[Bug 667352] [0.11] misc core API/bug comments/nitpicks

GStreamer (bugzilla.gnome.org) bugzilla at gnome.org
Thu Mar 15 02:54:01 PDT 2012


https://bugzilla.gnome.org/show_bug.cgi?id=667352
  GStreamer | gstreamer (core) | 0.11.x

Mark Nauwelaerts <mnauw> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |

--- Comment #5 from Mark Nauwelaerts <mnauw at users.sourceforge.net> 2012-03-15 09:53:55 UTC ---
Indeed, previously stated are resolved now, but there are some more where they
came from, so reopening (rather than a new bug with a same title/purpose,
though that is of course also possible if preferable).

Again, some may already be known and/or mentioned somewhere more or less, but
anyway ...:

* tag event is sticky event, presumably with the intent of persisting these and
delivering to downstream whenever possible.  It also allows for multiple ones,
which are distinguished by name.  However, afaik, each tag event has the same
name, which still only allows one event.  So things will not turn out as
persistent as expected/hoped, unless each element really sends all it has to in
one tag event (and also with subsequent updates).  Many (if not all) do not do
so, and might be somewhat cumbersome to arrange it so (and undocumented as
well).

* a GstVideoFrame holds a pointer to the buffer it is mapping which is not
refcounted.  That may be so intended and may work out well, but seems
potentially tricky.

* afaics gst_buffer_pool_config_get returns (a.o.) caps with an extra ref,
which seems not to be taken into account presently (documented or in code
anywhere).  It looks like this is not intended (and then needs some
adjustment).  But even if intended, a construction using gst_structure_id_get
with multiple fields is tricky since if it fails it is not known whether or not
it has succeeded enough for an extra ref to have already been taken.

* audio base classes in 0.10 use pbutils for auto-adding some codec tag info,
which is presently disabled in 0.11 (due to circular linkage maybe?), but needs
settling

* similarly, plc (packet loss concealment) used newsegment updates and is
presently not handled (maybe the new gap event is supposed to handle this?)

* w.r.t. some new events (stream-start, gap, stream-config), these appear
largely not used yet at present, and the semantics/documentation are not always
clear (e.g. how do caps/stream-config compete or compare, or stream-headers vs
codec-data for that matter)

-- 
Configure bugmail: https://bugzilla.gnome.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.


More information about the gstreamer-bugs mailing list