[ptr] Document maximum allocation size

[joshlf]

Partially addresses rust-lang/unsafe-code-guidelines#465

[rustbot]

r? @joshtriplett

(rustbot has picked a reviewer for you, use r? to override)

[RalfJung]

FCP passed in #116988 so you can now assume that null is 0 here.

[the8472]

These are fairly fundamental requirements of rust. Shouldn't they also go into the reference?

[RalfJung]

Yeah it's the usual issue where the stdlib docs is more likely to be actually visible but conceptually the reference is where this belongs. But given that this already defines "allocated object", this seems like a sensible place.

I think what is left here is uncontroversial:

It is guaranteed that an allocated object never spans more than isize::MAX bytes.
An allocated object cannot contain [null()] (i.e., the address with the numerical value 0) and
cannot contain the last (usize::MAX) byte of the address space.

But let's ping some people and nominate to see if anyone disagrees.
Cc @rust-lang/opsem @rust-lang/lang

r? @RalfJung

[the8472]

Yeah it's the usual issue where the stdlib docs is more likely to be actually visible but conceptually the reference is where this belongs.

There are those newfangled thingies called hyperlinks. 😉

Imo the reference should at least contain the requirement while std spells out the consequences of that. How it interacts with Layout, Allocator, pointer methods etc.

[RalfJung]

Where in the reference would you put this?

[the8472]

The type layout page already contains a section on size and alignment. The maximum allocation size could be specified there. Or maybe under the currently quite anemic memory model page.

[RalfJung]

The type layout page already contains a section on size and alignment. The maximum allocation size could be specified there.

No that's definitely the wrong page, this guarantee has nothing to do with types nor layouts.

Or maybe under the currently quite anemic memory model page.

Yeah that makes more sense. It's just hard to specify a memory model in a piecemeal way so it's somewhat unclear what one should say there.

[tmandry]

Discussed in lang team meeting.

@rfcbot merge

[rfcbot]

Team member @tmandry has proposed to merge this. The next step is review by the rest of the tagged team members:

• @joshtriplett
• @nikomatsakis
• @pnkfelix
• @scottmcm
• @tmandry

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

cc @rust-lang/lang-advisors: FCP proposed for lang, please feel free to register concerns.
See this document for info about what commands tagged team members can give me.

[RalfJung]

@bors r-

Sorry, one last nit -- @joshlf can you squash then commits?

[scottmcm]

Oh, good point Ralf. I didn't notice it's currently 9 commits.

[joshlf]

Done. You can see on the diff that the squash happened successfully without modifying the PR contents.

[scottmcm]

Thanks!

@bors r+ rollup

[bors]

📌 Commit 293b5cb has been approved by scottmcm

It is now in the queue for this repository.