[PATCH v2] quartz: Don't call memcpy with src=0 in FM2_WriteFilterData (Coverity)

[Alex Henrie]

Signed-off-by: Alex Henrie 
---
Passing NULL to memcpy causes a segfault in newer versions of GCC, see
https://www.imperialviolet.org/2016/06/26/nonnull.html

v2: Fixed email address
---
 dlls/quartz/filtermapper.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/dlls/quartz/filtermapper.c b/dlls/quartz/filtermapper.c
index 0d26198637..7a5b5aee72 100644
--- a/dlls/quartz/filtermapper.c
+++ b/dlls/quartz/filtermapper.c
@@ -148,8 +148,11 @@ static int add_data(struct Vector * v, const BYTE * pData, int size)
         LPBYTE pOldData = v->pData;
         v->capacity = (v->capacity + size) * 2;
         v->pData = CoTaskMemAlloc(v->capacity);
-        memcpy(v->pData, pOldData, v->current);
-        CoTaskMemFree(pOldData);
+        if (pOldData)
+        {
+            memcpy(v->pData, pOldData, v->current);
+            CoTaskMemFree(pOldData);
+        }
     }
     memcpy(v->pData + v->current, pData, size);
     v->current += size;

[Andrew Eikum]

On Mon, Aug 13, 2018 at 04:33:21PM -0600, Alex Henrie wrote:
> Signed-off-by: Alex Henrie
>
> Passing NULL to memcpy causes a segfault in newer versions of GCC, see
> https://www.imperialviolet.org/2016/06/26/nonnull.html
> v2: Fixed email address
>
> dlls/quartz/filtermapper.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/dlls/quartz/filtermapper.c b/dlls/quartz/filtermapper.c
> index 0d26198637..7a5b5aee72 100644
> --- a/dlls/quartz/filtermapper.c
> +++ b/dlls/quartz/filtermapper.c
> @@ -148,8 +148,11 @@ static int add_data(struct Vector * v, const BYTE * pData, int size)
>          LPBYTE pOldData = v->pData;
>          v->capacity = (v->capacity + size) * 2;
>          v->pData = CoTaskMemAlloc(v->capacity);
> -
>    memcpy(v->pData, pOldData, v->current);
> -
>    CoTaskMemFree(pOldData);
> -
>    if (pOldData)
> -
>    {
> -
>        memcpy(v->pData, pOldData, v->current);
> -
>        CoTaskMemFree(pOldData);
> -
>    }

Could all this logic be replaced by CoTaskMemRealloc?

Andrew

[Alex Henrie]

On Tue, Aug 14, 2018 at 7:46 AM Andrew Eikum  wrote:
> On Mon, Aug 13, 2018 at 04:33:21PM -0600, Alex Henrie wrote:
> > Signed-off-by: Alex Henrie
> >
> > Passing NULL to memcpy causes a segfault in newer versions of GCC, see
> > https://www.imperialviolet.org/2016/06/26/nonnull.html
> > v2: Fixed email address
> >
> > dlls/quartz/filtermapper.c | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/dlls/quartz/filtermapper.c b/dlls/quartz/filtermapper.c
> > index 0d26198637..7a5b5aee72 100644
> > --- a/dlls/quartz/filtermapper.c
> > +++ b/dlls/quartz/filtermapper.c
> > @@ -148,8 +148,11 @@ static int add_data(struct Vector * v, const BYTE * pData, int size)
> >          LPBYTE pOldData = v->pData;
> >          v->capacity = (v->capacity + size) * 2;
> >          v->pData = CoTaskMemAlloc(v->capacity);
> > -
> >    memcpy(v->pData, pOldData, v->current);
> > -
> >    CoTaskMemFree(pOldData);
> > -
> >    if (pOldData)
> > -
> >    {
> > -
> >        memcpy(v->pData, pOldData, v->current);
> > -
> >        CoTaskMemFree(pOldData);
> > -
> >    }
>
> Could all this logic be replaced by CoTaskMemRealloc?

Good point. I don't see why not.

-Alex