Revert a few more (int -> size_t) transitions, which could effect extensions (such as Itcl 3.4) which use internal Tcl header files. Better wait until 9.0 for this.

What we _can_ do is change some (internal) fields to 'unsigned': that doubles the epoch range without further danger. Thanks, Don, for pointing this out!

1 files changed, 3 insertions, 3 deletions

diff --git a/generic/tclObj.c b/generic/tclObj.c
index 1aa24f2..4ec0a57 100644
--- a/generic/tclObj.c
+++ b/generic/tclObj.c
@@ -344,17 +344,17 @@ typedef struct ResolvedCmdName {
 				 * reference (not the namespace that contains
 				 * the referenced command). NULL if the name
 				 * is fully qualified.*/
-    size_t refNsId;		/* refNsPtr's unique namespace id. Used to
+    unsigned long refNsId;		/* refNsPtr's unique namespace id. Used to
 				 * verify that refNsPtr is still valid (e.g.,
 				 * it's possible that the cmd's containing
 				 * namespace was deleted and a new one created
 				 * at the same address). */
-    size_t refNsCmdEpoch;	/* Value of the referencing namespace's
+    unsigned int refNsCmdEpoch;	/* Value of the referencing namespace's
 				 * cmdRefEpoch when the pointer was cached.
 				 * Before using the cached pointer, we check
 				 * if the namespace's epoch was incremented;
 				 * if so, this cached pointer is invalid. */
-    int cmdEpoch;		/* Value of the command's cmdEpoch when this
+    unsigned int cmdEpoch; /* Value of the command's cmdEpoch when this
 				 * pointer was cached. Before using the cached
 				 * pointer, we check if the cmd's epoch was
 				 * incremented; if so, the cmd was renamed,