2375. Is [iterator.requirements.general]/9 too broadly applied?

Section: 24.3.1 [iterator.requirements.general] Status: Resolved Submitter: Marshall Clow Opened: 2014-03-25 Last modified: 2021-06-23

Priority: 3

View all other issues in [iterator.requirements.general].

View all issues with Resolved status.

Discussion:

24.3.1 [iterator.requirements.general] p9 says:

Destruction of an iterator may invalidate pointers and references previously obtained from that iterator.

But the resolution of LWG issue 2360 specifically advocates returning *--temp; where temp is a local variable.

And 24.3.5.5 [forward.iterators] p6 says:

If a and b are both dereferenceable, then a == b if and only if *a and *b are bound to the same object.

which disallows "stashing" iterators (i.e, iterators that refer to data inside themselves).

So, I suspect that the restriction in p9 should only apply to input iterators, and can probably be moved into 24.3.5.3 [input.iterators] instead of 24.3.1 [iterator.requirements.general].

[2014-05-22, Daniel comments]

Given that forward iterators (and beyond) are refinements of input iterator, moving this constraint to input iterators won't help much because it would still hold for all refined forms.

[2021-06-23 Resolved by adoption of P0896R4 in San Diego. Status changed: New → Resolved.]

Proposed resolution: