m68k: write inline floating point immediates as hex consts, instead of scientific notation. the scientific notation format parsing depends on the libc the assembler was linked against, which can lead to varying results and compile errors

git-svn-id: trunk@48524 -

compiler/m68k/ag68kgas.pas

@@ -169,7 +169,7 @@ interface
       end;
 
 
-    function getopstr(var o:toper) : string;
+    function getopstr(size: topsize; var o:toper) : string;
       var
         i : tsuperregister;
       begin
@@ -220,10 +220,14 @@ interface
             getopstr:='#'+tostr(longint(o.val));
           top_realconst:
             begin
-              str(o.val_real,getopstr);
-              if getopstr[1]=' ' then
-                getopstr[1]:='+';
-              getopstr:='#0d'+getopstr;
+              case size of
+                S_FS:
+                  getopstr:='#0x'+hexstr(longint(single(o.val_real)),sizeof(single)*2);
+                S_FD:
+                  getopstr:='#0x'+hexstr(BestRealRec(o.val_real).Data,sizeof(bestreal)*2);
+              else
+                internalerror(2021020801);
+              end;
             end;
           else internalerror(200405021);
         end;
@@ -337,7 +341,7 @@ interface
                         sep:=':'
                       else
                         sep:=',';
-                      s:=s+sep+getopstr(taicpu(hp).oper[i]^);
+                      s:=s+sep+getopstr(taicpu(hp).opsize,taicpu(hp).oper[i]^);
                     end;
                 end;
            end;