linuxsampler-devel

[Linuxsampler-devel] Bug report and possible bug fix for disk streaming

[Andreas P. <and...@lk...>]

Hello,

I've found a bug in the streaming-from-disk code. I use a test gig with 
a short, mono, non-looped wave. The wave is 120976 samples long, and the 
problem is that not all of it are played when I press and hold a key. 
The disk thread reads the complete wave, but the audio thread does not 
render all samples - it seems as it stops after the cached 32768 samples.

The attached patch (one line) seems to fix the problem - but, I 
definetly don't have a full understanding of the Stream class and the 
interaction between the threads, so it would be nice if someone that 
has, could look at it.

/Andreas

Re: [Linuxsampler-devel] Bug report and possible bug fix for disk streaming

[Robert J. <rj...@sp...>]

Ah, great!

I reported an issue about clicking at the end of samples a few months back=
=20
(sadly I never got around to sending examples here), this sounds like it=20
could be the same. Looking forward to test it, though that will be a while=
=20
again...

/Robert


s=F6ndagen den 5 december 2004 11.37 skrev Andreas Persson:
> Hello,
>
> I've found a bug in the streaming-from-disk code. I use a test gig with
> a short, mono, non-looped wave. The wave is 120976 samples long, and the
> problem is that not all of it are played when I press and hold a key.
> The disk thread reads the complete wave, but the audio thread does not
> render all samples - it seems as it stops after the cached 32768 samples.
>
> The attached patch (one line) seems to fix the problem - but, I
> definetly don't have a full understanding of the Stream class and the
> interaction between the threads, so it would be nice if someone that
> has, could look at it.
>
> /Andreas

=2D-=20
http://spamatica.se/music/

Re: [Linuxsampler-devel] Bug report and possible bug fix for disk streaming

[Christian S. <sch...@so...>]

Es geschah am Sonntag 05 Dezember 2004 11:37 als Andreas Persson schrieb:
> Hello,
>
> I've found a bug in the streaming-from-disk code. I use a test gig with
> a short, mono, non-looped wave. The wave is 120976 samples long, and the
> problem is that not all of it are played when I press and hold a key.
> The disk thread reads the complete wave, but the audio thread does not
> render all samples - it seems as it stops after the cached 32768 samples.
>
> The attached patch (one line) seems to fix the problem - but, I
> definetly don't have a full understanding of the Stream class and the
> interaction between the threads, so it would be nice if someone that
> has, could look at it.


Index: linuxsampler/src/engines/gig/Stream.h
===================================================================
RCS file: /var/cvs/linuxsampler/linuxsampler/src/engines/gig/Stream.h,v
retrieving revision 1.2
diff -u -2 -r1.2 Stream.h
--- linuxsampler/src/engines/gig/Stream.h 27 Apr 2004 09:21:58 -0000 1.2
+++ linuxsampler/src/engines/gig/Stream.h 5 Dec 2004 10:28:18 -0000
@@ -59,5 +59,5 @@
 
             inline int GetReadSpace() {
-                return (pRingBuffer && State == state_active) ? 
pRingBuffer->read_space()  : 0;
+                return (pRingBuffer && State != state_unused) ? 
pRingBuffer->read_space()  : 0;
             }

Ouch, yes, that's definitely a bug. We must have forgotten that when the 
meaning of the stream states changed. This bug caused that already read data 
of the end of samples were overwritten by silence. The code to write the 
silence at the end (in Voice.cpp) should also be modified so that only as 
much silence is written as necessary. I will fix that.

Thanks for the report!

CU
Christian