platform/upstream/gstreamer.git
4 years agosystemclock: Don't divide by zero on Windows if high performance timers are not available
Sebastian Dröge [Mon, 9 Mar 2020 13:16:00 +0000 (15:16 +0200)]
systemclock: Don't divide by zero on Windows if high performance timers are not available

4 years agocaps: Don't assert in fixate() on EMPTY/ANY caps and document EMPTY/ANY behaviour...
Sebastian Dröge [Sat, 7 Mar 2020 09:09:05 +0000 (11:09 +0200)]
caps: Don't assert in fixate() on EMPTY/ANY caps and document EMPTY/ANY behaviour on more functions

fixate() will return empty caps if it gets empty caps passed and assert
early if any caps are provided as there's no meaningful way of fixating
any caps.

truncate() and simplify() will return the input caps in case of
any/empty caps as before, but slightly optimized and as documented
behaviour.

Also add tests for this and a few other operations behaviour on
empty/any caps.

4 years agogstaggregator: fix the prototype of sink_event_pre_queue
Mathieu Duponchelle [Wed, 4 Mar 2020 21:13:12 +0000 (22:13 +0100)]
gstaggregator: fix the prototype of sink_event_pre_queue

This is not an API breakage, as implementors are already
expected to return a GstFlowReturn

4 years agogst-launch: Follow up to missing s/g_print/gst_print/g
Seungha Yang [Tue, 3 Mar 2020 09:49:36 +0000 (18:49 +0900)]
gst-launch: Follow up to missing s/g_print/gst_print/g

Required to avoid broken log string on Windows but missed
in the commit of 493a3261a9757b5ade7aec289eb07221966f9eed

4 years agotests: info: Fix thread-id pattern matching on Windows
Seungha Yang [Sat, 29 Feb 2020 10:00:44 +0000 (19:00 +0900)]
tests: info: Fix thread-id pattern matching on Windows

The format modifier for thread-id prints hex value without "0x" prefix on Windows.

4 years agotracers: rusage: use thread-local storage for per-thread stats
Tim-Philipp Müller [Sun, 26 Jan 2020 00:56:44 +0000 (00:56 +0000)]
tracers: rusage: use thread-local storage for per-thread stats

.. instead of looking things up by thread id from a GHashTable,
which also happens to have no locking around insertion/lookup.

4 years agotracers: rusage: fix minor string leak in constructor
Tim-Philipp Müller [Sun, 26 Jan 2020 00:32:18 +0000 (00:32 +0000)]
tracers: rusage: fix minor string leak in constructor

4 years agogst-stats: Fix missing NULL checks
Johan Bjäreholt [Sat, 2 Nov 2019 10:49:25 +0000 (11:49 +0100)]
gst-stats: Fix missing NULL checks

gst-inspect-1.0 segfaults on tracing logs where it fails to find
element stats. So on the pipelines where we get the following WARNING
during execution will afterwards crash with a segfault as the
g_ptr_array has a index for it but it is just a NULL pointer.

WARN default gst-stats.c:444:do_message_stats: no element stats found for ix=X

An example of an pipeline which can reproducibly create a trace log
where this occurs would be this

GST_DEBUG="GST_TRACER:7" GST_TRACERS="stats;rusage;latency" gst-launch-1.0 videotestsrc num-buffers=120 ! autovideosink &> trace.log
gst-stats-1.0 trace.log

4 years agobasesink: Improve clarity of latency query maths debug message
Olivier Crête [Mon, 24 Feb 2020 20:24:44 +0000 (15:24 -0500)]
basesink: Improve clarity of latency query maths debug message

Add the equation to the debug message to make it easier for non-GStreamer
experts to understand why their pipeline has latency.

4 years agotests: Maintain compatibility with GLib 2.48
Nirbheek Chauhan [Wed, 26 Feb 2020 11:50:04 +0000 (17:20 +0530)]
tests: Maintain compatibility with GLib 2.48

That's the minimum version of GLib we require right now.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/514

4 years agoclocksync: Add new clocksync element
Jan Schmidt [Mon, 24 Feb 2020 17:47:35 +0000 (04:47 +1100)]
clocksync: Add new clocksync element

The clocksync element is a generic element that can be
placed in a pipeline to synchronise passing buffers to the
clock at that point. This is similar to 'identity sync=true',
but because it isn't GstBaseTransform-based, it can process
GstBufferLists without breaking them into separate GstBuffers

4 years agogst-inspect: Add define guard for g_log_writer_supports_color()
Seungha Yang [Wed, 26 Feb 2020 13:29:43 +0000 (22:29 +0900)]
gst-inspect: Add define guard for g_log_writer_supports_color()

g_log_writer_supports_color() was introduced since GLib 2.50.0
which is slightly higher version than our minimum required GLib version.

