In cases where we actually try to allocate a Shmem instance from a user-provided buffer size, we currently throw an AbortError instead of a RangeError. The WebGPU spec. states the definition of GPUBuffer that we should use a RangeError in all OOM cases:

Let data be ? CreateByteDataBlock(size).

Note: This may result in a RangeError being thrown. For consistency and predictability: …

Step to reproduce

1. Create a 64-bit (but not 32-bit) build of Firefox based on changes from bug 1838694 or later.

2. Run ./mach wpt '/_mozilla/webgpu/chunked/1/cts.https.html?q=webgpu:api,operation,buffers,map_oom:mappedAtCreation:*'. Observe that oom=true test cases fail with an error like as follows:

   EXPECTATION FAILED: subcase: usage=1
   THREW AbortError, instead of RangeError: AbortError: GPUDevice.createBuffer: Unable to allocate shmem of size 9007199254740984
   shouldThrow@https://gpuweb.github.io/cts/out/common/framework/fixture.js:266:50
   @https://gpuweb.github.io/cts/out/webgpu/api/operation/buffers/map_oom.spec.js:42:7
   

   Alternatively, one can currently use https://gpuweb.github.io/cts/standalone/?q=webgpu:api,operation,buffers,map_oom:mappedAtCreation:oom=true;* to reproduce.

Depends on D181691

https://hg.mozilla.org/mozilla-central/rev/f27563eb1417