msi: Avoid a potential NULL pointer dereference (coccicheck).

[Marcus Meissner]

On Thu, May 20, 2010 at 12:36:21PM +0200, Michael Stefaniuc wrote:
> Paul Vriens wrote:
> > On 05/20/2010 01:15 AM, Michael Stefaniuc wrote:
> >> The last "goto done" is for si == NULL. When MSI_GetSummaryInformationW
> >> returns NULL there is a crash.
> >> ---
> >>   dlls/msi/msi.c |    3 ++-
> >>   1 files changed, 2 insertions(+), 1 deletions(-)
> >>
> >> diff --git a/dlls/msi/msi.c b/dlls/msi/msi.c
> >> index 3170e6d..9c08d1b 100644
> >> --- a/dlls/msi/msi.c
> >> +++ b/dlls/msi/msi.c
> >> @@ -551,7 +551,8 @@ static UINT MSI_ApplicablePatchW( MSIPACKAGE
> >> *package, LPCWSTR patch )
> >>
> >>   done:
> >>       msiobj_release(&patch_db->hdr );
> >> -    msiobj_release(&si->hdr );
> >> +    if (si)
> >> +        msiobj_release(&si->hdr );
> >>       return r;
> >>   }
> >>
> > 
> > Hi Michael,
> > 
> > This one is mentioned by Coverity (#970). Marcus marked this one as
> > 'FALSE' with the remark:
> > 
> > "hdr is at position 0, so this will be NULL and msiobj_release handles it."
> > 
> > Thoughts?
> 
> Do the compilers treat the addressof operator on the struct member as
>   si + FIELD_OFFSET(MSISUMMARYINFO, hdr)
> ?
> Then yes, no dereference happens and the result of the above calculation
> is NULL which is fine as input for msiobj_release().

Yes, they do.

Ciao, Marcus