WINMM problem with Half-Life

[Mike McCormack]

Hi Eric,

After implementing the ICCVID codec, the intro AVI in Half-Life plays 
nicely, however when it finishes, WinMM gets stuck in a loop while 
cleaning up in MCIAVI_mciPlay.

I think it's trying to wait for the sound to finish, however I'm not 
sure how to fix the problem.  The patch below makes it work, but I'm not 
sure what bad side effects it will have...

Any ideas?

Mike

Index: dlls/winmm/mciavi/mciavi.c
===================================================================
RCS file: /cvstrees/crossover/office/wine/dlls/winmm/mciavi/mciavi.c,v
retrieving revision 1.1.1.10
diff -u -r1.1.1.10 mciavi.c
--- dlls/winmm/mciavi/mciavi.c	10 Jan 2004 00:11:58 -0000	1.1.1.10
+++ dlls/winmm/mciavi/mciavi.c	20 Jan 2004 16:22:23 -0000
@@ -503,12 +503,14 @@
     }
 
     if (wma->lpWaveFormat) {
+#if 0
        while (wma->dwEventCount != nHdr - 1)
         {
             LeaveCriticalSection(&wma->cs);
             Sleep(100);
             EnterCriticalSection(&wma->cs);
         }
+#endif
 
 	/* just to get rid of some race conditions between play, stop and pause */
 	waveOutReset(wma->hWave);

[Eric Pouech]

Mike McCormack a écrit :
> Hi Eric,
>
> After implementing the ICCVID codec, the intro AVI in Half-Life plays 
> nicely, however when it finishes, WinMM gets stuck in a loop while 
> cleaning up in MCIAVI_mciPlay.
>
> I think it's trying to wait for the sound to finish, however I'm not 
> sure how to fix the problem.  The patch below makes it work, but I'm not 
> sure what bad side effects it will have...
>
> Any ideas?
> as an ugly hack, does setting the dwEventCount in the WINE_MCIAVI struct 
> as volatile help ? (private_mciavi.h).

A+

[Mike McCormack]

Eric Pouech wrote:
> > Any ideas?
>
> as an ugly hack, does setting the dwEventCount in the WINE_MCIAVI struct 
> as volatile help ? (private_mciavi.h).

No, that didn't help, the one below does:

Mike


diff -u -r1.1.1.10 mciavi.c
--- dlls/winmm/mciavi/mciavi.c  10 Jan 2004 00:11:58 -0000      1.1.1.10
+++ dlls/winmm/mciavi/mciavi.c  21 Jan 2004 02:51:15 -0000
@@ -503,7 +503,7 @@
      }

      if (wma->lpWaveFormat) {
-       while (wma->dwEventCount != nHdr - 1)
+       while (wma->dwEventCount < (nHdr - 1) )
          {
              LeaveCriticalSection(&wma->cs);
              Sleep(100);

[Mike McCormack]

Mike McCormack wrote:
> Eric Pouech wrote:
> > > Any ideas?
> >
> > as an ugly hack, does setting the dwEventCount in the WINE_MCIAVI 
> > struct as volatile help ? (private_mciavi.h).
>
> No, that didn't help, the one below does:

Well, it sort of does... but with that patch, sometimes westuck in the 
critical section.  I guess the question is why does dwEventCount count 
further than expected?

Mike

[Eric Pouech]

> Well, it sort of does... but with that patch, sometimes westuck in the 
> critical section.  I guess the question is why does dwEventCount count 
> further than expected?
> yup. can you post a -debugmsg +mciavi trace
> A+

[Mike McCormack]

Here you go.

wine --debugmsg +mciavi,+winmm,+seh hl.exe >& hl-winmm.log

Pressed ^C after it hung.

Mike

Eric Pouech wrote:
> yup. can you post a -debugmsg +mciavi trace
> A+

[Eric Pouech]

> > yup. can you post a -debugmsg +mciavi trace
> > A+
>
> does this fix it ?
> A+

Index: dlls/winmm/mciavi/mmoutput.c
===================================================================
RCS file: /home/cvs/cvsroot/wine/wine/dlls/winmm/mciavi/mmoutput.c,v
retrieving revision 1.11
diff -u -r1.11 mmoutput.c
--- dlls/winmm/mciavi/mmoutput.c	5 Jan 2004 23:07:27 -0000	1.11
+++ dlls/winmm/mciavi/mmoutput.c	23 Jan 2004 20:38:26 -0000
@@ -563,7 +563,10 @@
 	ResetEvent(wma->hEvent);
 	if (InterlockedDecrement(&wma->dwEventCount) < 0 ||
 	    !wma->lpAudioIndex[wma->dwCurrAudioBlock].dwOffset)
+        {
+            InterlockedIncrement(&wma->dwEventCount);
 	    break;
+        }
 
 	mmioSeek(wma->hFile, wma->lpAudioIndex[wma->dwCurrAudioBlock].dwOffset, SEEK_SET);
 	mmioRead(wma->hFile, waveHdr[whidx].lpData, wma->lpAudioIndex[wma->dwCurrAudioBlock].dwSize);
@@ -573,7 +576,6 @@
 	waveOutWrite(wma->hWave, &waveHdr[whidx], sizeof(WAVEHDR));
 	wma->dwCurrAudioBlock++;
     }
-    InterlockedIncrement(&wma->dwEventCount);
 }
 
 LRESULT MCIAVI_PaintFrame(WINE_MCIAVI* wma, HDC hDC)

[Mike McCormack]

Hi Eric,

Yes, I have confirmed that this patch fixes the problem.

Thanks!

Mike

Eric Pouech wrote:
> does this fix it ?
> A+