Catch more socket errors.
Rework how sockets are managed in the GSource, wake up the maincontext instead
of adding/removing the sockets from the source.
Add callback for when the tunnel connection is lost. Some clients (Quicktime
Player) close the POST connection in tunneled mode and reopen the socket when
needed.
See #612915
Point g-ir-scanner to the .la file of our library, which hopefully
makes it find the right dependencies in all cases (ie. our locally
built libgstreamer and not the system-installed one). This is also
how it's done in Gtk+ and how it's documented in the wiki, see
http://live.gnome.org/GObjectIntrospection/AutotoolsIntegrationFixes#603710.
Use new girdir and typlibdir from core .pc files, so we can figure
out the right includes to pass to the gobject-introspection tools,
whether core is installed in the same prefix as gobject-introspection
or in a different prefix or uninstalled. This also keeps us from adding
bogus paths to the includes that only work if core is uninstalled.
Also add some missing includes/pkgs where needed.
Since we no longer use an array of error messages, there is no reason
to clamp the error code, which allows us to simplify the code some more
and also to actually report the correct error code for unknown errors.
2 goals in the refactoring:
- Put the error messages closer to their enum values, so that it's easy
to see which error belongs to which value.
- Make gcc not complain with -Wformat-nonliteral
Be careful when allocating the amount of bytes specified in the Content-Length
because it can be an insanely huge value. Try to allocate the memory but fail
gracefully with a nice error when the allocation failed.
Use send() instead of write() so that we can pass the MSG_NOSIGNAL flags to
avoid crashing with SIGPIPE when the remote end is not listening to us anymore.
Fixes#601772
I also renamed glib_enum_prefix to glib_gen_prefix as we also use that for the
marshallers. Also rename the rtsp-marshal.list to work with the unified prefix.
gstrtspconnection.c:gst_rtsp_connection_receive() can hang when an error occured
on a socekt. Fix this problem by checking for error on 'other' socket after poll
return.
Fixes#596159
The new API to send messages using GstRTSPWatch will first try to send the
message immediately. Then, if that failed (or the message was not sent
fully), it will queue the remaining message for later delivery. This avoids
unnecessary context switches, and makes it possible to keep track of
whether the connection is blocked (the unblocking of the connection is
indicated by the reception of the message_sent signal).
This also deprecates the old API (gst_rtsp_watch_queue_data() and
gst_rtsp_watch_queue_message().)
API: gst_rtsp_watch_write_data()
API: gst_rtsp_watch_send_message()
With gst_rtsp_connection_set_http_mode() it is possible to tell the
connection whether to allow HTTP messages to be supported. By enabling HTTP
support the automatic HTTP tunnel support will also be disabled.
API: gst_rtsp_connection_set_http_mode()
The error_full callback is similar to the error callback, but allows for
better error handling. For read errors a partial message is provided to
help an RTSP server generate a more correct error response, and for write
errors the write queue id of the failed message is returned.
Rewrote read_line() to support LWS (Line White Space), the method used by
RTSP (and HTTP) to break long lines. Also added support for \r and \n as
line endings (in addition to the official \r\n).
From RFC 2068 section 4.2: "Multiple message-header fields with the same
field-name may be present in a message if and only if the entire
field-value for that header field is defined as a comma-separated list
[i.e., #(values)]." This means that we should not split other headers which
may contain a comma, e.g., Range and Date.
Due to the odd syntax for WWW-Authenticate (and Proxy-Authenticate) which
allows commas both to separate between multiple challenges, and within the
challenges themself, we need to take some extra care to split these headers
correctly.
