Convert min/max macros to std::min/max

[th1000s]

Added explicit <float> or <int> template argument where needed.
Add -DNOMINMAX define for Windows.

---

Commit message was:

These only cause trouble in stdlib headers, such as:

/usr/include/c++/12/limits:1828:13: error: expected unqualified-id before ‘noexcept’
 1828 |       max() _GLIBCXX_USE_NOEXCEPT { return __LDBL_MAX__; }
      |             ^~~~~~~~~~~~~~~~~~~~~
Descent3/d3music/musicapi.cpp: In function ‘void D3MusicSetVolume(float)’:
Descent3/d3music/musicapi.cpp:167:57: error: ‘std::numeric_limits<float>::min’ cannot be used as a function
  167 |   const float kEpsilon = std::numeric_limits<float>::min();
      |                          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~

plus there is the whole NOMINMAX situation on Windows.

And don't introduce <algorithm> just for that.

Automated via:

$ sed -i 's/_\?_\?min *(\(.*\),/D3_min(\1,/' `git grep -l min`
$ sed -i 's/_\?_\?max *(\(.*\),/D3_max(\1,/' `git grep -l max`

plus remove duplicates from lib/Macros.h + lib/linux/linux_fix.h and fix netgames/dmfc/dmfcstats.cpp (2x min on one line) manually.

---

Or, any other name, maybe just min_ / max_?

[kevinbentley]

Is there a reason not to use std::min? numeric_limits::min is the minimum value a float can store, whereas std::min(a,b) is the lesser of the two numbers.

[winterheart]

Agree, would be better use std::min / std::max and get rid min / max macros altogether. In code there also _min, __min, _max, __max variations.

[th1000s]

numeric_limits::min()

That might just be the pre-processor replacing stuff where it should not.

So far I only saw <string> from the stdlib, so I thought I'd avoid <algorithm> which famously increases compile times.

However, moving to std::max() triggers a lot of no matching function for call to ‘max(int, float)’ (or float, ushort, int, ubyte in various combinations). For the time being renaming the macro is the quicker way to get Linux to compile.

[winterheart]

Well, there need a casting into desired type. Since there anyway a lot of renaming, why don't go little further and use standardized functions?

[th1000s]

Renaming the macro was mostly done to get the code to compile, small steps :)

Fully switching to std::max() is less automated and requires different casts -- max(int, float) should not become max(static_cast<double>(a), static_cast<double>(b)). And even with the correct types and just float(a) C-style casts, that makes the call sites a bit unwieldy.

Instead, maybe replace the macro with a custom d3_max() template using std::common_type?

[kevinbentley]

You can specialize std::min like this:

int a = 123;
float b = 100.01;
float x = std::min<float>(a,b);

IMO, that's the best way to do it.

[th1000s]

It is now templated where needed. <float> was only added (in a somewhat automated way) where at least one component was a float / double according to gcc, I also converted a few 1.0 to 1.0f or added <int> manually.

[Lgt2x]

@th1000s can you please rebase on master branch ?