Tracking issue for core::ptr::dangling/_mut (feature ptr_dangling)

[bluss]

Tracking issue for core::ptr::dangling/_mut (feature ptr_dangling)

• (To be) Implemented in PR Add functions core::ptr::dangling/_mut<T>() -> *const/*mut T #45527

[nagisa]

One issue I can see with this function is that it might return pointers that actually point somewhere even on systems where the 0th page is considered to be always unmapped.

[petrochenkov]

As I understand it the point of the function is to return something non-null and aligned, so it doesn't matter whether it's a valid pointer or not.
We can just specify that you are not allowed to dereference the pointer returned by ptr::dangling without guaranteeing anything about it validity (i.e. equality to other existing valid pointers).

[nagisa]

That seems fine. I was concerned because IIRC the PR specifically said something about them being valid and the documentation for the function also said something similar.

…

On Oct 30, 2017 11:13 PM, "Vadim Petrochenkov" ***@***.***> wrote: As I understand it the point of the function is to return something non-null and aligned, so it doesn't matter whether it's a valid pointer or not. We can just specify that you are not allowed to dereference the pointer returned by ptr::dangling without guaranteeing anything about it validity (i.e. equality to other existing valid pointers). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#45557 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AApc0se4c7fVhlH3da-gXfTz85UsMLrvks5sxjv4gaJpZM4QIJlZ> .

[nagisa]

Ah the mornings. The PR description had a proposal to name the function invalid_... And the description specifically claims it will not point to any object. I'd like the documentation changed so it doesn't attempt to guarantee the function will always return an address that's not pointing to a valid object.

…

On Oct 31, 2017 7:53 AM, "Simonas Kazlauskas" ***@***.***> wrote: That seems fine. I was concerned because IIRC the PR specifically said something about them being valid and the documentation for the function also said something similar. On Oct 30, 2017 11:13 PM, "Vadim Petrochenkov" ***@***.***> wrote: > As I understand it the point of the function is to return something > non-null and aligned, so it doesn't matter whether it's a valid pointer or > not. > We can just specify that you are not allowed to dereference the pointer > returned by ptr::dangling without guaranteeing anything about it > validity (i.e. equality to other existing valid pointers). > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#45557 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AApc0se4c7fVhlH3da-gXfTz85UsMLrvks5sxjv4gaJpZM4QIJlZ> > . >

[bluss]

If you want changes in the PR #45527, please write about them over there.

[SimonSapin]

This feature never landed and the implementation PR is closed. Closing this as well, feel free to reopen as appropriate.