>Well, as I understand funtion DoubleToNum it converts data to valid double one before calling DLL. The following is the code I use:
>
>set proce to clsheap addi && Ed clsheap class
>
>ascmx = replicate(chr(0),8*10)
>xdata =DoubleToNum(ascmx) && Is this OK ?
>
>retval=MyDll(,,@xdata)
>
>The C0000005 error appear after that.

No, what you want to do is initialize a buffer to the length of the array (10 Doubles at 8 bytes/double is 80 bytes, and then parse the results after the return:

cBuffer = REPL(CHR(0), 80)
*
*  blah, blah woof, woof
*
retval =MyDLL(,,@cBuffer)
LOCAL i, ArrayOfDoubles[10], nPosn
FOR i = 0 TO 9
   nPosn = i*8 + 1
   ArrayOfDoubles[i+1] = DoubleToNum(SUBST(cBuffer,nPosn,8))
ENDFOR

That should populate the memvar array ArrayOfDoubles with the 10 values returned.


>
>Must I convert LONG data type too?
>

If the Long is embedded in a structumemllocation, you need to convert from the string to the numeric value using DWORDToNum():

nLongInteger = DWORDToNum(SUBST(cStructureHoldingLong,nPositionofLong,4))

If the Long is passed as a parameter rather than being embedded in a structure, then it will be handled as a 32 bit integer value without the need to convert. If you need to convert an integer to the string representing the Integer to concatenate into a structure, you'd use NumToDWORD():

cDWORDString = NumToDWORD(nMyIntegerValue)

See my postings on the Wiki (htp://fox.wikis.com) entitled "Pointers on Pointers" for more details.
>