I added an assert in my patch attached to bug 20503, but it's possible
to trigger this condition through a buggy program too. An assert
doesn't seem appropriate in that case.
--Juan
[0009-Add-an-error-if-the-ref-count-is-invalid-when-releas.patch]
From 1c924c31c52028b747fd51464fdd129c6564360a Mon Sep 17 00:00:00 2001
From: Juan Lang <juan.lang@...>
Date: Tue, 3 Nov 2009 16:46:50 -0800
Subject: [PATCH] Add an error if the ref count is invalid when releasing a context
---
dlls/crypt32/context.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/dlls/crypt32/context.c b/dlls/crypt32/context.c
index b8ad220..e0c1932 100644
--- a/dlls/crypt32/context.c
+++ b/dlls/crypt32/context.c
@@ -173,7 +173,10 @@ BOOL Context_Release(void *context, size_t contextSize,
BOOL ret = TRUE;
if (base->ref <= 0)
+ {
+ ERR("%p's ref count is %d\n", context, base->ref);
return FALSE;
+ }
if (base->type == ContextTypeLink)
{
/* The linked context is of the same type as this, so release
--
1.6.0.6
