2000-07-25 00:53 UTC+0100 Victor Szakats <info@szelvesz.hu>

harbour/ChangeLog

@@ -1,3 +1,8 @@
+2000-07-25 00:53 UTC+0100 Victor Szakats <info@szelvesz.hu>
+
+   * source/vm/itemapi.c
+     + Some NOTEs added to the Eval API functions.
+
 2000-07-25 00:33 UTC+0100 Victor Szakats <info@szelvesz.hu>
 
    * source/rtl/setkey.c

harbour/source/vm/itemapi.c

@@ -98,6 +98,16 @@ BOOL hb_evalNew( PEVALINFO pEvalInfo, PHB_ITEM pItem )
 /* NOTE: CA-Cl*pper is buggy and will not check if more parameters are
          added than the maximum (9). [vszakats] */
 
+/* NOTE: CA-Cl*pper NG suggest that the Items passed as parameters should/may 
+         be released by the programmer explicitly. But in fact hb_evalRelease()
+         will automatically release all of them. The sample programs in the 
+         NG are doing it that way. Releasing the parameters explicitly in 
+         Harbour will cause an internal error, while it will be silently 
+         ignored (?) in CA-Cl*pper. This is due to the different internal 
+         handling of the Items, but IIRC it causes leak in CA-Clipper. All in 
+         all, don't release the eval parameter Items explicitly to make both 
+         Harbour and CA-Clipper happy. [vszakats] */
+
 BOOL hb_evalPutParam( PEVALINFO pEvalInfo, PHB_ITEM pItem )
 {
    HB_TRACE(HB_TR_DEBUG, ("hb_evalPutParam(%p, %p)", pEvalInfo, pItem));
@@ -153,6 +163,10 @@ PHB_ITEM hb_evalLaunch( PEVALINFO pEvalInfo )
    return pResult;
 }
 
+/* NOTE: CA-Clipper NG states that hb_evalLaunch() must be called at least 
+         once and only once before calling hb_evalRelease(). Harbour doesn't 
+         have these requirements. [vszakats] */
+
 BOOL hb_evalRelease( PEVALINFO pEvalInfo )
 {
    HB_TRACE(HB_TR_DEBUG, ("hb_evalRelease(%p)", pEvalInfo));