Signed-off-by: Alistair Leslie-Hughes leslie_alistair@hotmail.com --- dlls/oleaut32/usrmarshal.c | 3 +++ 1 file changed, 3 insertions(+)
diff --git a/dlls/oleaut32/usrmarshal.c b/dlls/oleaut32/usrmarshal.c index 163b48f6f4..07dabde174 100644 --- a/dlls/oleaut32/usrmarshal.c +++ b/dlls/oleaut32/usrmarshal.c @@ -1905,6 +1905,9 @@ void CALLBACK ITypeInfo_ReleaseFuncDesc_Proxy( SHORT param; TRACE("(%p, %p)\n", This, pFuncDesc);
+ if (!pFuncDesc) + return; + for(param = 0; param < pFuncDesc->cParams; param++) free_embedded_elemdesc(pFuncDesc->lprgelemdescParam + param); if(param)
Is it possible to add a test for this? And ideally a bug reference too.
From: wine-devel wine-devel-bounces@winehq.org On Behalf Of Nikolay Sivov Sent: Wednesday, September 09, 2020 5:46 AM Subject: Re: [PATCH] oleaut32: ITypeInfo_ReleaseFuncDesc_Proxy handle NULL parameter
Is it possible to add a test for this? And ideally a bug reference too.
Agreed, this fix seems strange in isolation. Why only ReleaseFuncDesc and not ReleaseVarDesc or ReleaseTypeAttr? And why should the proxy tolerate NULL, but not direct (in-apartment) calls?
Not saying it's definitely wrong, but it seems incomplete if freeing NULL is supposed to be OK.