P3820 R1 Fix constexpr uncaught_exceptions and current_exception

[wg21bot]

P3820R0 Split constexpr uncaught_exceptions into distinct runtime and consteval functions (Lénárd Szolnoki)

[wg21bot]

P3820R1 Fix constexpr uncaught_exceptions and current_exception (Lénárd Szolnoki)

[inbal2l]

2025-09-09 Library Evolution Telecon

P3820R1: Fix constexpr uncaught_exceptions and current_exception (fix for C++26)

(Discussed alongside with: "P3818R1: constexpr exception fix for potentially constant initialization")
See also: #2428

2025-09-09 Library Evolution Telecon Minutes

Champion: Lenard Szolnoki
Chair: Andreas Weis
Minute Taker: Bronek Kozicki

Summary

POLL: Considering that we are already in the NB comment phase for C++26, adopt the solution laid out in P3818R1 (special handling for potentially constant initialization for both uncaught_exceptions and current_exception) as a solution to the problem for C++26.

SF	F	N	A	SA
5	6	2	4	3

Attendance: 22
Author's Position: SF (P3818)
Outcome: No consensus.

POLL: Considering that we are already in the NB comment phase for C++26, remove constexpr from uncaught_exceptions and make current_exception conditionally constant evaluated solution to the problem for C++26 (like P3820, but without the consteval_ functions, requires a new paper).

SF	F	N	A	SA
2	2	2	9	5

Attendance: 22
Author's Position: SF (P3820)
Outcome: Consensus against.

POLL: Remove the constexpr from uncaught_exceptions and current_exception for C++26

SF	F	N	A	SA
7	9	3	2	0

Attendance: 22
Author's Position: A (From author of proposal in the reflector - for removing the "constexpr")
Outcome: Consensus in favor.

Next Steps

Few design directions were discussed, but the feeling of the room was that this is too much of a large change at this stage.LEWG recommended removing the “constexpr” from “uncaught_exceptions” and “current_exception“ for C++26, but was in favor of exploring design solutions for C++29. We asked for a paper, we will take a final vote (by a paper or an LWG issue) during the NB comments time (ACTION: Fix wording addressing "thread" before P3818 seen again).

UPDATE: A paper reflecting the desired change was published post the meeting: "P3842R0: A conservative fix for constexpr uncaught_exceptions() and current_exception()"

See also: #2428