赵诗涵
I agree - subscripts from 0 to 226 actually iterate 227 times.
> In what situation is `R__HAS_STD_SPAN` not sufficient? If `ROOT` has been configured to build in C++20 mode and properly installed, it must be set in `RConfigure.h`. If it is...
> In practice, changing the test as this PR does introduces a ODR (one definition rule) violation: i.e. some of the code use one version of `std::span` and some of...
If I understand correctly, this basically involves ABI compatibility. According to the GCC [document (ABI policy)](https://gcc.gnu.org/onlinedocs/libstdc++/manual/abi.html), in C++, ABI is a mixture of standard library API and compiler ABI. Suppose...
> ... this is only the case once the support is deemed stable. To my knowledge even this is not guaranteed, for instance mangling has changed between standard versions. As...
Anyway, back to this PR: the motivation is the same as https://github.com/root-project/root/commit/ff95b036a75e9c9b33223907ba32862dce0e55f7. At least I want to confirm whether it is appropriate.
CI completed with MSVC "heap overflow".
> Apologies for letting this sit way too long. I think what we need is: > > * get rid of `R__HAS_STD_SPAN` (incl the CMake code) > * remove the...
> Closing this now that #11874 is merged. Note that we're not there yet; some parts for full C++20 support are still missing. I expect that 6.28/04 will finally provide...