Original commit message from CVS:
* gst/playback/gstdecodebin2.c: (gst_decode_bin_finalize):
Don't leak mutex.
* tests/check/elements/playbin.c:
(test_sink_usage_video_only_stream),
(test_suburi_error_unknowntype), (test_suburi_error_invalidfile),
(test_suburi_error_wrongproto), (test_missing_urisource_handler),
(test_missing_suburisource_handler),
(test_missing_primary_decoder), (playbin_suite):
Run all tests once with decodebin and once with decodebin2.
One test does not pass yet with decodebin2.
Original commit message from CVS:
* gst/playback/gstplaybasebin.c: (setup_subtitle),
(gen_source_element), (gst_play_base_bin_change_state):
Attempt at a better error message in case we don't have the required
URI handler installed; post missing-plugin message also when we're
missing an URI handler for the subtitle URI; clean up properly also
when an error occurs and we never made it to PAUSED state.
* tests/check/elements/playbin.c: (GST_START_TEST),
(playbin_suite):
Check that we're also getting a missing-plugin messsage for a
missing subtitle URI handler (and clean up properly).
Original commit message from CVS:
* gst/playback/Makefile.am:
* gst/playback/gstplaybasebin.c: (string_arr_has_str),
(unknown_type), (setup_subtitle), (gen_source_element):
* gst/playback/gstplaybin.c: (plugin_init):
Post missing-plugin messages on the bus for missing sources and
missing decoders/demuxers/depayloaders; fix error code used when
we're missing an URI handler source; for media types that we are not
handling on purpose at the moment, don't print "don't know how to
handle xyz" messages to the terminal or post missing-plugin
messages on the bus.
* tests/check/elements/playbin.c: (create_playbin),
(GST_START_TEST), (gst_codec_src_uri_get_type),
(gst_codec_src_uri_get_protocols), (gst_codec_src_uri_get_uri),
(gst_codec_src_uri_set_uri), (gst_codec_src_uri_handler_init),
(gst_codec_src_init_type), (gst_codec_src_base_init),
(gst_codec_src_create), (gst_codec_src_class_init),
(gst_codec_src_init), (plugin_init), (playbin_suite):
Add some tests for the missing-plugin stuff.
Original commit message from CVS:
* gst/subparse/gstsubparse.c: (parse_mdvdsub):
* gst/subparse/gstsubparse.h:
Remove spurious 1000 subtrahend when calculating the timestamp from
the frame number and the frame rate . Also, use the frames/second
value specified in the first line of the file, if one is specified
there. Should fix#357503.
* tests/check/elements/subparse.c: (do_test),
(test_tmplayer_do_test), (test_microdvd_do_test), (GST_START_TEST),
(subparse_suite):
Add some basic unit tests for the microdvd subtitle format.
Original commit message from CVS:
* tests/check/elements/gdpdepay.c: (cleanup_gdpdepay),
(setup_gdpdepay_streamheader):
* tests/check/elements/gdppay.c: (cleanup_gdppay),
(setup_gdppay_streamheader):
Fix the dp tests, but activating the pads for the streamheader tests
too and cleaning up conditionaly
Original commit message from CVS:
* gst/playback/gstdecodebin.c: (type_found):
Special-case the text/plain media type: we only want to recognise it
as a 'raw' decoded media type if it comes from a demuxer or subtitle
parser, but not if the entire stream is of text/plain type. If the
entire stream is text/plain, we should just error out.
This fixes playback of audio files with lyrics in totem. Totem can't
distinguish between text files and subtitle files and passes any
.txt file with the same basename as the main file to playbin as
suburi, and playbin will then throw a 'subtitle found, but no video
stream' error, which isn't entirely helpful. See #380342.
Also, with this change we'll show a slightly more correct error
message in case totem passes a playlist file to us (although a
custom error message wording instead of the default text would
probably not be a bad idea either).
Same problem also needs to be fixed for playbin+decodebin2.
* tests/check/Makefile.am:
* tests/check/elements/decodebin.c: (src_handoff_cb),
(decodebin_new_decoded_pad_cb), (GST_START_TEST),
(decodebin_suite):
Add simple unit test for decodebin for the above.
