* generic/tclObj.c (TclContinuationsEnter): [Bug 2895323]: Updated

	comments to describe when the function can be entered for the same
	Tcl_Obj* multiple times. This is a continuation of the 2009-11-10
	entry where a memory leak was plugged, but where not sure if that
	was just a band-aid to paper over some other error. It isn't, this
	is a legal situation.


FossilOrigin-Name: 60381f421b3c15abb25aa19cde6591e263d33bb4

1 files changed, 20 insertions, 6 deletions

diff --git a/generic/tclObj.c b/generic/tclObj.c
index 7a37fd3..56db483 100644
--- a/generic/tclObj.c
+++ b/generic/tclObj.c
@@ -13,7 +13,7 @@
  * See the file "license.terms" for information on usage and redistribution of
  * this file, and for a DISCLAIMER OF ALL WARRANTIES.
  *
- * RCS: @(#) $Id: tclObj.c,v 1.139.2.6 2009/11/10 16:46:16 dgp Exp $
+ * RCS: @(#) $Id: tclObj.c,v 1.139.2.7 2009/12/10 19:13:15 andreas_kupries Exp $
  */
 
 #include "tclInt.h"
@@ -574,12 +574,26 @@ TclContinuationsEnter(Tcl_Obj* objPtr,
     ContLineLoc* clLocPtr = 
 	(ContLineLoc*) ckalloc (sizeof(ContLineLoc) + num*sizeof(int));
 
-    if (newEntry == 0) {
+    if (!newEntry) {
 	/*
-	 * Somehow we're entering ContLineLoc data for the same value more
-	 * than one time.  Not sure whether that's expected, or a sign of
-	 * trouble, but at a minimum, we should take care not to leak the
-	 * old entry.
+	 * We're entering ContLineLoc data for the same value more than one
+	 * time. Taking care not to leak the old entry.
+	 *
+	 * This can happen when literals in a proc body are shared. See for
+	 * example test info-30.19 where the action (code) for all branches of
+	 * the switch command is identical, mapping them all to the same
+	 * literal. An interesting result of this is that the number and
+	 * locations (offset) of invisible continuation lines in the literal
+	 * are the same for all occurences.
+	 *
+	 * Note that while reusing the existing entry is possible it requires
+	 * the same actions as for a new entry because we have to copy the
+	 * incoming num/loc data even so. Because we are called from
+	 * TclContinuationsEnterDerived for this case, which modified the
+	 * stored locations (Rebased to the proper relative offset). Just
+	 * returning the stored entry and data would rebase them a second
+	 * time, or more, hosing the data. It is easier to simply replace, as
+	 * we are doing.
 	 */
 
 	ckfree((char *) Tcl_GetHashValue(hPtr));