Andrew de Quincey wrote: > On Monday 27 October 2003 22:46, Emard wrote: > > Soma idea about CRC and network problems associated with it; > > > > can we verify that the crc calculation arithmetic is carried > > out properly (i.e. take a look at another's author code doing the > > same thing). > > > > Typical error in CRC's is improper usage of signed and unsigned > > arithmetic, shifts with and without sign or bit rotations with carry > > and similar stuff regarding boundary conditions. > > > > From the experience I have seen, if crc is enabled, some > > packets that should have passed are rejected. > > But lots do work with CRC enabled.... sounds like a boundary condition.
Maybe the CRC check fails because some TS packets have errors? We could add a check for the transport_error_indicator to dvb_dmx_swfilter_section_packet(). I'm not a network specialist and could be wrong here, but IIRC ethernet has a CRC for the full packet, but IP only has a header checksum. So without CRC you could silently get wrong data. Sometimes wrong data is better then no data, so maybe we should add a flag to NET_ADD_IF which allows one to specify if the CRC check is wanted. (Not that this would solve any real problems, but it might be nice for testing.) Johannes -- Info: To unsubscribe send a mail to [EMAIL PROTECTED] with "unsubscribe linux-dvb" as subject.
