Use case: we want to block the source pad of a leaky queue and
drop the buffer that causes the block. If we return PROBE_DROP
then the buffer gets dropped, but we get called again. If we
return PROBE_OK we can't easily drop the buffer. If we just
replace the item into the GstPadProbeInfo structure with NULL,
GStreamer will push a NULL buffer to the next element when we
unblock the pad probe. This patch ensures it doesn't do that.
https://bugzilla.gnome.org/show_bug.cgi?id=734342
Also skip gst_pipeline_get_clock() and gst_pipeline_set_clock() from the
bindings as they are confused with gst_element_*_clock().
API: gst_pipeline_get_pipeline_clock()
https://bugzilla.gnome.org/show_bug.cgi?id=744442
Only print interesting caps features, don't
append (memory:SystemMemory) to all caps,
which makes them much more unwieldy and
harder to read. Also use internal function
to get caps features so that our printing
has no side effects on the caps.
https://bugzilla.gnome.org/show_bug.cgi?id=746809
Don't unwrap strings that start but don't finish with a double quote. If a
string is delimited by two quotes we unescape them and any special characters
in the middle (like \" or \\). If the first character or the last character
aren't a quote we assume it's part of an unescaped string.
Moved some deserialize_string unit tests because we don't try to unwrap strings
missing that second quote anymore.
https://bugzilla.gnome.org/show_bug.cgi?id=688625
Do not do any checks for the start/stop in the new
gst_segment_to_running_time_full() method, we can let this be done by
the more capable gst_segment_clip() method. This allows us to remove the
enum of results and only return the sign of the calculated running-time.
We need to put the old clipping checks in the old
gst_segment_to_running_time() still because they work slightly
differently than the _clip methods.
See https://bugzilla.gnome.org/show_bug.cgi?id=740575
Add a clip argument to gst_segment_to_running_time_full() to disable
the checks against the segment boundaries. This makes it possible to
generate an extrapolated running-time for timestamps outside of the
segment.
See https://bugzilla.gnome.org/show_bug.cgi?id=740575
Add a helper method to get a running-time with a little more features
such as detecting if the value was before or after the segment and
negative running-time.
API: gst_segment_to_running_time_full()
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=740575
The position in the segment is relative to the start but the offset
isn't, so subtract the start from the position when setting the offset.
Add unit test for this as well.
gstbuffer.c:522:58: error: implicit conversion from enumeration type 'GstBufferFlags' to
different enumeration type 'GstBufferCopyFlags' [-Werror,-Wenum-conversion]
if (!gst_buffer_copy_into (copy, (GstBuffer *) buffer, flags, 0, -1))
~~~~~~~~~~~~~~~~~~~~ ^~~~~
gstbuffer.c:534:46: error: implicit conversion from enumeration type 'GstBufferCopyFlags' to
different enumeration type 'GstBufferFlags' [-Werror,-Wenum-conversion]
return gst_buffer_copy_with_flags (buffer, GST_BUFFER_COPY_ALL);
~~~~~~~~~~~~~~~~~~~~~~~~~~ ^~~~~~~~~~~~~~~~~~~
./gstbuffer.h:433:31: note: expanded from macro 'GST_BUFFER_COPY_ALL'
...((GstBufferCopyFlags)(GST_BUFFER_COPY_METADATA | GST_BUFFER_COPY_MEMORY))
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Don't stop the pool in set_config(). Instead, let the controlling
element manage it. Most of the time, when an active pool is being
configured is because the caps didn't change.
https://bugzilla.gnome.org/show_bug.cgi?id=745377
A variant of gst_buffer_copy that forces the underlying memory
to be copied.
This is added to avoid adding an extra reference to a GstMemory
that might belong to a bufferpool that is trying to be drained.
The use case is when the buffer copying is done to release the
old buffer and all its resources.
https://bugzilla.gnome.org/show_bug.cgi?id=745287
Shouldn't take the lock while unreferencing messages, because that may cause
more messages to be sent, which will try to take the lock and cause the app to
hang.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=728777
gst_bin_sync_children_states() will iterate over all the elements of a bin and
sync their states with the state of the bin. This is useful when adding many
elements to a bin and would otherwise have to call
gst_element_sync_state_with_parent() on each and every one of them.
https://bugzilla.gnome.org/show_bug.cgi?id=745042
Instead of always shortening the __FILE__ path, even if the
log message is not actually printed, which might happen if
the log level is activated but the category is not, only
shorten the path if we're actually going to output it and
if it looks like it needs shortening. Log handlers had no
guarantee that they would get a name instead of a path
anyway on any architecture, so it shouldn't be a problem.
https://bugzilla.gnome.org/show_bug.cgi?id=745213
A single unlinked pad can make the latency query fail across the
pipeline, which is probably not desirable. Instead, we return a default
anything goes value.
