Get rid CL_REQUIRE/CL_PROVIDE(cl_F_mostneg), it is not really necessary.

Move static variables most_negative_[SFDL]F into most_negative_LF() function
(which is the only user of those variables) in order to avoid possible static
order initialization problems. While at it, make most_negative_LF() function
static.

include/cln/float.h

@@ -559,7 +559,6 @@ extern const cl_F most_positive_float (float_format_t f);
 
 // Returns the smallest (most negative) floating point number in float format f.
 extern const cl_F most_negative_float (float_format_t f);
-//CL_REQUIRE(cl_F_mostneg)
 
 // Returns the least positive floating point number (i.e. > 0 but closest to 0)
 // in float format f.

src/float/misc/cl_F_mostneg.cc

@@ -3,8 +3,6 @@
 // General includes.
 #include "cl_sysdep.h"
 
-CL_PROVIDE(cl_F_mostneg)
-
 // Specification.
 #include "cln/float.h"
 
@@ -19,6 +17,15 @@ CL_PROVIDE(cl_F_mostneg)
 
 namespace cln {
 
+static inline const cl_LF most_negative_LF (uintC len)
+{
+	var Lfloat erg = allocate_lfloat(len,LF_exp_high,-1);
+	fill_loop_up(&TheLfloat(erg)->data[0],len,~(uintD)0);
+	return erg;
+}
+
+const cl_F most_negative_float (float_format_t f)
+{
 	// Exponent so groß wie möglich, Mantisse 1...1, Vorzeichen -.
 
 	static const cl_SF most_negative_SF =
@@ -34,15 +41,6 @@ static const cl_DF most_negative_DF =
 		encode_DF(-1,DF_exp_high-DF_exp_mid,bit(DF_mant_len-32+1)-1,bitm(32)-1);
 #endif
 
-inline const cl_LF most_negative_LF (uintC len)
-{
-	var Lfloat erg = allocate_lfloat(len,LF_exp_high,-1);
-	fill_loop_up(&TheLfloat(erg)->data[0],len,~(uintD)0);
-	return erg;
-}
-
-const cl_F most_negative_float (float_format_t f)
-{
 	floatformatcase((uintC)f
 	,	return most_negative_SF;
 	,	return most_negative_FF;
@@ -53,4 +51,3 @@ const cl_F most_negative_float (float_format_t f)
 
 }  // namespace cln
 
-CL_PROVIDE_END(cl_F_mostneg)