4 years agoDon't use glib format modifiers with sscanf or printf
Nirbheek Chauhan [Tue, 25 Feb 2020 13:43:59 +0000 (19:13 +0530)]
Don't use glib format modifiers with sscanf or printf

We do not have a way to know the format modifiers to use with string
functions provided by the system. `G_GUINT64_FORMAT` and other string
modifiers only work for glib string formatting functions. We cannot
use them for string functions provided by the stdlib. See:
https://developer.gnome.org/glib/stable/glib-Basic-Types.html#glib-Basic-Types.description

F.ex.:

```
 ../tools/gst-stats.c:921:11: error: too many arguments for format [-Werror=format-extra-args]
   printf ("Number of Buffers passed: %" G_GUINT64_FORMAT "\n", num_buffers);
           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../tools/gst-stats.c:922:11: error: unknown conversion type character 'l' in format [-Werror=format=]
   printf ("Number of Events sent: %" G_GUINT64_FORMAT "\n", num_events);
           ^~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86_64/include/glib-2.0/glib/gtypes.h:32,
                 from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86_64/include/glib-2.0/glib/galloca.h:32,
                 from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86_64/include/glib-2.0/glib.h:30,
                 from ../gst/gst.h:27,
                 from ../tools/tools.h:28,
                 from ../tools/gst-stats.c:30:
/builds/nirbheek/cerbero/cerbero-build/dist/windows_x86_64/lib/glib-2.0/include/glibconfig.h:69:28: note: format string is defined here
 #define G_GUINT64_FORMAT "llu"
                            ^
```

and

```
../tests/misc/netclock-replay.c: In function 'main':
../tests/misc/netclock-replay.c:98:23: error: unknown conversion type character 'l' in format [-Werror=format=]
     if (sscanf (line, "%" G_GUINT64_FORMAT " %" G_GUINT64_FORMAT " %"
                       ^~~
In file included from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86/include/glib-2.0/glib/gtypes.h:32,
                 from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86/include/glib-2.0/glib/galloca.h:32,
                 from /builds/nirbheek/cerbero/cerbero-build/dist/windows_x86/include/glib-2.0/glib.h:30,
                 from ../tests/misc/../../libs/gst/net/gstntppacket.c:38,
                 from ../tests/misc/netclock-replay.c:31:
/builds/nirbheek/cerbero/cerbero-build/dist/windows_x86/lib/glib-2.0/include/glibconfig.h:69:28: note: format string is defined here
 #define G_GUINT64_FORMAT "llu"
                            ^
```

This is needed for upgrading glib inside Cerbero which builds with
`-Werror` on Windows:
https://gitlab.freedesktop.org/gstreamer/cerbero/merge_requests/419

4 years agodebugutils: skip "parent" property for elements when dumping pipeline graph
Tim-Philipp Müller [Wed, 19 Feb 2020 18:49:07 +0000 (18:49 +0000)]
debugutils: skip "parent" property for elements when dumping pipeline graph

Seems unnecessary to print the parent name for every
element in the pipeline graph, it's clear from the
graph what the parent element is and it's hard to
imagine a case where this is useful info rather than
just distracting spam. So far this was only done for
pads, but we should just do it for everything.

4 years agobaseparse: fix memory leak
Matus Gajdos [Thu, 19 Dec 2019 10:28:13 +0000 (11:28 +0100)]
baseparse: fix memory leak

A buffer to be skipped wasn't unref'd in gst_base_parse_chain().

Fixes #406

4 years agoleak tracer: Initialize GValue
Olivier Crête [Mon, 27 Jan 2020 19:46:18 +0000 (14:46 -0500)]
leak tracer: Initialize GValue

4 years agoleaks: Do not trace refs for object we do not follow
Thibault Saunier [Thu, 13 Feb 2020 20:53:29 +0000 (17:53 -0300)]
leaks: Do not trace refs for object we do not follow

When the user sets filters, we should not trace ref counts of object that
are not traced. This optimizes the tracer by potentially avoiding
generating useless backtraces.

4 years agobaseparse: Set the private duration before posting a duration-changed message
Zebediah Figura [Mon, 10 Feb 2020 22:35:06 +0000 (16:35 -0600)]
baseparse: Set the private duration before posting a duration-changed message

Otherwise an application cannot rely on a subsequent call to e.g. gst_pad_query_duration() succeeding.

4 years agobus: Make setting/replacing/clearing the sync handler thread-safe
Sebastian Dröge [Wed, 12 Feb 2020 10:32:05 +0000 (12:32 +0200)]
bus: Make setting/replacing/clearing the sync handler thread-safe

Previously we would use the object lock only for storing the sync
handler and its user_data in a local variable, then unlock it and only
then call the sync handler. Between unlocking and calling the sync
handler it might be unset and the user_data be freed, causing it to be
called with a freed pointer.

To prevent this add a refcounting wrapper struct around the sync
handler, hold the object lock while retrieving it and increasing the
reference count and only actually free it once the reference count
reaches zero.

As a side-effect we can now also allow to actually replace the sync
handler. Previously it was only allowed to clear it after initially
setting it according to the docs, but the code still allowed to clear it
and then set a different one.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/506

4 years agodocs: Fix bold markdown syntax for GST_DEBUG_NO_COLOR
Seungha Yang [Thu, 13 Feb 2020 06:38:15 +0000 (15:38 +0900)]
docs: Fix bold markdown syntax for GST_DEBUG_NO_COLOR

Fixing markdown syntax

4 years agocaps: keep ANY caps empty internally
Henry Wilkes [Mon, 27 Jan 2020 11:58:57 +0000 (11:58 +0000)]
caps: keep ANY caps empty internally

Keep the ANY caps empty internally when appending and merging
caps/structures. Previously, an ANY caps could end up containing
internal structures, which could be fetched by the user, and gave the
caps a non-zero length.

Also, made sure that `gst_caps_set_features_simple` frees the features
if caps is empty.

