'Anonymous' reported a problem when using Fl_Value_Input. It seemed,

that contrary to the man pages, a fractional 'step' value above 1
would not create a floating point input.

In fact it does, but the output was formatted wrong so that the 
digits after the decimal point were never rendered.

I changed the Fl_Valuator::format(double v) function how I beleive
it should format the output correctly, but as already stated by
the original author in the source code, this is a hack that should
be fixed by providing a 'precission' setting for valuators.

Anyway. My fix makes sure that all digits right of the decimal 
point are always rendered, so that the step value will show full
precision. This gives a much better behavior in respect to steps values
like 2.5, 3.75, etc., but also leads to 8 digits after the decimal point
for step(1.0/3.0)... .

I suggest that we keep this change (hence the commit), risking that 
rendering of valuator text will change in a few cases (odd step() values).



git-svn-id: file:///fltk/svn/fltk/branches/branch-1.1@4091 ea41ed52-d2ee-0310-a9c1-e6b18d33e121

CHANGES

@@ -1,5 +1,7 @@
 CHANGES IN FLTK 1.1.7
 
+	- Fl_Valuator would not format text output with decimal 
+          point when the step value was fractional, but above 1
 	- Documentation fixes (STR #648, STR #692, STR #730, STR
 	  #744, STR #745)
 	- fl_filename_relative() didn't compare drive letters in

src/Fl_Valuator.cxx

@@ -118,10 +118,13 @@ double Fl_Valuator::increment(double v, int n) {
 int Fl_Valuator::format(char* buffer) {
   double v = value();
   // MRS: THIS IS A HACK - RECOMMEND ADDING BUFFER SIZE ARGUMENT
-  if (!A) return snprintf(buffer, 128, "%g", v);
-  int i, X;
-  double ba = B / A;
-  for (X = 1, i = 0; X < ba; X *= 10) i++;
+  if (!A || !B) return snprintf(buffer, 128, "%g", v);
+  int i;
+  double ab = A/B;
+  for (i=0; i<8; i++) {
+    if ((ab-floor(ab))<1e-9) break;
+    ab *= 10.0;
+  }
   return sprintf(buffer, "%.*f", i, v);
 }