(Previous discussion (C++17): #16684)

With package managers shipping newer versions of compilers and older releases of operating systems going EOL, it occurs that at some point in the future it will be almost uncontroversial to switch to C++20.

See the new features here: https://en.cppreference.com/w/cpp/compiler_support#cpp20

I think noteworthy are:

• Lambdas in unevaluated contexts: It will be possible to use Assert(...) (a lambda) inside another lambda. (Already fixed by util/check: Don't use a lambda for Assert/Assume #24714 )
• Designated initializers: Named arguments for constructors, see also Add syscall sandboxing using seccomp-bpf (Linux secure computing mode) #20487 (comment) (Already fixed by Use designated initializers #24531)
• std::span: Replacing our Span, see also c++20: Opt-in to modeling view and borrowed_range for Span #23226
• Epoch for std::chrono: Being able to remove ChronoSanityCheck from util: remove Boost posix_time usage from GetTime* #21110
• Remove gmtime, refactor: Replace GetTimeMicros by SystemClock #27233 (comment)
• Some better deduction guides (e.g.

  bitcoin/src/util/overloaded.h

  Line 18 in 16d698c

  //! Explicit deduction guide (not needed as of C++20)

  )
• "Key equivalence stuff" (?) from Switch BlockMap to use an unordered_set under the hood #19677
• consteval refactor: Remove c_str from util/check #26960 (comment)
• concepts util: Allow std::byte and char Span serialization #27927 (comment)
• tracepoints: Disables -Wgnu-zero-variadic-macro-arguments to compile without warnings #27401
• fuzz: Flatten all FUZZ_TARGET macros into one #28065 (comment)
• serialize changes: net: Use serialization parameters for CAddress serialization #25284 (comment) net: Use serialization parameters for CAddress serialization #25284 (comment) and Discussion: Upgrading to C++20 #23363 (comment)

Overall this doesn't look like massive improvements, so switching to C++20 is probably low(est) priority.