After the split of libdvdcss from libdvdread,
the stream callback code was moved to libdvdcss.

This went largely unnoticed because most consumers
played their DVD‌ ISO images on the systems with libdvdcss
installed.

However, I tested Kodi by playing unencrypted DVD‌ ISO
and noticed the inability to play it.

This code is inspired by stream_* functions in libdvdcss.

Signed-off-by: Vasyl Gello <vasek.gello@gmail.com>

Steve Lhomme authored 2 years ago and Jean-Baptiste Kempf committed 2 years ago

LLVM ends up using the size of an unsigned int if it's used in the bitfield.
In other places the bitfields are using unsigned char which always uses the
proper size in the packed structures.

/!\ btni_t fields are kept with unsigned int as unsigned char cannot hold
10 bits. btni_t is not actually read as a packed structure. So we keep the
current format for backward compatibility.

Steve Lhomme authored 2 years ago and Jean-Baptiste Kempf committed 2 years ago

LLVM ends up using the size of an unsigned int if it's used in the bitfield.
In other places the bitfields are using unsigned char which always uses the
proper size in the packed structures.

* multichannel_ext_t would end up having a size of 27 instead of 24
* cell_playback_t would end up having a size of 26 instead of 24
* pgci_srp_t would end up having a size of 11 instead of 8
* playback_type_t would end up having a size of 4 instead of 1

Close #8

This was authorized by Eric Smith, original author of VOBDUMP, by email
on May 15, 2022

Steve Lhomme authored 3 years ago and Jean-Baptiste Kempf committed 2 years ago

According to [1] these 16 bits represent the "end address relative to command table".
This corresponds to the last_byte check in other places.

[1] http://dvd.sourceforge.net/dvdinfo/pgc.html#cmd

Steve Lhomme authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

It may not be in the end of the buffer anymore.

The byte value used was always equivalent to
> state->start[state->byte_position] << state->bit_position

Also constify the read buffer.

Steve Lhomme authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

The comments say it exists in the wild.

calloc() may or may not return NULL in that case. Just use NULL but assume it's
not an error.

to compute a canonical path

Steve Lhomme authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

We were allocating 4x the number of PTTs. We should use the same value as
nr_of_ptts.

It is not checked to be zero as most other field so may contain useful values.

Brad Smith authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

a.out support was obsolete 18 years ago.

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Brad Smith authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Mathias Couder authored 3 years ago and Jean-Baptiste Kempf committed 3 years ago

And notably add VideoLAN Code of Conduct and improve lisibility

This issues a warning for each packed struct each time the header gets
included

Actually a work-around, tbh, but it'll take more time to fix.