We now have GstTestClock-based tests that validate the same logic,
without inducing spurious timing failures / overly relying on sleeps.
Fixes: #346Fixes: #347Fixes: #348
Co-authored by: Thibault Saunier <tsaunier@igalia.com>
Scale the number of threads used in the stress tests according to
the number of cores/cpus. We want some contention, but we also
don't want too much contention, as some operating systems are
better at handling 100 threads running wild on a single core
than others.
Fixes a race where an entry is set to BUSY in
gst_system_clock_id_wait_jitter() and is UNSCHEDULED before
gst_system_clock_id_wait_jitter_unlocked() starts processing it. The
wakeup added by gst_system_clock_id_unschedule() must be cleaned up.
Two stress tests are added. One test that triggers the specific issue
described above. The second stresses the code path where a wait is
rescheduled because the poll returned early.
https://bugzilla.gnome.org/show_bug.cgi?id=761586
when an entry being waited on in the async thread is unscheduled, clear the
wakeup queue so we can continue waiting on other entries.
When an entry being waited on in the async thread is unlocked because an earlier
entry was added to the list, set the entry to OK again. This makes sure that
only the entries being waited on have the BUSY flag set and wake up the timer
poll when they are unscheduled.
Add some cleanups to the system clock tests, to free all the memory and
unschedule/unref all clock IDs we allocate.
Use a mutex in one test to avoid potential threading problems on multicore
machines.
Original commit message from CVS:
* gst/gst.c:
_gst_trace_on is already provided by gsttrace.h, no need to declare
it ourselves.
* docs/libs/gstreamer-libs-sections.txt:
Add 'buffers', 'check_cond' and 'check_mutex' from libgstcheck
and remove strange tcase_add_test which is outputting a warning.
* libs/gst/check/gstcheck.c:
* libs/gst/check/gstcheck.h:
Properly declare 'buffers', 'check_cond', 'check_mutex' extern
and define them in gstcheck.c instead of having every .c file whcih
includes gstcheck.h be defining its own copy and relying on symbol
interposing to marry them all, which doesn't work on Solaris.
* tests/check/elements/identity.c: (GST_START_TEST):
Don't define 'buffers' locally, it comes from libgstcheck.
* tests/check/generic/sinks.c: (send_buffer):
Fix type of variable (GstFlowReturn, not GstStateChangeReturn)
* tests/check/gst/gststructure.c: (GST_START_TEST):
* tests/check/gst/gstsystemclock.c: (GST_START_TEST):
* tests/check/gst/gstutils.c: (GST_START_TEST):
* tests/check/gst/gstvalue.c: (GST_START_TEST):
Add a bunch of casts to make various constants fit the types
they're being assigned to.
Original commit message from CVS:
* docs/gst/gstreamer-sections.txt:
* gst/gstclock.h:
* tests/check/gst/gstsystemclock.c: (GST_START_TEST),
(gst_systemclock_suite):
Start merging in the easy bits of #361155, the monotonic clock patch.
This one adds a few handy macros with docs and a testsuite.
Original commit message from CVS:
* tests/check/gst/gstsystemclock.c:
Cleanup the test a little (use gst-logging and not g_message). Improve
test to check if a wait reached the target.
Original commit message from CVS:
* tests/check/gst/gstsystemclock.c: (store_callback),
(GST_START_TEST):
Improve test_async_order to also work when both timers are already
expired when we get scheduled to check it.
Original commit message from CVS:
* tests/check/gst/gstsystemclock.c: (GST_START_TEST),
(mixed_thread), (mixed_async_cb), (gst_systemclock_suite):
Unref some more to make valgrind happy.
Original commit message from CVS:
* gst/gstsystemclock.c: (gst_system_clock_id_wait_jitter_unlocked),
(gst_system_clock_id_wait_jitter),
(gst_system_clock_id_wait_async), (gst_system_clock_id_unschedule):
Fix anoying regression that survived a few releases. When adding an
async entry while blocking on a sync entry, the sync entry will unblock
but still be busy, so it should continue to wait instead of returning
_BUSY to the app.
Add some comments here and there.
* tests/check/gst/gstsystemclock.c: (mixed_thread),
(mixed_async_cb), (GST_START_TEST), (gst_systemclock_suite):
Add testcase for this.
Original commit message from CVS:
Patch by: René Stadler <mail at renestadler dot de>
* gst/gstclock.c: (gst_clock_id_wait):
Make period ids add the interval to the origial requested time instead
of the possibly updated time which can be wrong when there are multiple
waiters for the same id. Fixes#382592.
* gst/gstsystemclock.c: (gst_system_clock_async_thread),
(gst_system_clock_id_wait_jitter_unlocked),
(gst_system_clock_id_wait_jitter):
Fix restart in the async notify thread when an async entry is added to
the front of the list. Fixes#381492.
* tests/check/gst/gstsystemclock.c: (store_callback),
(notify_callback), (GST_START_TEST), (gst_systemclock_suite):
Added test for multiple async waits.
Added test for async wait order.
Original commit message from CVS:
* gst/gstclock.c: (gst_clock_entry_new),
(gst_clock_id_compare_func), (gst_clock_id_wait),
(gst_clock_id_wait_async), (gst_clock_id_unschedule),
(gst_clock_init), (gst_clock_get_internal_time),
(gst_clock_set_master), (do_linear_regression),
(gst_clock_add_observation), (gst_clock_set_property):
* gst/gstclock.h:
Review docs.
Small cleanups.
Fix a possible segfault when the window-size is made smaller.
Calculate jitter before performing the clock wait. Ideally
the clock implementation should calculate jitter but we need
API breakage for that.
* gst/gstsystemclock.c: (gst_system_clock_init):
Docs review.
* libs/gst/base/gstbasesink.c: (gst_base_sink_do_sync):
Remove leftover else
* tests/check/gst/gstsystemclock.c: (GST_START_TEST),
(gst_systemclock_suite):
Added check to test GST_CLOCK_DIFF.
Original commit message from CVS:
Added GstBin test.
Added GstSystemClock test.
Implemented clock distribution code in GstBin.
Implemented iterate sinks method for future use.
Rearranged gstelement.h
Fix GstIterator comparison bug.
Moved some code to GstPipeline, mostly clocking related.
