Is there any reason to not to use HB_IS_OBJECT instead of HB_IS_ARRAY here ?

[tfonrouge]

core/src/vm/classes.c

Line 4893 in 06725f4

if( HB_IS_ARRAY( pObject ) )

I have an case that which in certain conditions I get a segfault error in my destructor class procedure when accessing a data member which is contained inside a member in the current object, below is the debug trace output:

libharbour.3.4.0.dylib was compiled with optimization - stepping may behave oddly; variables may not be available.
Process 28239 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x74)
    frame #0: 0x0000000100fecc54 libharbour.3.4.0.dylib`hb_clsParentInstanceOffset(pClass=0x0000000000000000, uiParentCls=19) at classes.c:743 [opt]
   740 	{
   741 	   HB_USHORT uiCount = pClass->uiSuperClasses;
   742 	
-> 743 	   while( uiCount )
   744 	   {
   745 	      if( pClass->pSuperClasses[ --uiCount ].uiClass == uiParentCls )
   746 	         return pClass->pSuperClasses[ uiCount ].uiOffset;
Target 0: (Arel_Darwin_x86_64) stopped.
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x74)
  * frame #0: 0x0000000100fecc54 libharbour.3.4.0.dylib`hb_clsParentInstanceOffset(pClass=0x0000000000000000, uiParentCls=19) at classes.c:743 [opt]
    frame #1: 0x0000000101009f6d libharbour.3.4.0.dylib`HB_FUN_msgGetData at classes.c:4908 [opt]
    frame #2: 0x0000000100fd1e0f libharbour.3.4.0.dylib`hb_vmSend(uiParams=<unavailable>) at hvm.c:6119 [opt]
    frame #3: 0x0000000100fd9ecb libharbour.3.4.0.dylib`hb_xvmSend(uiParams=0) at hvm.c:9621 [opt]
    frame #4: 0x00000001000acded Arel_Darwin_x86_64`HB_FUN_TBASETABLE_DESTROY + 301

The following Harbour code halts when reaches "masterSource:linkedObjField", the member linkedObjField is supposed to be a object:

METHOD PROCEDURE Destroy() CLASS TBaseTable

    LOCAL table
    LOCAL masterSource

    masterSource := ::masterSource

    IF HB_ISOBJECT( masterSource )
        IF hb_isObject( masterSource:linkedObjField ) .AND. HB_IsObject( masterSource:linkedObjField:table ) .AND. masterSource:linkedObjField:table == self
            masterSource:linkedObjField := nil

The reason is that inside of the function "HB_FUNC_STATIC( msgGetData )", the value for "pObject->item.asArray.value->uiClass" is 0 ( zero ), because the item passed ( pObject ) is not an object anymore, it's a simple array.

And when the function hb_clsParentInstanceOffset() is called, his 1st parameter is NULL

         nIndex += hb_clsParentInstanceOffset( s_pClasses[ uiObjClass ],
                                               pMethod->uiSprClass );

and then program stops when accessing the uiSuperClasses member in the pClass (nulled) var:

static HB_USHORT hb_clsParentInstanceOffset( PCLASS pClass, HB_USHORT uiParentCls )
{
   HB_USHORT uiCount = pClass->uiSuperClasses;

   while( uiCount )

Can be also needed to check here for a NULL pClass value ?

In my local repo I changed, in this case, HB_IS_ARRAY for HB_IS_OBJECT and the issue was fixed, however there are other places when a similar scenario is found.