Original commit message from CVS:
* ext/pango/gsttextoverlay.c: (gst_text_overlay_init),
(gst_text_overlay_text_pad_unlink), (gst_text_overlay_text_event),
(gst_text_overlay_video_event), (gst_text_overlay_pop_text),
(gst_text_overlay_text_chain), (gst_text_overlay_video_chain),
(gst_text_overlay_change_state):
* ext/pango/gsttextoverlay.h:
Some textoverlay fixes: for one, in the video chain function,
actually wait for a text buffer to come in if there is none at the
moment and there should be one; also, deal more gracefully with
incoming buffers that do not have a timestamp or duration; discard
text buffer when not needed any longer. Fixes#341681.
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/textoverlay.c:
(notgst_check_setup_src_pad2), (notgst_check_teardown_src_pad2),
(setup_textoverlay), (buffer_is_all_black), (create_black_buffer),
(create_text_buffer), (cleanup_textoverlay), (GST_START_TEST),
(test_video_waits_for_text_send_text_newsegment_thread),
(test_video_waits_for_text_shutdown_element),
(test_render_continuity_push_video_buffers_thread),
(textoverlay_suite):
Add some unit tests for textoverlay.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/ffmpegcolorspace.c:
(ffmpegcolorspace_suite):
Enable ffmpegcolorspace test now that the RGBA32 issue is fixed
(for now not for valgrinding though, since it takes too long).
Original commit message from CVS:
* gst/videotestsrc/Makefile.am:
* tests/check/Makefile.am:
Make sure our checks and the videotestsrc plugin link against the
local uninstalled gst libs and not any installed gst libs that
might happen to exist as well.
* tests/check/elements/adder.c: (message_received),
(test_event_message_received), (test_play_twice_message_received):
* tests/check/elements/ffmpegcolorspace.c: (GST_START_TEST):
Fix compiler warnings when compiling against core with disabled
debugging system.
Original commit message from CVS:
* tests/check/elements/audiorate.c: (test_injector_base_init),
(test_injector_class_init), (test_injector_chain),
(test_injector_init), (probe_cb), (do_perfect_stream_test),
(GST_START_TEST), (audiorate_suite):
More tests for audiorate: inject buffers to check behaviour when
buffers overlap.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/audiorate.c: (probe_cb), (got_buf),
(do_perfect_stream_test), (GST_START_TEST), (audiorate_suite):
Add some basic unit tests for audiorate. Disabled at the moment
since it doesn't pass yet (see bug #363119).
Original commit message from CVS:
* gst/subparse/gstsubparse.c: (subrip_fix_up_markup),
(parse_subrip), (handle_buffer):
Add missing closing tags for markup and fix broken markup,
otherwise pango won't render anything (fixes#357531). Also,
make sure the text we send out is always NUL-terminated
(better safe than sorry etc.).
* tests/check/elements/subparse.c: (test_srt_do_test),
(test_srt):
Some more tests for .srt incl. tests for the above stuff.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/subparse.c: (buffer_from_static_string),
(setup_subparse), (teardown_subparse), (test_srt_do_test),
(GST_START_TEST), (subparse_suite):
Add very simple unit test for subparse.
Original commit message from CVS:
* gst/playback/gstplaybasebin.c: (subbin_startup_sync_msg),
(setup_source):
Catch async errors when starting up the subtitle bin, so we can
stop waiting and continue with the main film instead of hanging
forever. Fixes#339366.
* tests/check/elements/playbin.c: (playbin_suite):
Enable unit test for the above.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/playbin.c: (GST_START_TEST),
(gst_red_video_src_uri_get_type),
(gst_red_video_src_uri_get_protocols),
(gst_red_video_src_uri_get_uri), (gst_red_video_src_uri_set_uri),
(gst_red_video_src_uri_handler_init),
(gst_red_video_src_init_type), (gst_red_video_src_base_init),
(gst_red_video_src_create), (gst_red_video_src_class_init),
(gst_red_video_src_init), (plugin_init), (playbin_suite):
Some small and basic unit tests for playbin; not very useful yet,
but at least a start.
