[Zodb-checkins] CVS: Zope3/src/zodb/btrees - BucketTemplate.c:1.16

[Tim Peters]

Update of /cvs-repository/Zope3/src/zodb/btrees
In directory cvs.zope.org:/tmp/cvs-serv20148/src/zodb/btrees

Modified Files:
	BucketTemplate.c 
Log Message:
bucket_set():  Expanded the comment on the gcc-wng-squashing initialization
hack.


=== Zope3/src/zodb/btrees/BucketTemplate.c 1.15 => 1.16 ===
--- Zope3/src/zodb/btrees/BucketTemplate.c:1.15	Wed Apr 16 11:21:50 2003
+++ Zope3/src/zodb/btrees/BucketTemplate.c	Wed Apr 16 11:59:04 2003
@@ -296,11 +296,18 @@
 {
     int i, cmp;
     KEY_TYPE key;
-    VALUE_TYPE value = {0}; /* Subtle: VALUE_TYPE can be various types,
-			       including a 6-char array,
-			       but {0} is always a valid initializer.
-			       Without an initialization, GCC warns about
-			       this variable being used uninitialized. */
+
+    /* Subtle:  there may or may not be a value.  If there is, we need to
+     * check its type early, so that in case of error we can get out before
+     * mutating the bucket.  But because value isn't used on all paths, if
+     * we don't initialize value then gcc gives a nuisance complaint that
+     * value may be used initialized (it can't be, but gcc doesn't know
+     * that).  So we initialize it.  However, VALUE_TYPE can be various types,
+     * including int, PyObject*, and char[6], so it's a puzzle to spell
+     * initialization.  It so happens that {0} is a valid initializer for all
+     * these types.
+     */
+    VALUE_TYPE value = {0};	/* squash nuisance warning */
     int result = -1;    /* until proven innocent */
     int copied = 1;