Hi guys,
These changes were made by me in December. The main goal was to move
memory allocation functions from RT context.
Frankly speaking, I forgot to make different tests with random buffer
sizes and sample rates and did not informe you about these changes.
So all recent recording (and may be playing?)wave bugs are from me :(.
26.01.2016 21:43, Tim E. Real пишет:
> On January 26, 2016 04:34:20 PM Robert Jonsson wrote:
>> Hey guys,
>>
>> Just a heads up while I remember it.
>>
>> Has anybody been trying to record something with current git?
>> I got a new microphone, the type you connect directly to usb, not the best
>> but very handy.
>>
>> I got clicking in the recordings though and was blaming the mic, the
>> drivers and jack but finally it dawned on me that it seems to be MusE
>> adding clicks.
>> I need to test more and try to narrow down when this started occuring, a
>> qualified guess would be the new and shiny wave drawing.
>>
>> A test that you can perform if you'd like is to setup one Input and connect
>> it to two Wave tracks and start recording to both. After recording a minute
>> or two, compare the two files.
>> Easiest way I found to do it is:
>> xxd file1 out1.hex
>> xxd file2 out2.hex
>> diff out1.hex out2.hex
>>
>> If they are identical diff will stay silent.
>> I did one recording where they are identical, but then a longer one with
>> three tracks where the files had differences.
>> I will test this more.
>>
>> Regards,
>> Robert
> As it happens I am in the wave section right now.
>
> I am surprised it hasn't crashed on you.
> I have (locally) fixed several problems and crashes.
>
> During testing, one of the first things I noticed was after
> using the wave editor 'functions' to edit a wave, there were
> 'spikes' left in the edited wave that weren't in the original.
> And it would occasionally crash.
>
> The following was changed for two reasons:
> 'n' may not be evenly divisible by writeSegSize.
> And size_t is an unsigned. So 'n' was 'underflowing' but
> actually 'rolling over' so that "while(n > 0)" was failing.
> This would either cause a crash or cause invalid or incomplete
> data to be written ('spikes').
>
> size_t SndFile::write(int srcChannels, float** src, size_t n)
> {
> size_t wrFrames = 0;
>
> if(n <= writeSegSize)
> wrFrames = realWrite(srcChannels, src, n);
> else
> {
> while(n > 0)
> {
> size_t nrWrote = realWrite(srcChannels, src, writeSegSize,
> wrFrames);
> wrFrames += nrWrote;
> n -= nrWrote;
> }
> }
> return wrFrames;
> }
>
> Changed to:
>
> size_t SndFile::write(int srcChannels, float** src, size_t n)
> {
> size_t wrFrames = 0;
>
> if(n <= writeSegSize)
> wrFrames = realWrite(srcChannels, src, n);
> else
> {
> while(1)
> {
> size_t sz = (n - wrFrames) < writeSegSize ? (n - wrFrames) :
> writeSegSize;
> size_t nrWrote = realWrite(srcChannels, src, sz, wrFrames);
> if(nrWrote == 0) // Nothing written?
> break;
> wrFrames += nrWrote;
> if(wrFrames >= n)
> break;
> }
> }
> return wrFrames;
> }
>
>
> Another thing I found was this:
>
> void SndFile::setFormat(int fmt, int ch, int rate)
> {
> ...
> writeBuffer = new float [writeSegSize * std::max(2, ch)];
> }
>
> The writeBuffer allocation was moved out of there and into
> SndFile::openWrite() because there was a crash with some operations
> because writeBuffer was not allocated because setFormat() had not
> been called beforehand.
>
>
> Since these functions are responsible for recording,
> these may be part of your problem.
>
> But also being a USB mic there may be problems there.
> I've seen it recommended to use 3 Jack periods with USB,
> not sure why, but LAD and LAU always have discussions
> about USB. Maybe ask there if my fixes don't solve it.
>
>
> /Lots/ of wave fixes here, will try to commit soon.
>
> Tim.
>
>
>
>
> ------------------------------------------------------------------------------
> Site24x7 APM Insight: Get Deep Visibility into Application Performance
> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
> Monitor end-to-end web transactions and take corrective actions now
> Troubleshoot faster and improve end-user experience. Signup Now!
> http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140
> _______________________________________________
> Lmuse-developer mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/lmuse-developer
--
Regards,
Andrew
------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140
_______________________________________________
Lmuse-developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/lmuse-developer