4 years agocaps: fix is_strictly_equal
Henry Wilkes [Tue, 21 Jan 2020 19:02:48 +0000 (19:02 +0000)]
caps: fix is_strictly_equal

Fixed gst_caps_is_strictly_equal() to take into account whether either of
the caps are ANY caps. Previously, two ANY caps could be considered not
strictly equal if one of them still contained some remnant *internal*
structure (this can happen if an ANY caps has emerged from an append or
merge operation). Also, an ANY caps with no remnant internal structures
was considered strictly equal to an EMPTY caps. Similarly, a non-ANY caps
was considered strictly equal to an ANY caps if its remnant internal
structures happened to match.

Also changed gst_caps_is_fixed to take into account that an ANY caps
should not be considered fixed even if it contains a single remnant
internal fixed structure. This affects gst_caps_is_equal(), which uses a
separate method if both caps are fixed. Previously, this meant that a
non-ANY fixed caps was considered equal to an ANY caps if it contained a
single matching remnant internal structure.

Added some tests for these two equality methods, which covers the above
examples, as well as asserts existing behaviour.

Fixes #496

4 years agobasetransform: Make gst_base_transform_reconfigure() public
Sebastian Dröge [Mon, 10 Feb 2020 10:58:47 +0000 (12:58 +0200)]
basetransform: Make gst_base_transform_reconfigure() public

This has the same function as the negotiate() functions in various other
base classes and is required to be able to completely re-implement
submit_input_buffer() in subclasses.

4 years agobaseparse: Don't set meaningless buffer dts from segment->start
Thibault Saunier [Tue, 7 Jan 2020 20:12:54 +0000 (17:12 -0300)]
baseparse: Don't set meaningless buffer dts from segment->start

When we do not have any information about DTSs we shouldn't try to make
them up, moreover after seeking `segment->start` has nothing to do with
the next buffer timing (and is probably after the actual buffer timestamp)
and since, since https://gitlab.freedesktop.org/gstreamer/gstreamer/commit/fa8312472f08d468677d188d5cf1ad52c5b5b0a0
we do:

```
if (buffer->dts > buffer->dts)
    buffer->pts = buffer->dts
```

we end up setting `buffer->pts = segment->start` which is plain
broken and leads to downstream decoder accept the first buffer
as it will be inside the segment (its pts==segment->start) which
basically means accurate seeking behaves mostly the same way as
keyframe seeks.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/492

4 years agosystemclock: No need to check for CLOCK_TAI in the meson
Olivier Crête [Fri, 27 Dec 2019 17:36:10 +0000 (12:36 -0500)]
systemclock: No need to check for CLOCK_TAI in the meson

POSIX defines CLOCK_MONOTONIC to always be a macro, so I think
it's safe to assume that CLOCK_TAI will also be.

4 years agoGstSystemClock: Add GST_CLOCK_TYPE_TAI
Ederson de Souza [Fri, 13 Dec 2019 19:07:40 +0000 (11:07 -0800)]
GstSystemClock: Add GST_CLOCK_TYPE_TAI

GST_CLOCK_TYPE_TAI is GStreamer abstraction for CLOCK_TAI. Main
motivation for this patch is support for transmission offloading features
- when network packets are timestamped with the time they are deemed to
be actually transmitted. Linux API for that requires that time to be
in CLOCK_TAI coordinate.

With GST_CLOCK_TYPE_TAI, applications can use CLOCK_TAI directly on
their pipelines, avoiding the need to cross timestamp packet times. By
leveraging system's CLOCK_TAI, applications also don't need to keep track
of leap seconds - less burden for them. Just keep system's CLOCK_TAI
accurate and use it.

4 years agobin: Don't consider having a group-id or being STREAM_START if we have not a single...
Sebastian Dröge [Fri, 24 Jan 2020 21:56:32 +0000 (23:56 +0200)]
bin: Don't consider having a group-id or being STREAM_START if we have not a single STREAM_START message

This would cause us to set GST_GROUP_ID_INVALID as group-id in the
aggregated STREAM_START message if there are no sinks at all or none of
them have a STREAM_START message, which is simply wrong.

If we have not a single STREAM_START message then the bin should not be
considered STREAM_START.

4 years agoevent/message: Don't allow setting invalid group ids
Sebastian Dröge [Fri, 24 Jan 2020 15:52:49 +0000 (17:52 +0200)]
event/message: Don't allow setting invalid group ids

They are optional on STREAM_START messages/events but if available
should have at least a valid value.

For STREAM_GROUP_DONE events don't allow creating it with an invalid
group id as this does not make any sense.

4 years agoaggregator: Initialize source pad segment position to -1 when resetting
Sebastian Dröge [Thu, 23 Jan 2020 17:27:14 +0000 (19:27 +0200)]
aggregator: Initialize source pad segment position to -1 when resetting

This allows start-time selection in gst_aggregator_pad_chain_internal()
to actually work as that code assumes it to be -1 for actually
overriding the value.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/500

4 years agobin: Fix deep-element-removed log message
Jan Alexander Steffens (heftig) [Thu, 9 Jan 2020 19:07:06 +0000 (20:07 +0100)]
bin: Fix deep-element-removed log message

child and bin were switched.

https://gitlab.freedesktop.org/gstreamer/gstreamer/merge_requests/354

4 years agodocs: Document the new 'redirect-location' error message detail field
Thibault Saunier [Tue, 3 Sep 2019 21:14:49 +0000 (17:14 -0400)]
docs: Document the new 'redirect-location' error message detail field

4 years agoparse: add support for presets
Stefan Sauer [Tue, 30 Dec 2014 10:48:26 +0000 (11:48 +0100)]
parse: add support for presets

Add new parse syntax: @preset="<preset-name>" to load presets.
Fixes #86

4 years agogst-launch: handle ERROR messages in the sync handler
Mathieu Duponchelle [Thu, 26 Dec 2019 14:08:09 +0000 (15:08 +0100)]
gst-launch: handle ERROR messages in the sync handler

Errors causing the pipeline to fail going from NULL to PAUSED
were not displayed, and the pipeline was not dumped either in
those cases.

