Set the caps right after allocation of the buffer because we know the buffer is
writable then and we are correctly negotiated. Since ffmpeg keeps around
references to frames, making the buffer metadata writable where it was done
before pushing will always end up with a copy and that makes the sink do a slow
memcpy all the time.
Set caps on buffers right after we allocate them to avoid refcounting problems
and having to make the buffer metadata writable for no good reason.
Don't unmap the memory with a 0 size or we would modify the memory size when
it's not needed.
FFMpeg parsing and decoding calls require to additionally allocate bytes
at the end of the input bitstream and this padding must be initialized
to zero.
https://bugzilla.gnome.org/show_bug.cgi?id=595590
Sometimes the parser loses track of timestamps and starts to reuse old
timestamp. Feed it some dummy data and clear some context variables to work
around the problem.
When estimating the frame duration, the diff between two incomming timestamps
should be scaled by the amount of frames in the interval. Improves duration
estimation and DTS interpolation.
Use the diff between input timestamps to estimate the duration when no duration
is set on input buffers. Only do this when there are no reordered input
timestamps. Improves interpolation in DTS mode when no input duration is set.
When we flush the parser cache, we only need to clear the bytes of the cache,
not the complete state of the cache. In the case of H264 this doesn't require
the parser to receive a new SPS/PPS after a DISCONT buffer.
Add function to reset the timestamp tracking.
Check for reordered timestamps on the input buffers and assume PTS input
timestamps when we see reordered timestamps.
Recover from an occasionally wrong input timestamp by also tracking the output
timestamps. When we detect a reordered output timestamp, assume DTS input
timestamps again.
Fixes#611500
The buffer durations were not being reordered along with the timestamp
and offset of the buffers, resulting in buffers using the duration of the
latest incoming frame instead of their original frame.
Fixes#611398
When we have an input width/height that should be used for clipping, only
perform the clipping if the rectangle is smaller than the actual picture size.
Fixes#330681
Make check for vdpau decoders more generic. There might be vdpau
decoders we don't expect when using an external ffmpeg version,
and we want those blacklisted as well (e.g. ffdec_mpeg4_vdpau).
Resetting default values is currently very complex in libavcodec, so
we only call it when needed (i.e. when a context was previously used).
Shaves off 10% of the setup of a decoder.
When we are dropping frames because of QoS disable the DTS interpolation because
we won't be able to update the timestamps and end up setting the wrong
timestamps. Instead, simply use the timestamps from ffmpeg.
This now uses ffmpeg functionality to keep random metadata next to
the buffers and to get the correct offset for a frame, similar to how
timestamps are handled.
Fixes bug #578278.
Takes codec frame delay into account (roughly the same way it does for timestamps for reordered frames) to produce frames with correct offsets.
A special hack to allow trailing frame with timestamp=segment.stop to be displayed.
Fixes bug #578278.
After a DISCONT, mark the next frame with DISCONT but don't wait for a new
keyframe. This greatly improves performance on lossy networks or currupted
frames as the decoder can usually continue and conceil errors up to the next
keyframe.
Avoid an infinite loop consuming buffer timestamp info when
the video frames contain only GST_CLOCK_TIME_NONE timestamps.
Add some debug logging in the timestamp tracking paths.
Fixes: #585845
If the same instance of the plugin is asked to be initialised more that once,
instances after the first one do not register the elements properly and the
elements become not usable.
For example, if you call gst_update_registry (), is not possible to create
elements after the call since the plugin is asked to be initialised again and
does not register the elements.
Fixes#584291
The patch from Bug #580796 hacked around existing infrastructure to handle
timestamps as DTS (as in all AVI files) causing the logic to be disabled.
Properly hook the timestamp handling into the existing infrastructure to handle
these cases too, partially reverting a26b94d92c
and moving some stuff around.
Refixes #580796.
ffmpeg only tells us on a per-decoded-buffer basis if the stream is
interlaced or not. When we see a change, we force negotiation.
We can't detect that in our get_buffer() (when doing downstream allocation),
because at that point the interlaced flags aren't set on the outgoing
buffer.
