HI all,

Is there any way to play ASF file to contain any compressed data (to say
H264) with Windows Media Player having usual DirectShow decoder (not DMO).

If so, what software has to be installed (f.e. WMF Runtime or WM Player )?
Thanks in advance.

Regards,
Dmitry Vergeles
www.solveigmm.com

WMP7+ will read and parse the ASF using the WMASFReader and
decode the audio and video streams using any
DirectShow-compatible decoder, that is a DirectShow
transform filter, a DMO or a VCM/ACM codec. This works as
long as the correct media type is associated with the stream
and the stream is either audio or video.

No special software is required but the usual DirectShow and
WMF runtimes, which are already installed if WMP works.

Is there a specific problem you are having?

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

Hello Alessandro,

May be I've missed some key things from my last dealing with Winows Media
7.1 SDK. In this key could you please refer to a documentation article that
says
about decoders for ASF playing in windows media technologies may be
DirectShow
filters?

In the article-  "Microsoft Windows Media and DirectShow: Options for
Application Developers"
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnwm...
addingwindowsmediasupportwiththewindowsmediaformat.asp

there is a section " Reading Windows Media files with the Windows Media
Format
SDK"  that says -

the process of reading and decoding Windows Media is supported directly
through the Windows Media Format SDK using codecs that are implemented as
either Audio Codec Manager (ACM) or Video Codec Manager (VCM) installable
codecs, or DirectX Media Objects (DMOs).

Moreover, since WMP player 6.4 had a different behaveour with  WMP 7+ while
playing Windows Media, it was possible to play ASF with usuall DirectShow
decoder with 6.4 version but from WMP 7.0  only DMO decoders were used.

So, it herhaps that my concept are obsolete...

Yes I can't play ASF with h264 both by WMPlayer 9.0 and 10. It says that
there is no appropreated codec.
But if I render it in GraphEdit it is played properly.
What cause for the problem may be?

Regards,
Dmitry Vergeles
www.solveigmm.com

There is none. Read below.

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnwm...

Actually, the URL is:

http://msdn.microsoft.com/library/en-us/dnwmt/html/wm_ds_options.asp

As far I understand it, this article refers to the WMReader
object of the WMFSDK and not the WMASFReader filter in
DirectShow. The WMReader object delivers decompressed data
by default and it internally decompresses the stream using
either DMOs or ACM/VCM codecs (this last bit is new to me,
but let's believe what the article says).

WMP7+ uses the WMASFReader source filter which delivers
compressed data instead, which is then decoded by a
DirectShow filter in the playback graph (since you can only
add filters to a graph). This filter can be the DMOWrapper,
the ACMWrapper or the AVIDecompressor, but it should be
possible to use any other filter, as your successful
attempts in GraphEdit and WMP6.4 show.

There are 3 known behaviors:

1. (WMP6.4) RenderFile() on a FilterGraphNoThread using the
WMSourceFilter

2. (GE) RenderFile() on a FilterGraph using the WMASFReader

3. (WMP7+) semi-manual graph building in a
FilterGraphNoThread using the WMASFReader

If your decoding filter works in GE and WMP6.4 (and you have
both WMP9 and DX9 installed, otherwise there may be other
issues), then the problem is how WMP7+ messes with the
graph, which is mostly unknown, and the solution would be to
find out what WMP7+ does that prevents your filter from
working.

If you're game, I can provide techniques (which require
C/C++) and tools to help troubleshoot this and I can help
you do it, but there is no guarantee and no foreseeable time
frame, unless someone from MSFT barges in and tells us the
solution.

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

Alessandro, first of all thank you for the so detailed explanation..

Hmm.. It sounds

I think indeed!

One more thing is that except GraphEdit, the ASF files are played with
DSPlay sample to be shipped with WMFSDK 9.

It sounds to be designing :)
I'm ready to try...

[Sorry for the text/richtext, but it is better to preserve the code and the links.]

DSPlay behaves exactly like GE.

I'll explain about hooking and logging later. As a convention, I prefix hook function names with 'new' and the original function names with 'old'.