In addition, dumping the pipeline from the sync handler means
the dump matches exactly the state of the pipeline at the
moment the error was posted.

4 years agogst-inspect: Increase array size for printing rank name
Seungha Yang [Sun, 22 Dec 2019 12:13:00 +0000 (21:13 +0900)]
gst-inspect: Increase array size for printing rank name

Now the rank value can be MAX_INT (2147483647)

4 years agopluginfeature: Allow updating initial rank of plugin feature
Seungha Yang [Thu, 15 Aug 2019 11:56:40 +0000 (20:56 +0900)]
pluginfeature: Allow updating initial rank of plugin feature

Introducing "GST_PLUGIN_FEATURE_RANK" environment variable in order for users
to adjust rank of plugin(s) via environment.

A "feature" and "rank" key-value pair should be separable by ":",
and each key-value pair is recognized per "," delimiters. The rank
can be a numerical value or one of pre-defined rank values
such as "NONE", "MARGINAL", "SECONDARY", and "PRIMARY" in case-insensitive manner.

In addition to pre-defined { NONE, MARGINAL, SECONDARY, PRIMARY },
"MAX" can be passed to key value used to ensure having a higher rank
than other plugin features.

Example)
- GST_PLUGIN_FEATURE_RANK=qtdemux:256,h264parse:NONE
Set rank of qtdemux plugin to 256 (primary) and 0 (none) for h264parse.

4 years agogstinfo: Add new API for getting debug log lines
Nirbheek Chauhan [Thu, 29 Aug 2019 18:53:09 +0000 (00:23 +0530)]
gstinfo: Add new API for getting debug log lines

If you're using a custom log handler, you had to reverse-engineer the
debug log format and create your own format function. Now, you can
call `gst_debug_log_get_line()` and it will return a string (without
ANSI escape color codes) representation instead.

This is useful in situations when you need to log the ordinary
gst_debug log to a resource that can't be opened as a `FILE` handle.

Also includes a test.

4 years agotests: remove system-dependent tests
Mathieu Duponchelle [Fri, 20 Dec 2019 13:01:02 +0000 (14:01 +0100)]
tests: remove system-dependent tests

We now have GstTestClock-based tests that validate the same logic,
without inducing spurious timing failures / overly relying on sleeps.

Fixes: #346
Fixes: #347
Fixes: #348

Co-authored by: Thibault Saunier <tsaunier@igalia.com>

4 years agotests-clock: Fix race in test_late_crank
Thibault Saunier [Fri, 20 Dec 2019 13:53:21 +0000 (10:53 -0300)]
tests-clock: Fix race in test_late_crank

There was a case where we started waiting on the clock before setting
the clock time, leading to the wait succeeding instead of being late:

    gsttestclock.c:1073:F:testclock:test_late_crank:0: '1 * GST_SECOND' (1000000000) is not equal to 'context.jitter' (-4000000000)

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/426

Co-authored by: Mathieu Duponchelle <mathieu@centricular.com>

4 years agobin: Add method to find elements by factory name
Niels De Graef [Fri, 15 Nov 2019 14:49:32 +0000 (15:49 +0100)]
bin: Add method to find elements by factory name

A common use case of a dynamically built pipeline is that you want to
(conditionally) find a certain element, e.g. the `rtpbin`s in a
`uridecodebin`. If that element has a fixed name inside its parent bin
(and only has a single instance) this can be easily done by
`gst_bin_get_by_name()`.

If there are multiple instances of the element however, you can only use
`gst_bin_iterate_all_by_interface()`, but this doesn't work if you don't
have the specific `GType` (which is often the case, due to plugins being
dynamically loaded). As such, another fallback could be to use the
well-known name of the element's factory (in case of our example, this
is of course `"rtpbin"`).

4 years agogstreamer: use of g_value_dup_string
Stéphane Cerveau [Wed, 18 Dec 2019 14:57:35 +0000 (15:57 +0100)]
gstreamer: use of g_value_dup_string

Use helper method to get string from GValue.

4 years agotests: fix pipelines_parse_launch.delayed_link flakiness
Mathieu Duponchelle [Fri, 13 Dec 2019 17:21:32 +0000 (18:21 +0100)]
tests: fix pipelines_parse_launch.delayed_link flakiness

Fixes #345

There were two causes for the flakiness, one much rarer than
the other.

The test sets up a source with a sometimes pad added during
the transition of a wrapper bin from READY to PAUSED.

It runs 4 iterations, the last of which makes it so the
negotiation fails.

In that case, the intention as correctly presented by the following
comment:

/* [..] ie, the pipeline should create ok but fail to change state */

However the implementation of run_delayed_test was neither calling
get_state on the pipeline (it called it on the wrapper bin), nor
checking that the return of get_state was FAILURE (it actually
checked that it was not).

This led to an obvious race condition, and was fixed by calling
get_state on the pipeline, then checking that in this specific
case (expect_link == FALSE), the state change has actually failed.

