[patch] EIT decoder problem

[Johann Hanne]

Hi,

there seems to be a problem with the EIT decoder.

The current SVN version contains the following code in src/tables/eit.c:
--
    /* Check if we have all the sections */
    b_complete = 0;
    for(i = 0; i <= p_eit_decoder->i_last_section_number; i++)
    {
      if(!p_eit_decoder->ap_sections[i])
        break;

      if(i == p_eit_decoder->i_last_section_number)
        b_complete = 1;
    }
--
I.e. if it has not received *every* section_number from 0 to 
last_section_number, it will discard everything. This seems to be the correct 
thing for every PSI table but the EIT.  ETSI EN 300 468 V1.5.1 (2003-01) says 
the following for the section_number of an EIT:
--
section_number: This 8-bit field gives the number of the section. The 
section_number of the first section in the sub_table shall be "0x00". The 
section_number shall be incremented by 1 with each additional section with 
the same table_id, service_id, transport_stream_id, and original_network_id. 
In this case, the sub_table may be structured as a number of segments. Within 
each segment the section_number shall increment by 1 with each additional 
section, but a gap in numbering is permitted between the last section of a 
segment and the first section of the adjacent segment.
--
The point seems to be "a gap in numbering is permitted". To be honest, I don't 
understand the whole paragraph: As far as I understand, PSI tables are put 
into sections, and sections are split into TS packets. Sections have a 
maximum of 1024/4096 bytes and thus a PSI table which is too large may be put 
into more than one section (this is what section_number and 
last_section_number is all about). However, I don't understand what "segment" 
means and specifically not "section of a segment". Can somebody give me a 
hint, please?

Whatsoever, I've added the following debug code to src/dvbpsi.c:
--
         if(p_section->b_syntax_indicator)
          {
            p_section->i_extension =   (p_section->p_data[3] << 8)
                                     | p_section->p_data[4];
            p_section->i_version = (p_section->p_data[5] & 0x3e) >> 1;
            p_section->b_current_next = p_section->p_data[5] & 0x1;
            p_section->i_number = p_section->p_data[6];
            p_section->i_last_number = p_section->p_data[7];
            p_section->p_payload_start = p_section->p_data + 8;

            /* new code >>> */
            if (((((uint16_t)(p_data[1] & 0x1f) << 8) | p_data[2]) == 0x12) &&
                p_section->i_table_id == 0x50 &&
                p_section->i_extension == 28106) {
              printf("i_table_id=0x%02x, i_extension=%d, i_number=%d, 
i_last_number=%d\n",
                     p_section->i_table_id,  p_section->i_extension,
                     p_section->i_number, p_section->i_last_number);
            }
/* <<< new code */
--
0x50 is the subtable for the scheduled events of the current TS, 28106 is the 
service id of a german TV station (ARD/Das Erste). When I feed the EIT PID 
0x12 to libdvbpsi now, this gives the following output:
--
...
i_table_id=0x50, i_extension=28106, i_number=56, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=64, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=72, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=80, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=88, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=96, i_last_number=248
i_table_id=0x50, i_extension=28106, i_number=104, i_last_number=248
...
--
So there are indeed gaps in section_number. The attached patch modifies 
src/tables/eit.c so that accepts the gaps. It's not perfect, but it works.

Comments, please?

Cheers,
  Johann
-------------- next part --------------
A non-text attachment was scrubbed...
Name: libdvbpsi-eit.patch
Type: text/x-diff
Size: 1412 bytes
Desc: not available
Url : http://mailman.videolan.org/pipermail/libdvbpsi-devel/attachments/20070407/e4b9f8a8/attachment.patch