The DupEncodingIntRep() routine is broken. It fails to set the typePtr

field of the duplicated Tcl_Obj to indicate that the intrep is that of
the &encodingType.  The impact must be that refcounting of Tcl_Encodings
are out of balance, at least in the (unusual?) case where "encoding" values
need duplication.  I hesitate to actually commit this fix until I have a
test demonstrating the need for it, and a chance to see the impact on
actual encoding operations.  The lifetime management of encodings has a
history of being tricky and raising controversy.  More work required.

1 files changed, 1 insertions, 0 deletions

diff --git a/generic/tclEncoding.c b/generic/tclEncoding.c
index 4edebcf..32055a3 100644
--- a/generic/tclEncoding.c
+++ b/generic/tclEncoding.c
@@ -355,6 +355,7 @@ DupEncodingIntRep(
     Tcl_Obj *dupPtr)
 {
     dupPtr->internalRep.twoPtrValue.ptr1 = Tcl_GetEncoding(NULL, srcPtr->bytes);
+    dupPtr->typePtr = &encodingType;
 }
 
 /*