Original commit message from CVS:
2007-12-18 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch-marshal.list:
* gst/switch/gstswitch.h (struct _GstStreamSelectorClass):
* gst/switch/gstswitch.c (enum, gst_selector_pad_class_init)
(gst_selector_pad_get_property)
(gst_selector_pad_get_running_time)
(gst_stream_selector_class_init, gst_segment_get_timestamp)
(gst_segment_set_stop, gst_segment_set_start)
(gst_stream_selector_set_active_pad, gst_stream_selector_block)
(gst_stream_selector_push_pending_stop)
(gst_stream_selector_switch): Change so that the signals and
properties deal in running time, not buffer time. Document the
signals more. Change uint64 in API to int64, to reflect what's in
GstSegment.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.c (gst_selector_pad_chain): Return OK when
a buffer is ignored, not NOT_LINKED. No sense in making a source
element error out; at least fdsrc considers NOT_LINKED to be a
fatal error. Patch 11/12. There is no patch 12/12. Foo.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch-marshal.list:
* gst/switch/gstswitch.h (struct _GstStreamSelectorClass):
* gst/switch/gstswitch.c (gst_stream_selector_class_init)
(gst_stream_selector_block): Make the block() signal return the
last stop time of the active pad. Patch 10/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.c (gst_selector_pad_get_property)
(gst_selector_pad_class_init, gst_stream_selector_class_init)
(gst_stream_selector_get_property): Expose 'last-stop-time' as a
pad property, not an element property.
(gst_selector_pad_chain): Mark the last_stop time as timestamp +
duration, not timestamp. Patch 9/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.c (gst_stream_selector_change_state)
(gst_stream_selector_block, gst_stream_selector_switch): Use the
cond mechanism instead of blocked pads. Patch 8/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.h (struct _GstStreamSelector):
* gst/switch/gstswitch.c (gst_stream_selector_wait)
(gst_selector_pad_chain, gst_stream_selector_init)
(gst_stream_selector_dispose): Add infrastructure for new blocking
mechanism that does not use gst_pad_set_blocked, which does not
work on sink pads. Patch 7/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.h (struct _GstStreamSelector): Add some
state variables.
* gst/switch/gstswitch.c (gst_stream_selector_push_pending_stop)
(gst_selector_pad_chain): Push any pending stop event.
(gst_stream_selector_set_active_pad)
(gst_stream_selector_set_property): Factor out setting the active
pad to a function. Close the segment of the previous active pad if
told to do so via a stop_time != GST_CLOCK_TIME_NONE.
(gst_stream_selector_switch): Implement switch vmethod. Patch 5/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.c (gst_stream_selector_block): Implement
the block() signal. This implementation will be replaced in future
patches, however. Patch 4/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.h (struct _GstStreamSelectorClass):
* gst/switch/gstswitch.c (gst_stream_selector_class_init): Add
`block' and `switch' signals.
* gst/switch/Makefile.am:
* gst/switch/gstswitch-marshal.list: Add foo to generate a
marshaller for the `switch' signal. Patch 2/12.
Original commit message from CVS:
2007-12-17 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.h:
* gst/switch/gstswitch.c: Replace with files from
gststreamselector.[ch], registered as the "switch" plugin, with
"GstSwitch" types. Patch 1/12.
Original commit message from CVS:
2007-12-05 Andy Wingo <wingo@pobox.com>
* gst/switch/gstswitch.c (gst_switch_set_property): Don't push
buffers from app thread when unsetting `queue-buffers', it's
dangerous and the chain function will do it for us anyway.
Original commit message from CVS:
* gst/switch/gstswitch.c (gst_switch_chain, gst_switch_set_property):
If all information is known at time of setting start-time
property, send new segments then.
Original commit message from CVS:
* examples/switch/switcher.c (main):
* gst/switch/gstswitch.c (gst_switch_chain):
Make switch more reliable and also not lock up when
sink pad caps change.
