dlls/ntdll/reg.c -- bugfix

[Gerald Pfeifer]

I am by far an expert of that code, but the current

 keyinfo.ClassLength = max( 0, wine_server_reply_size(reply) - reply->namelen );

there is the same as

 keyinfo.ClassLength = wine_server_reply_size(reply) - reply->namelen;

and I assume this is not the expected outcome.

The reason is that wine_server_reply_size returns data_size_t which, like 
reply->namelen is unsigned as defined in include/wine/server_protocol.h,
so the difference of these two always will be unsigned and thus larger
than 0.

Again, I am no an expert here, but I believe the following patch 
establishes the original intent of this code.  If this is the case,
this indeed fixes a genuine bug.

Gerald

ChangeLog:
Fix computation in enumerate_key().

Index: dlls/ntdll/reg.c
===================================================================
RCS file: /home/wine/wine/dlls/ntdll/reg.c,v
retrieving revision 1.79
diff -u -3 -p -r1.79 reg.c
--- dlls/ntdll/reg.c	17 Oct 2007 14:06:21 -0000	1.79
+++ dlls/ntdll/reg.c	22 Nov 2007 20:29:19 -0000
@@ -271,7 +271,7 @@ static NTSTATUS enumerate_key( HANDLE ha
                     fixed_size = (char *)keyinfo.Name - (char *)&keyinfo;
                     keyinfo.LastWriteTime = modif;
                     keyinfo.TitleIndex = 0;
-                    keyinfo.ClassLength = max( 0, wine_server_reply_size(reply) - reply->namelen );
+                    keyinfo.ClassLength = max( 0, (int)wine_server_reply_size(reply) - (int)reply->namelen );
                     keyinfo.ClassOffset = keyinfo.ClassLength ? fixed_size + reply->namelen : -1;
                     keyinfo.NameLength = reply->namelen;
                     memcpy( info, &keyinfo, min( length, fixed_size ) );