Re: [Audacity-devel] 24 bit bug

[Markus M. <me...@me...>]

Am Son, 2003-05-25 um 02.40 schrieb Tino Meinen:
> This might be just me being stupid:
> 
> 1. Start up audacity
> 2. Generate silence
> 3. Set sample format to 24 bit
> 4. Set sample format to 32 bit
> 
> What you get is white noise with the samples randomly being assinged to
> 0 and 1.

In SimpleBlockFile.cpp, line 260, it reads

for( int i = 0; i < framesRead; i++ )
   intPtr[i] = (unsigned int)intPtr[i];

This should be

for( int i = 0; i < framesRead; i++ )
   if (intPtr[i] < 0)
      intPtr[i] = (int)(((unsigned int)intPtr[i] >> 8) | 0xFF000000);
   else
      intPtr[i] = (unsigned int)intPtr[i] >> 8;

!!! BUT ONLY FOR X86 MACHINES (I suppose).

I can't test how this should be for machines with other byte orders.

A simple and platform independent fix would be:

for( int i = 0; i < framesRead; i++ )
   intPtr[i] = intPtr[i] / 256;

This is optimized to a shift operation anyway (at least on MS Visual
C++).


Markus