You need to create a C++ project for a DLL. This is the pseudo-code:

----------------------------------------
HMODULE hQuartzDLL = NULL;

BOOL WINAPI DllMain(
 HINSTANCE hModule,
 DWORD dwReason,
 LPVOID lpReserved)
{
 switch(dwReason)
 {
  case DLL_PROCESS_ATTACH:
   OpenLogChannel();
   HookFunction(ole32.dll,CoCreateInstance);
  break;

  case DLL_PROCESS_DETACH:
   FreeFunction(ole32.dll,CoCreateInstance);
   FreeLibrary(hQuartzDLL);
   CloseLogChannel();
  break;
 }
 return TRUE;

HRESULT WINAPI newCoCreateInstance(
 REFCLSID rclsid,
 LPUNKNOWN pUnkOuter,
 DWORD dwClsContext,
 REFIID riid,
 LPVOID* ppv)
{
 HRESULT hr = S_OK;

 if(S_OK == (hr = oldCoCreateInstance(...))
  && IsEqualCLSID(rclsid,CLSID_FilterGraphNoThread)
  && NULL != (hQuartzDLL = LoadLibrary(quartz.dll)) {
  IGraphBuilder* pGraph
   = (*(IUnknown**)ppv)->QueryInterface(IGraphBuilder);
  HookGraphBuilder(pGraph);
  pGraph->Release();
 }

 return hr;

----------------------------------------

You need to inject this DLL inside the WMP process before it builds the graph: a good idea is to inject it when the process is created.

The easiest way to hook functions and inject DLLs into processes is to use Detours, a library from http://research.microsoft.com/sn/detours/ (the free 1.5 version is ok). It is possible to do it without, but it's not nearly as easy.

The withdll.exe sample can be used to create a process with an injected DLL (just remember to use the absolute path of the DLL or to put the DLL somewhere in the system %PATH%).

Detours offers 2 ways to hook functions:

----------------------------------------
DETOUR_TRAMPOLINE(
 HRESULT WINAPI oldCoCreateInstance(...args...),
 CoCreateInstance);
HRESULT WINAPI newCoCreateInstance(...args...)
{
 ...

DetourFunctionWithTrampoline(
    (PBYTE)oldCoCreateInstance,
    (PBYTE)newCoCreateInstance);
----------------------------------------

and

----------------------------------------
HRESULT (STDMETHODCALLTYPE *oldInterface_Method)
(
 IInterface* pThis
 ...args...
) = NULL;
HRESULT STDMETHODCALLTYPE newInterface_Method
(
 IInterface* pThis
 ...args...
)
{
 ...

*(LPBYTE*)&oldInterfaceMethod
 = DetourFunction(
  ((LPBYTE**)(pInterface))[0][METHOD_OFFSET],
  (LPBYTE)newInterface_Method);

----------------------------------------

The first way is useful to hook API entrypoints, as shown, the second is more useful with interface methods (notice that an interface method always has a hidden first argument that is the this pointer to the interface, as shown).

Since I still haven't found a way to get a pointer to a virtual method in C++ that works, I just cast it to something the C++ compiler won't pester me about and get to it walking the vtable myself, as shown. You need to know the METHOD_OFFSET in the vtable, which you can find out easily by counting the methods in the interface definition (inluding inherited interfaces) so QueryInterface() in IUnknown (and thus in any COM interface) has offset 0 and, for example, RenderFile() in IGraphBuilder() is at offset 13.

Since C is much easier to deal with when it comes to vtables, you may choose to mix C and C++ (in which case I'll explain how to get a pointer to a method in C).

Now you know how to inject a DLL, hook an API entrypoint or hook an interface method.

To log, you can use any way you like. I write to a text file because this kind of I/O has no risk of a deadlock like GUI operations (since both the GUI and the graph will use the message pump, unless you use different synchronized threads) and is easier that network I/O, like pipes or sockets or whatever, and shared memory (however, network I/O and shared memory would allow you to write an external logging UI that shows the log in real-time). To implement logging to a file, you need an extern file handle and an extern mutex to prevent write operations in different threads to overlap.