Add a new function new_aligned_buffer() which creates a GstBuffer of
the requested size/caps, with the memory being allocated/freed by ffmpeg's
av_malloc/av_free which guarantees properly aligned memory.
Added a can_allocate_aligned internal property which we use to figure out
whether downstream can provide us with 128bit aligned buffers.
We simply allocate the memory using ffmpeg's av_malloc which provides us
with properly memalign'ed data.
This avoids write-outside-of-bounds when sse/altivec code is being used.
We should post a STREAM DECODE error message on the bus when we return
GST_FLOW_ERROR, otherwise the user ends up seeing an ugly internal flow
error message, which isn't very nice.
The problem is that the ffmpeg aac decoder fails... but still accepts
the following buffers as if nothing happened. But because some things
were not properly set in the internal code, all hell breaks loose.
For a given AVCodec, when the sample_fmts field is non-NULL, that means that
that codec can only handle a specific set of SampleFormat.
With this patch, we now look for its presence and create the proper pad template
caps.
Fixes#569441
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ff_aud_caps_new),
(gst_ffmpeg_codecid_to_caps), (gst_ffmpeg_smpfmt_to_caps),
(gst_ffmpeg_codectype_to_caps), (gst_ffmpeg_caps_to_smpfmt),
(gst_ffmpeg_caps_to_codecid), (av_smp_format_depth):
* ext/ffmpeg/gstffmpegcodecmap.h:
Add mapping for EAC3 and QCELP audio codecs.
Add conversion functions for all available audo SampleFormat.
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(gst_ffmpegdec_setcaps), (gst_ffmpegdec_negotiate),
(clip_audio_buffer), (gst_ffmpegdec_audio_frame):
Remove assumptions that we can only handle stereo 16bit signed integer
audio, and store the depth locally.
Original commit message from CVS:
reviewed by: Edward Hervey <edward.hervey@collabora.co.uk>
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_finalize):
Fix check for memory to free.
Fixes#560644
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_avpicture_fill):
Initialize some more variables.
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(alloc_output_buffer):
Disable direct rendering for h264, some functions just seem to read from
invalid memory.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(gst_ffmpegdec_get_buffer), (get_output_buffer):
Enable direct rendering.
Add some more debug info about image strides.
Original commit message from CVS:
Based on a patch by: Alexis Ballier <aballier at gentoo dot org>
* configure.ac:
* ext/ffmpeg/gstffmpeg.c:
* ext/ffmpeg/gstffmpeg.h:
* ext/ffmpeg/gstffmpegaudioresample.c:
* ext/ffmpeg/gstffmpegcfg.c: (gst_ffmpeg_flags_get_type),
(gst_ffmpeg_cfg_init):
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_codecid_to_caps),
(gst_ffmpeg_caps_to_pixfmt), (gst_ffmpeg_caps_with_codecid):
* ext/ffmpeg/gstffmpegcodecmap.h:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(gst_ffmpegdec_register):
* ext/ffmpeg/gstffmpegdeinterlace.c:
* ext/ffmpeg/gstffmpegdemux.c:
* ext/ffmpeg/gstffmpegenc.c: (gst_ffmpegenc_getcaps),
(gst_ffmpegenc_setcaps), (gst_ffmpegenc_register):
* ext/ffmpeg/gstffmpegmux.c:
* ext/ffmpeg/gstffmpegprotocol.c: (gst_ffmpegdata_seek):
* ext/libpostproc/gstpostproc.c:
* ffmpegrev:
Update ffmpeg/swscale snapshot to the latest revision and adjust
to API changes. Fixes bug #556405.
Require libavutil for swscale too when building with an external
ffmpeg and fix includes for external ffmpeg.
Original commit message from CVS:
Patch by: Robin Stocker <robin at nibor dot org>
* ext/ffmpeg/gstffmpegdec.c:
(gst_ffmpegdec_add_pixel_aspect_ratio):
If both, the decoder and the demuxer, provide a non-1:1 PAR
prefer the one of the demuxer instead of the one of the decoder.
