Re: [PATCH 1/4] oleaut32: Add tests for safearray unmarshaling into passed variant

[Dmitry Timoshkov]

Hi Piotr,

Piotr Caban <piotr(a)codeweavers.com> wrote:
> @@ -1316,11 +1316,19 @@ static void test_marshal_VARIANT(void)
>      ok(*wirev, "wv[6] %08x\n", *wirev); /* win2k: this is lpsa. winxp: this is (char*)lpsa + 1 */
>      wirev++;
>      check_safearray(wirev, lpsa);
> -    VariantInit(&v2);
> +    mem = CoTaskMemAlloc(1024);
> +    lpsa_copy = lpsa2 = SafeArrayCreate(VT_I8, 1, &sab);
> +    CoTaskMemFree(mem);
> +    mem = lpsa2->pvData;
> +    V_VT(&v2) = VT_UI4 | VT_ARRAY | VT_BYREF;
> +    V_ARRAYREF(&v2) = &lpsa2;

Looks like the result of CoTaskMemAlloc/CoTaskMemFree is not used.

-- 
Dmitry.

[PATCH 1/4] oleaut32: Add tests for safearray unmarshaling into passed variant

[Piotr Caban]

Hi Dmitry,

On 01/13/16 03:47, Dmitry Timoshkov wrote:
> Piotr Caban <piotr(a)codeweavers.com> wrote:
> > @@ -1316,11 +1316,19 @@ static void test_marshal_VARIANT(void)
> >       ok(*wirev, "wv[6] %08x\n", *wirev); /* win2k: this is lpsa. winxp: this is (char*)lpsa + 1 */
> >       wirev++;
> >       check_safearray(wirev, lpsa);
> > -    VariantInit(&v2);
> > +    mem = CoTaskMemAlloc(1024);
> > +    lpsa_copy = lpsa2 = SafeArrayCreate(VT_I8, 1, &sab);
> > +    CoTaskMemFree(mem);
> > +    mem = lpsa2->pvData;
> > +    V_VT(&v2) = VT_UI4 | VT_ARRAY | VT_BYREF;
> > +    V_ARRAYREF(&v2) = &lpsa2;
>
> Looks like the result of CoTaskMemAlloc/CoTaskMemFree is not used.
> I've added it intentionally so there's a bigger chance a different 
> memory is allocated if safe array is reallocated (I'm trying to show 
> that memory is reused without freeing and allocating it again). Of 
> course there's no guarantee for that. I can remove it if it's preferable.

Thanks,
Piotr

[PATCH 1/4] oleaut32: Add tests for safearray unmarshaling into passed variant

[Dmitry Timoshkov]

Piotr Caban <piotr(a)codeweavers.com> wrote:
> > > @@ -1316,11 +1316,19 @@ static void test_marshal_VARIANT(void)
> > >       ok(*wirev, "wv[6] %08x\n", *wirev); /* win2k: this is lpsa. winxp: this is (char*)lpsa + 1 */
> > >       wirev++;
> > >       check_safearray(wirev, lpsa);
> > > -    VariantInit(&v2);
> > > +    mem = CoTaskMemAlloc(1024);
> > > +    lpsa_copy = lpsa2 = SafeArrayCreate(VT_I8, 1, &sab);
> > > +    CoTaskMemFree(mem);
> > > +    mem = lpsa2->pvData;
> > > +    V_VT(&v2) = VT_UI4 | VT_ARRAY | VT_BYREF;
> > > +    V_ARRAYREF(&v2) = &lpsa2;
> >
> > Looks like the result of CoTaskMemAlloc/CoTaskMemFree is not used.
> > I've added it intentionally so there's a bigger chance a different 
> > memory is allocated if safe array is reallocated (I'm trying to show 
> > that memory is reused without freeing and allocating it again). Of 
> > course there's no guarantee for that. I can remove it if it's preferable.

I'd suggest to add a comment about this so that noone would try to remove
"dead code".

-- 
Dmitry.

[PATCH 1/4] oleaut32: Add tests for safearray unmarshaling into passed variant

[Piotr Caban]

On 01/13/16 08:19, Dmitry Timoshkov wrote:
> Piotr Caban <piotr(a)codeweavers.com> wrote:
> > > > @@ -1316,11 +1316,19 @@ static void test_marshal_VARIANT(void)
> > > >        ok(*wirev, "wv[6] %08x\n", *wirev); /* win2k: this is lpsa. winxp: this is (char*)lpsa + 1 */
> > > >        wirev++;
> > > >        check_safearray(wirev, lpsa);
> > > > -    VariantInit(&v2);
> > > > +    mem = CoTaskMemAlloc(1024);
> > > > +    lpsa_copy = lpsa2 = SafeArrayCreate(VT_I8, 1, &sab);
> > > > +    CoTaskMemFree(mem);
> > > > +    mem = lpsa2->pvData;
> > > > +    V_VT(&v2) = VT_UI4 | VT_ARRAY | VT_BYREF;
> > > > +    V_ARRAYREF(&v2) = &lpsa2;
> > >
> > > Looks like the result of CoTaskMemAlloc/CoTaskMemFree is not used.
> > > I've added it intentionally so there's a bigger chance a different
> > > memory is allocated if safe array is reallocated (I'm trying to show
> > > that memory is reused without freeing and allocating it again). Of
> > > course there's no guarantee for that. I can remove it if it's preferable.
>
> I'd suggest to add a comment about this so that noone would try to remove
> "dead code".
> I've sent a fixed version of the patch. I have added FADF_STATIC feature 
> and removed no longer needed CoTaskMemAlloc call.

Thanks,
Piotr