Age | Commit message (Collapse) | Author |
|
oc-mr1-dev
|
|
The video stream stop sequence can encounter a race conidtion
in case pending buffers arrive during the same time frame. Any
incoming stream callbacks that manage to execute during the
window between 'mDataQ' flush and 'flushFreeBatchBufQ' will modify
the 'mBufsStaged' counter without the knowledge of the top
'QCamera3HardwareInterface' instance. This makes subsequent calls
'aggregateBufToBatch' prone to batch overflow errors.
To resolve this, call 'flushFreeBatchBufQ' before flushing 'mDataQ'.
This way calls to 'aggregateBufToBatch' from failing stream
callbacks should not be successfull in modifiying 'mBufsStaged'.
Bug: 65549208
Test: Camera CTS
Change-Id: Iee55ebd57e66ca359fa1eaa21fcc86021f2b920b
|
|
All vendor supported tags should be advertised in the respective
available lists.
Remove initial capability optimization. The vendor tag ops are not
initialized at this point which will lead to errors when trying to
append vendor tags.
Bug: 63817523
Test: Camera CTS
Change-Id: I417963660815b8cfe5678e8e017981f462037f0d
|
|
Test: compare with RAW10 output
Bug: 65552901
Change-Id: I5104571fd18ec6df2ecebb0433db4aaaec6e4be8
QCamera2: HAL3: Fix wrong conversion for first 2 pixels
Test: Camera CTS
Bug: 65552901
Change-Id: I3ee00beb9ebef811e74e95134ead886e3da6d476
|
|
am: f137d40cd7
Change-Id: I0b6a6eadd785d7c29bd9e290c60d2be4c465c440
|
|
am: c0610d4f9a
Change-Id: I1d407758c588539599224591370d1cf2c12a454c
|
|
Bug: 36716469
CRs-Fixed: 2053027
Change-Id: I6699c3682eec419fb6efb0583e9909e5739476a0
(cherry picked from commit a5eedde871c8e87784014e541c4aec247e954dc2)
|
|
4bb1ccf65e am: f0c2aafa9d
am: d02729a700
Change-Id: I45b619563213ce99c2fb37aee70f28d91b216644
|
|
am: 4bb1ccf65e
Change-Id: I13a479a02e71114b888ec03f447dcd9cbb08ebf6
|
|
am: fe4bd8accd
Change-Id: I14d04f45db92819eed1031e142314379c8e73575
|
|
This is never used by anything in the system.
Test: Build
Bug: 37512375
Change-Id: Ibfe65476b1891ad66d090e567cd90c5954b91f90
|
|
am: d36c3de0ef
Change-Id: I62a084ba2aaae02c72e0a098230bad8d5fd15e4e
|
|
Currently reprocess input buffers are unmapped upon JPEG encode
completion. This is problematic because after HAL returns input buffer
after CPP processing, it doesn't own the buffer any more.
The sequence of events leading to the issue:
1) Reprocess request 1 enters, input buffer FD is N
2) HAL caches the input buffer
3) CPP runs on input buffer, returns buffer to framework.
Cache entry remains. JPEG encoding starts
4) Framework frees buffer; FD N is now reusable
5) Reprocess request 2 enters, new buffer also gets FD N
6) HAL finds buffer in cache
7) CPP tries to run, but uses stale cache entry, crashes
8) (if no crash) JPEG encoding completes, cache cleaned up
Bug: 62721870
Test: Camera CTS
Change-Id: I22a06fad9ead03af917e547a9ee4d49d8db2e2d0
|
|
Test: Use QCamera3HWI
Change-Id: I62338f00cd4a970b5f9d615c42ba32da05a7f690
|
|
Offline reprocessing failures should return appropriate status
to camera service otherwise the request will not complete and
camera will go in a bad state.
Bug: 37559776
Test: Complete Camera/Camera2 CTS
Change-Id: Ifbc59f05836a9cab2c80dd7865555478c96888f5
|
|
Return error code -EINVAL instead of 0 when the stream format is
unsupported.
Additionally check and return respective error in case stream rotation
is invalid.
Bug: 35372309
Test: adb shell
/data/nativetest/VtsHalCameraProviderV2_4TargetTest/VtsHalCameraProviderV2_4TargetTest
Complete Camera/Camera2 CTS tests
Change-Id: If6aee12c68054243a90d655bbaf0e95c36887682
|
|
In case there is dropped metadata for certain frame number, HAL should
make sure:
1. partial_result field of the capture_result isn't 0.
2. the pending CaptureRequest for the frame number needs to be returned.