Fixes bug #556336.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ff_vid_caps_new),
(gst_ff_aud_caps_new), (gst_ffmpeg_codecid_to_caps),
(gst_ffmpeg_codectype_to_caps):
* ext/ffmpeg/gstffmpegcodecmap.h:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_negotiate),
(gst_ffmpegdec_register):
* ext/ffmpeg/gstffmpegenc.c: (gst_ffmpegenc_getcaps),
(gst_ffmpegenc_register):
Add some more width/height/channels/rate limitations to caps
to cater for more automagic negotiation. Addresses #532422.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_frame),
(gst_ffmpegdec_chain):
If we have a parser and we did not consume any of the bytes of a new
buffer, make sure we submit the buffer again with its original timestamp
instead of a -1 timestamp. Fixes various h264 cases with reordered
frames.
If we have a discont and a timestamp but the first buffer after the
discont did not produce any data, make sure we set the timestamp on the
next buffer instead. Fixes initial timestamp on realaudio in many cases.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_finalize),
(gst_ffmpegdec_get_buffer), (gst_ffmpegdec_frame):
Finalizing a decoder that was never used shouldn't trigger an assertion.
Add debug messages for the two other g_assert_if_reached().
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c:
If ffmpeg reports 0 bytes of input data consumed, don't break out
unless it also didn't produce any output. Fixes the audio in #377400
and doesn't break anything else I've tested.
Enable the mp3 parser, and set mp3 and mpeg2-video decoding autoplug
at marginal level, as they seem to both work fine now.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
Explicitely disable the AAC decoders as they don't work very well
and we have better alternatives. Fixes bug #534392.
Original commit message from CVS:
Patch by:
Hans de Goede <j dot w dot r degoede at hhs dot nl>
* configure.ac:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_avpicture_fill):
* ext/ffmpeg/gstffmpegcodecmap.h:
* ext/ffmpeg/gstffmpegdec.c: (get_output_buffer):
* ext/ffmpeg/gstffmpegdemux.c: (gst_ffmpegdemux_loop):
* ext/ffmpeg/gstffmpegmux.c: (gst_ffmpegmux_collected):
Use av_picture_copy() instead of libswscale to copy pictures. This
removes the swscale dependency and is faster. Fixes bug #534390.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(gst_ffmpegdec_register):
Previous commit in fact broke playback for standard wmv3.
Instead make both ffdec_vc1 and ffdec_wmv3 accept any wmv3 variant and
figure out the proper codecid when opening the ffmpeg decoder.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(gst_ffmpegdec_video_frame), (gst_ffmpegdec_register):
Bump the priority of VC1 decoder so that it goes before the WMV3
decoder. This allows proper auto-pluggin with decodebin/playbin.
Fixes#531857
Original commit message from CVS:
2008-03-18 Andy Wingo <wingo@pobox.com>
* ext/ffmpeg/gstffmpegdec.c (gst_ffmpegdec_sink_event): Only drain
if we've already set up a codec.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_sink_event):
When we receive a newsegment event, we must drain any pending frames
because they belong to the previous segment. This fixes some cases of
very large timestamps when doing segment seeks.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(get_output_buffer), (gst_ffmpegdec_video_frame),
(gst_ffmpegdec_chain):
Work around an ffmpeg bug where it always returns 0 timestamps.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_class_init),
(gst_ffmpegdec_setcaps), (check_keyframe),
(gst_ffmpegdec_video_frame), (gst_ffmpegdec_sink_event),
(gst_ffmpegdec_set_property):
Detect DTS or PTS as timestamps. This is done by tracking frame
reordering on the output and making sure that timestamps don't go
backwards. Fixes#482660.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_video_frame):
Don't blindly copy input timestamp to output timestamp but prefer the
one attached to the picture when we can.