The second, rarer race condition is at set_state time. When we
don't expect the link to succeed, the return of set_state may
either be FAILURE or ASYNC, depending on timing. This was fixed
by taking expect_link into account when checking the return value
of set_state.

Co-authored by: Thibault Saunier <tsaunier@igalia.com>

4 years agopluginloader: handle fsync interrupted by signal (EINTR)
Peter Seiderer [Thu, 12 Dec 2019 10:39:56 +0000 (11:39 +0100)]
pluginloader: handle fsync interrupted by signal (EINTR)

According to [1] EINTR is a possible errno for fsync(),
so handle it as all other EINTR (do/while(errno == EINTR)).

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
4 years agoregistry: handle fsync interrupted by signal (EINTR)
Peter Seiderer [Thu, 12 Dec 2019 10:37:56 +0000 (11:37 +0100)]
registry: handle fsync interrupted by signal (EINTR)

According to [1] EINTR is a possible errno for fsync(),
so handle it as all other EINTR (do/while(errno == EINTR)).

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
4 years agofilesink: handle fsync interrupted by signal (EINTR)
Peter Seiderer [Thu, 12 Dec 2019 10:07:07 +0000 (11:07 +0100)]
filesink: handle fsync interrupted by signal (EINTR)

According to [1] EINTR is a possible errno for fsync() and it happens in
reality on linux (video writing via splitmuxsink with robust muxing enabled
on a cifs mounted network share), so handle it as all other EINTR
(do/while(errno == EINTR)).

Fixes:

  GError.message: Error while writing to file "vidoe_001.mp4". GError.domain: 2372 GError.code: 10 from: FileSink debug: gstfilesink.c(849): gst_file_sink_render (): /GstPipeline:Pipeline/GstSplitMuxSink:SplitMuxSink/GstBin:QueueBin/GstFileSink:FileSink: Interrupted system call

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
4 years agoRemove deprecated GTimeVal
Olivier Crête [Tue, 10 Dec 2019 22:06:02 +0000 (17:06 -0500)]
Remove deprecated GTimeVal

GTimeVal won't work past 2038

4 years agodevice, elementfactory: relax floating requirement
Mathieu Duponchelle [Tue, 10 Dec 2019 12:31:50 +0000 (13:31 +0100)]
device, elementfactory: relax floating requirement

Using g_assert() is a bit too extreme, as it will abort the whole
program unless G_DISABLE_ASSERTS is true.

Switch to g_critical()

4 years agogstcheck: remove bogus refcount asserts
Mathieu Duponchelle [Tue, 10 Dec 2019 08:42:37 +0000 (09:42 +0100)]
gstcheck: remove bogus refcount asserts

As soon as gstcheck potentially calls out to code it does not
control, such as gst_element_request_pad, all assertions about
pad refcounts go out the window.

4 years agoaggregator: fix logging in new update_segment API
Mathieu Duponchelle [Fri, 6 Dec 2019 10:40:44 +0000 (11:40 +0100)]
aggregator: fix logging in new update_segment API

4 years agoaggregator: add method to update srcpad segment
Mathieu Duponchelle [Thu, 5 Dec 2019 12:44:33 +0000 (13:44 +0100)]
aggregator: add method to update srcpad segment

4 years agobus: Clean up #ifdefs to compile with debugging enabled in all combinations
Sebastian Dröge [Thu, 5 Dec 2019 07:54:32 +0000 (09:54 +0200)]
bus: Clean up #ifdefs to compile with debugging enabled in all combinations

Thanks to Roland Jon for finding this.

4 years agodevice, elementfactory: don't enforce floating status
Mathieu Duponchelle [Wed, 4 Dec 2019 19:12:02 +0000 (20:12 +0100)]
device, elementfactory: don't enforce floating status

The reference we receive when calling g_object_new should be
floating, but we can't force it at our level.

Switch from g_object_force_floating() to a simple assertion.

See https://gitlab.freedesktop.org/gstreamer/gst-python/issues/27

4 years agotestclock: added single clock id process function
Tulio Beloqui [Wed, 19 Jun 2019 11:45:54 +0000 (13:45 +0200)]
testclock: added single clock id process function

Co-authored-by: Havard Graff <hgr@pexip.com>
4 years agobus: Use new GSource dispose function
Sebastian Dröge [Mon, 21 Oct 2019 14:56:14 +0000 (17:56 +0300)]
bus: Use new GSource dispose function

Without this it is possible that we have a GSource with reference count
0 stored in the GstBus that is currently in the process of being
destroyed. gst_bus_remove_watch() might then access it, increase its
reference count to 1 again, call GSource API on it and then unref it,
which will then finalize it a second time.

The dispose function allows the GSource to be resurrected until it
returned so the above would be safe now.

This caused some spurious crashes during shutdown in various
applications.

4 years agoMeson: Add 'coretracers' feature option
Xavier Claessens [Tue, 3 Dec 2019 20:40:59 +0000 (15:40 -0500)]
Meson: Add 'coretracers' feature option

This was the only plugin still built when using
-Dauto_features=disabled, besides coreelements.

4 years agogstharness: don't push the event to the queue before processing
Håvard Graff [Tue, 3 Dec 2019 11:23:01 +0000 (11:23 +0000)]
gstharness: don't push the event to the queue before processing

The application might pull and unref it by the time the code gets
around to check it for EOS.