Perhaps we should also be emitting a gst_message_new_latency() when a
PLAYING element has one of its pads linked.
https://bugzilla.gnome.org/show_bug.cgi?id=745197
This reverts commit 1911554cff.
This breaks the functionality of GST_PAD_FLAG_NEED_PARENT, the reason for this
flag is that if a pad is removed from a running element, you don't want
functions (such as chain or event) to be called on the pad without a parent set.
This can happen if you remove a request or sometimes pad from a running element.
I don't see the code that caused this in tsdemux, but if it needs to unset
the flag on remove, it should do it itself and then make sure that the parent
exists in any pad function.
Revive message in dispose handler before we signal the bus thread,
otherwise the bus thread might be woken up and unref the message
before we had a chance to revive it yet.
Async message delivery (where the posting thread gets blocked
until the message has been processed and/or freed) was pretty
much completely broken.
For one, don't use GMutex implementation details to check
whether a mutex has been initialized or not, esp. not
implementation details that don't hold true any more with
newer GLib versions where atomic ops and futexes are used
(spotted by Josep Torras). This led to async message
delivery no longer blocking with newer GLib versions on
Linux.
Secondly, after async delivery don't free mutex/GCond
embedded inside the just-freed message structure.
Use a new (private) mini object flag to signal GstMessage
that the message being freed is part of an async delivery
on the bus so that the dispose handler can keep the message
alive and the bus can free it once it's done cleaning up
stuff.
If an element implements wrongly the URI query and set the uri to NULL and if
the element calls gst_pad_create_stream_id at some point, it will lead to crash
as the uri is not supposed to be NULL in the gst_pad_create_stream_id_internal
function.
https://bugzilla.gnome.org/show_bug.cgi?id=744520
Before we just took the values from the first pad that succeded the query,
now we accumulate the results of every sinkpad properly and return that
result.
Add GST_SEEK_FLAG_TRICKMODE_KEY_UNITS and
GST_SEEK_FLAG_TRICKMODE_NO_AUDIO, and rename GST_SEEK_FLAG_SKIP
to GST_SEEK_FLAG_TRICKMODE (with backwards compat define).
Do the same for the corresponding SEGMENT flags.
https://bugzilla.gnome.org/show_bug.cgi?id=735666
Make host IPs in square brackets store only the IP, i.e. strip the brackets.
Strip leading whitespace characters in URIs.
Fail parsing when host part does not match any valid formats from RFC3986.
https://bugzilla.gnome.org/show_bug.cgi?id=743195
The problem was that the macro was always used with 'ret' as the defaultval
argument.
This would result in the macro eventually expanding to
if (G_UNLIKELY (ret != ret && ret != GST_FLOW_OK))
... ret != ret will always fail, and therefore we'd never call the
following line.
Instead of that, store the previous value locally for comparision
Add domain checks for the input values, and a variable precision
calculation that loops if necessary to ensure we never overflow
accumulators and then silently produce garbage results.
Make the (non-public) linear regression function available for
unit testing by putting it in a separate source file the test
can include. Add a unit test that the new regression function
produces sensible results for several inputs taken from real-world
captures.
gst_clock_add_observation_unapplied() adds a new master/slave clock
observation and runs the regression without activating the new
calibration results.
gst_clock_adjust_with_calibration() uses directly passed calibration
parameters, instead of using the clock's current calibration,
allowing for calculations using pending or old calibration params
Otherwise the struct is going to be copied, which is not very efficient. And
also has the nice side effect that modifications of the struct might be
done in a copy, and we later use the original struct without the changes.
Caused LATENCY queries to always return the initialization values in one of my
tests, instead of the actual values reported by child elements.
API: GST_LEVEL_MAX
By compiling gstreamer (or plugins) with GST_LEVEL_MAX defined, only
the debugging statements at or below that level will be compiled in.
This allows compiling in some debugging (like errors and warnings) which
helps in debugging, but without the full cpu/memory overhead of all debugging
enabled.
Pools are allowed to change the size in order to adapt padding. So
don't check the size. Normally pool will change the size without
failing set_config(), but it they endup changing the size before
the validate method may fail on a false positive.
https://bugzilla.gnome.org/show_bug.cgi?id=741420
Add a function to check if the preset iface implementation is editable and
document this from the implementers perspective.