Add new variables for the output timestamp and duration to make the code
a little more clear.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_video_frame):
Initialize hurry_up to 0 to fix "might be used uninitialized"
compiler warning.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_base_init),
(gst_ffmpegdec_setcaps), (gst_ffmpegdec_video_frame):
* ext/ffmpeg/gstffmpegenc.c: (gst_ffmpegenc_base_init):
When doing QoS, don't drop the frame before decoding because we might
drop an important reference frame, just make the decoder hurry_up on
this frame.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_class_init),
(gst_ffmpegdec_init), (get_output_buffer), (gst_ffmpegdec_chain),
(gst_ffmpegdec_change_state), (gst_ffmpegdec_set_property),
(gst_ffmpegdec_get_property):
Add padding to input data before feeding it to ffmpeg. Also add option
to disable this (although it does not seem to cause slowdown).
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(alloc_output_buffer), (gst_ffmpegdec_get_buffer),
(gst_ffmpegdec_release_buffer), (gst_ffmpegdec_negotiate),
(get_output_buffer):
Change the pad_alloc calculations for weird clipped sizes, refactor the
code a bit.
Add support for some different refcounting algorithm.
Direct rendering still disabled by default.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_class_init):
Disable direct-rendering by default until buffer allocation works
correctly.
Rename the alias of the direct rendering property from 'direct' to
'direct-rendering'.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps),
(gst_ffmpegdec_get_buffer):
Disable direct rendering for h264 since it does not always work.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_set_property),
(gst_ffmpegdec_get_property):
Implement get/set for the new property too.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_class_init),
(gst_ffmpegdec_init), (gst_ffmpegdec_close), (gst_ffmpegdec_open),
(gst_ffmpegdec_setcaps), (gst_ffmpegdec_get_buffer),
(gst_ffmpegdec_release_buffer), (get_output_buffer),
(gst_ffmpegdec_video_frame), (gst_ffmpegdec_audio_frame),
(gst_ffmpegdec_frame), (gst_ffmpegdec_change_state),
(gst_ffmpegdec_set_property), (gst_ffmpegdec_get_property):
Reenable pad_alloc, seem to work now.
Added property to easily disable it later on.
Remove some old code that tried hard to break the get_buffer
functions. Fixes#321662.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_video_frame),
(gst_ffmpegdec_chain):
Remove some more overly clever code that does nothing but mess up
timestamps.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_drain),
(gst_ffmpegdec_sink_event), (gst_ffmpegdec_chain):
Flush delayed frames on DISCONT if we have them.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_video_frame),
(gst_ffmpegdec_chain):
Flush on DISCONT because ffmpeg does not reliably tell us about
keyframes.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
* ext/ffmpeg/gstffmpegdemux.c: (gst_ffmpegdemux_register):
Don't register the WavPack demuxer and decoder. They don't work,
we have better ones and the output of the demuxer/input of the
decoder is in a different format than what audio/x-wavpack of the
wavpack plugin is (it seems that the demuxer strips of the wavpack
headers from every frame).
This fixes typefinding of Wavpack files again, as the ffmpeg
typefinder was preffered for some reason and gave
application/x-gst_ff-wv instead of audio/x-wavpack.
Original commit message from CVS:
Patch by: Tuomas Kulve <tuomas dot kulve at movial dot fi>
* ext/ffmpeg/gstffmpegdec.c:
Fix the return type of the setcaps function. Fixes#485033.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_codecid_to_caps),
(gst_ffmpeg_caps_with_codecid), (gst_ffmpeg_get_codecid_longname):
Add Real Video 3.0 and remove some bogus caps.
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(gst_ffmpegdec_video_frame):
When skipping because of QoS, pretend we consumed all bytes for the
frame so that the pcache's timestamp is updated correctly.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
Don't set rank of sp5x video decoder to NONE.
Fixes#466221
Re-order the cases of the switch statement so that the default case
is last.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_caps_with_codecid):
Improve debugging of codec data.
realvideo caps are underspecified, use fields of alternative variant
before we settle this.
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(gst_ffmpegdec_video_frame):
Add more debugging of used parsers.
Setup the realvideo slices correctly before calling the decoder, fixed
realvideo in matroska.
