I started looking at this some time ago in #4297. I was probably too sneaky though. Still not complete, there is quite lot of things that break and I don't have much time for it now. The main problem is clashing operators.

I started looking at this some time ago in #4297. I was probably too sneaky though. Still not complete, there is quite lot of things that break and I don't have much time for it now. The main problem is clashing operators.

I can close this PR in favor of yours. I just wanted to see what needs to be done and it seems moving up is non-trivial.

I started looking at this some time ago in #4297. I was probably too sneaky though. Still not complete, there is quite lot of things that break and I don't have much time for it now. The main problem is clashing operators.

I can close this PR in favor of yours. I just wanted to see what needs to be done and it seems moving up is non-trivial.

OK. Yes, it does not seem as easy as was the move to C++17.

There is not much missing for C++20 support. It looks like there is just one small failure left.

Hmm, the error is in boost, that might be hard to fix unless it is our bad usage of boost of course :-). Seems to be related to a particular version as it fails just in the sanitizers build.

What are your thoughts on switching? Do you think it is time for that? We don't get many goodies until we bump minimal required GCC, but even some things from GCC 9 are nice. It would also unlock C++20 for downstreams that have newer compilers. On the other hand, we would be having a more experimental setup, with GCC that is far from stable on C++20.

Hmm, the error is in boost, that might be hard to fix unless it is our bad usage of boost of course :-). Seems to be related to a particular version as it fails just in the sanitizers build.

The boost version should be the same for Ubuntu 20.04 but this compiler is Clang. Maybe that is the reason.

What are your thoughts on switching? Do you think it is time for that? We don't get many goodies until we bump minimal required GCC, but even some things from GCC 9 are nice. It would also unlock C++20 for downstreams that have newer compilers. On the other hand, we would be having a more experimental setup, with GCC that is far from stable on C++20.

I am happy to switch, there are some folks that have been waiting for us to upgrade to C++20 already. We can try to be conservative with the features we introduce and keep Ubuntu 20.04 as the minimum supported version until it hits EOL.

Hmm, the error is in boost, that might be hard to fix unless it is our bad usage of boost of course :-). Seems to be related to a particular version as it fails just in the sanitizers build.

Cherrypicking #4663 fixes the issue. There were two more small fixes required which I can fork into a separate PR.

Hmm, the error is in boost, that might be hard to fix unless it is our bad usage of boost of course :-). Seems to be related to a particular version as it fails just in the sanitizers build.

Cherrypicking #4663 fixes the issue. There were two more small fixes required which I can fork into a separate PR.

Personally I'm OK with the operator fixes to be there, but the boost should be separate.

I've just notice that docs/C++.md still mentions C++11. No point fixing it to 17 now, but should be fixed when we switch to C++20.

Personally I'm OK with the operator fixes to be there, but the boost should be separate.

Yes, the changes were for testing purposes, I will move them into separate PRs.

It looks like there is very little that can be done to fix the compilation issue sans fixing boost. I am not sure why this pops up with clang only though.

@jafingerhut @vlstill Any concerns or objections? I am planning to pull the trigger and switch the next week.

@jafingerhut @vlstill Any concerns or objections? I am planning to pull the trigger and switch the next week.

No concerns or objections from me. It looks like you have a good variety of reviews from other folks, which is good.