Re: [PATCH 2/2] d3dxof: Use sizeof(GUID) instead of hardcoding the value.

[Rico Schüller]

Why is there a "static guid" used in GetType? What happens if you query:
hr = IDirectXFileData_GetType(lpDirectXFileData, &clsid_type);
hr = IDirectXFileData_GetType(lpDirectXFileData2, &clsid_type2);
Are thy both containing the value to clsid_type2 then? Do clsid_type and 
clsid_type2 match in that case to the same pointer both pointing to the 
same static guid? I don't see a test for this. Why doesn't *pguid = 
&This->pobj->type; do the job in that case?

While trying to check the cases above, I got the following problem...
Under which circumstances is the test_dump test run? It looks like the 
tests are never run because there is no objects.txt, see:
     hFile = CreateFileA("objects.txt", GENERIC_READ, FILE_SHARE_READ, 
NULL, OPEN_EXISTING, 0, NULL);
     if (hFile == INVALID_HANDLE_VALUE)
       return;
     CloseHandle(hFile);

All machines (http://test.winehq.org/data/tests/d3dxof:d3dxof.html) seem 
to run 107 tests or skip all of them, which is the same on my machine (I 
got 107 tests), so none of them is running the whole test_dump ... How 
is this supposed to work?

Cheers
Rico

On 21.10.2012 14:50, Christian Costa wrote:
> @@ -647,7 +647,7 @@ static HRESULT WINAPI IDirectXFileDataImpl_GetType(IDirectXFileData* iface, cons
>     if (!pguid)
>       return DXFILEERR_BADVALUE;
> - memcpy(&guid, &This->pobj->type, 16);
> -
>
> memcpy(&guid, &This->pobj->type, sizeof(GUID));
>  *pguid = &guid;
>
> return DXFILE_OK;

[PATCH 2/2] d3dxof: Use sizeof(GUID) instead of hardcoding the value.

[Christian Costa]

Le 21/10/2012 16:21, Rico Schüller a écrit :
> Why is there a "static guid" used in GetType? What happens if you query:
> hr = IDirectXFileData_GetType(lpDirectXFileData, &clsid_type);
> hr = IDirectXFileData_GetType(lpDirectXFileData2, &clsid_type2);
> Are thy both containing the value to clsid_type2 then? Do clsid_type 
> and clsid_type2 match in that case to the same pointer both pointing 
> to the same static guid? I don't see a test for this.

I think because some brain dead apps write at this location. it's to 
preserve the object value.
That said it has been a long time I wrote this code. I don't remember 
and it's curious I didn't left a comment for that.
I will check if I can find something.
> Why doesn't *pguid = &This->pobj->type; do the job in that case?

Indeed. This can be changed. Let's discard this patch.
That said few time ago I have problem with this in some d3drm code and 
had to change to use memcpy.
I didn't investigate much tough.
> While trying to check the cases above, I got the following problem...
> Under which circumstances is the test_dump test run? It looks like the 
> tests are never run because there is no objects.txt, see:
>     hFile = CreateFileA("objects.txt", GENERIC_READ, FILE_SHARE_READ, 
> NULL, OPEN_EXISTING, 0, NULL);
>     if (hFile == INVALID_HANDLE_VALUE)
>       return;
>     CloseHandle(hFile);
>
> All machines (http://test.winehq.org/data/tests/d3dxof:d3dxof.html) 
> seem to run 107 tests or skip all of them, which is the same on my 
> machine (I got 107 tests), so none of them is running the whole 
> test_dump ... How is this supposed to work?

It's for debug. But it does not check much thing. Only display.
It's very usefull to compare outputs with both native and builtin.