----------------------------------------
FILE* f = NULL;
HANDLE m = NULL;

OpenLogChannel()
{
 f = fopen(...);
 m = CreateMutex(NULL,FALSE,NULL);

CloseLogChannel()
{
 fclose(f);
 CloseHandle(m);
LogMessage(LPCSTR format, ...)
{
 va_list va;
 va_start(va,format);
 WaitForSingleObject(m,INFINITE);
  vfprintf(f,format,va);
 ReleaseMutex(m);
----------------------------------------

This way LogMessage() works like printf() and it is very flexible. You can use the Unicode variants if you prefer a Unicode build.

And now all the building pieces are in place so, what should you log? That is, what should you do inside the routine I called HookGraphBuilder()?

For example, you could hook IGraphBuilder::RenderFile() in order to dump the graph contents before and after the actual call to the original method. You can dump the graph yourself or save it to a GRF file:

http://msdn.microsoft.com/library/en-us/directshow/htm/saveafiltergra...

You can parse the GRF file afterwards if GE can't open it:

http://msdn.microsoft.com/library/en-us/directshow/htm/directshowgrap...

Here is my sample parser:

http://groups-beta.google.com/group/microsoft.public.win32.programmer...

You can hook IFilterGraph::AddFilter(), IGraphBuilder::AddSourceFilter() and IFilterGraph2::AddSourceFilterForMoniker() to log whether WMP adds filters on its own.

You can also set your IAMFilterGraphCallback and IAMGraphBuilderCallback callbacks to be notified by the graph of errors in rendering the graph. If you do, call IObjectWithSite::GetSite() before IObjectWithSite::SetSite() to see whether WMP installed its own callbacks: if it did, cache the references and call them when your callbacks are called.

You can also ask the graph to log what it does on its own with IGraphBuilder::SetLogFile().

You can do all of this for a file that works and one that doesn't, to see what's different. You can do the same with WMP7+ and GE and WMP6.4... If you spy GE or DSPlay or other programs, in newCoCreateInstance() you should also accept CLSID_FilterGraph in addition to CLSID_FilterGraphNoThread().

This was a long message to write :-)

If you need support when you try this out, just ask on this same thread.

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

wow, it looks to be impressive..

One may assume :))

So, I'll try it ASAP and notify the results.

Regards,
Dmitry Vergeles
www.solveigmm.com

  "Alessandro Angeli [MVP::DigitalMedia]" <nob...@nowhere.in.the.net> wrote in message news:u$AeuffNFHA.688@TK2MSFTNGP10.phx.gbl...
  [Sorry for the text/richtext, but it is better to preserve the code and the links.]

  Dmitry Vergeles (SolveigMM) wrote:

  >> 2. (GE) RenderFile() on a FilterGraph using the
  >> WMASFReader
  >> 3. (WMP7+) semi-manual graph building in a
  >> FilterGraphNoThread using the WMASFReader
  >
  > One more thing is that except GraphEdit, the ASF files
  > are played with DSPlay sample to be shipped with WMFSDK 9.

  DSPlay behaves exactly like GE.

  > It sounds to be designing :)
  > I'm ready to try...

  I'll explain about hooking and logging later. As a convention, I prefix hook function names with 'new' and the original function names with 'old'.

  You need to create a C++ project for a DLL. This is the pseudo-code:

  ----------------------------------------
  HMODULE hQuartzDLL = NULL;

  BOOL WINAPI DllMain(
   HINSTANCE hModule,
   DWORD dwReason,
   LPVOID lpReserved)
  {
   switch(dwReason)
   {
    case DLL_PROCESS_ATTACH:
     OpenLogChannel();
     HookFunction(ole32.dll,CoCreateInstance);
    break;

    case DLL_PROCESS_DETACH:
     FreeFunction(ole32.dll,CoCreateInstance);
     FreeLibrary(hQuartzDLL);
     CloseLogChannel();
    break;
   }
   return TRUE;
  }