4 years agobaseparse: Don't copy invalid DTS to the PTS
Vivia Nikolaidou [Thu, 28 Nov 2019 11:09:45 +0000 (13:09 +0200)]
baseparse: Don't copy invalid DTS to the PTS

We were checking to make sure the buffer's DTS wouldn't be after its
PTS. However, the check would also trigger when DTS is NONE, which is
e.g. in the case of some broken cameras.

Fixes #470

4 years agotracers: Don't leak temporary GstStructure
Edward Hervey [Wed, 27 Nov 2019 14:47:32 +0000 (15:47 +0100)]
tracers: Don't leak temporary GstStructure

CID: 1455462

4 years agoGstBuffer: size-related optimization
Edward Hervey [Wed, 21 Nov 2018 15:14:58 +0000 (16:14 +0100)]
GstBuffer: size-related optimization

Avoid calling generic function when it's possible to directly
return/get sizes

4 years agoGstBuffer: Inline fast-path for merged memory
Edward Hervey [Wed, 21 Nov 2018 15:13:48 +0000 (16:13 +0100)]
GstBuffer: Inline fast-path for merged memory

4 years agodocs: mention gst_parse_bin_from_description() in gst_parse_launch() docs
Tim-Philipp Müller [Wed, 27 Nov 2019 09:41:36 +0000 (09:41 +0000)]
docs: mention gst_parse_bin_from_description() in gst_parse_launch() docs

4 years agodatetime: Add constructor for timestamps in microseconds
Linus Svensson [Fri, 22 Nov 2019 15:04:20 +0000 (16:04 +0100)]
datetime: Add constructor for timestamps in microseconds

4 years agobaseparse: Make sure PTS >= DTS
Vivia Nikolaidou [Fri, 11 Oct 2019 14:33:42 +0000 (17:33 +0300)]
baseparse: Make sure PTS >= DTS

If, for example, we are accumulating rounding errors from the buffer
duration when calculating the PTS/DTS, it can happen that the buffer
thinks it should be presented before it's decoded. In that case we just
clamp the DTS.

4 years agogstbuffer: update documentation
Stéphane Cerveau [Mon, 18 Nov 2019 00:15:31 +0000 (00:15 +0000)]
gstbuffer: update documentation

remove unclear documentation about GST_BUFFER_FLAG_MARKER

4 years agogst-launch: Disable printing current position by default when stdout is not a tty
Seungha Yang [Tue, 12 Nov 2019 02:24:45 +0000 (11:24 +0900)]
gst-launch: Disable printing current position by default when stdout is not a tty

... and add new option to force-enable printing position even if stdout
is not a tty.

4 years agostructure: add gst_structure_take
Havard Graff [Sun, 3 Nov 2019 11:55:13 +0000 (12:55 +0100)]
structure: add gst_structure_take

(╯°□°)╯︵ ┻━┻

4 years agoharness: fixed race condition on forward pad while forwarding sticky events to sink...
Tulio Beloqui [Tue, 20 Aug 2019 11:57:09 +0000 (13:57 +0200)]
harness: fixed race condition on forward pad while forwarding sticky events to sink harness

Co-authored-by: Camilo Celis <camilo@pexip.com>
Co-authored-by: Havard Graff <hgr@pexip.com>
4 years agohotdoc: Add missing json escaping
Thibault Saunier [Tue, 12 Nov 2019 22:15:34 +0000 (19:15 -0300)]
hotdoc: Add missing json escaping

Fixes https://gitlab.freedesktop.org/gstreamer/gst-docs/issues/50

4 years agoevent: Fix gir warning
Wonchul Lee [Tue, 12 Nov 2019 06:19:28 +0000 (15:19 +0900)]
event: Fix gir warning

It fixes below gir warnings.
../subprojects/gstreamer/gst/gstevent.c:2246: Warning: Gst:
gst_event_new_instant_rate_sync_time: unknown parameter
'rate_multiplier' in documentation comment, should be 'rate'
../subprojects/gstreamer/gst/gstevent.c:2296: Warning: Gst:
gst_event_parse_instant_rate_sync_time: unknown parameter
'rate_multiplier' in documentation comment, should be 'rate'

4 years agogst/parse: define pure-parser depending on bison version
Víctor Manuel Jáquez Leal [Mon, 26 Aug 2019 10:48:28 +0000 (12:48 +0200)]
gst/parse: define pure-parser depending on bison version

After release bison 2.5 the declaration %pure-parser was deprecated
in favor of %define api.pure

Nonetheless, until bison 3.4, the declaration was treated as backward
compatibility, but now bison shows a warning:

  warning: deprecated directive, use ‘%define api.pure’

The patch's approach is to handle both directives according with the
used bison's version, by string replacement at source configuration
stage.

4 years agopad: clear sticky event tag upon stream-start
Nayana Topolsky [Thu, 21 Feb 2019 12:29:31 +0000 (13:29 +0100)]
pad: clear sticky event tag upon stream-start

When playing gapless there were situations when some sticky events
like tags were stuck at some pad and then revived much later.
Therefore it is better to clear them upon stream-start.

Fixes #360

4 years agotaglist: Fix broken empty set character in code
Seungha Yang [Thu, 30 May 2019 13:29:23 +0000 (22:29 +0900)]
taglist: Fix broken empty set character in code

Previous one was not a valid ASCII empty set character.
'tig' and 'git log -p' couldn't represent it as expected.

4 years agogst-launch: Add support printing current position of pipeline
Seungha Yang [Thu, 30 May 2019 11:53:34 +0000 (20:53 +0900)]
gst-launch: Add support printing current position of pipeline

By default, gst-launch will print the current position of pipeline (with duration if available).
To disable it, use "--no-position" option.

