Use C++20 for programs that generate the HTML lists

[jwakely]

This currently also includes the unmerged commit from #444, because it modifies a line of code added by that commit. That should disappear from this PR once the other PR is merged.

As of version 19.1.0, Clang's libc++ doesn't provide chrono::clock_cast or chrono::parse so maybe it's too soon to make this change.

[jwakely]

As of version 19.1.0, Clang's libc++ doesn't provide chrono::clock_cast or chrono::parse so maybe it's too soon to make this change.

And the ubuntu-latest runner-image for GitHub Actions is ubuntu-22.04 which only has clang-15, which isn't good enough. There is an ubuntu-24.04 image with a newer clang, but that image is still in beta.

And the GCC version in ubuntu-latest is insufficient to build the C++20 code too.

[jwakely]

chrono::parse requires GCC 14, but there's a workaround.

[jwakely]

chrono::clock_cast isn't supported by libc++, but there's a workaround.

[jwakely]

There are several places in the code that should be converted to use views::filter

[jwakely]

There are several places in the code that should be converted to use views::filter

e.g.

   // This construction calls out for filter-iterators
//   std::multiset<lwg::issue, order_by_first_tag> pending_issues;
   std::vector<lwg::issue> pending_issues;
   for (auto const & elem : issues ) {
      if (predicate(elem)) {
         pending_issues.emplace_back(elem);
      }
   }

   sort(begin(pending_issues), end(pending_issues), order_by_section{section_db});

But I'm not sure if would actually be faster/simpler/better to insert into a multiset from a filter_view than the current code.

[jwakely]

There are several places in the code that should be converted to use views::filter

e.g.

   // This construction calls out for filter-iterators
//   std::multiset<lwg::issue, order_by_first_tag> pending_issues;
   std::vector<lwg::issue> pending_issues;
   for (auto const & elem : issues ) {
      if (predicate(elem)) {
         pending_issues.emplace_back(elem);
      }
   }

   sort(begin(pending_issues), end(pending_issues), order_by_section{section_db});

But I'm not sure if would actually be faster/simpler/better to insert into a multiset from a filter_view than the current code.

Maybe for C++23 one day, just because everything above can be replaced by:

auto pending_issues = issues | std::views::filter(predicate) | std::ranges::to<std::multiset<lwg::issue, order_by_section>>(order_by_section{section_db});

[Dani-Hub]

Some weeks ago I considered a similar change locally (Only switching to C++20 without anything else) to see how mingw can handle these changes. I now tested your branch successfully on my system and it seems to work well. I appreciate this course of action.