  HRESULT WINAPI newCoCreateInstance(
   REFCLSID rclsid,
   LPUNKNOWN pUnkOuter,
   DWORD dwClsContext,
   REFIID riid,
   LPVOID* ppv)
  {
   HRESULT hr = S_OK;

   if(S_OK == (hr = oldCoCreateInstance(...))
    && IsEqualCLSID(rclsid,CLSID_FilterGraphNoThread)
    && NULL != (hQuartzDLL = LoadLibrary(quartz.dll)) {
    IGraphBuilder* pGraph
     = (*(IUnknown**)ppv)->QueryInterface(IGraphBuilder);
    HookGraphBuilder(pGraph);
    pGraph->Release();
   }

   return hr;
  }

  ----------------------------------------

  You need to inject this DLL inside the WMP process before it builds the graph: a good idea is to inject it when the process is created.

  The easiest way to hook functions and inject DLLs into processes is to use Detours, a library from http://research.microsoft.com/sn/detours/ (the free 1.5 version is ok). It is possible to do it without, but it's not nearly as easy.

  The withdll.exe sample can be used to create a process with an injected DLL (just remember to use the absolute path of the DLL or to put the DLL somewhere in the system %PATH%).

  Detours offers 2 ways to hook functions:

  ----------------------------------------
  DETOUR_TRAMPOLINE(
   HRESULT WINAPI oldCoCreateInstance(...args...),
   CoCreateInstance);
  HRESULT WINAPI newCoCreateInstance(...args...)
  {
   ...
  }

  DetourFunctionWithTrampoline(
      (PBYTE)oldCoCreateInstance,
      (PBYTE)newCoCreateInstance);
  ----------------------------------------

  and

  ----------------------------------------
  HRESULT (STDMETHODCALLTYPE *oldInterface_Method)
  (
   IInterface* pThis
   ...args...
  ) = NULL;
  HRESULT STDMETHODCALLTYPE newInterface_Method
  (
   IInterface* pThis
   ...args...
  )
  {
   ...
  }

  *(LPBYTE*)&oldInterfaceMethod
   = DetourFunction(
    ((LPBYTE**)(pInterface))[0][METHOD_OFFSET],
    (LPBYTE)newInterface_Method);

  ----------------------------------------

  The first way is useful to hook API entrypoints, as shown, the second is more useful with interface methods (notice that an interface method always has a hidden first argument that is the this pointer to the interface, as shown).

  Since I still haven't found a way to get a pointer to a virtual method in C++ that works, I just cast it to something the C++ compiler won't pester me about and get to it walking the vtable myself, as shown. You need to know the METHOD_OFFSET in the vtable, which you can find out easily by counting the methods in the interface definition (inluding inherited interfaces) so QueryInterface() in IUnknown (and thus in any COM interface) has offset 0 and, for example, RenderFile() in IGraphBuilder() is at offset 13.

  Since C is much easier to deal with when it comes to vtables, you may choose to mix C and C++ (in which case I'll explain how to get a pointer to a method in C).

  Now you know how to inject a DLL, hook an API entrypoint or hook an interface method.

  To log, you can use any way you like. I write to a text file because this kind of I/O has no risk of a deadlock like GUI operations (since both the GUI and the graph will use the message pump, unless you use different synchronized threads) and is easier that network I/O, like pipes or sockets or whatever, and shared memory (however, network I/O and shared memory would allow you to write an external logging UI that shows the log in real-time). To implement logging to a file, you need an extern file handle and an extern mutex to prevent write operations in different threads to overlap.

  ----------------------------------------
  FILE* f = NULL;
  HANDLE m = NULL;

  OpenLogChannel()
  {
   f = fopen(...);
   m = CreateMutex(NULL,FALSE,NULL);
  }
  CloseLogChannel()
  {
   fclose(f);
   CloseHandle(m);
  }
  LogMessage(LPCSTR format, ...)
  {
   va_list va;
   va_start(va,format);
   WaitForSingleObject(m,INFINITE);
    vfprintf(f,format,va);
   ReleaseMutex(m);
  }
  ----------------------------------------

  This way LogMessage() works like printf() and it is very flexible. You can use the Unicode variants if you prefer a Unicode build.

