[mythtv] [mythtv-commits] Ticket #2664: Backend SEGFAULT on invalid PMT

[Janne Grunau]

On Monday 13 November 2006 16:32, Daniel Kristjansson wrote:
> On Mon, 2006-11-13 at 14:37 +0100, Janne Grunau wrote:
> >
> > This is unfortunately not true. The packet assembler has bugs left.
> > Several people (including me) still see bogus EIT packets though at
> > a much lower rate.
>
> Hmmm, that sucks. But I don't think it is related to this
> problem, PAT's and PMT's are usually in one or two TS packets
> which I think we handle ok.

I'm not sure. The errors were the same as in #2147 and [11072] fixed 
them.

> > Starting end of october/begin of november Stuart Morgen and I
> > started to see more pes packet assembler warnings ("Started
> > pespaket
> > but !payloadstart",...). Strange thing is that I haven't found a
> > related change set since 0.20 release.
>
> I don't think anything has changed in MythTV,

There is just one innocent looking change in the related classes 
(DVBSignalMonitor/Recorder. MPEGStreamData).

> but it's possible 
> something changed in the transmission or DVB drivers to give us
> packets that we don't handle correctly.

At least a possibility since I updated DVB drivers in that time frame. 
Changed transmissions are more unlikely since a simultaneous chande in 
th UK and Germany are unlikely (though I don't know if the use the same 
equipment). 

> > I'm not sure if our handling of DVB SI packets is part of the
> > problem. We handle them currently in the assembler as PES packets
> > which they clearly aren't. They have for example no HAS_CRC flag
> > since they have all crc.
>
> I plan to address this when I fix #799, fixing both will require
> me to study the MPEG standards with respect to PES packets to fix
> intelligently (i.e. both efficiently and correctly). But for now
> couldn't we just make the HasCRC() method virtual and have it
> always return true for SI sections?

I will try that. Some more work is probably necessary since the CRC is 
afaik done on pes packet level.

Janne