Re: [Audacity-devel] Zoom Issue

[<Mar...@ao...>]

Yippee!
 
I finally found where the bug is (I think, 'proof' below), and (as  
suspected) it is an overflow of an int.  The problem is  with wxScrollBar; I don't know 
how to fix it, and need some ideas.
Let's see if I found it though:
 
The problem is with the horizontal scroll bar.  In  Project.cpp (about li 
1003) we set mViewInfo.sbarTotal which subsequently  gets used to set the range 
of a wxScrollBar, which is an int.
 
>From what I can figure out, mViewInfo.sbarTotal is the total number of  
pixels it would take to display the waveform, at this zoom level, if they were  
available (ie you had a very very large monitor, if you zoom in).  That is  
mViewInfo.total(s)*mViewInfo.zoom(px/s) = pixels.
 
However as we zoom in this overflows, causing the problem.
 
mViewInfo.zoom is limited to 6000000px/s, in AudacityProject::Zoom (that's  
about 136 pixels per sample at CD sampling rates, which seems adequate).   The 
overflow level of an int (here) is 2147483647 and so this give a maximum  
mViewInfo.total as 2147483647/6000000s = 357.91s before we get a problem at  
maximum zoom, pretty close to the (experimental) 357.95s that I reported the  
minimum length the error would occur at a couple of days ago.
 
mViewInfo.total is the length of the audio, in seconds; for you about 7481s  
(2h04m41s).  Assuming you are working a CD sampling rate of 44100 samples/s  
you should see the error occurring when you zoom in beyond about 6.5s  of 
visible audio.  Can you confirm this?
 
TTFN
Martyn
PS I tried limiting sbarTotal to the max int value but then we can't drag  
the scrollbar to the end of the waveform, so that's not a very good idea,  
although it stops the jumping.  It would be better for us if wx supported  long 
long for this parameter.
 
In a message dated 29/03/2007 16:12:14 GMT Daylight Time, ud...@xs...  
writes:
Mar...@ao... wrote:
> I am rather ignorant with regards  anything other than VS8 on  Audacity.  
Are 
> you compiling the  code yourself?  

I did in the past (to fix other issues or get  features early). Current
build is a Fedora build. I am on Linux.  FC6.

> If so, compiling with  Unicode and Debug set 
> will  probably give you the assert message (not that it's  there any  
more!).

I could give that a try. If that happens, what then?
I can  read the source to a certain degree, I can change stuff but I do
not have  enough overview to know what I am looking for or doing.

>   
> - 'at 3307500 samples it crashes' type reports  are much more  useful than 
> 'about 1-2 hours'.  I hope this does not   offend.

I mean to say that the files are that long. Current file is  2h04m41s+35
cdda frames.
If I zoom in around the 2h04m mark I can easily  reproduce the issue.
Also near the beginning of the file or halfway it  happens as far as I
remember.


Kind regards,
Udo