  And now all the building pieces are in place so, what should you log? That is, what should you do inside the routine I called HookGraphBuilder()?

  For example, you could hook IGraphBuilder::RenderFile() in order to dump the graph contents before and after the actual call to the original method. You can dump the graph yourself or save it to a GRF file:

  http://msdn.microsoft.com/library/en-us/directshow/htm/saveafiltergra...

  You can parse the GRF file afterwards if GE can't open it:

  http://msdn.microsoft.com/library/en-us/directshow/htm/directshowgrap...

  Here is my sample parser:

  http://groups-beta.google.com/group/microsoft.public.win32.programmer...

  You can hook IFilterGraph::AddFilter(), IGraphBuilder::AddSourceFilter() and IFilterGraph2::AddSourceFilterForMoniker() to log whether WMP adds filters on its own.

  You can also set your IAMFilterGraphCallback and IAMGraphBuilderCallback callbacks to be notified by the graph of errors in rendering the graph. If you do, call IObjectWithSite::GetSite() before IObjectWithSite::SetSite() to see whether WMP installed its own callbacks: if it did, cache the references and call them when your callbacks are called.

  You can also ask the graph to log what it does on its own with IGraphBuilder::SetLogFile().

  You can do all of this for a file that works and one that doesn't, to see what's different. You can do the same with WMP7+ and GE and WMP6.4... If you spy GE or DSPlay or other programs, in newCoCreateInstance() you should also accept CLSID_FilterGraph in addition to CLSID_FilterGraphNoThread().

  This was a long message to write :-)

  If you need support when you try this out, just ask on this same thread.

  --

  // Alessandro Angeli
  // MVP :: Digital Media
  // a dot angeli at psynet dot net

WMP 7+ always uses the WMF SDK to do decoding, which is restricted to VCM or DMO decoders. So hooking the DLLs below might be informative, but I'm not sure if you'll find anything too helpful. Maybe you'll find some way to hack the system, but it won't be an official, supported way to do things.

The "official" way to do what you want is to write the codec as a DMO.

Thanks,
Geoff

--
This posting is provided "AS IS" with no warranties, and confers no rights.

  "Dmitry Vergeles (SolveigMM)" <dark...@elecard.net.ru> wrote in message news:uYdtJynNFHA.1476@TK2MSFTNGP09.phx.gbl...
  wow, it looks to be impressive..
  >>This was a long message to write :-)
  One may assume :))

  So, I'll try it ASAP and notify the results.

  Regards,
  Dmitry Vergeles
  www.solveigmm.com

    "Alessandro Angeli [MVP::DigitalMedia]" <nob...@nowhere.in.the.net> wrote in message news:u$AeuffNFHA.688@TK2MSFTNGP10.phx.gbl...
    [Sorry for the text/richtext, but it is better to preserve the code and the links.]

    Dmitry Vergeles (SolveigMM) wrote:

    >> 2. (GE) RenderFile() on a FilterGraph using the
    >> WMASFReader
    >> 3. (WMP7+) semi-manual graph building in a
    >> FilterGraphNoThread using the WMASFReader
    >
    > One more thing is that except GraphEdit, the ASF files
    > are played with DSPlay sample to be shipped with WMFSDK 9.

    DSPlay behaves exactly like GE.

    > It sounds to be designing :)
    > I'm ready to try...

    I'll explain about hooking and logging later. As a convention, I prefix hook function names with 'new' and the original function names with 'old'.

    You need to create a C++ project for a DLL. This is the pseudo-code:

    ----------------------------------------
    HMODULE hQuartzDLL = NULL;

    BOOL WINAPI DllMain(
     HINSTANCE hModule,
     DWORD dwReason,
     LPVOID lpReserved)
    {
     switch(dwReason)
     {
      case DLL_PROCESS_ATTACH:
       OpenLogChannel();
       HookFunction(ole32.dll,CoCreateInstance);
      break;

      case DLL_PROCESS_DETACH:
       FreeFunction(ole32.dll,CoCreateInstance);
       FreeLibrary(hQuartzDLL);
       CloseLogChannel();
      break;
     }
     return TRUE;
    }

