The experiment deployment of TclDuplicatePureObj() has been reverted from

the trunk.  This branch attempts to prepare a reversion for the
core-8-branch also, via a cherrypick and resolution of conflicts.

Starting an experiment in one new release to reject it
in the next is at best odd.  Also, valgrind test runs of core-8-branch
show memory leaks with TclDuplicatePureObj() on the stack trace.  This
checkin shows a clean valgrind run.

There are three test failures on this checkin, which I hope someone versed
in the ArithSeries work can resolve.  They are the tests:

lseq-3.14
lseq-4.19
lseq-bug-54329e39c7

merge bug-bc7ddc7944 (revert use of TclDuplicatePureObj)

1 files changed, 2 insertions, 8 deletions

diff --git a/generic/tclIORChan.c b/generic/tclIORChan.c
index 3b1573b..0af76bf 100644
--- a/generic/tclIORChan.c
+++ b/generic/tclIORChan.c
@@ -2285,10 +2285,7 @@ NewReflectedChannel(
     rcPtr->mode = mode;
     rcPtr->interest = 0;		/* Initially no interest registered */
 
-    rcPtr->cmd = TclDuplicatePureObj(interp, cmdpfxObj, &tclListType);
-    if (!rcPtr->cmd) {
-	return NULL;
-    }
+    rcPtr->cmd = TclListObjCopy(NULL, cmdpfxObj);
     Tcl_IncrRefCount(rcPtr->cmd);
     rcPtr->methods = Tcl_NewListObj(METH_WRITE + 1, NULL);
     while (mn <= (int)METH_WRITE) {
@@ -2425,10 +2422,7 @@ InvokeTclMethod(
      * before the channel id.
      */
 
-    cmd = TclDuplicatePureObj(NULL, rcPtr->cmd, &tclListType);
-    if (!cmd) {
-	return TCL_ERROR;
-    }
+    cmd = TclListObjCopy(NULL, rcPtr->cmd);
     Tcl_ListObjIndex(NULL, rcPtr->methods, method, &methObj);
     Tcl_ListObjAppendElement(NULL, cmd, methObj);
     Tcl_ListObjAppendElement(NULL, cmd, rcPtr->name);