iOS simulator shader compilation failure, subsequent crash

[warmenhoven]

As a result of c37a880, when I run in the iOS simulator (iPhone 16 Pro Max, iOS 18.0), the following shader is unable to compile; I've also attached a log:

#version 450

layout(local_size_x = 16, local_size_y = 8) in;

layout(set = 0, binding = 3, rgba8) writeonly uniform image2D uImage;
layout(set = 0, binding = 4, std430) readonly buffer SSBO
{
	uint packed_pixels[];
};

layout(set = 0, binding = 0, std140) uniform UBO
{
	uvec2 resolution;
	uint word_stride;
};

vec3 rgb565_to_rgb888(uint word)
{
	return vec3((uvec3(word) >> uvec3(11, 5, 0)) & uvec3(31, 63, 31)) / vec3(31.0, 63.0, 31.0);
}

void main()
{
	// We work on two horizonal pixels in parallel since we cannot rely on 16-bit storage.
	uvec2 first_input_pixel = gl_GlobalInvocationID.xy;
	uvec2 first_output_pixel = first_input_pixel * uvec2(2, 1);

	if (all(lessThan(first_output_pixel, resolution)))
	{
		uint word = packed_pixels[first_input_pixel.y * word_stride + first_input_pixel.x];
		uint lo_word = word & 0xffffu;
		uint hi_word = word >> 16u;
		vec3 lo = rgb565_to_rgb888(lo_word);
		vec3 hi = rgb565_to_rgb888(hi_word);

		imageStore(uImage, ivec2(first_output_pixel), vec4(lo, 1.0));
		if (first_output_pixel.x + 1u < resolution.x)
			imageStore(uImage, ivec2(first_output_pixel) + ivec2(1, 0), vec4(hi, 1.0));
	}
}

The compiler error is:

[mvk-error] VK_ERROR_INITIALIZATION_FAILED: Shader library compile failed (Error code 3):
program_source:27:71: error: cannot reserve 'buffer' resource location at index 9
kernel void main0(constant UBO& _49 [[buffer(9)]], const device SSBO& _65 [[buffer(9)]], texture2d<float, access::write> uImage [[texture(0)]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]])
                                                                      ^
.
[mvk-error] VK_ERROR_INVALID_SHADER_NV: Compute shader function could not be compiled into pipeline. See previous logged error.

cross.log

Later, it leads to the following crash:

(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = signal SIGSEGV
  * frame #0: 0x0000000133a606e0 MoltenVK`MVKPipeline::getDescriptorSetCount(this=0x0000000000000000) at MVKPipeline.h:146:44
    frame #1: 0x0000000133a60430 MoltenVK`MVKResourcesCommandEncoderState::encodeMetalArgumentBuffer(this=0x000000016ce012f0, stage=kMVKShaderStageCompute) at MVKCommandEncoderState.mm:669:29
    frame #2: 0x0000000133a63ef0 MoltenVK`MVKComputeResourcesCommandEncoderState::encodeImpl(this=0x000000016ce012f0, (null)=0) at MVKCommandEncoderState.mm:1163:2
    frame #3: 0x000000013395456c MoltenVK`MVKCommandEncoderState::encode(this=0x000000016ce012f0, stage=0) at MVKCommandEncoderState.h:93:9
    frame #4: 0x0000000133954828 MoltenVK`MVKCommandEncoder::finalizeDispatchState(this=0x000000016ce00450) at MVKCommandBuffer.mm:774:28
    frame #5: 0x0000000133ab4b24 MoltenVK`MVKCmdDispatch::encode(this=0x0000600000caaf10, cmdEncoder=0x000000016ce00450) at MVKCmdDispatch.mm:47:14
    frame #6: 0x00000001339521a0 MoltenVK`MVKCommandEncoder::encodeCommandsImpl(this=0x000000016ce00450, command=0x0000600000caaf10) at MVKCommandBuffer.mm:379:18
    frame #7: 0x00000001339519ac MoltenVK`MVKCommandEncoder::encodeCommands(this=0x000000016ce00450, command=0x0000600002981260) at MVKCommandBuffer.mm:372:3
    frame #8: 0x00000001339517c0 MoltenVK`MVKCommandEncoder::encode(this=0x000000016ce00450, mtlCmdBuff=0x000060000370dd40, pEncodingContext=0x000060000370de48) at MVKCommandBuffer.mm:346:5
    frame #9: 0x0000000133951a80 MoltenVK`MVKCommandBuffer::submit(this=0x0000000103db72b0, cmdBuffSubmit=0x000060000370de00, pEncodingContext=0x000060000370de48) at MVKCommandBuffer.mm:240:11
    frame #10: 0x0000000133a2b138 MoltenVK`MVKQueueFullCommandBufferSubmission<1ul>::submitCommandBuffers(this=0x000060000370de00) at MVKQueue.mm:659:59
    frame #11: 0x0000000133a28bfc MoltenVK`MVKQueueCommandBufferSubmission::execute(this=0x000060000370de00) at MVKQueue.mm:479:2
    frame #12: 0x0000000133a2590c MoltenVK`execute(qSubmit=0x000060000370de00) at MVKQueue.mm:72:98
    frame #13: 0x0000000133a25840 MoltenVK`MVKQueue::submit(this=0x0000000103e0ea30, qSubmit=0x000060000370de00) at MVKQueue.mm:99:10
    frame #14: 0x0000000133a263a0 MoltenVK`VkResult MVKQueue::submit<VkSubmitInfo>(this=0x0000000103e0ea30, submitCount=1, pSubmits=0x000000016ce030c0, fence=0x0000600003308b40, cmdUse=kMVKCommandUseQueueSubmit) at MVKQueue.mm:138:28
    frame #15: 0x00000001338d24a8 MoltenVK`vkQueueSubmit(queue=0x0000000103e0ea48, submitCount=1, pSubmits=0x000000016ce030c0, fence=0x0000600003308b40) at vulkan.mm:434:24