place "place projection" definition in glossary

[tshepang]

The term is defined in 2 separate places, and the definitions do not match.

[ehuss]

I'm not sure these are intended to be the same.

@RalfJung Is the definition of place projection intended to be the same for these two scenarios (closure capture and UB pointer arithmetic)?

[RalfJung]

Yeah this changes the opsem page to suddenly include pointer deref's and that doesn't make sense in that context (since it is talking about pointer arithmetic, and a deref isn't pointer arithmetic).

[tshepang]

since this is about pointer arithmetic, should field expression be removed from the more narrow definiton of place expression

[RalfJung]

We should probably reword the UB page to use a different term, since it only refers to a few specific place projections.

[tshepang]

would this work

A tuple index expression or an array/slice index expression that violates the (safety) requirements of in-bounds pointer arithmetic.

[RalfJung]

You forgot "field expression".

I would suggest something like #2253.

[rustbot]

☔ The latest upstream changes (possibly #2203) made this pull request unmergeable. Please resolve the merge conflicts.

[tshepang]

You forgot "field expression".

I actually thought it was accidental, since I assumed that, given offset works only with integers, then only APIs taking integers were connected to this rule.

[RalfJung]

Please don't make silent assumptions. Instead, ask for clarification. :)

It was very much deliberate. (*ptr).field is UB if field has offset N and that offset goes beyond the bounds of the allocation ptr is derived from.