[vorbis-dev] Re: Comments on your code

[Monty]

> On Fri, Jan 26, 2001 at 05:49:08PM -0800, Monty wrote:
> > Do you still have this problem using decoder_example,
> > vorbisfile_example or XMMS?
> 
> decoder_example shows the same problem, after playing around for
> a while, I'm convinced that the problem is in libvorbis somewhere.

This would indeed place the problem in libvorbis.

> As you can see, the dump is essentially the same as when it
> crashed in ogg123 and I went on to `make debug' on both the libvorbis
> library and the examples -- the crash still occurs but the malloc
> that causes it looks honest enough, this means that something is
> writing into blocks that it shouldn't be and is screwing up the memory
> allocation lists.

Agreed.  Have you tried it with Electric Fence? 

> There is a noticable pattern to the way it crashes.
> 
> I guess I should learn how to use electric-fence but not this minute.

Oh, answers my question.  Actually Electric Fence is as easy as
-lefence in the link line.  There are three possible environment
variables to set to change the exact nature of the testing; run with
one of EF_PROTECT_ABOVE, EF_PROTECT_BELOW or EF_PROTECT_FREE set to 1.

> Seems like ogg123 is blameless, I'm betting on something that writes
> beyond malloced space somewhere in libvorbis. I'm stumped why it should
> hit the Alpha and not the i386. I have another, newer Alpha that I
> can try this on and see if the same thing happens (could be somehow
> dependent on the system libraries).

Is this in only a specific file or in all of them?  

BTW, I recently tested the beta 4 code on a 21164 based box and all
was well.  Is this your first Ogg install?  Be very careful that the
Ogg and Vorbis libraries on the system match; we've had header changes
and mixing library builds will produce this behavior.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.