P2370 Stacktrace from exception

[wg21bot]

P2370R0 Stacktrace from exception (Andrei Nekrashevich, Antony Polukhin)

[wg21bot]

P2370R1 Stacktrace from exception (Andrei Nekrashevich, Antony Polukhin)

[cor3ntin]

Mailing review 26-07-2021-13-09-2001

Lots of interest for this paper to land in 23 along the already approved stacktrace proposal.
A R1 was provided during the review addressing most of the concerns and feedback offered during the review.
A core expert offered some wording review

We probably still want to hear from implementers directly about the potential issues pertaining to ABI stability.

Given all the feedbacks were addressed and reviewers are happy with the general design and wording,
I recommend that this paper is reviewed during a LEWG telecon.

[brycelelbach]

Deferring to C++26 due to lack of time.

[eyalroz]

@brycelelbach : Can you elaborate a little more about this? While the "infrastructure" is very nice, the killer feature for many is for existing code, and the standard library especially, to throw exceptions with stack traces.

[NicolBolas]

@brycelelbach : "Lack of time" means lack of time. It means the committee doesn't have the bandwidth to work through the issues of the proposal (while still doing what's already on their plate) before C++23 needs to be feature complete.

[eyalroz]

@NicolBolas : I think you meant to address me. Anyway, there seem to be issues with the proposal. I was hoping for a few words about what issues have come up.

[wg21bot]

P2370R2 Stacktrace from exception (Andrei Nekrashevich, Antony Polukhin)

[brycelelbach]

2022-06-21 Library Evolution Telecon

P2370R2: Stacktrace from exception

2022-06-21 Library Evolution Telecon Minutes

P2490R0: Zero-overhead exception stacktraces

2022-06-21 Library Evolution Telecon Minutes

Chair: Bryce Adelstein Lelbach

Minute Taker: Ben Craig

Champion: Antony Polukhin (P2370) & Ed Catmur (P2490)

Start: 2022-06-21 11:51 Eastern

Does P2370 have:

• Examples?
  • Yes
• Field experience?
  • Yes. The field experience indicates that this can be done without
    an ABI break.
• Performance considerations?
  • Yes. Performance overheads were a source of discussion with this proposal.
• Discussion of prior art?
  • Yes.
• Wording?
  • Yes.
• Feature test macro?
  • Yes.

This needs to be seen by Evolution.

P2380R2 proposes capturing the stacktrace at the throw side and allows you to globally or per-thread enable/disable capturing stacktraces.

P2490R0 proposes to capture the stacktrace at entering the catch block rather than at the throw side.

End: 12:17

Summary

P2380R2 (Stacktrace from exceptions) and P2490R0 (Zero-overhead exception stacktraces) propose mechanisms for attaching stacktraces to exceptions.
P2380R2 proposes capturing the stacktrace at the throw side and allows you to globally or per-thread enable/disable capturing stacktraces.
P2490R0 proposes to capture the stacktrace at entering the catch block rather than at the throw side.

It became apparent during Library Evolution's review of these papers that they should be seen by Evolution first.
Once Evolution has decided upon a design direction for this feature, Library Evolution can review the Standard Library exposure of the feature.

Next Steps

Send P2380R2 (Stacktrace from exceptions) and P2490R0 (Zero-overhead exception stacktraces) to Evolution.

[eyalroz]

It became apparent during Library Evolution's review of these papers that they should be seen by Evolution first.

Can you elaborate slightly on why that is? i.e. will changes to the language be necessary?

[jwakely]

How to enable/disable stacktraces and how to access them from an exception are properties of the exception handling mechanism, which is a language feature. One of the proposals has an implementation in a compiler and involved changes to the compiler and language runtime, not just new library APIs.