-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
clarify safety documentation of ptr::swap and ptr::copy #114794
Conversation
r? @m-ou-se (rustbot has picked a reviewer for you, use r? to override) |
library/core/src/intrinsics.rs
Outdated
/// * `src` must remain valid for reads even after `dst` is written, and vice versa. | ||
/// (In other words, there cannot be aliasing restrictions on the use of these pointers.) | ||
/// |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't quite understand what this meant in practice until I checked out #81005.
Maybe there's a different way of wording this?
Would it work to not add this bullet point, but instead add "during the entire call to swap()" to the first two bullet points? (Maybe with a clarifying note saying that this means that the src and dst regions may not overlap.)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
They may overlap though, the pointers just must allow overlapping access. So for instance passing ptr
and ptr.add(N)
is fine even if things overlap. However passing &mut *ptr
twice generates two fresh mutable references and now if things overlap, that lead to UB due to aliasing violations.
It's hard to precisely say this while the aliasing model is still being determined.
add "during the entire call to swap()" to the first two bullet points?
If we do that I think we have to say what happens during the call. So we could say something like
src
must be [valid] for reads ofcount * size_of::<T>()
bytes, and must remain valid even ifdst
is written to forcount * size_of::<T>()
bytes.
But that seems very similar to the current wording, so I am not sure if it is clear enough.
Not sure if this wording is any better, but it's the best I could come up with so far. |
@m-ou-se I think I replied to all your comments -- what's the next step for this PR? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good, thanks!
Small nit: Would it be better to say even when
or even while
instead of even if
? (The if
makes it sound a bit like those things only happen conditionally.)
r=me either way.
My mother tongue uses the same word for "if" and "when" so this is probably just me missing a connotation that will be obvious to native speakers. I changed the wording. @bors r=m-ou-se |
@bors rollup |
clarify safety documentation of ptr::swap and ptr::copy Closes rust-lang#81005
…iaskrgr Rollup of 3 pull requests Successful merges: - rust-lang#114794 (clarify safety documentation of ptr::swap and ptr::copy) - rust-lang#115397 (Add support to return value in StableMIR interface and not crash due to compilation error) - rust-lang#115559 (implied bounds: do not ICE on unconstrained region vars) Failed merges: - rust-lang#115532 (Implement SMIR generic parameter instantiation) r? `@ghost` `@rustbot` modify labels: rollup
Closes #81005