[prev in list] [next in list] [prev in thread] [next in thread] 

List:       pykde
Subject:    [PyKDE] Leaking in virtual handler
From:       James Emerton <ephelon () gmail ! com>
Date:       2005-04-20 21:09:16
Message-ID: c9424d940504201409bbbad34 () mail ! gmail ! com
[Download RAW message or body]

I have found a memory leak owing to how SIP handles virtual methods.=20
If a type conversion is done by sipParseResult(), a new object is
allocated but never deleted.

I have an interface that returns a QString.  If I implement this in
python and return a python string, a conversion to QString is
performed by creating a new QString instance.  This instance  is never
deallocated.

I have attached a patch which appears to correct this.  I am not
entirely comfortable with the way I am detecting whether sipResOrig is
the same as the pointer contained in sipResObj.

James

["gencode.patch" (application/octet-stream)]

--- gencode.orig	Wed Apr 20 11:50:54 2005
+++ gencode.c	Wed Apr 20 13:42:29 2005
@@ -4194,7 +4194,8 @@ static void generateVirtualHandler(sipSp
 
 		generateResultType(&res_noconstref,fp);
 
-		prcode(fp," *sipResOrig;\n");
+		/* James - initialise to NULL */
+		prcode(fp," *sipResOrig = NULL;\n");
 	}
 
 	/* Call the method. */
@@ -4264,11 +4265,26 @@ static void generateVirtualHandler(sipSp
 		);
 
 	/* Make a copy if needed. */
+	/* James - clean up the original to plug memory leak 
+	     What if this was not allocated with new? */
 	if (copy)
+	{
 		prcode(fp,
 "	else\n"
 "		sipRes = *sipResOrig;\n"
-			);
+				);
+		if( res->atype == class_type )
+			prcode(fp,
+"\n"
+"	if( sipResOrig && sipGetCppPtr((sipWrapper*)sipResObj, sipClass_%T) != sipResOrig )\n"
+"		delete sipResOrig;\n"
+				,res);
+		else
+			prcode(fp,
+"\n"
+"	delete sipResOrig;\n"
+				);
+	}
 
 	if (isTransferVH(vhd))
 		prcode(fp,


_______________________________________________
PyKDE mailing list    PyKDE@mats.imk.fraunhofer.de
http://mats.imk.fraunhofer.de/mailman/listinfo/pykde


[prev in list] [next in list] [prev in thread] [next in thread]