Original commit message from CVS:
Patch by: James "Doc" Livingston <doclivingston at gmail com>
* ext/vorbis/Makefile.am:
* ext/vorbis/vorbis.c: (plugin_init):
* ext/vorbis/vorbisparse.c: (gst_vorbis_parse_class_init),
(vorbis_parse_parse_packet), (vorbis_parse_chain):
* ext/vorbis/vorbisparse.h:
* ext/vorbis/vorbistag.c: (gst_vorbis_tag_base_init),
(gst_vorbis_tag_class_init), (gst_vorbis_tag_init),
(gst_vorbis_tag_parse_packet):
* ext/vorbis/vorbistag.h:
Add new vorbistag element which derives from vorbisparse
and is essentially the same as well, only that it implements
the GstTagSetter interface and can modify the stream's
vorbiscomment on the fly (#335635).
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/vorbistag.c: (setup_vorbistag),
(cleanup_vorbistag), (buffer_probe), (start_pipeline),
(get_buffer), (stop_pipeline), (_create_codebook_header_buffer),
(_create_audio_buffer), (GST_START_TEST), (vorbistag_suite):
Add unit test for new vorbistag element.
Original commit message from CVS:
* tests/check/elements/adder.c: (adder_suite):
Don't set timeout to 6 seconds when we're running
in valgrind ... (and how is 6 seconds longer than
the default anyway?)
Original commit message from CVS:
* ext/gnomevfs/gstgnomevfssrc.c:
Fix misleading docs addition.
* tests/check/elements/videotestsrc.c: (check_rgb_buf):
Get rid of compiler warning the right way.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/.cvsignore:
* tests/check/elements/ffmpegcolorspace.c: (rgb_format_to_caps),
(create_rgb_conversions), (rgb_conversion_free),
(right_shift_colour), (fix_expected_colour), (check_rgb_buf),
(got_buf_cb), (GST_START_TEST), (ffmpegcolorspace_suite):
Add unit test for ffmpegcolorspace (RGB <=> RGB only so far),
but disable for now since it doesn't pass (something wrong with
RGBA somewhere).
Original commit message from CVS:
patch by: Michael Smith <msmith at fluendo dot com>
* gst/tcp/gstmultifdsink.c: (is_sync_frame),
(gst_multi_fd_sink_client_queue_buffer),
(gst_multi_fd_sink_new_client):
* tests/check/elements/multifdsink.c: (GST_START_TEST),
(multifdsink_suite):
Fix implementation of sync-method 'next-keyframe'
Original commit message from CVS:
* tests/check/elements/videotestsrc.c: (check_rgb_buf):
Returning a return value often helps. In this case, we
don't need the return value anyway, so just get rid of it.
Should make build bots much happier.
Original commit message from CVS:
* gst/videotestsrc/videotestsrc.c: (paintinfo_find_by_structure),
(paint_get_structure), (gst_video_test_src_get_size),
(gst_video_test_src_smpte), (gst_video_test_src_snow),
(gst_video_test_src_unicolor), (paint_setup_AYUV),
(paint_hline_AYUV), (paint_setup_ARGB8888), (paint_setup_ABGR8888),
(paint_setup_RGBA8888), (paint_setup_BGRA8888), (paint_hline_str4):
* gst/videotestsrc/videotestsrc.h:
Add support for AYUV and the various RGBA formats. Initialise
fields of paintinfo structs allocated on the stack.
* tests/check/elements/videotestsrc.c: (right_shift_colour),
(fix_expected_colour), (check_rgb_buf), (got_buf_cb),
(GST_START_TEST), (videotestsrc_suite):
Add unit tests for videotestsrc's RGB output.
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (GST_START_TEST):
Fix leaks. Wait for state transitions that might happen ASYNC, as well
as some that won't.
Original commit message from CVS:
* tests/check/elements/gdpdepay.c: (gdpdepay_suite):
I forgot to include the file containing the #define :)
Now includes "config.h"
Original commit message from CVS:
* gst/gdp/gstgdpdepay.c: (gst_gdp_depay_init):
proxying get/set caps is the wrong thing to do, since we really
do change caps quite fundamentally
* tests/check/elements/gdpdepay.c:
* tests/check/elements/gdppay.c:
remove declaration of buffers, it's already done in gstcheck.h
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (get_float_mc_caps),
(get_int_mc_caps), (GST_START_TEST), (audioconvert_suite):
Patch from #347221 adding a test for audioconvert
channel remappings.
