Re: [libdc] Corrupted frames under OS X - libdc1394 2.0.1

[Mark <mar...@go...>]

On 28.02.2008, at 03:00, David Moore wrote:

> On Wed, 2008-02-27 at 15:39 -0500, David Moore wrote:
>> It is
>> not possible for the image data to be corrupted on the bus, since  
>> there
>> are checksums to verify the integrity of each packet.
>>
>
> Ah, I am wrong there!
>
> I went back and looked at the OHCI specs, and in fact the CRC is
> checked, but the packet data is passed to us even if it has been
> detected to be corrupt.  We are responsible for looking at the error
> codes for _each_ packet and making sure they are valid.
>
> For obscure reasons, this only applies to Mac OS.  The technique Linux
> uses for capturing packets _does_ automatically discard them if they  
> are
> corrupt.  That's why I hadn't seen this before.
>
> So in fact, these corruptions you are seeing are probably a result of
> bus traffic, just like the skipped packets.  I can weed these out by
> checking the error codes for each packet, which I am currently  
> ignoring.
>
> I will whip up some code for this and check it into SVN hopefully  
> later
> tonight.  This change will only apply to Mac OS.

Are bad packets automatically re-transmitted or does that mean that  
from now on a frame with bad packets will be just dropped? If the  
later is the case, it would be great to know that somehow - maybe  
capture_dequeue could just deliver the bad frame with a return code of  
DC1394_BAD_FRAME instead of DC1394_SUCCESS.

Mark