Re: fusion: improve UninstallAssembly stub slightly, fix typo in test

[James Hawkins]

On Mon, May 26, 2008 at 9:24 PM, Dan Kegel  wrote:
> Improve stub slightly so callers (like our test) looking for a
> disposition find one.
> Also fix a minor typo in the test.

diff --git a/dlls/fusion/asmcache.c b/dlls/fusion/asmcache.c
index 7da79f4..5697433 100644
--- a/dlls/fusion/asmcache.c
+++ b/dlls/fusion/asmcache.c
@@ -151,6 +151,9 @@ static HRESULT WINAPI
IAssemblyCacheImpl_UninstallAssembly(IAssemblyCache *iface
 {
     FIXME("(%p, %d, %s, %p, %p) stub!\n", iface, dwFlags,
           debugstr_w(pszAssemblyName), pRefData, pulDisposition);
+    /* caller might test disposition to see why it failed, so come up
with some excuse... */
+    if (pulDisposition)
+       *pulDisposition = IASSEMBLYCACHE_UNINSTALL_DISPOSITION_STILL_IN_USE;

return E_NOTIMPL;
 }

I disagree with this change.  If you want to fix the valgrind warning,
the test should set the disposition before calling UninstallAssembly.
This disposition is misleading anyway.  Hypothetically, a user of this
API could sit in a loop waiting for the assembly to not be in use.