Original commit message from CVS:
* tests/check/elements/audioresample.c: (test_reuse),
(audioresample_suite):
Add test case for bug #342789 fixed below.
Original commit message from CVS:
* gst/audioresample/gstaudioresample.c:
(gst_audioresample_class_init), (gst_audioresample_init),
(audioresample_start), (audioresample_stop),
(gst_audioresample_set_property), (gst_audioresample_get_property):
Implement GstBaseTransform::start and ::stop so that audioresample
can clear its internal state properly and be reused insted of
causing non-negotiated errors with playbin under some circumstances
(#342789).
* tests/check/elements/audioresample.c: (setup_audioresample),
(cleanup_audioresample):
Need to set element state here so that ::start and ::stop are
called.
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (set_channel_positions),
(get_float_mc_caps), (get_int_mc_caps):
* tests/check/elements/audioresample.c:
* tests/check/elements/audiotestsrc.c: (GST_START_TEST):
* tests/check/elements/videorate.c:
* tests/check/elements/videotestsrc.c: (GST_START_TEST):
* tests/check/elements/volume.c:
* tests/check/elements/vorbisdec.c:
* tests/check/pipelines/vorbisenc.c: (GST_START_TEST):
Don't busy-wait in tests; this was causing test timeouts very
frequently when running under valgrind.
Original commit message from CVS:
* gst/tcp/README:
* gst/tcp/gstmultifdsink.c: (gst_multi_fd_sink_init),
(gst_multi_fd_sink_remove_client_link),
(gst_multi_fd_sink_client_queue_caps),
(gst_multi_fd_sink_client_queue_buffer),
(gst_multi_fd_sink_handle_client_write),
(gst_multi_fd_sink_render):
* gst/tcp/gstmultifdsink.h:
make multifdsink properly deal with streamheader:
- streamheader is taken from caps
- buffers marked with IN_CAPS are not sent
- streamheaders are sent, on connection, from the caps of the
buffer where the client gets positioned to
- further streamheader changes are done every time the client
will receive a buffer with different caps
* tests/check/elements/multifdsink.c: (GST_START_TEST),
(gst_multifdsink_create_streamheader):
add tests for this
Original commit message from CVS:
* gst/gdp/gstgdppay.c: (gst_gdp_pay_class_init),
(gst_gdp_pay_init), (gst_gdp_buffer_from_caps),
(gst_gdp_pay_buffer_from_buffer), (gst_gdp_buffer_from_event),
(gst_gdp_pay_reset_streamheader), (gst_gdp_pay_chain),
(gst_gdp_pay_sink_event), (gst_gdp_pay_set_property),
(gst_gdp_pay_get_property):
add crc-header and crc-payload properties
don't error out on some things that are recoverable
* tests/check/elements/gdppay.c: (GST_START_TEST), (gdppay_suite):
add test for crc
Original commit message from CVS:
* gst/volume/gstvolume.c: (volume_choose_func),
(volume_update_real_volume), (gst_volume_class_init),
(gst_volume_init), (volume_process_float), (volume_process_int16),
(volume_process_int16_clamp), (volume_set_caps),
(volume_transform_ip), (plugin_init):
* gst/volume/gstvolume.h:
rewrite the passthrough check, split _int16 and _int16_clamp, fix
another property desc., remove unused param from process function
* tests/check/elements/volume.c: (volume_suite):
reactivate the passthrough test
Original commit message from CVS:
* tests/check/elements/adder.c: (test_event_message_received),
(test_play_twice_message_received), (GST_START_TEST),
(adder_suite):
Added check to show that #339935 is fixed with ongoing
adder and collectpads fixes.
