = cCallback;
	callbackInfo.userdata1 = (void*)lambda;
	callbackInfo.userdata2 = nullptr;
	return wgpuBufferMapAsync(m_raw, static_cast<WGPUMapMode>(mode), offset, size, callbackInfo);
}

Are Future even usable at this point ? Thanks in advance for your help.

ajvp

[melody-rs]

Futures aren't really implemented yet, but your callbacks will still be called though! You can instead set a flag in your callback that indicates that the future completed as a workaround.
Something like this:

struct OnBufferMappedContext {
    bool operationEnded = false;
    bool mappingSucceeded = false;
};

auto OnBufferMapped = [](WGPUMapAsyncStatus status, WGPUStringView message) {
    OnBufferMappedContext& context = *reinterpret_cast<OnBufferMappedContext*>(userdata1);
    context.operationEnded = true;
    if (status == WGPUMapAsyncStatus_Success) {
        std::cout << "Buffer mapped" << std::endl;
        context.mappingSucceeded = true;
    }
    else {
        std::cerr << "Could not map buffer: " << status << std::endl;
        std::cerr << message.data << std::endl;
    }
};

OnBufferMappedContext mapContext {};
buffer2.mapAsync(wgpu::MapMode::Read, 0, 16 * sizeof(float), wgpu::CallbackMode::AllowProcessEvents, OnBufferMapped);

while (!mapContext.operationEnded) {
    instance.processEvents();
    std::cout << "Waiting..." << std::endl;
}

buffer2.unmap();

Though I'd make sure to use atomics since iirc callbacks can be called on any thread.

[nahahahah]

Seems good to me. Sorry for the answering delay. It will do for now. Do you have any idea of the status of Futures at the moment ?

Thanks,

ajvp