Creating tiny reusable buffers runs out of memory

[domoritz]

What version of Bun is running?

1.0.19+906f86d6f

What platform is your computer?

Darwin 23.4.0 arm64 arm

What steps can reproduce the bug?

> bun repl
Welcome to Bun v1.0.19
Type ".help" for more information.
[!] Please note that the REPL implementation is still experimental!
    Don't consider it to be representative of the stability or behavior of Bun overall.
> const a = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const b = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const c = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const d = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const e = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const f = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const g = new ArrayBuffer(0, {maxByteLength: 4294967295})
undefined
> const h = new ArrayBuffer(0, {maxByteLength: 4294967295})
1 | "use strict";void 0;
2 | const h = new ArrayBuffer(0, {
    maxByteLength: 4294967295
});
              ^
RangeError: Out of memory

>

What is the expected behavior?

The program should not run out of memory. https://stackoverflow.com/questions/77029625/memory-allocated-for-resizeable-javascript-arraybuffer explains that the memory should not be allocated and my buffers are all of size 0.

What do you see instead?

No response

Additional information

No response

[Electroid]

Interesting, it’s possible that instead of setting the max length, it is actually allocating that much up front.

[domoritz]

That could be the case but that would be problematic since that somewhat defeats the purpose of a resizable buffer.

[domoritz]

Can you confirm that this is a bug? We are making an Apache Arrow release soon and it's using reusable buffers by default. If this isn't a bug, I would want to revert the feature asap. Right now, we don't have many bun users so this issue isn't a release blocker.

[domoritz]

Oh, looks like Safari has the same issue so this might be a blocker.