    HRESULT WINAPI newCoCreateInstance(
     REFCLSID rclsid,
     LPUNKNOWN pUnkOuter,
     DWORD dwClsContext,
     REFIID riid,
     LPVOID* ppv)
    {
     HRESULT hr = S_OK;

     if(S_OK == (hr = oldCoCreateInstance(...))
      && IsEqualCLSID(rclsid,CLSID_FilterGraphNoThread)
      && NULL != (hQuartzDLL = LoadLibrary(quartz.dll)) {
      IGraphBuilder* pGraph
       = (*(IUnknown**)ppv)->QueryInterface(IGraphBuilder);
      HookGraphBuilder(pGraph);
      pGraph->Release();
     }

     return hr;
    }

    ----------------------------------------

    You need to inject this DLL inside the WMP process before it builds the graph: a good idea is to inject it when the process is created.

    The easiest way to hook functions and inject DLLs into processes is to use Detours, a library from http://research.microsoft.com/sn/detours/ (the free 1.5 version is ok). It is possible to do it without, but it's not nearly as easy.

    The withdll.exe sample can be used to create a process with an injected DLL (just remember to use the absolute path of the DLL or to put the DLL somewhere in the system %PATH%).

    Detours offers 2 ways to hook functions:

    ----------------------------------------
    DETOUR_TRAMPOLINE(
     HRESULT WINAPI oldCoCreateInstance(...args...),
     CoCreateInstance);
    HRESULT WINAPI newCoCreateInstance(...args...)
    {
     ...
    }

    DetourFunctionWithTrampoline(
        (PBYTE)oldCoCreateInstance,
        (PBYTE)newCoCreateInstance);
    ----------------------------------------

    and

    ----------------------------------------
    HRESULT (STDMETHODCALLTYPE *oldInterface_Method)
    (
     IInterface* pThis
     ...args...
    ) = NULL;
    HRESULT STDMETHODCALLTYPE newInterface_Method
    (
     IInterface* pThis
     ...args...
    )
    {
     ...
    }

    *(LPBYTE*)&oldInterfaceMethod
     = DetourFunction(
      ((LPBYTE**)(pInterface))[0][METHOD_OFFSET],
      (LPBYTE)newInterface_Method);

    ----------------------------------------

    The first way is useful to hook API entrypoints, as shown, the second is more useful with interface methods (notice that an interface method always has a hidden first argument that is the this pointer to the interface, as shown).

    Since I still haven't found a way to get a pointer to a virtual method in C++ that works, I just cast it to something the C++ compiler won't pester me about and get to it walking the vtable myself, as shown. You need to know the METHOD_OFFSET in the vtable, which you can find out easily by counting the methods in the interface definition (inluding inherited interfaces) so QueryInterface() in IUnknown (and thus in any COM interface) has offset 0 and, for example, RenderFile() in IGraphBuilder() is at offset 13.

    Since C is much easier to deal with when it comes to vtables, you may choose to mix C and C++ (in which case I'll explain how to get a pointer to a method in C).

    Now you know how to inject a DLL, hook an API entrypoint or hook an interface method.

    To log, you can use any way you like. I write to a text file because this kind of I/O has no risk of a deadlock like GUI operations (since both the GUI and the graph will use the message pump, unless you use different synchronized threads) and is easier that network I/O, like pipes or sockets or whatever, and shared memory (however, network I/O and shared memory would allow you to write an external logging UI that shows the log in real-time). To implement logging to a file, you need an extern file handle and an extern mutex to prevent write operations in different threads to overlap.

    ----------------------------------------
    FILE* f = NULL;
    HANDLE m = NULL;

    OpenLogChannel()
    {
     f = fopen(...);
     m = CreateMutex(NULL,FALSE,NULL);
    }
    CloseLogChannel()
    {
     fclose(f);
     CloseHandle(m);
    }
    LogMessage(LPCSTR format, ...)
    {
     va_list va;
     va_start(va,format);
     WaitForSingleObject(m,INFINITE);
      vfprintf(f,format,va);
     ReleaseMutex(m);
    }
    ----------------------------------------

    This way LogMessage() works like printf() and it is very flexible. You can use the Unicode variants if you prefer a Unicode build.