API: gst_preset_is_editable()
gstdebugutils.c: In function 'gst_debug_bin_to_dot_data':
gstdebugutils.c:683:530: error: 'return' with no value, in function returning non-void [-Werror]
g_return_if_fail (GST_IS_BIN (bin));
TRUE is 1, but every other non-zero value is also considered true. Comparing
for equality with TRUE would only consider 1 but not the others.
Also normalize booleans in a few places.
The documentation states that gst_element_send_event is to "send an event
to an element".
Therefore we *send* upstream events to a source pad and downstream events
to a sink pad
Previously, dropping a query from a pad probe would deem the
query succeeded, and the caller might then assume the query's
results are valid, and thus dereference an invalid object
such as a GstCaps.
We now assume dropped queries did not succeed. Dropped events
and buffers are still deemed a success.
Added back after previous revert, as it's been double checked.
https://bugzilla.gnome.org/show_bug.cgi?id=740003
Previously, dropping a query from a pad probe would deem the
query succeeded, and the caller might then assume the query's
results are valid, and thus dereference an invalid object
such as a GstCaps.
We now assume dropped queries did not succeed. Dropped events
and buffers are still deemed a success.
If a task thread is calling pause on it self and the
controlling/"main" thread stops the task, it could end in a race
where gst_task_func loops and then checks for paused after the
controlling thread just changed the task state to stopped.
Hence the task would actually call func again even though it was
both paused and stopped.
https://bugzilla.gnome.org/show_bug.cgi?id=740001
It's architecture dependent and should not be placed into the include
directory as the assumption is that all those headers are architecture
independent.
https://bugzilla.gnome.org/show_bug.cgi?id=739767
The GST_INFO ("initialized GStreamer succesfully") is currently at the end of
gst_init_check which isn't guaranteed to be run since GStreamer can be
initialized by using init_pre and init_post directly from GOptionContext like
gst-launch does. Ensure this message is displayed by moving it to init_post.
Use some macros to make our value functions setup a bit
tidier, and micro-optimise a few reallocs by setting an
initial size for the global type arrays.
Remove a useless assert (we just instantiated this type). Drop the free'ing of
the extension array. As we just created the instance this is always NULL.
Running two threads, one executing the timer and one unscheduling it, the
unscheduled status set by the second thread is sometimes overwritten by the
first one.
https://bugzilla.gnome.org/show_bug.cgi?id=737999
These are actually not true.
gsturi.c: In function '_gst_uri_string_to_table.constprop':
gsturi.c:1316:27: error: 'pct_kv_sep' may be used uninitialized in this function [-Werror=maybe-uninitialized]
for (next_sep = strcasestr (value, pct_kv_sep); next_sep;
^
gsturi.c:1283:24: error: 'pct_part_sep' may be used uninitialized in this function [-Werror=maybe-uninitialized]
next_sep = strcasestr (next_sep + 1, pct_part_sep)) {
^
We're not actually doing anything differently anywhere when
we detect that we're running under valgrind, so let's not
print that confusing message that makes people wonder how
they can switch it off so they can valgrind the normal
code paths. Seeing that we're not doing that nor have done
so in the last 10 years we might just as well remove the
entire check actually.
I was confused by the existence of `gst_buffer_get_meta` as it suggested
to me that you should only attach one of any type of GstMeta to a buffer.
It's perfectly fine to attach multiple from a single API so I'm
documenting that here.
https://bugzilla.gnome.org/show_bug.cgi?id=737129
Use user_data to pass the log_file handle to the logger-function.
If one wants to change the log target (e.g. GST_DEBUG_FILE), simply call
gst_debug_remove_log_function() and re-add the handler with the new log-target
using gst_debug_add_log_function ().
Inactive pads should at all times have the flushing flag set. This means
that when we get a flush-stop on an inactive pad we must ignore it.
On sinkpads, make this more explicit. We used to not clear the flush
flag but remove the events and then return an error because the flushing
flag was set. Now just simply refuse the event without doing anything.
On srcpads, check that we are trying to push a flush-stop event and
refuse it. We would allow this and mark the srcpad as non-flushing
anymore.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=735357
Don't rely on g_source_remove() because it operates on the main
context. If a signal watch was added to a new thread-default context
g_source_remove() would have no effect. So simply use
g_source_destroy() to avoid this problem.