Original commit message from CVS:
* examples/switch/switcher.c (my_bus_callback, switch_timer,
last_message_received, main):
* gst/switch/gstswitch.c (gst_switch_release_pad,
gst_switch_request_new_pad, gst_switch_chain, gst_switch_event,
gst_switch_set_property, gst_switch_get_property,
gst_switch_get_linked_pad, gst_switch_getcaps,
gst_switch_bufferalloc, gst_switch_dispose, gst_switch_init):
* gst/switch/gstswitch.h (switch_mutex, GST_SWITCH_LOCK,
GST_SWITCH_UNLOCK):
Add an extra lock to protect against certain variables instead of
using the object lock. Fix case where caps are different in the
sink pads causes deadlock. Update example to use different caps
on each sink pad.
Original commit message from CVS:
* ext/libmms/gstmms.h:
No reason to use gpointers instead of typed pointes here as far as I
can see.
* ext/mythtv/gstmythtvsrc.c:
* ext/neon/gstneonhttpsrc.c:
* gst/switch/gstswitch.c:
Don't use gtk-doc magic markers for things that aren't meant to be
parsed by gtk-doc. Makes gtk-doc complain a bit less.
Original commit message from CVS:
* gst/switch/gstswitch.c (ARG_ACTIVE_SOURCE, ARG_STOP_VALUE,
ARG_LAST_TS, parent_class, gst_switch_release_pad,
gst_switch_request_new_pad, gst_switch_chain, gst_switch_event,
gst_switch_set_property, gst_switch_get_property, gst_switch_getcaps,
gst_switch_dispose, gst_switch_init, gst_switch_class_init):
* gst/switch/gstswitch.h (previous_sinkpad, nb_sinkpads, stop_value,
current_start, last_ts):
Allow application to provide a stop timestamp, so a new segment
update can be sent before switching.
Original commit message from CVS:
* gst/switch/gstswitch.c: (gst_switch_release_pad),
(gst_switch_request_new_pad), (gst_switch_poll_sinkpads),
(gst_switch_loop), (gst_switch_get_type):
whoever that was: DO NOT IMPORT PRIVATE SYMBOLS THAT ARE NOT IN
HEADERS. Had to be said.
Original commit message from CVS:
2004-02-15 Julien MOUTTE <julien@moutte.net>
* gst/switch/gstswitch.c: (gst_switch_loop): More fixes for
correct data refcounting.
Original commit message from CVS:
2004-02-15 Julien MOUTTE <julien@moutte.net>
* gst/switch/gstswitch.c: (gst_switch_change_state),
(gst_switch_class_init): Cleaning the sinkpads correctly on state
change, mostly the EOS flag.
Original commit message from CVS:
2004-02-14 Julien MOUTTE <julien@moutte.net>
* gst-libs/gst/play/play.c: (gst_play_connect_visualization): Disable
visualization until i find a way to fix switch correctly.
* gst-libs/gst/riff/riff-read.c: (gst_riff_peek_head): Fix a bug when
EOS arrives.
* gst/switch/gstswitch.c: (gst_switch_release_pad),
(gst_switch_request_new_pad), (gst_switch_poll_sinkpads),
(gst_switch_loop), (gst_switch_dispose), (gst_switch_class_init):
Reworked switch to get a more correct behaviour with events and refing
of data stored in sinkpads.
* gst/switch/gstswitch.h: Adding an eos flag for every sinkpad so that
we don't pull from a pad in EOS.
Original commit message from CVS:
2004-01-25 Julien MOUTTE <julien@moutte.net>
* gst-libs/gst/play/gstplay.c: (gst_play_pipeline_setup),
(gst_play_identity_handoff), (gst_play_set_location),
(gst_play_set_visualization), (gst_play_connect_visualization): Another
try in visualization implementation. Still have an issue with switch
blocking when pulling from video_queue and only audio comes out of
spider.
* gst/switch/gstswitch.c: (gst_switch_release_pad),
(gst_switch_poll_sinkpads), (gst_switch_class_init): Implementing pad
release method. And check if the pad is usable before pulling.
Original commit message from CVS:
2004-01-20 Julien MOUTTE <julien@moutte.net>
* gst/switch/gstswitch.c: (gst_switch_request_new_pad),
(gst_switch_init): Fixed switch element : proxying link and setting
caps from src to sink on request.
Original commit message from CVS:
Adding a new plugin: switch.
It takes N input and only has 1 output. You can "switch" the forwarded input through properties ("nb_sources", "active_source") and i will probably add tuner interface support soon.
It should be able to handle any kind of data passing through it.
It is still a work in progress don't consider it usable for production yet.
