From: Patrick Steinhardt <ps@pks.im>
Date: Wed, 29 Oct 2025 09:07:59 GMT
Subject: Re: [PATCH 03/14] hash: use uint32_t for object_id algorithm
Message-ID: <aQHZbx-2JCnCgL_d@pks.im>
In-Reply-To: <aQFg2tvS6Wf7WGd4@fruit.crustytoothpaste.net>

On Wed, Oct 29, 2025 at 12:33:30AM +0000, brian m. carlson wrote:
> On 2025-10-28 at 09:16:57, Patrick Steinhardt wrote:
> > An alternative would be to introduce an enum and set up bindgen so that
> > we can pull this enum into Rust. I'd personally favor that over using an
> > uint32_t as it conveys way more meaning. Have you considered this?
> 
> That would lead to problems because we zero-initialize some object IDs
> (and you see later in the series what problems that causes) and that
> will absolutely not work in Rust, since setting an enum to an invalid
> value is undefined behaviour.

We could of course try and represent the uninitialized state with a
third enum state. But it would probably make things awfully unergonomic
all over the place :/

Patrick