Original commit message from CVS:
2007-07-25 Julien MOUTTE <julien@moutte.net>
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_video_frame):
* ext/ffmpeg/gstffmpegdemux.c: (gst_ffmpegdemux_loop): Handle
buffers with invalid duration correctly even in the case where
we have a demuxer AND a decoder from ffmpeg linked.
Fixes: #460274.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (clip_video_buffer),
(gst_ffmpegdec_video_frame):
Be a bit smarter when clipping the frame without a duration, we don't
want the frame to extend to the end of the segment in this case.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_init),
(gst_ffmpegdec_open), (gst_ffmpegdec_save_incoming_values),
(gst_ffmpegdec_get_best_values), (gst_ffmpegdec_video_frame),
(gst_ffmpegdec_sink_event):
Fix timestamping for out-of-order incoming buffers. Instead of blindly
copying the incoming buffer timestamps on the outgoing buffers we cache
the latest 2 incoming buffer timestamps and duration and make a wise
choice as to what the outgoing buffer timestamp and duration should be.
Fixes#342962
Original commit message from CVS:
Patch by: Yves Lefebvre <ivanohe at abacom dot com>
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_setcaps):
Draw edges for h264. Fixes#364139.
Set the error resilience flag just in case...
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_base_init),
(gst_ffmpegdec_register):
* ext/ffmpeg/gstffmpegdemux.c: (gst_ffmpegdemux_base_init),
(gst_ffmpegdemux_register):
* ext/ffmpeg/gstffmpegenc.c: (gst_ffmpegenc_base_init),
(gst_ffmpegenc_register):
* ext/ffmpeg/gstffmpegmux.c: (gst_ffmpegmux_base_init),
(gst_ffmpegmux_register):
Don't use GINT_TO_POINTER and GPOINTER_TO_INT with GTypes, this
can break horribly if sizeof(GType) happens to be bigger than
sizeof(int), because GPOINTER_TO_INT might then chop off some bits
of our GType (the reason this seems to works nevertheless is the
put-current-type-also-as-value-0-into-the-hash-table-as-fallback
hack used in the current code).
In any case, instead of just fixing this, let's not use a hash table
with GTypes here at all. g_type_{set|get}_qdata() seems to do the
job just as well.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
My console is not your TODO list.
If it were your TODO list, you would have DONE it by now.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(gst_ffmpegdec_chain):
When H264 is presented without codec_data, use a parser to frame it
by assuming that no codec_data implies it is unpacketised.
Always use the return value from the parser to advance the input
buffer position.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_formatid_to_caps),
(gst_ffmpeg_formatid_get_codecids):
Added proper mime types for mxf and gxf formats.
Added known configuration for asf muxer. Implies allowing the ffmpeg asf
muxer to take WMV and WMA.
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_chain):
Most demuxers/formats don't give keyframe information on encoded
audio streams. We therefore don't discard them, and use the clipping
functionnality to drop the decoded buffers we didn't need.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
* ext/ffmpeg/gstffmpegenc.c: (gst_ffmpegenc_register):
Don't leak caps in some cases and make build bots happy again.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
wc1/wmv3 ffmpeg decoder now works. Bumping rank to GST_RANK_MARGINAL.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegcodecmap.c: (gst_ffmpeg_codecid_to_caps),
(gst_ffmpeg_get_codecid_longname):
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_register):
Seems like the ffmpeg folks finally understood that VC9 does not
exist, and that it's in fact VC1 (aka WMV9 or WMV3, but that's yet
another debate).
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_open),
(check_keyframe), (gst_ffmpegdec_video_frame),
(gst_ffmpegdec_chain):
Add some more timestamp debugging.
Fixed wrong duration whn not using a parser.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_chain):
Do early keyframe check again, feeding bad data in ffmpeg is not a good
idea after all.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_chain):
Disable early keyframe check, it causes crashes and is not really
needed.
Original commit message from CVS:
* ext/ffmpeg/gstffmpegdec.c: (gst_ffmpegdec_frame),
(gst_ffmpegdec_chain):
Even more timestamp fixage.
Also keep track of pending durations.
don't lose our timestamp when joining the cache.
