GitLab

Needed because hideResumePanel() decreases the height.
Close #14566
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Close #7950
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Otherwise, goes through PES handlers which screams
about invalid header.

current event is relative to programs, not to first eit callback

This patch reworks a bit the way we handle textures & surfaces.
The surface is the size of the decoder buffer (i_width/i_height)
The texture is the size of the visible picture (i_visible_*)
StretchRect is now copying only the valuable parts of the decoder
buffer (ie. it includes offsets & visible_*)
Fix #16597

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

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

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

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

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

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

this is also done in ffmpeg/libav
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

`rtsp_get_answers` would crash/have unexpected behavior in the following
scenarios:

  - remote server sent more than 256 headers (off-by-one write)
  - remote server sent one of the following headers without payload
    - Server
    - Session
    - CSeq

In order to fix the issue a bunch of checks have been introduced, mostly
boiling down checking to see that `sscanf` succeeds in reading data into
the destination variable.

A few warnings (`msg_Warn`) have been introduced to help diagnostic
servers that misbehave.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fixed bounds check in `rtsp_unschedule_field` and `rtsp_free_answers`. Since
both functions share a common goal a (`static`) helper function has been
introduced.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The previous code would loop forever (in two places) if this function is ever
invoked, it would also read out-of-bounds if the `scheduled` queue was full.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The previous code would write out-of-bounds if the answered queue was full,
since the code expects there to be at least one NULL value among the fields
(something which is not guaranteed).
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

The previous code would read out-of-bounds if the scheduled queue was full,
since the code expects there to be at least one NULL value among the fields
(something which is not guaranteed).
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

If `psz_buffer` points to a string equivalent to just "RTSP/1.0", we
would read 2 bytes outside the buffer when (the removed) `memcpy` was
invoked.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

- `strchr` can return `NULL`
- `data` is a pointer to a buffer which has a length that depends on
  the previous read of `Content-Header`.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Given the below (example) correspondance with a remote-server, the
module will iterate endlessly since we never move to the next line to
process.

% nc -l -p 8080 <<EOF
> RTSP/1.0 200 OK
> CSeq: 1
> Server: Real
> RealChallenge1: foobar
>
> RTSP/1.0 200 OK
> CSeq: 1
> ETag: hello world
> Content-Length: 5
>
> m=123
>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

If the remote server does not include the `ETag` header when in the
response to the DESCRIBE request, the module would crash due to usage of
`strlen(session_id)` later on, where `session_id` would be  `NULL`.

% netcat -l -p 8080 <<EOF
> RTSP/1.0 200 OK
> CSeq: 1
> Server: Real
> RealChallenge1: foobar
>
> RTSP/1.0 200 OK
> CSeq: 2
> Content-Length: 10
>
> helloworld
EOF
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

If the remote server yields an error on the `DESCRIBE` request, while
also including an error message the module would crash due to an invalid
free.

% netcat -l -p 8080 <<EOF
> RTSP/1.0 200 OK
> CSeq: 1
> Server: Real
> RealChallenge1: DEADBEEF
>
> RTSP/1.0 199 OK
> CSeq: 2
> Alert: I like turtles
>
EOF

The reason being that `alert` will point to the middle of allocated
memory. Given the sourroundings the original author probably forgot to
`strdup` the message - even though that is very unnecessary.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The module would crash if the remote server did not include `RealChallenge1` in
the headers of the initial reply after establishing a connection. The below is a
sample testcase:

% nc -l -p 8080 <<EOF
> RTSP/1.0 200 OK
> Server: Real
>
> EOF
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

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

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

Close #15630
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This keystore is used to store credentials on memory if no others keystore is
found or if the user doesn't want to store them permanently. This keystore is
owned by libvlc and is shared with all vlc objects. It's released when the
libvlc instance is released.