| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
to the component while in executing->idle state.
Change-Id: Ia207f0a112025ed0e2c4d62f404afbbfa694cbab
related-to-bug: 11557463, 11573543
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
ACodec has initiated the transition from exectuing->idle and is waiting
for the component to return all buffers so they can be freed.
It's counter productive and with some vendors harmful to submit additional
buffers to the component from this point on.
Change-Id: I332be254a1faf94902c77b5792a373ab447e44d3
related-to-bug: 11573543
|
|/
|
|
|
|
|
|
|
|
|
|
|
| |
The failure happens when vendor code sends EMPTY_BUFFER_DONE
twice on the same buffer. The second time through, we see a slot
that has already been emptied, and get very confused.
This should be handled with an error, not an assertion, so we now
log a stern complaint and carry on.
Bug 11557463
Change-Id: I76e53ff08924bca81afb6980df8e59a6012ff8fe
|
|
|
|
|
|
|
|
|
| |
Repeating the most-recent buffer appears to fail when the producer
disconnects. This adds a NULL check to avoid crashing.
Bug 10607920
Change-Id: Ied036004425f046fcf9c26303823681f3fddb99e
|
|
|
|
|
|
|
| |
if no new frame has been delivered by surface flinger within the timeout
interval.
Change-Id: I282f1b726dfe5646b178d7858d6f5d4f5a264fde
|
|
|
|
| |
Change-Id: I626bac6df4fc3d8478046193f06ecc7ea60dd3a8
|
|
|
|
|
|
|
|
|
| |
The state transition check was too strict, and we were crashing
mediaserver inappropriately.
Bug 9819944
Change-Id: I1482ed1cfee37088d4893ee81cf1b2b950d2e930
|
|\ |
|
| |
| |
| |
| | |
Change-Id: I3c5d4be2a2e8783fbf98b3e268fd02658f71dc7d
|
|/
|
|
|
|
| |
i.e. feed no more input frames to the encoder while suspended.
Change-Id: I51391e18c1517548e869f8ddece19f4af37e78f9
|
|
|
|
|
|
| |
Bug 7900302
Change-Id: I30b9cca783e0a48f77035b745b7d5e20edf10f27
|
|
|
|
|
|
|
|
|
|
|
|
| |
Update BufferQueue and ConsumerBase users to new BufferQueue
API, to allow BufferQueue slots to be reused. Buffer consumers
generally now need to track the unique frameNumber belonging to
each frame acquired if they are using BufferQueue directly.
Otherwise, they can simply track the graphicBuffer.
Change-Id: I30ee3158cf40fb10bbd085241646d5f1128ee480
Signed-off-by: Lajos Molnar <lajos@google.com>
Related-to-bug: 7093648
|
|\ |
|
| |
| |
| |
| |
| | |
Bug: 8988871
Change-Id: I24a2620d356f3bb6207ed834c79fd3e48b429695
|
|/
|
|
|
|
|
|
|
| |
BufferQueue item timestamps are in nanoseconds, MediaCodec buffer
timestamps are in microseconds. Adjust.
Bug 8988357
Change-Id: Ib12824d7eea0316a14543c756e3a127c0866db6b
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Various fixes:
- Set the maximum number of BQ buffers we're allowed to
acquire equal to the actual number of codec buffers. That
way we keep the codec as full as possible, and never try to
acquire more than we're allowed from the BufferQueue.
- Actually use "end of stream sent" flag.
- Name the BufferQueue (for debug messages).
Bug 8359403
Change-Id: I3b8c1f679bbebf6a89e623e13ca029eda7f657ba
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Assorted tweaks:
- Allow signalEndOfInputStream() before ACodec is in Executing
state (added message to two more states).
- Return an error if signalEndOfInputStream() is called a second
time on the same stream.
- Require AndroidOpaque color format in createInputSurface().
- Disallow dequeueInputBuffer() after an input surface has been
created (boolean flag in MediaCodec tracks it).
- Discard input surface when encoder is re-configure()ed (drop
OMXNodeInstance's ref when we go back to Loaded).
Bug 7991062
Change-Id: Iff30f3036e14eb5a2f6536910dcf11aba33031ee
|
|
Also, renamed a CHECK_INTERFACE macro that was clashing with the
Binder version.
Bug 7991062
Change-Id: If5e6ed0a06d9f67975497676e4b05abe3aa3d6c0
|