4 years agogst-launch: Port to the direct use of GMainLoop
Seungha Yang [Wed, 29 May 2019 11:22:54 +0000 (20:22 +0900)]
gst-launch: Port to the direct use of GMainLoop

... instead of custom event loop.
This can make it easy to use GMainLoop related APIs in code.

4 years agogst-launch: Remove meaningless global variable
Seungha Yang [Wed, 29 May 2019 11:24:06 +0000 (20:24 +0900)]
gst-launch: Remove meaningless global variable

4 years agopipeline: Instant rate change handling
Jan Schmidt [Thu, 7 Feb 2019 12:59:51 +0000 (23:59 +1100)]
pipeline: Instant rate change handling

Implement aggregation of INSTANT_RATE_REQUEST messages and sending of
INSTANT_RATE_SYNC_TIME events.

4 years agobasesink: Add support for instant-rate-change events
Sebastian Dröge [Tue, 15 May 2018 15:42:25 +0000 (18:42 +0300)]
basesink: Add support for instant-rate-change events

Post instant-rate-request message when receiving an instant-rate-change
event, and handle the incoming instant-rate-sync-time events from the
pipeline.

4 years agoevent/message: Add new instant-rate-sync-time event and instant-rate-request message
Sebastian Dröge [Mon, 14 May 2018 20:14:24 +0000 (23:14 +0300)]
event/message: Add new instant-rate-sync-time event and instant-rate-request message

4 years agoevent: Add new GST_EVENT_INSTANT_RATE_CHANGE and GST_SEEK_FLAGS_INSTANT_RATE_CHANGE
Sebastian Dröge [Wed, 9 May 2018 12:28:13 +0000 (15:28 +0300)]
event: Add new GST_EVENT_INSTANT_RATE_CHANGE and GST_SEEK_FLAGS_INSTANT_RATE_CHANGE

A seek with that flag set must be non-flushing, not change the playback
direction and start/stop position. A seek handler will then send the new
GST_EVENT_INSTANT_RATE_CHANGE event downstream for downstream elements
to immediately apply the new playback rate before the new in-band segment
event arrives.

4 years agoelementfactory: add GST_ELEMENT_FACTORY_TYPE_HARDWARE
Víctor Manuel Jáquez Leal [Sat, 2 Nov 2019 14:06:28 +0000 (15:06 +0100)]
elementfactory: add GST_ELEMENT_FACTORY_TYPE_HARDWARE

This new symbol matches with the elements within "Hardware" class.

4 years agoqueue2: Use g_object_notify_by_pspec
Niels De Graef [Thu, 31 Oct 2019 10:06:48 +0000 (11:06 +0100)]
queue2: Use g_object_notify_by_pspec

`g_object_notify()` actually takes a global lock to look up the
`GParamSpec` that corresponds to the given property name. It's not a
huge performance hit, but it's easily avoidable by using the
`_by_pspec()` variant.

4 years agotee: First deactivate the pad and then remove it when releasing pads
Sebastian Dröge [Thu, 24 Oct 2019 22:41:27 +0000 (01:41 +0300)]
tee: First deactivate the pad and then remove it when releasing pads

This reverts a96002bb28c21b30fb9338a4620ad20504c70aa5, which is not
necessary anymore. If we release the pad after removing it then none of
the deactivation code will actually be called because the pad has no
parent anymore, and we require a parent on the pad for deactivation to
happen.

This can then, among other things, cause a streaming thread to be still
stuck in a pad probe because the pad was never flushed, and waiting
there forever because now the pad will actually never be flushed anymore.

4 years agotee: Check for the removed pad flag also in the slow pushing path
Sebastian Dröge [Thu, 24 Oct 2019 22:39:50 +0000 (01:39 +0300)]
tee: Check for the removed pad flag also in the slow pushing path

If a pad is currently being released we don't want to forward the
FLUSHING flow return but instead consider it as NOT_LINKED. FLUSHING
would also cause upstream to be FLUSHING.

This part was missed in a3c4a3201a705eb1934ceeea34d1ca42d4571c07 and
resulted in a different (and wrong) workaround in
a96002bb28c21b30fb9338a4620ad20504c70aa5.

4 years agotee: Lock mutex before reading the removed flag of the pads
Sebastian Dröge [Thu, 24 Oct 2019 22:39:05 +0000 (01:39 +0300)]
tee: Lock mutex before reading the removed flag of the pads

Otherwise we're not guaranteed to read the very latest value that
another thread might've written in there when the pad was released, and
could instead work with an old value.

4 years agobin: Drop need-context messages without source instead of crashing
Sebastian Dröge [Mon, 30 Sep 2019 08:34:51 +0000 (11:34 +0300)]
bin: Drop need-context messages without source instead of crashing

4 years agomeson: build gir even when cross-compiling if introspection was enabled explicitly
Tim-Philipp Müller [Thu, 17 Oct 2019 11:13:35 +0000 (12:13 +0100)]
meson: build gir even when cross-compiling if introspection was enabled explicitly

This can be made to work in certain circumstances when
cross-compiling, so default to not building g-i stuff
when cross-compiling, but allow it if introspection was
enabled explicitly via -Dintrospection=enabled.

Fixes #454 and #381.