    And now all the building pieces are in place so, what should you log? That is, what should you do inside the routine I called HookGraphBuilder()?

    For example, you could hook IGraphBuilder::RenderFile() in order to dump the graph contents before and after the actual call to the original method. You can dump the graph yourself or save it to a GRF file:

    http://msdn.microsoft.com/library/en-us/directshow/htm/saveafiltergra...

    You can parse the GRF file afterwards if GE can't open it:

    http://msdn.microsoft.com/library/en-us/directshow/htm/directshowgrap...

    Here is my sample parser:

    http://groups-beta.google.com/group/microsoft.public.win32.programmer...

    You can hook IFilterGraph::AddFilter(), IGraphBuilder::AddSourceFilter() and IFilterGraph2::AddSourceFilterForMoniker() to log whether WMP adds filters on its own.

    You can also set your IAMFilterGraphCallback and IAMGraphBuilderCallback callbacks to be notified by the graph of errors in rendering the graph. If you do, call IObjectWithSite::GetSite() before IObjectWithSite::SetSite() to see whether WMP installed its own callbacks: if it did, cache the references and call them when your callbacks are called.

    You can also ask the graph to log what it does on its own with IGraphBuilder::SetLogFile().

    You can do all of this for a file that works and one that doesn't, to see what's different. You can do the same with WMP7+ and GE and WMP6.4... If you spy GE or DSPlay or other programs, in newCoCreateInstance() you should also accept CLSID_FilterGraph in addition to CLSID_FilterGraphNoThread().

    This was a long message to write :-)

    If you need support when you try this out, just ask on this same thread.

    --

    // Alessandro Angeli
    // MVP :: Digital Media
    // a dot angeli at psynet dot net

Geoff, you're being a bit cryptic :-) However, I had a
second look at what WMP9 does and it was *very* informative:
it uses an undocumented private "WMRenderer Source Filter"
which delivers uncompressed samples. So it behaves very
differently from both the global (and documented) WM source
filters used by GraphEdit, WMP6.4 and WMP7, which deliver
compressed samples to be decompressed by whatever decoding
filter there is (including, but not limited to, wrapped
VCM/ACM/DMO decoders).

"WMRenderer Source Filter" (*)
|
+-> "WMPlayer Time Compression DMO" (*) (LPCM)
|   |
|   +-> "WMPlayer Spectrum Analyzer DMO" (*)
|       |
|       +-> DirectSound Audio Renderer
|
+-> "WMPlayer Video Processing DMO" (*) (YV12)
    |
    +-> Video Mixing Renderer 7

(* = undocumented private filters/DMOs; their name is quoted
because it is the instance's name and not the class'
friendly name, which is unknown since they have no entry in
the registry)

Since the source filter is not the WMASFReader and the
decoding is done *inside* the source filter, it would be
very hard to hack it to make it use a general DirectShow
filter.

The WMP, WMF *and* DirectShow SDKs should clearly state that
WMP9+ does not use either the WMSourceFilter or the
WMASFReader but it uses an internal source filter that
behaves in a very different way, that is it performs the
decoding internally using WMF instead of externally,
deferring it to DirectShow like previous WMP versions and
other DirectShow-based applications do, and so it is limited
to installed VCM/ACM/DMO decoders and incapable of taking
advantage of DirectShow filters as one would expect.

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

Hi Allesandro,

At last I have found the time to realize the things you have described me.
I have some troubles (probably with my understanding).
I wrote my dll  (denote it my_library.dll) and was trying to trace by debugger (MS VC++ 6.0) DllEntryPoint of my DLL.

I set the following parameters  :

Executable for debug session:
E:\PROJECTS\Misk\detours\samples\bin\withdll.exe

Program arguments:
-d:E:\PROJECTS\Misk\my_library\Debug\my_library.dll C:\Program Files\Windows Media Player\wmplayer.exe

When I start debugging, withdll.exe runs wmplayer.exe but doesn't call my DllMain routine.
Moreover I have a look at withdll sources and haven't found an obvious calling of LoadLibrary routine...

