Must copy static const D3DVERTEXELEMENT9 into dynamically alllocated D3DVERTEXELEMENT9 * before pointer passed correctly in Wine?

[misha680]

Sorry for reposting - apparently my original nabble link did not work :(

---

Dear All:

Sorry to bother... I am working on a D3DXCreateMesh/ID3DXMesh patch, and I
encountered a very strange error.

Specifically, it seems that when passing a LPD3DVERTEXELEMENT9 * parameter
to D3DXCreateMesh, 
a) in Windows - one can create 

    static const D3DVERTEXELEMENT9 decl1[3] = {
        {0, 0, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
D3DDECLUSAGE_POSITION, 0},
        {0, 0xC, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
D3DDECLUSAGE_NORMAL, 0},
        D3DDECL_END(), };

and pass (LPD3DVERTEXELEMENT9 *)&decl1 directly.

b) in Wine tests - I must first allocate a D3DVERTELEMENT9 *, then call
HeapAlloc, and finally copy my original decl1 and pass &decl (pointer to
newly allocated structure) for it to work. Any ideas?

Here is my git diff
http://wine.1045685.n5.nabble.com/file/n2805603/patch patch 

Thx
Misha

-- 
View this message in context: http://wine.1045685.n5.nabble.com/Must-copy-static-const-D3DVERTEXELEMENT9-i...
Sent from the Wine - Devel mailing list archive at Nabble.com.

[Nicolas Le Cam]

2010/9/7 misha680 <misha680(a)gmail.com>
> Sorry for reposting - apparently my original nabble link did not work :(
>
> ---
>
> Dear All:
>
> Sorry to bother... I am working on a D3DXCreateMesh/ID3DXMesh patch, and I
> encountered a very strange error.
>
> Specifically, it seems that when passing a LPD3DVERTEXELEMENT9 * parameter
> to D3DXCreateMesh,
> a) in Windows - one can create
>
> static const D3DVERTEXELEMENT9 decl1[3] = {
>        {0, 0, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
> D3DDECLUSAGE_POSITION, 0},
>        {0, 0xC, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
> D3DDECLUSAGE_NORMAL, 0},
>        D3DDECL_END(), };
>
> and pass (LPD3DVERTEXELEMENT9 *)&decl1 directly.
>
> b) in Wine tests - I must first allocate a D3DVERTELEMENT9 *, then call
> HeapAlloc, and finally copy my original decl1 and pass &decl (pointer to
> newly allocated structure) for it to work. Any ideas?
>
> Here is my git diff
> http://wine.1045685.n5.nabble.com/file/n2805603/patch patch
>
> Thx
> Misha
>
> --
> View this message in context:
> http://wine.1045685.n5.nabble.com/Must-copy-static-const-D3DVERTEXELEMENT9-i...
> Sent from the Wine - Devel mailing list archive at Nabble.com.
>
> Hi Misha,

It's currently a stub in wine [1] and tests are using a static const
D3DVERTEXELEMENT9 array as parameter [2] so it should be because of your
modified D3DXCreateMesh implementation which isn't visible in your patch.

[1]
http://source.winehq.org/git/wine.git/?a=blob;f=dlls/d3dx9_36/mesh.c#l586
[2]
http://source.winehq.org/git/wine.git/?a=blob;f=dlls/d3dx9_36/tests/mesh.c#l...

-- 
Nicolas Le Cam

[misha680]

My apologies if my first patch was unclear.

Here is a slightly modified git diff
http://wine.1045685.n5.nabble.com/file/n2806011/patch patch 

along with the relevant make test output
http://wine.1045685.n5.nabble.com/file/n2806011/maketestoutput
maketestoutput 

The relevant part of the make test output:
../../../tools/runtest -q -P wine -M d3dx9_36.dll -T ../../.. -p
d3dx9_36_test.exe.so mesh.c && touch mesh.ok
0x684e7c86 (nil)
0x32fcf8 0x123170

And the relevant statements are, in test/mesh.c:

    D3DVERTEXELEMENT9 *decl;

    static const D3DVERTEXELEMENT9 decl1[3] = {
        {0, 0, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
D3DDECLUSAGE_POSITION, 0},
        {0, 0xC, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT,
D3DDECLUSAGE_NORMAL, 0},
        D3DDECL_END(), };

    hr = D3DXCreateMesh(1, 3, D3DXMESH_MANAGED, (LPD3DVERTEXELEMENT9
*)&decl1, NULL, &d3dxmesh);
    todo_wine ok(hr == D3DERR_INVALIDCALL, "Got result %x, expected %x
(D3DERR_INVALIDCALL)\n", hr, D3DERR_INVALIDCALL);

    decl = HeapAlloc(GetProcessHeap(), 0, sizeof(decl1));
    if (!decl)
    {
        skip("Couldn't create declaration copy\n");
        return;
    }
    memcpy(decl, decl1, sizeof(decl1));

    hr = D3DXCreateMesh(1, 3, D3DXMESH_MANAGED, &decl, NULL, &d3dxmesh);
    todo_wine ok(hr == D3DERR_INVALIDCALL, "Got result %x, expected %x
(D3DERR_INVALIDCALL)\n", hr, D3DERR_INVALIDCALL);

    return;

and in mesh.c:

HRESULT WINAPI D3DXCreateMesh(DWORD numfaces, DWORD numvertices, DWORD
options, CONST LPD3DVERTEXELEMENT9 *declaration,
                              LPDIRECT3DDEVICE9 device, LPD3DXMESH *mesh)
{
    printf("%p %p\n", declaration, *declaration);

Now

LPD3DVERTEXELEMENT9 is a D3DVERTEXELEMENT9 *,
http://msdn.microsoft.com/en-us/library/bb172630%28VS.85%29.aspx

so LPD3DVERTEXELEMENT9 * is a D3DVERTEXELEMENT9 **, and to get at the
underlying D3DVERTEXELEMENT9 array, I believe I need to access *declaration,
e.g., (*declaration)[0], [1], etc.

Notably, if I use the HeapAlloc + memcpy approach, the test _fail_ in
Windows (I only tried once fyi), so at the moment it seems like one thing
works in Wine, another in Windows.

Thank you
Misha
-- 
View this message in context: http://wine.1045685.n5.nabble.com/Must-copy-static-const-D3DVERTEXELEMENT9-i...
Sent from the Wine - Devel mailing list archive at Nabble.com.

[Henri Verbeet]

On 7 September 2010 14:43, misha680 <misha680(a)gmail.com> wrote:
> LPD3DVERTEXELEMENT9 is a D3DVERTEXELEMENT9 *,
> http://msdn.microsoft.com/en-us/library/bb172630%28VS.85%29.aspx
>
> so LPD3DVERTEXELEMENT9 * is a D3DVERTEXELEMENT9 **, and to get at the
> Well yes, but why do you think that makes sense as an argument to
> D3DXCreateMesh()? Did you check the prototype against the SDK?

[Misha Koshelev]

On Tue, 2010-09-07 at 15:56 +0200, Henri Verbeet wrote:
> On 7 September 2010 14:43, misha680 <misha680(a)gmail.com> wrote:
> > LPD3DVERTEXELEMENT9 is a D3DVERTEXELEMENT9 *,
> > http://msdn.microsoft.com/en-us/library/bb172630%28VS.85%29.aspx
> >
> > so LPD3DVERTEXELEMENT9 * is a D3DVERTEXELEMENT9 **, and to get at the
> > Well yes, but why do you think that makes sense as an argument to
> > D3DXCreateMesh()? Did you check the prototype against the SDK?

Thank you. Will take a look later today.

Misha