Our idea to make the function "FloatWidth" safer is as follows.
-----
int Float Width(double floatIn) {
  if (floatIn <= 0.0) return -1;             // ADDED, BUT please make the condition more validly
  double float_exponent = fabs( log10( floatIn ) );
  ++float_exponent;
  return (int)float_exponent;
-----


The actual processing flow with Test_SPAM is as follows.
(assuming using AmberTools23 with updates 1-6 are applied)

1. In the function "DataIO_Std:WriteDataNormal" (DataIO_Std.cpp, line 1220), in if-block (lines 1240-1260)
     x_col_format.SetCoordFormat( maxFrames, Xdim.Min(), Xdim.Step(), xcol_width, xcol_precision ); // line 1259
   is called with Xdim.Step()==0.0 (*).
   (*) With Test_SPAM, the if-block is executed twice: 1st time Xdim.Step()==1.0, and 2nd time Xdim.Step()==0.0.

2. In the function "TextFormat::SetCoordFormat" (TextFormat.cpp, line 52), in if-block (lines 61-65)
   In case of step==0.0,
     FloatWidth( step ); // line 62
   is called.

3. In the function "FloatWidth" (StringRoutines.cpp, line 98), "2147483647" is generated.
   As floatIn==0.0, the value of float_exponent in lines 99 and 100 is as follows, respectively.
     float_exponent = fabs( log10( floatIn ) ) = fabs(log10(0.0)) = fabs(-Infinity) = +Infinity
     ++float_exponent (= +Infinity)
   The return value (int)float_exponent (= (int)(+Infinity)) depends on the platform, i.e.,
     On Intel: -2147483648
     On ARM:   +2147483647

4. In the function "TextFormat::SetCoordFormat" (TextFormat.cpp, line 52) again
   In if-statement (line 63),
     On Intel: as prec_exp_width < 0, col_precision is not changed
     On ARM:   as prec_exp_width = +2147483647, col_precision is now +2147483647

5. In the function "DataIO_Std:WriteDataNormal" (DataIO_Std.cpp, line 1220) again
     On Intel: x_col_format.fmt()=%8.3f
     On ARM:   x_col_format.fmt()=%8.2147483647f, which may cause segmentation fault