Codechange: use the fmt library for simpler debug formats

[rubidium42]

Motivation / Problem

The format codes for printf are complicated and annoying to work with. When changing something from C-string to std::string you need to explicitly add .c_str(), but also for some types you need to "inject" PRINTF_SIZE.

Preferably we would use something in the standard library, but formatting has landed in C++20 so... almost there but not quite yet. There is however the "fmt" library that has the same API as C++20 just in a different namespace, so we can use that as a stop-gap measure until we go to C++20. See https://fmt.dev/7.1.3/api.html#compatibility-with-c-20-std-format

Description

The vendoring of the "fmt" library and full conversion of all DEBUG statements to Debug.

Limitations

• ~9000 lines of extra code due to the vendoring of the library.
• There is a set of fmt::printf functions, however they do not support things like %ld so using that as a migration step is out of the question.
• The error messages on wrong formats can be very unhelpful, though it does have compile time validation of the format strings.
• There are a lot of inconsistencies in the debug messages. Those are not in scope here, unless someone wants to patch that on top of this. That I'll gladly accept.

Checklist for review

Some things are not automated, and forgotten often. This list is a reminder for the reviewers.

• The bug fix is important enough to be backported? (label: 'backport requested')
• This PR affects the save game format? (label 'savegame upgrade')
• This PR affects the GS/AI API? (label 'needs review: Script API')
  • ai_changelog.hpp, gs_changelog.hpp need updating.
  • The compatibility wrappers (compat_*.nut) need updating.
• This PR affects the NewGRF API? (label 'needs review: NewGRF')
  • newgrf_debug_data.h may need updating.
  • PR must be added to API tracker

[glx22]

Ok, for targets using vcpkg you can add fmt to the install list. For MinGW it's also fmt. Don't know for Linux and Emscripten.

[LordAro]

Isn't it a header-only library anyway? Might as well just vendor it

[rubidium42]

Ok, for targets using vcpkg you can add fmt to the install list. For MinGW it's also fmt. Don't know for Linux and Emscripten.

I tried looking for that, but couldn't find it in the source code. So I guess it is in some helper repository. In any case, I have not bothered with that as the decision for "vendoring" it or not needs to be made first. If we are going to vendor it, then adding the library to those lists has been quite futile.

Isn't it a header-only library anyway? Might as well just vendor it

You can use it as header-only library or as normal library. For the former you need to add an extra define before the include, though that should not be a problem.

[TrueBrain]

This would make life so much better in many ways (how often I get warnings because I used %d instead of %lu .. I cannot even start to tell you about).

PS: I vote to vendor this library; it is small enough after all.
PPS: your description is pretty much out-of-date :P