Additionally the source_id was removed from GstBusPrivate because it
was redundant with the signal watch GSource also stored in that
structure.
https://bugzilla.gnome.org/show_bug.cgi?id=734716
This function is not really pad or slow for the common case of requesting a
pad with the name of the template. It is only slower if you to name your pads
directly instead of letting the element handle it.
Also there's no reason to deprecate it in favor of a more complicated function
for the common case.
Previously gst_element_link_pads_full() forgot to unreference or release
request pads in several error cases. Also comments were added mentioning
why releasing is not necessary in some places.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=733119
default_alloc_buffer() calls gst_buffer_new_allocate() but does not check for
failed allocation.
This patch makes default_alloc_buffer() return an error (GST_FLOW_ERROR) if
buffer allocation fails.
https://bugzilla.gnome.org/show_bug.cgi?id=733974
SetEvent() seems to not call SetLastError(0) internally, so checking last
error after calling SetEvent() may return the error from an earlier W32 API
call. Fix this by calling SetlastError(0) explicitly.
Currently WAKE_EVENT() code is cramped into a macro and doesn't look to be
entirely correct. Particularly, it does not check the return value of
SetEvent(), only the thread-local W32 error value. It is likely that SetEvent()
actually just returns non-zero value, but the code mistakenly thinks that the
call has failed, because GetLastError() seems to indicate so.
https://bugzilla.gnome.org/show_bug.cgi?id=733805
We can now create and free a buffer list with one slice alloc/free
call in most cases, instead of one slice alloc/free for the list,
one slice alloc/free for the GArray, and one malloc/free for the
GArray array. In practice we know the max size of our buffer list
from the start, so can avoid reallocs.
https://bugzilla.gnome.org/show_bug.cgi?id=732284
When serializing GstStructures from events in GDP it will add a taglist
as a GstStructure field, having the compare function allows comparison of
GstStructures to check if the serialized/deserialized version matches the
original one, among other cases.
https://bugzilla.gnome.org/show_bug.cgi?id=733131
* GstGlobalDeviceMonitor was renamed to GstDeviceMonitor
* Expand GST_MESSAGE_DEVICE to the full enum value names
* Correct the incorrect references to the GstDeviceProvider interfaces
* Describe caps arguments for gstcheck interface
* Add missing docs for GstNetAddressMeta and its add function
* Add docs for toc helper macros
* Avoid refering to GstValueList type as done elsewhere
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732786
GST_MESSAGE_ANY was considered a long by pygobject and gjs, and thus
couldn't be used in gst_bus_poll() and similar APIs as they expect an
int-typed enum.
Just use 0xffffffff instead for now.
https://bugzilla.gnome.org/show_bug.cgi?id=732633
Support for (nullable) was added to G-I at the same time as nullable
return values. Previous versions of G-I will not mark return values as
nullable, even when an (allow-none) annotation is present, so it is
not necessary to add (allow-none) annotations for compatibility with
older versions of G-I.
https://bugzilla.gnome.org/show_bug.cgi?id=730957
Boxed types can't be derived from, and we don't support
deriving from our special fundamental types (the code
checks for GType equality in most places.
Currently we leak the internal representation of them as two GValues that
contain a fraction. Without this we could store fraction ranges as
data[0] = (min_n << 32) | (min_d)
data[1] = (max_n << 32) | (max_d)
and wouldn't require an additional allocation per range.
Otherwise negative values will sets all of the 64 bits due to two's
complement's definition of negative values.
Also add a test for negative int ranges.
Both gst_value_intersect and gst_value_subtract will call
gst_value_compare if one of their arguments isn't a list.
gst_value_compare will then re-do a check to see if one of
the arguments is a list (for the special case of comparing a unitary
value with a list of length 1).
The problem is that the various G_VALUE_HOLDS represent an expensive
amount of calling gst_value_compare (almost half of it) to see if
the provided arguments are list. These checks can be done without
when we know that the arguments aren't lists.
* Create a new "nolist" gst_value_compare which avoids that special
case comparision
Benchmarks:
valgrind/callgrind: average speedup in instruction calls for
gst_value_intersect and gst_value_subtract is around 56% (Makes 63%
of the calls it used to take previously)
tests/benchmarks/capsnego: With default settings (depth 4, children 3
607 elements), time taken for transition from READY to PAUSED:
Before : 00.391519153
After : 00.220397492
56% of the time previously used, +77% speedup
https://bugzilla.gnome.org/show_bug.cgi?id=731756
First handle all miniobjects before we attempt to dereference the first
field pointer and look at the GType. With the recent glib change to
speed up G_IS_OBJECT, this causes crashes on miniobjects otherwise.
