GitLab

This reverts commit 705a427f.

Some recording have a still SCR or a broken/offset SCR.
Detects if SCR - PTS is far above limits and rejects it.

dequeueOutputBuffer() can fail when flushing.

The previous implementation would potentially read the "element" right
after the last one in the subsequently used array (as priv->i_title
denotes the number of titles present).

These changes also include a check to make sure that var_GetInteger(
p_input, "title" ) does not result in a negative value (as such can be
used to read from outside the array, but on the other side).
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

As the function is not guaranteed to succeed we should check the return-value
in order to prevent potentially using a handle which is not valid.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

The previous implementation assumed that the remote entity would not
ask for authentication unless we have stated that we have the
capability for such, we should however not trust this assumption given
that it would (if it happens) cause problems with the username /
password being NULL.
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>

It is clearer to initialize the relevant data-member directly than to
use memset on the entire structure.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The order, and number, of data-members inside struct iovec is not
guaranteed according to the POSIX specification; as such these changes
make sure that we initialize what we expect to initialize by using
designated-initializers.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fixes "QPixmap::grabWidget is deprecated, use QWidget::grab() instead".
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Users of vlc_tls_t coalesce sends in user-space correctly. That being
the case, the Nagle algorithm just increases latency for no benefits.

Also remove duplicated error message (net_ConnectTCP prints a better
one).

Send the whole HTTP request in a single TCP frame.

This was not implemented properly and of questionable usefulness.

...with insecure HTTP.