There's no good reason to do this; we're not protecting anything that isn't
already protected by GStreamer locks.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
We don't do this in pull mode (and in general never have; avidemux and wavparse
are two examples of GStreamer elements that don't flush the upstream pad when
deactivating).
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Synchronization around flushing (on both sides) is tricky here, and the
solution used by this patch means that a "kernel" thread can be blocked on a
"user" thread. However, since it's not a real kernel thread and can't bring
down the rest of the "kernel", I don't think this is a concern.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
They depend on a valid TEB and can't be called from non-Wine threads.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This noticeably improves stream startup time. The process of typefinding and
negotiation takes about 70 ms for a simple test file on one machine, and this
is currently done twice. This patch replaces the second instance with a simple
seek, which takes less than 10 ms.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This is a fatal error condition for GStreamer, but should not cause errors for
DirectShow.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
It does not matter whether we use PAUSED or PLAYING, but we should at least be
consistent.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Instead of propagating GStreamer flush events to corresponding DirectShow pins.
This is mainly to avoid more callbacks from GStreamer and further separate the
Win32 and Unix code.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This is a rather large and complex change. It comprises several parts:
(1) Instead of directly sending EOS, segment, and sample events to the
downstream DirectShow sink, first queue them in a local buffer (i.e.
"pin->event").
(2) Spawn a separate thread per source pin (i.e. "stream_thread") which consumes
said events and sends them downstream.
(3) When flushing or stopping, explicitly wait for this thread to pause or stop
(respectively).
There are a few reasons for this:
(1) It reduces the number of Unix -> PE callbacks we need to make, easing PE
conversion. This is not a great advantage *a priori*, and may not be worth a
similar dedicated "handler" thread for most modules, but winegstreamer is
different—we were already marshalling these calls onto another thread, and
now that marshalling can go away (almost).
(2) Because GStreamer can only do pad negotiation (and hence autoplugging) while
running (in contrast to DirectShow, which can do it while stopped), we
currently have to renegotiate every time the pipeline is started. Most
applications don't start the graph more than once, but even that requires
two negotiations, and startup time is demonstrably too high. It would be
nice to keep the graph in PAUSED state all of the time, but this is
difficult to do without more fine-grained control over the streaming thread.
[In particular, we cannot reliably wait for all samples to be delivered
except by stopping the GStreamer pipeline.]
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This is the message session relies on succeeding.
Signed-off-by: Derek Lesho <dlesho@codeweavers.com>
Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
GStreamer doesn't necessarily signal all error when setting the
PLAYING or PAUSED state. If Wine just waits on no-more-pads, it risks
a deadlock if an error is signaled after gst_element_get_state was
called.
Signed-off-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This matches the behaviour of GstBaseSrc and GstBaseSink. In particular, we
now return FALSE for unhandled source events.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
GST_EVENT_TAG is a downstream-only event.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
gst_pad_get_current_caps() is not guaranteed to return the updated caps when
GST_EVENT_CAPS is handled by the event handler, or even after
gst_pad_event_default(). The actual implementation of gst_pad_send_event() in
fact sets it later. Avoid a potential race condition by using our own caps
pointer.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
"gstdemux" is not a very specific name, or accurate (it does not always demux,
nor is it the only object which may demux using GStreamer). Also, while it
does not risk colliding with GStreamer definitions as-is (lacking an
underscore), I am inclined to avoid using "gst" or "gstreamer" as a prefix for
our definitions, simply for the sake of clarity.
"parser" is the common term for both DirectShow filters and GStreamer elements
which "drive" the pipeline, i.e. explicitly pull data from upstream and push
data downstream. While it's arguably still not specific enough (i.e. should be
specifically identified as a DirectShow parser), I think that the differing
terminology in Media Foundation (where the counterpart is a "media source")
makes the choice of a shorter name justifiable.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
It's entirely unnecessary; only renderers and live sources care about this.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This information is already printed by GStreamer via GST_MESSAGE_ERROR.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This reverts commit fd25ba65e0.
The relevant commit breaks visual correctness in many applications, for the sake
of performance. As such I am rather inclined to revert it for 6.0, especially as
a full solution will probably involve a revert of this patch along with upstream
changes to GStreamer.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=49535
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Otherwise it is used uninitialised if a failure occurs.
Signed-off-by: Gijs Vermeulen <gijsvrm@gmail.com>
Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Gstreamer stores the first component in the most significant bits, same as DirectShow.
Signed-off-by: Anton Baskanov <baskanov@gmail.com>
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
According to the GStreamer documentation, the contents of "buffer" should be
considered undefined on failure, which means that the caller has no way of
freeing a newly allocated buffer.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
From GStreamer documentation: "If the handler returns GST_BUS_DROP, it should
unref the message, else the message should not be unreffed by the sync handler."
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This reverts commit ddd43e0a80.
This function returns void before 1.12.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=49647
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This allows "Zero Escape: Nine Hours, Nine Persons, Nine Doors" to start.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Primarily so that we can connect using video caps that can't be translated into
DirectShow caps. They will later be converted into translatable caps once the
DirectShow source pin is connected.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=47642
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This reduces the time videoconvert takes to process I420 -> YV12 from 15 ms to 0.8 ms on my system.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
These fields don't prevent us from connecting, but they may force videoconvert
to do an expensive conversion where not necessary.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
The mpegaudioparse element will not send a duration-changed message if no
Xing or VBRI headers are present.
Signed-off-by: Akihiro Sagawa <sagawa.aki@gmail.com>
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This can help mpegaudioparse report a usable duration, and presumably also
other elements built from baseparse.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Some application filters enumerate types without a format block.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Some callbacks may be pushed during pipeline initialization and before
the dispatch thread is actually started and waiting for them, which
sometimes causes some hangs.
This fixes Fallout 3 and Panzer Corps -and possibly others- intro videos
hanging forever.
Signed-off-by: Rémi Bernon <rbernon@codeweavers.com>
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
If the stride is not equal to the width, this calculation will be invalid.
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This fixes background music in Tomb Raider II.
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This check is not equivalent to the existence of "their_src" for the
MPEG-1 splitter, since that always exposes a source audio pin.
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
In particular, pass the correct buffer size for the format we have chosen.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=48570
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
In the unlikely event that we are hotplugging not from an IAsyncReader, we
already have the ACM wrapper and l3codeca.acm available to do the same thing.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>