P4049 R1 Relaxing and extending std::copy

[wg21bot]

P4049R0 Relaxing and extending std::copy (Giuseppe D'Angelo)

[foonathan]

SG9 saw this on Monday in Brno: https://wiki.isocpp.org/NotesSG9P4049R0

Poll: Ignoring the contiguous iterator case, we want Option A, well-defined but "garbled" behavior for overlapping.

SF	F	N	A	SA
0	0	1	2	2

Attendance: 10
Author: SA
Consenus against.

Poll: Ignoring the contiguous iterator case, we want Option B, erroneous behavior for overlapping.

SF	F	N	A	SA
4	1	0	1	0

Attendance: 10
Author: SF
Strong consensus in favor

A: I don't see how implementations will take advantage of this.

Poll: For contiguous and trivially copyable, we want erronoeus but the behavior that "mmemove" would do for overlapped access (not in the paper).

SF	F	N	A	SA
2	1	1	2	0

Attendance: 10
Author: A
No consensus.

Poll: For contiguous and trivially copyable, we want well-defined behavior namely the behavior that "memmove" would do for overlapped access (option E1).

SF	F	N	A	SA
1	3	2	0	0

Attendance: 10
Author: SF
Consensus.

Poll: We want to apply the contiguous and trivially copyable well-defined behavior (what "memmove" would do but using the regular assignment) also for non-trivially copyable types (option E2)

SF	F	N	A	SA
2	2	2	0	0

Attendance: 10
Author: SF
Strong consensus.

Poll: Forward P4049R0 with option B-E2 (as proposed in the paper) to LEWG.

SF	F	N	A	SA
3	2	1	0	0

Attendance: 10
Author: SF

Strong consensus.

[rarutyun]

SG1 reviewed P4049R0 in Brno, 2026

Poll: SG1 forwards P4049R0 to LEWG for C++29 as the fix for the missing precondition in copy_n algorithms, and continues to think that it is ok for C++ parallel algorithm preconditions to continue to strengthen from sequential algorithm preconditions as needed to enable parallelism.

SF	F	N	A	SA
7	4	1	0	0

Outcome: Unanimous consent