Fix more corner-cases like [0e92c404f19ede5b2eb06e6db27647d3138cc56|0e92c404f1]: The only place where a type of &tclByteArrayType can be trusted is when determining its length, because the character length of a (UTF-8) string is always equal to the byte length of the byte array.

1 files changed, 18 insertions, 0 deletions

diff --git a/generic/tclInt.h b/generic/tclInt.h
index 00b246b..2353450 100644
--- a/generic/tclInt.h
+++ b/generic/tclInt.h
@@ -3668,6 +3668,24 @@ MODULE_SCOPE void	TclDbInitNewObj(Tcl_Obj *objPtr, CONST char *file,
 
 /*
  *----------------------------------------------------------------
+ * Macro that encapsulates the logic that determines when it is safe to
+ * interpret a string as a byte array directly. In summary, the object must be
+ * a byte array and must not have a string representation (as the operations
+ * that it is used in are defined on strings, not byte arrays). Theoretically
+ * it is possible to also be efficient in the case where the object's bytes
+ * field is filled by generation from the byte array (c.f. list canonicality)
+ * but we don't do that at the moment since this is purely about efficiency.
+ * The ANSI C "prototype" for this macro is:
+ *
+ * MODULE_SCOPE int	TclIsPureByteArray(Tcl_Obj *objPtr);
+ *----------------------------------------------------------------
+ */
+
+#define TclIsPureByteArray(objPtr) \
+	(((objPtr)->typePtr==&tclByteArrayType) && ((objPtr)->bytes==NULL))
+
+/*
+ *----------------------------------------------------------------
  * Macro used by the Tcl core to compare Unicode strings. On big-endian
  * systems we can use the more efficient memcmp, but this would not be
  * lexically correct on little-endian systems. The ANSI C "prototype" for