4 years agoRemove autotools build system
Tim-Philipp Müller [Sun, 9 Jun 2019 00:34:04 +0000 (01:34 +0100)]
Remove autotools build system

4 years agocore: Avoid usage of deprecated API
Edward Hervey [Thu, 10 Oct 2019 13:53:16 +0000 (15:53 +0200)]
core: Avoid usage of deprecated API

GTimeval and related functions are now deprecated in glib.
Replacement APIs have been present since 2.26

4 years agoCheck buffer size before checking buffer data
Xavier Claessens [Mon, 23 Sep 2019 15:19:07 +0000 (11:19 -0400)]
Check buffer size before checking buffer data

If the expected size is bigger than the actual buffer size, it would
memcmp random memory which could lead to crashes instead of proper error
reporting.

4 years agodataurisrc: Do not include trailing `\0` into buffer
Xavier Claessens [Tue, 24 Sep 2019 14:09:08 +0000 (10:09 -0400)]
dataurisrc: Do not include trailing `\0` into buffer

4 years agoharness: Add gst_harness_pull_until_eos()
Xavier Claessens [Tue, 24 Sep 2019 14:06:51 +0000 (10:06 -0400)]
harness: Add gst_harness_pull_until_eos()

4 years agodocumentation: fix a number of typos
Aaron Boxer [Sun, 6 Oct 2019 15:12:11 +0000 (11:12 -0400)]
documentation: fix a number of typos

4 years agogstdirectcontrolbinding: Fix integer comparison
Jordan Petridis [Fri, 4 Oct 2019 17:01:46 +0000 (20:01 +0300)]
gstdirectcontrolbinding: Fix integer comparison

i is declared as gint but then compared against `n_values` guint
in the for loop below.

4 years agogst: Don't pass miniobjects to GST_DEBUG_OBJECT() and similar macros
Sebastian Dröge [Mon, 30 Sep 2019 08:49:35 +0000 (11:49 +0300)]
gst: Don't pass miniobjects to GST_DEBUG_OBJECT() and similar macros

The argument must be at least a GObject according to the GstLogFunction
definition, and while the default C log function handles miniobjects
just fine this is crashing bindings and user-supplied log functions that
(rightfully) don't expect anything but GObjects.

4 years agogstvalue: use value_nick for serialization
Mathieu Duponchelle [Sat, 7 Sep 2019 02:36:18 +0000 (04:36 +0200)]
gstvalue: use value_nick for serialization

not value_name . This was causing incorrect launch lines to be
displayed by gst-device-monitor, and the deserialization code
below works with nicks.

4 years agodeviceprovider: set the bus to non-flushing before calling klass->start
Mathieu Duponchelle [Mon, 9 Sep 2019 22:28:45 +0000 (00:28 +0200)]
deviceprovider: set the bus to non-flushing before calling klass->start

Not posting DEVICE_ADDED messages while a device provider is being
started makes things awkward for applications, as they have to call
get_devices() after starting the monitor.

This requires redundant code on the application side, and as far as
I understand also could cause race conditions, when a device gets
added between the calls to gst_device_monitor_start() and
gst_device_monitor_get_devices(), causing the application to "see"
the same device twice.

4 years agoelement: Enforce that elements created by gst_element_factory_create/make() are floating
Sebastian Dröge [Thu, 12 Sep 2019 07:09:18 +0000 (10:09 +0300)]
element: Enforce that elements created by gst_element_factory_create/make() are floating

Bindings might have a hard time making sure that the reference is indeed
still floating after returning here.

See https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/444

4 years agodevice: Enforce that elements created by gst_device_create_element() are floating
Sebastian Dröge [Thu, 12 Sep 2019 07:08:39 +0000 (10:08 +0300)]
device: Enforce that elements created by gst_device_create_element() are floating

Bindings might have a hard time making sure that the reference is indeed
still floating after returning here.

See https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/444

4 years agodevice: gst_device_create_element() is `transfer floating`, not `transfer full`
Sebastian Dröge [Thu, 12 Sep 2019 07:03:08 +0000 (10:03 +0300)]
device: gst_device_create_element() is `transfer floating`, not `transfer full`

Fixing the annotation fixes leaking of the created element in all
bindings using GObject-Introspection.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/issues/444

4 years agobase: GstBaseSrc/GstBaseSink::get_caps: add (nullable) to `filter`
Sebastiano Barrera [Tue, 10 Sep 2019 10:31:40 +0000 (12:31 +0200)]
base: GstBaseSrc/GstBaseSink::get_caps: add (nullable) to `filter`

The virtual method named `get_caps` in both `GstBaseSrc` and
`GstBaseSink` has a `filter` parameter which can be `NULL` (the
default implementation in GstBaseSrc already considers the case).
Before this commit, there was no gtk-doc annotation representing this
fact, which caused the corresponding entry in the GIR file to also
miss this fact.

This caused bugs in other places, such inducing the Vala compiler to
introduce a wrongly assert on `(filter != NULL)` in every
implementation of the `get_caps` method implemented in Vala.

4 years agoDon't pass default GLib marshallers for signals
Niels De Graef [Mon, 26 Aug 2019 05:34:30 +0000 (07:34 +0200)]
Don't pass default GLib marshallers for signals

By passing NULL to `g_signal_new` instead of a marshaller, GLib will
actually internally optimize the signal (if the marshaller is available
in GLib itself) by also setting the valist marshaller. This makes the
signal emission a bit more performant than the regular marshalling,
which still needs to box into `GValue` and call libffi in case of a
generic marshaller.

Note that for custom marshallers, one would use
`g_signal_set_va_marshaller()` with the valist marshaller instead.