Re: [PATCH v2 3/3] d2d1: Implement d2d_effect_{Get, Set}Input{, Count}().

[Henri Verbeet]

On Wed, 4 Aug 2021 at 04:26, Ziqing Hui  wrote:
> On 8/3/21 11:53 PM, Henri Verbeet wrote:
> > > static HRESULT STDMETHODCALLTYPE d2d_effect_SetInputCount(ID2D1Effect *iface, UINT32 count)
> > >  {
> > > - FIXME("iface %p, count %u stub!\n", iface, count);
> > > - struct d2d_effect *effect = impl_from_ID2D1Effect(iface);
> > > - unsigned int i;
> > > - return E_NOTIMPL;
> > > - TRACE("iface %p, count %u.\n", iface, count);
> > > -
> > > - if (count < effect->min_inputs || count > effect->max_inputs)
> > > -
> > >    return E_INVALIDARG;
> > > -
> > > - if (count != effect->inputs_count)
> > > - {
> > > -
> > >    if (count < effect->inputs_count)
> > > -
> > >    {
> > > -
> > >        for (i = count; i < effect->inputs_count; ++i)
> > > -
> > >        {
> > > -
> > >            if (effect->inputs[i])
> > > -
> > >                ID2D1Image_Release(effect->inputs[i]);
> > > -
> > >        }
> > > -
> > >    }
> > > -
> > >    effect->inputs_count = count;
> > > -
> > >    HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, effect->inputs, sizeof(*effect->inputs) * count);
> > > - }
> > > -
> > > - return S_OK;
> > >
> > > }
> >
> > That doesn't generally work; the memory returned by HeapReAlloc() may
> > not be the same as what was originally passed in. More broadly, it's
> > not clear to me that we gain much from avoiding d2d_array_reserve()
> > here.
>
> I need to use HeapReAlloc(HEAP_ZERO_MEMORY) to initialize the newly allocated memory to zero.
>
> Which way is better? HeapReAlloc(HEAP_ZERO_MEMORY) or d2d_array_reserve() + memset(0)?

I'd use memset().