Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
But unit tests time out after pad gets blocked,
so still something missing somewhere.
|
|
GstProbeType, GstProbeReturn and GstActivateMode -> GstPad*
|
|
... and not a pad block
|
|
* Use GST_UNSTABLE_API
* Remove unused variables
* Properly disable tests that require videomixer if not present
* Clarify failures with ERROR messages
* Use new probe API
Compiles ... but most tests fail due to caps issues (which is also present
in 0.10)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Still needs more checks though
|
|
|
|
|
|
|
|
|
|
|
|
Should lower the risks of timing out
|
|
|
|
So the output tells us which test is actually running. Also use
GST_CHECK_MAIN() while we're at it.
|
|
Enables some more tests / code paths in an uninstalled setup. Also
remove duplicate GST_PLUGIN_DIRS.
https://bugzilla.gnome.org/show_bug.cgi?id=628943
|
|
not available
Fixes 'make check' in uninstalled setup where it currently can't
find the -good plugins.
https://bugzilla.gnome.org/show_bug.cgi?id=628943
|
|
Fix a race between no_more_pads_object_cb and compare_relink_single_node.
|
|
Add a test for a race between no_more_objects_cb and compare_relink_single_node
in gnlcomposition.
|
|
Make gnl_composition_remove_object return immediately FALSE when removing an
object not contained in the composition.
|
|
When removing an object from a composition, there might still be some
internal refcounting still present since it's not in NULL.
Also, use ASYNC_DONE message to detect for pre-roll
|
|
We're not viewing them anyway, so might as well take the cheapest
to use.
|
|
|
|
|
|
|
|
|
|
This should lower the amount of time required to run the tests
|
|
|
|
|
|
|
|
We already depend on a much recent version.
|
|
The reason behind this is that updating the lists and those properties
are decently simple/fast to do it at that time. The internal pipeline
will still not be updated if update == False.
|
|
When update is set to False, the internal pipeline is never updated,
nor are the stop/duration properties of the composition.
This allows for massive modifications with a much lower overhead.
|
|
We depend on a higher version already.
|
|
|
|
outside the active segment.
|
|
fixed by #537361
Original commit message from CVS:
* tests/check/gnloperation.c: (GST_START_TEST):
Adapt test to 'fixed' behaviour of videomixer fixed by #537361
|
|
messages so that we only switch segment when all...
Original commit message from CVS:
* gnl/gnlcomposition.c: (add_message), (flush_messages),
(replace_message), (has_message), (dump_messages),
(gnl_composition_handle_message), (no_more_pads_object_cb),
(update_pipeline):
Properly accumulate segment_start/segment_done messages so that we only
switch segment when all segment_done have been emitted.
Inspired by the logic in gstbin.
* gnl/gnloperation.c: (get_unlinked_sink_ghost_pad),
(remove_sink_pad), (synchronize_sinks),
(gnl_operation_request_new_pad), (gnl_operation_release_pad):
Methods to properly remove request pads.
* tests/check/common.h:
Remove a leak
|
|
Original commit message from CVS:
* tests/check/common.h:
* tests/check/complex.c: (GST_START_TEST):
* tests/check/gnlsource.c: (GST_START_TEST):
* tests/check/simple.c: (GST_START_TEST):
Round of gst-indent and more utilities
* tests/check/gnloperation.c: (GST_START_TEST), (gnonlin_suite):
New test for multiple-input-pad operations. Needs more work
|
|
specific media-start and media-duration.
Original commit message from CVS:
* tests/check/common.h:
Add a bigger function for creating a video source with specific
media-start and media-duration.
* tests/check/simple.c: (GST_START_TEST):
Let's check with non-obvious media-start/duration sources.
|
|
elements are being removed.
Original commit message from CVS:
* gnl/gnlcomposition.c: (unblock_child_pads), (unblock_childs),
(unlock_child_state), (lock_child_state),
(gnl_composition_change_state), (object_pad_removed),
(gnl_composition_add_object), (gnl_composition_remove_object):
Unblock child source pads when the pads OR the elements are being
removed.
Also unblock those pads when going to READY/NULL.
Lock the state of gnlobjects being added, else it screws up state
changes of the composition.
Properly unref items when using GstIterators.
* gnl/gnloperation.c: (gnl_operation_class_init),
(gnl_operation_finalize), (get_unused_static_sink_pad),
(add_sink_pad):
memleak fixes.
* tests/check/common.h:
* tests/check/simple.c: (GST_START_TEST):
memleak fixes.
|
|
adding a default source.
Original commit message from CVS:
* gnl/gnlcomposition.c: (gnl_composition_add_object):
Recalculate start/stop/duration of composition when adding a default
source.
* tests/check/complex.c: (GST_START_TEST):
Modify test so it also test for the case where the composition starts
with a default source.
|
|
calculating the highest priority. This f...
Original commit message from CVS:
* gnl/gnlcomposition.c: (convert_list_to_tree):
Take into account the priority of default sources when calculating the
highest priority. This fixes #362399
* tests/check/gnloperation.c: (GST_START_TEST), (gnonlin_suite):
Unit test for fixed behaviour.
|
|
now look at all objects under the highest ...
Original commit message from CVS:
* gnl/gnlcomposition.c: (gnl_composition_ghost_pad_set_target),
(next_stop_in_region_above_priority), (convert_list_to_tree),
(get_stack_list), (get_clean_toplevel_stack),
(gnl_composition_remove_object):
Better check for the 'next stop' of a stack. It will now look at all
objects under the highest priority (and not the lowest as previously).
Fixes #356380.
Set new pads as active.
Add some debug.
* gnl/gnlobject.c: (gnl_object_ghost_pad_full):
* gnl/gnloperation.c: (gnl_operation_add_element), (add_sink_pad):
Activate new pads.
* tests/check/gnloperation.c: (GST_START_TEST), (gnonlin_suite):
Added new test (test_pyramid_operations) to test issues fixed for
#356380.
* tests/check/simple.c: (GST_START_TEST):
Unref sinkpads before testing some issues.
|