Hello again everyone.Would those of you affected by bug 900364 (https://bugs.launchpad.net/mixxx/+bug/900364) regarding corrupt audio when using the microphone on Linux please compile PortAudio with the attached patch and see if it corrects the problem? (Patch is attached to the bug report as well.)
Sincerely, Sean M. Pappalardo "D.J. Pegasus" Mixxx Developer - Controller Specialist
--- Begin Message ---On Sunday 13 October 2013 23:31, RJ Ryan wrote: > The cause was indeed identified. The line that Clemens pointed out from the > ALSA hostapi in PortAudio fully explains the observations we've seen with > this bug over the years. (for example, the screenshot you just sent out is > exactly as expected given that line of code and a device with a minimum of > 2 input channels). Well I class that more as useful speculation. Clemens was probably not aware that Portaudio is coded with the intention of handling the cases where (hostChans != userChans), so finding they can differ is not a big surprise. However, it was clear from my test that in some cases it is not being done correctly. ....read on.... On another machine with ice1712 card, where hw:0 can only be opened for capture with 12 channels, it is fine to record with any number of channels from 1 to 12 and Portaudio does adjust fine, as intended. I have tracked the issue down, I believe, to inside the buffer processing. When the formats do not differ (and there is no de-interleaving to do) the input conversion is skipped, and this thwarts the mechanism for input channel adaption, that uses the input data stride, since data is not then transferred. When the formats do differ, correct channel adaption occurs. It is possible that Alsa hw devices are the only users of this adaption mechanism, even though it is in the common code. This can be remedied by adding a check for matching channel numbers when deciding whether to skip the conversion. Ross, Phil, can you check the reasonableness of this approach, since I am not very familiar with the buffer processing etc? The attached patch to pa_process.c resolves the issues identified in my earlier tests. Perhaps others would like to test if it resolves all the cases of bad mono recording, or if there is still another issue. Also bear in mind it may not be the final solution. "pa_process-chans-no-skip.diff" Regards Alan--- svn-working-copy/src/common/pa_process.c 2013-08-13 21:35:27.000000000 +0100 +++ lib-src/portaudio-v19/src/common/pa_process.c 2013-10-16 08:46:44.000000000 +0100 @@ -743,8 +744,10 @@ destSampleStrideSamples = bp->inputChannelCount; destChannelStrideBytes = bp->bytesPerUserInputSample; - /* process host buffer directly, or use temp buffer if formats differ or host buffer non-interleaved */ - if( bp->userInputSampleFormatIsEqualToHost && bp->hostInputIsInterleaved && bp->hostInputChannels[0][0].data) + /* process host buffer directly, or use temp buffer if formats differ or host buffer non-interleaved, + * or if the number of channels differs between the host (set in stride) and the user */ + if( bp->userInputSampleFormatIsEqualToHost && bp->hostInputIsInterleaved + && bp->hostInputChannels[0][0].data && bp->inputChannelCount == hostInputChannels[0].stride) { userInput = hostInputChannels[0].data; destBytePtr = (unsigned char *)hostInputChannels[0].data;_______________________________________________ Portaudio mailing list portau...@music.columbia.edu http://music.columbia.edu/mailman/listinfo/portaudio
--- End Message ---
------------------------------------------------------------------------------ November Webinars for C, C++, Fortran Developers Accelerate application performance with scalable programming models. Explore techniques for threading, error checking, porting, and tuning. Get the most from the latest Intel processors and coprocessors. See abstracts and register http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk
_______________________________________________ Get Mixxx, the #1 Free MP3 DJ Mixing software Today http://mixxx.org Mixxx-devel mailing list Mixxx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mixxx-devel
