Re: RegSetValueEx problem?

[Alexandre Julliard]

gerard patel  writes:
> I have browsed server/registry.c and it seems that the registry
> stores a terminating null for strings; so if you have in the registry
> file a string like "0123456789", it get in memory a length of 11.
> So I guess that the fix would be to fix the length by -1 somewhere
> in the path server->ntdll->advapi

No, the current behavior is correct for NT; what you want is a version
check to get the Win95 behavior. Something like the patch below should
do it (and of course WineMine should be fixed to work on NT too).

Index: dlls/advapi32/registry.c
===================================================================
RCS file: /opt/cvs-commit/wine/dlls/advapi32/registry.c,v
retrieving revision 1.24
diff -u -r1.24 registry.c
--- dlls/advapi32/registry.c	2001/03/23 19:12:01	1.24
+++ dlls/advapi32/registry.c	2001/04/09 20:49:35
@@ -640,8 +640,12 @@
 {
     UNICODE_STRING nameW;

-    if (count && is_string(type))
+    if (GetVersion() & 0x80000000)  /* win95 */
     {
+        if (type == REG_SZ) count = strlenW( (WCHAR *)data ) + sizeof(WCHAR);
+    }
+    else if (count && is_string(type))
+    {
         LPCWSTR str = (LPCWSTR)data;
         /* if user forgot to count terminating null, add it (yes NT does this) */
         if (str[count / sizeof(WCHAR) - 1] && !str[count / sizeof(WCHAR)])
@@ -663,7 +667,11 @@
     WCHAR *dataW = NULL;
     NTSTATUS status;

-    if (count && is_string(type))
+    if (GetVersion() & 0x80000000)  /* win95 */
+    {
+        if (type == REG_SZ) count = strlen(data) + 1;
+    }
+    else if (count && is_string(type))
     {
         /* if user forgot to count terminating null, add it (yes NT does this) */
         if (data[count-1] && !data[count]) count++;

RegSetValueEx problem?

[davep]

On Mon, 09 Apr 2001, you wrote:
> gerard patel  writes:
> > I have browsed server/registry.c and it seems that the registry
> > stores a terminating null for strings; so if you have in the registry
> > file a string like "0123456789", it get in memory a length of 11.
> > So I guess that the fix would be to fix the length by -1 somewhere
> > in the path server->ntdll->advapi
>
> No, the current behavior is correct for NT; what you want is a version
> check to get the Win95 behavior. Something like the patch below should
> do it (and of course WineMine should be fixed to work on NT too).

{patch snipped}

I have now been able to build and run WineMine under NT4, and it works
perfectly. On exit each parameter is saved to the registry as 16 bytes including
the training garbage (not visible in Regedit of course) and when restarted the
exact same 16 bytes are reloaded from the registry.

It does therefore seem that there is something wrong with Wne's implementation.