Failing MODULE_DllProcessAttach causes other module left uninitialized

[Uwe Bonnes]

Hallo,

in loader/module.c:MODULE_DllProcessAttach the recursion to attach ends as
soon as _one_ initialization fails. This leaves other modules
uninitialized. I had this problem with xnview and native shell (without
native shfolder). The missing SHGetFolderPathA caused COMDLG32_DllEntryPoint
to return FALSE and in cause  WSVFW32_LibMain wasn't called. Later this
caused MULTIMEDIA_GetIData to not find it's IData and cause a crash. The
missing HGetFolderPathA wasn't the culprit.

I think the original error was obscured to much here. Either we judge a
failing InitDLL as error and abort there, or in MODULE_DllProcessAttach we
should recurse without looking at the result.

What about appended patch? 
-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
Index: wine/loader/module.c
===================================================================
RCS file: /home/wine/wine/loader/module.c,v
retrieving revision 1.150
diff -u -r1.150 module.c
--- wine/loader/module.c	2 Apr 2002 19:47:30 -0000	1.150
+++ wine/loader/module.c	9 Apr 2002 14:06:26 -0000
@@ -212,17 +212,14 @@
     wm->flags |= WINE_MODREF_MARKER;
 
     /* Recursively attach all DLLs this one depends on */
-    for ( i = 0; retv && i < wm->nDeps; i++ )
+    for ( i = 0; i < wm->nDeps; i++ )
         if ( wm->deps[i] )
-            retv = MODULE_DllProcessAttach( wm->deps[i], lpReserved );
+            MODULE_DllProcessAttach( wm->deps[i], lpReserved );
 
     /* Call DLL entry point */
+    retv = MODULE_InitDLL( wm, DLL_PROCESS_ATTACH, lpReserved );
     if ( retv )
-    {
-        retv = MODULE_InitDLL( wm, DLL_PROCESS_ATTACH, lpReserved );
-        if ( retv )
-            wm->flags |= WINE_MODREF_PROCESS_ATTACHED;
-    }
+      wm->flags |= WINE_MODREF_PROCESS_ATTACHED;
 
     /* Re-insert MODREF at head of list */
     if ( retv && wm->prev )

[Alexandre Julliard]

Uwe Bonnes <bon(a)elektron.ikp.physik.tu-darmstadt.de> writes:
> I think the original error was obscured to much here. Either we judge a
> failing InitDLL as error and abort there, or in MODULE_DllProcessAttach we
> should recurse without looking at the result.

A failed InitDLL is an error, we can't simply ignore it and
continue. We have to unload the dll and its dependencies in that case;
but that's not entirely trivial to do.

-- 
Alexandre Julliard
julliard(a)winehq.com