Original commit message from CVS:
* gst/adder/gstadder.c: (gst_adder_get_type):
Make it easier to copy&paste
* gst/volume/Makefile.am:
* gst/volume/gstvolume.c: (volume_update_real_volume),
(gst_volume_set_volume), (gst_volume_set_mute),
(gst_volume_class_init), (volume_process_int16), (volume_set_caps),
(volume_transform_ip), (volume_update_mute),
(volume_update_volume):
* gst/volume/gstvolume.h:
Add own debug category, move duplicate code to helper function, fix
property texts, add more comments and prepare ffor liboil-goodness
* tests/check/Makefile.am:
* tests/check/elements/volume.c: (GST_START_TEST), (volume_suite):
add test for mute and passtrough case, be a bit more verbose to track
failure
* tests/check/generic/states.c: (GST_START_TEST):
catch elements that fail to instantiate
Original commit message from CVS:
* ext/theora/theoradec.c:
* ext/theora/theoraenc.c:
* ext/theora/theoraparse.c: (gst_theora_parse_class_init):
* gst/audiorate/gstaudiorate.c:
make more debug catagories static
* tests/check/Makefile.am:
* tests/check/elements/adder.c: (message_received),
(test_event_message_received), (GST_START_TEST),
(test_play_twice_message_received), (adder_suite):
added test case for using element twice, extra bonus points for anyone
who can make these test run reliably
Original commit message from CVS:
* tests/check/elements/alsa.c: (test_device_property_probe):
Fix test case: don't try to free NULL GValueArray when there
are no devices.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/alsa.c: (test_device_property_probe),
(alsa_suite), (main):
Add simple test that runs a device property probe on alsasrc,
alsasink and alsamixer. Disable valgrind check for now (too
many leaks in libasound, and valgrind ignored my suppressions
additions).
Original commit message from CVS:
* gst/audioconvert/gstchannelmix.c: (gst_channel_mix_fill_others):
Fix#341696: crash when mixing L+R+C to mono or stereo.
* tests/check/Makefile.am:
* tests/check/elements/audioconvert.c: (set_channel_positions),
(get_float_mc_caps), (get_int_mc_caps), (GST_START_TEST),
(audioconvert_suite):
Add test for the above, including some generic framework bits for
testing multichannel things.
Original commit message from CVS:
* gst/tcp/gstmultifdsink.c: (gst_multi_fd_sink_remove_client_link),
(gst_multi_fd_sink_new_client), (gst_multi_fd_sink_stop):
remove stupid example from docs - it should come with a simple
C program instead.
Clean up/fix docs
* tests/check/elements/multifdsink.c: (wait_bytes_served),
(fail_if_can_read), (GST_START_TEST),
(gst_multifdsink_create_streamheader), (multifdsink_suite):
add a test for changing streamheader which exposes a bug in
multifdsink
Original commit message from CVS:
* tests/check/elements/multifdsink.c: (wait_bytes_served),
(GST_START_TEST), (fail_unless_read), (multifdsink_suite):
add two more tests, one doing streamheader
Original commit message from CVS:
* gst/tcp/gstmultifdsink.c: (gst_multi_fd_sink_stop):
clean up the bufqueue when shutting down
* tests/check/Makefile.am:
* tests/check/elements/multifdsink.c: (setup_multifdsink),
(cleanup_multifdsink), (GST_START_TEST), (multifdsink_suite),
(main):
add a test for the leak that was just fixed
Original commit message from CVS:
* gst/adder/gstadder.c: (gst_adder_setcaps),
(gst_adder_query_duration), (gst_adder_query), (forward_event),
(gst_adder_src_event), (gst_adder_sink_event),
(gst_adder_class_init), (gst_adder_finalize),
(gst_adder_request_new_pad), (gst_adder_collected):
* gst/adder/gstadder.h:
Updated some docs. Added comments and FIXMEs all over the place.
Improve debugging info.
Fix leak on finalize by not calling the parent.
Implement duration query.
Make event forwarding threadsafe.
Correctly send NEWSEGMENT at start and after flush.
Handle EOS correctly.
Post error when not negotiated.
* tests/check/elements/adder.c: (GST_START_TEST):
Added FIXME in the test.
Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/adder.c: (event_loop), (GST_START_TEST):
Disable the adder test, until the build-slaves posses the kindness to
either like it or to give valid reason for not doing so
Original commit message from CVS:
* tests/check/elements/adder.c: (event_loop), (GST_START_TEST),
(adder_suite):
Shuffle NULL state change around and raise timeout more
Original commit message from CVS:
* gst/typefind/gsttypefindfunctions.c: (mp4_find_box),
(mp4_type_find), (plugin_init):
Add typefind to distinguish between "audio/x-m4a" and new type
"video/mp4". Fixes#340375
* tests/check/elements/adder.c: (adder_suite):
Raise timeout to make buildbot happy
Original commit message from CVS:
* gst/adder/gstadder.c: (gst_adder_sink_event),
(gst_adder_request_new_pad), (gst_adder_change_state):
* gst/adder/gstadder.h:
* tests/check/Makefile.am:
* tests/check/elements/adder.c: (event_loop), (GST_START_TEST),
(adder_suite), (main):
Add sink-event handling to adder. It tries to merge incomming
newsegment-events. Added test to check if segment_done is comming
through.
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (verify_convert),
(GST_START_TEST):
interpret the out[] buffer in the order the bytes are actually
put in, which is LITTLE_ENDIAN, not BYTE_ORDER.
Other tests should use BYTE_ORDER since the array is filled in
with actual values
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (verify_convert),
(GST_START_TEST):
when a test fails, give an indication of which it is
Original commit message from CVS:
* gst/videorate/gstvideorate.c: (gst_video_rate_reset),
(gst_video_rate_swap_prev), (gst_video_rate_chain):
fix up docs
fix a leak when no caps negotiated
fix counting of input frames
* tests/check/elements/.cvsignore:
* tests/check/elements/videorate.c: (assert_videorate_stats),
(GST_START_TEST), (videorate_suite):
add tests for these
Original commit message from CVS:
* tests/check/elements/audioconvert.c: (get_float_caps),
(GST_START_TEST), (audioconvert_suite):
Added check for correct clipping when doing float samples
in audioconvert.
Original commit message from CVS:
Patch by: Edward Hervey <edward@fluendo.com>
* gst/videorate/gstvideorate.c: (gst_video_rate_chain):
* tests/check/Makefile.am:
* tests/check/elements/videorate.c: (assert_videorate_stats),
(setup_videorate), (cleanup_videorate), (GST_START_TEST),
(videorate_suite), (main):
Fix an infinite loop if frames are passed in with wrongly ordered
timestamps. Fixes#339013.
Original commit message from CVS:
* check/Makefile.am:
* check/elements/audiotestsrc.c: (setup_audiotestsrc),
(cleanup_audiotestsrc), (GST_START_TEST), (audiotestsrc_suite),
(main):
add a test for audiotestsrc, testing all waves. Even seems
leak-free at first glance, nice job Stefan
Original commit message from CVS:
* check/elements/audioconvert.c: (setup_audioconvert),
(cleanup_audioconvert), (get_int_caps), (verify_convert),
(GST_START_TEST), (audioconvert_suite):
clean up tests a little, fix some leaks.
Original commit message from CVS:
* check/Makefile.am:
have some tests be disabled for valgrinding
* check/elements/vorbisdec.c: (cleanup_vorbisdec),
(GST_START_TEST):
* ext/vorbis/vorbisdec.c: (vorbisdec_finalize):
Fix A Leak. Chain To Parent Finalize.
Original commit message from CVS:
2005-08-29 Andy Wingo <wingo@pobox.com>
* check/elements/audioconvert.c: Convert from native endian, not
little endian.
Original commit message from CVS:
* check/Makefile.am:
* check/elements/audioconvert.c: (setup_audioconvert),
(cleanup_audioconvert), (get_int_caps), (verify_convert),
(GST_START_TEST), (audioconvert_suite), (main):
add a test for audioconvert
* gst/audioresample/gstaudioresample.c:
* gst/audioresample/gstaudioresample.h:
set DURATION so that TIMESTAMP(a) + DURATION(a) == TIMESTAMP(b);
note that for buffers of 1/3 sec this means DURATION(c) is
one nanosecond more than for a and b