Could you please explain me what is wrong?

Dmitry

I don't think anything is wrong.

withdll.exe uses Detours' DetourCreateProcessWithDll() to
create another process for wmplayer.exe, completely
independent, and quits immediately after that. The created
process is however altered to cause wmplayer.exe to load
your DLL, that's why withdll.exe never loads it: it is
loaded in the spawned wmplayer.exe process. That was the
whole point: run wmplayer.exe and spy on it by loading a DLL
in *its* process space.

I think you will have a hard time using a debugger to cross
process boundaries, hence the easier file logging technique
:-)

Anyway, if you are trying this out of curiosity, go ahead
(it's an interesting exercise in system hacking :-)),
otherwise take a look at my reply to Geoff to have the
solution to your original problem served to you.

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

I understood about my debugging faults.
As for your reply to Geoff, you write:

Was you look as you described (with detours lib)?

As far as I understand that is the answer for my initial question.
And it sounds as: only VCM/DMO video decoders are used by WMP to play ASF.
Isn't it?

If so, there is one interesting thing have happened during this thread evolution.
We have performed some tests on various PCs and found two PCs where our ASF contained H264 was normally played by WMP with DirectShow decoder...
One PC had WMP9, another WMP10.

This fact, your and Geoff's conclusions have confused all finally :)

Regards,
Dmitry Vergeles
www.solveigmm.com

Yes.

With WMP9+, yes, only WMM codecs and DMOs. With any other
DirectShow-based application, including WMP6.4 and WMP7
(IIRC, but I don't know and WMP8), a filter should be fine.
That's because WMP9+ does not use DirectShow to decode the
streams but only to render them.

That's "interesting" :-) Let's see if I can send you a small
tool that will tell us what filters are used on those 2 PCs.
Did you install something different on each PC or do all of
them have the same configuration, apart from the version of
WMP?

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

"Alessandro Angeli [MVP::DigitalMedia]" <nob...@nowhere.in.the.net> wrote in
message news:%23AVsWfQPFHA.2132@TK2MSFTNGP14.phx.gbl...

I'm not quite sure how that could be working, but I certainly can't disprove
what you've seen with your own two eyes. In any case, I encourage you to
work towards a DMO solution as that is the officially supported solution.

Geoff

--
This posting is provided "AS IS" with no warranties, and confers no rights.

"Alessandro Angeli [MVP::DigitalMedia]" <nob...@nowhere.in.the.net> wrote in
message news:unLMiyUOFHA.576@TK2MSFTNGP15.phx.gbl...
*snip*
*snip*

Sorry for being cryptic... it is not my intention. As you say below WMP uses
its own private filter for rendering ASF files, and this filter always uses
codec decoders _inside_ the WMF SDK. Which means only ACM/VCM/DMO are
supported.

Also, a small correction... WMP 7 also uses the private filter. The big
change was between WMP 6.4 and WMP 7. It is possible that WMP 7 and WMP 9
have different behaviors in the codec area, though nothing springs to mind
immediately.

*snip*

*snip*

I agree that the documentation could be more clear in this area; the best I
can advise is to ask questions in this forum.

Also to anyone reading this discussion, I wrote a fair amount of the code in
question, so in theory I know what I'm talking about. But I often find that
the MVPs like Alessandro have a better grasp on how things work in the real
world. :-)

Geoff

--
This posting is provided "AS IS" with no warranties, and confers no rights.

Hi Allesandro,

After my investigation with help of the spy dll you've sent me I have to
admit that I was wrong.
Both PCs had VFW decoders installed that we've noticed after the spy dll
logs reading.
I'm sorry for you spent your time because of my carelessnes :)

Regards,
Dmitry

Never mind, at least the DLL was useful to you and the
result is coherent with what Geoff told us :-)

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net

Of course you're right: WMP7 behaves like WMP9 and doesn't
use the WMASFReader but its own source filter.

--

// Alessandro Angeli
// MVP :: Digital Media
// a dot angeli at psynet dot net