The transport stream loop started 2 bytes too far and the while() condition was
pointing to 2 bytes too soon. The code has been corrected to first add 2 bytes
to the current position, thereby skipping the length fields (which have already
been read). Now the loop starts at the correct position.

The member i_crc is not used anymore now the CRC32 can be checked and
generated by the dvbpsi_BuildPSISection() and dvbpsi_ValidPSISection()
functions.

From dvbpsi_packet_push() we can learn that for the TOT table,
with syntax_indicator '0' the p_section->p_payload_start pointer
points to UTC Time field (3 bytes after PSI section start).

To get the byte position of the descriptor loop one needs to add
2 bytes instead of 5, as was previously done.

If one needs to calculate the CRC32 outside of libdvbpsi, then they
should use the new dvbpsi_CalculateCRC32() function instead.

Rename AM_CONFIG_HEADER to AC_CONFIG_HEADERS

In the past the b_syntax_indicator field could be used to distinguish between
tables with a CRC32 and those without. However these days this assumption is no
longer valid and the check should be modified to include tables that have a CRC32
even when syntax_indcator is '0'.

CAT: The p_section->p_payload_start pointer already points to the fist descriptor for this PSI section.

According to ITU-T Rec H.222.0 and the code in dvbpsi_packet_push() the
p_section->p_payload_start pointer already points to the first descriptor for
this PSI table. Adding 5 bytes is plain wrong.

- document first argument of dvbpsi_bat_ts_descriptor_add()

The function dvbpsi_decoder_psi_sections_completed() is meant to detect if all
sections of a multi section psi table have arrived. After the rewrite in
commit-id 57c6cd63 it ignored gaps in the arrived sections. This means that
the function now expects that individual sections of multi section psi tables
arrive in order. However the ISO/IEC 13818-1 standard explains in Annex C, C 2ii
that it is allowed to transmit sections out of order.

 "The section_number field allows the sections of a particular table to be reassembled
  in their original order by the decoder. There is no obligation within this Recommendation
  | International Standard that sections must be transmitted in numerical order, but this
  is recommended, unless it is desired to transmit some sections of the table more frequently
  than others, e.g. due to random access considerations."

The p_decoder->p_sections linked list is sequentially ordered and is now changed to detect
a gap in the ordering. This is possible because the first section in a multi-section psi table
is numbered 0.

change 0x8C & 0x9C to 0xC8 & 0xC9
Signed-off-by: Michael Krufky <mkrufky@linuxtv.org>
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

The SDT field known as i_ts_id is named i_extension in the struct dvbpsi_psi_section_t.

Fixes in documentation for:
 src/descriptors/dr_4b.h   document dvbpsi_nvod_ref_s
 src/descriptors/dr_76.h   document CRID_* defines
 src/tables/atsc_eit.h     document all arguments of dvbpsi_atsc_NewEIT()
 src/tables/atsc_ett.h     document all arguments of dvbpsi_atsc_NewETT()
 src/tables/atsc_stt.h     document missing arguments of dvbpsi_atsc_DetachSTT()
 src/tables/atsc_vct.h     document fix for functions dvbpsi_atsc_InitVCT() and dvbpsi_atsc_NewVCT()
 src/tables/bat.h          document arguments of dvbpsi_bat_ts_add()
 src/table/eit.h           document arguments of dvbpsi_eit_init()
 src/tables/sdt.h          remove unused argument from function dvbpsi_sdt_init()
 src/tables/tot.h          remove unused argument from function dvbpsi_tot_init()

Some tables like EIT and ATSC EIT need to know what the value of table_id and
i_extension fields are for this PSI section. Therefor add this information to
the tables decoded structs.

The param struct was used after it was freed.

The p_prev->p_next was never set to NULL, but instead the p_pmt->p_next pointer was.
This resulted in never freeing additional PMT entries from the PAT.

When p->i_number == p_section->i_number AND p_prev == NULL, then the
first element of the list has to be replaced. Since it is the first
element we have to update the pointer to the list, which is pointing to
the first element
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

Decoding EIT event descriptors was off by 2 bytes in function dvbpsi_eit_sections_decode().

Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

When using -Werror and -Wcast-align the use of sockaddr casting triggers a warning
about the cast saying: 'cast increases required alignment of target type' on ARM
and MIPS platform. A safer way to cast back and forth between sockaddr and sockaddr_in
(or sockaddr_in6) is using sockaddr_storage.

Repack DVBPSI_DECODER_COMMON to be byte aligned and add some checks to
do clever casting.
The TOT and SIS table structure definition need to be packed explicitly for use on ARM and MIPS platforms.

Remove empty remark.

In addition, my compiler complains about some non-sensical byte
overflow checks, which break my build (with -Werror enabled):
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

It seems that there was a bug introduced in renaming message
dvbpsi_message in dvbpsi_private.h. This is masked (in my environment
at least) by the configure script concluding that variadic macros are
not supported (i.e. ac_cv_cpp_variadic_macros=no) if -Werror is
included in the C flags, as the fragment in configure.ac used to
determine this generates a warning and thus failure.
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>