EVALUATION

According to Alex we've lost our Beatnik audo engine experience which is unfortunate. I managed to narrow down the responsible changes between 142 & 150 outlined at:

http://jpsesvr.sfbay.sun.com:8080/ctetools/html/ViewDetail.jsp?index=2404
                                     

EVALUATION

The crash that occurs can be reproduced with a slightly modified version of the testcase that accepts 4 args: (bank, program, channel, note)

For several combinations of those args the test crashes though I have not been able to find a pattern.

The crash occurs in the following snippet of code: (I've marked line 835 which is what the debugger pinpoints as the last line in the stacktrace, see below)

for (a = MusicGlobals->Four_Loop; a > 0; --a)
           {
               zIndex1 = zIndex2 - (this_voice->previous_zFrequency >> 8);
               this_voice->previous_zFrequency += (this_voice->LPF_frequency - this_voice->previous_zFrequency) >> 3;
               amplitudeReverb = ((amplitudeL + amplitudeR) * this_voice->reverbLevel) >> 8;
               amplitudeChorus = ((amplitudeL + amplitudeR) * this_voice->chorusLevel) >> 8;

               for (inner = 0; inner < 4; inner++)
               {
                   b = source[cur_wave>>STEP_BIT_RANGE];
                   c = source[(cur_wave>>STEP_BIT_RANGE)+1];
                   sample = ((((INT32) (cur_wave & STEP_FULL_RANGE) * (c-b))>>STEP_BIT_RANGE) + b - 0x80) << 2;
                   sample = (sample * Xn + Z1value * Z1 + z[zIndex1 & MAXRESONANCE] * Zn) >> 16;
                   zIndex1++;
                   z[zIndex2 & MAXRESONANCE] = (INT16)sample;
                   zIndex2++;
                   Z1value = sample - (sample >> 9);
                   *destL += sample * amplitudeL;
                   destL[1] += sample * amplitudeR;
                   destL += 2;
                   *destReverb += sample * amplitudeReverb;
                   destReverb++;
                   *destChorus++ += sample * amplitudeChorus;
835-->                                     cur_wave += wave_increment;
               }
               amplitudeL += amplitudeLincrement;
               amplitudeR += amplitudeRincrement;
           }


Here's the stack trace:

00000000`05f1f520 00000000`059a8c9e jsound!PV_ServeStereoInterp2FilterFullBufferNewReverb(struct GM_Voice * this_voice = 0x00000000`00000000, void * threadContext = 0x00000000`00000000)+0x64a [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\genfiltersreverb.c @ 835]
00000000`05f1f5f0 00000000`059a6d67 jsound!PV_ServeStereoInterp2FilterFullBuffer(struct GM_Voice * this_voice = 0x00000000`0015f6a0, void * threadContext = 0x00000000`00000000)+0x70e [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\gensynthfilters.c @ 659]
00000000`05f1f6a0 00000000`059a7158 jsound!PV_ServeThisInstrument(struct GM_Voice * pVoice = 0x00000000`00000000, void * threadContext = 0x00000000`00000000)+0x5e7 [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\gensynth.c @ 1242]
00000000`05f1f710 00000000`059a738d jsound!PV_ServeStereoInstruments(void * threadContext = 0x00000000`00000006)+0xb8 [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\gensynth.c @ 1554]
00000000`05f1f760 00000000`059a74ab jsound!PV_ProcessSampleFrame(void * threadContext = 0x00000000`00000001, void * destinationSamples = 0x00000000`001b0480)+0x19d [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\gensynth.c @ 1910]
00000000`05f1f7a0 00000000`059b4b85 jsound!HAE_BuildMixerSlice(void * threadContext = 0x00000000`00000006, void * pAudioBuffer = 0x00000000`059c2160, int bufferByteLength = 5, int sampleFrames = 1758128)+0x7b [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\engine\gensynth.c @ 1643]
00000000`05f1f800 00000000`0599126c jsound!PV_AudioWaveOutFrameThread(void * threadContext = 0x00000000`01d2a060)+0x175 [e:\1.4.2\j2se\src\windows\native\com\sun\media\sound\engine\hae_api_winos.c @ 852]
00000000`05f1f9a0 00000000`01d3ccf7 jsound!Java_com_sun_media_sound_MixerThread_runNative(struct JNINativeInterface_ ** e = 0x00000000`00000000, struct _jobject * thisObj = 0x00000000`00000000, int64 frameProc = 0)+0xc [e:\1.4.2\j2se\src\share\native\com\sun\media\sound\mixerthread.c @ 61]
                                     

EVALUATION

Changing this to p3. It is one of several sound (midi in fact) issues that have cropped up as a result of amd64 support. I don't feel this needs to be a p1 since midi on x64 is pretty fragged and has been from the start. Working with QA to gfind the root cause of this one.
                                     

EVALUATION

I've cast a glance at the issue
The issue is not related with cr 6498661 (the issue affects sampled recording only, midi classes don't use it)
I've checked similar routine for playback, there is no such problem there.

Further research requires much time. As far as I understand the reason of the crash is in MixerSequencer or in MixerSynth native implementation

Regards
Alex