🚀 go-pugleaf

I have been unsuccessful in writing a DLL function that returns either a
float or a double.  I cannot find anything in the documentation I've got
(including the MS Developer Network CDROM) that discusses problems or
solutions for doing this.  I have tried several memory model and floating
point (/FP_) options.

I have tried to determine what is happening via CodeView.  I believe that
floating point return values are placed at some address for retrieval by the
calling code (unlike integer returns which are conveyed via AX {and DX} if
necessary).  Does the DS!=SS behavior come into play here?

In the calling function:
     SS = 1197
     DS = 1197

In the DLL function:
     SS = 1197
     DS = 1D67

In the DLL function , the following code is executed (immediately before the
return):
    FLD DWORD PTR [BP-08]
    INT    3D ;fwait
    MOV    BX,WORD PTR [BP+06]
    NOP
    FSTP   DWORD PTR [BX]
    INT    3D ;fwait
    MOV    AX,WORD PTR [BP+06]
    MOV    DX,SS
    POP    DI
    POP    SI
    LEA    SP,WORD PTR [BP-02]
    POP    DS
    POP    BP
    DEC    BP
    RETF   0006

Upon return from the function, the following code is executed:
    MOV    BX,AX
    MOV    AX,WORD PTR [BX]
    MOV    DX,WORD PTR [BX+02]
    MOV    WORD PTR [BP-011C],AX
    MOV    WORD PTR [BP-011A],DX


If I manually and temporlarly change (within the debugger) the DS register
to use the DS of the DLL, I can successfully retrieve the stored value.

Is there some way that I can force the DLL to place the value in an
accessible location and coerce the calling code to locate it?

---
   walt.gaber@drd.com                                       (918) 743-3013
   DRD Corp., 5506 South Lewis Ave., Tulsa, OK 74105        (918) 745-9037 fax