Test: Verified no error if metadata is dropped
Bug: 29699751
Change-Id: Iab51f4493c14bb0bee641baa8d7c8b74819a94e5
|
|
If usage flag has combined bits from different consumers, fail
configure_streams if camera device doesn't support it.
Camera device take into consideration of below 3 factors:
- stream format
- stream dataspace, and
- HAL interface restriction
Test: Camera CTS
Bug: 33777818
Change-Id: Ie2b2ad44de74ad908269bc23c8ce8ce065153a8f
|
|
|
|
By sending non-empty metadata only for the last result in a batch, the
number of memcpy within HAL and between HAL and cameraserver is reduced.
Test: HFR CTS
Bug: 35775704
Change-Id: If5ba901e61ff2be6feba946b1f3978eb47bd3c5e
|
|
Test: camera.msm8992 compiles
Bug: 36135110
Change-Id: Idc902d7b9351e2377c02747d34196d2f6c611643
|
|
|
|
The open_legacy check will trigger camera HAL allocating
more memory.
Test: dumpsys meminfo -> check the cameraserver memory usage
Bug: 35058355
Change-Id: I8faab7629be33fed614de6890891f38fdf91e39a
|
|
am: ebac1398dc
Change-Id: I9a764ad5fce0a6e6acd1294aebb59f49bdd887f5
|
|
am: 49d31d2e16
Change-Id: I4c5d8147a89988cd4e9b3579f105538fbcd989df
|
|
2165f4582c am: 792656012b
am: 059145d3d1
Change-Id: Ica50db420917952da21fc1d65b16fc49c1c9340b
|
|
am: 792656012b
Change-Id: I3c66523cdbd6d69b2e232cb7e356927e3a7be793
|
|
am: 2165f4582c
Change-Id: I8f5832cd74175e9084aff764f1fdb03a11c40c25
|
|
Only calls channel->stop() if channel is not nullptr.
Bug: 32706326
Change-Id: I8935eb4e16edd928d0f67ab161fe77f4c180ea7e
|
|
Since libstagefrighthw will not copying headers anymore
Bug: 33241851
Test: build bullhead target
Change-Id: Ib7e11e360220e7caba354027ae44485e5652f5d9
|
|
Test: Compile
Bug: 33811945
Change-Id: I073e10a68d2e86e6c6a312e189952756374ceafb
|
|
am: a260c90a93
Change-Id: If59a1e2d70701fd55b529ef5a0cc4964093b6284
|
|
Wall time may jump forward/backward. Especially when it jumps forward,
pthread_cond_timedwait may return early with TIMEOUT.
Switch to MONOTONIC time base instead.
Bug: 33110475
Change-Id: I7801c8a0c85dff25ed431c3aec63eecf3dc37b52
|
|
am: 66c8c04e7b
Change-Id: I83792687cfdd8e8c41ad266281d0b55234e29908
|
|
When metadata is dropped due to lower performance scenarios,
report ERROR_RESULT to framework instead of ENODEV
as the metadata error can be recoverable.
Bug: 31008299
Change-Id: I14b6e45171692b966e30cd17730e2c0c56c2babd
|
|
am: 5aab601d6f
Change-Id: I8c783d4e0df2a2ba6d73241b3ab3eb885050c139
|
|
am: a484356f38
Change-Id: Ic98a8fb806b1ee1952aeabff90695f641ff122aa
|
|
Bug: http://b/31289077
Bug: http://b/27171986
Test: remove liblog from libcutils and recompile
Change-Id: Icb2bc17d6e952145984c1b54b36c14191603e1c4
|
|
am: d651f40d73
Change-Id: I42cc8006dc8f6aa5372551eebb63ead7d2de0528
|
|
am: ce52d752d4
Change-Id: I50bf052b2a5c8565dfd3a2e19aa588500595e6c7
|
|
system/core/include is included in the global include path using
-isystem, which hides log format warnings when built with gcc.
Fix the warnings in preparation for switching from -isystem to -I.
Test: lunch aosp_angler-userdebug && m -j
Bug: 31492149
Change-Id: Ie75967aa30bae2e9bebbd65efe0ff570a80203fd
|
|
Only set timestamp source to be REALTIME if sensor sync is available, in
other words, if timestamp is calibrated.
Bug: 30471814
Change-Id: I507dc9f6a66ff2b7c1a627432dadd8e60f78c48e
|
|
When the phone powers up for several days,
the timestamp (in ns, nano sec) of frames will exceed the
precision of float data type.
It will lead to incorrect timestamp calculation in
HFR batch mode. Use double instead
Ack-by: Gaoxiang Chen
BUG=24842044
Change-Id: I63598e56178b980fa817c1c77f77f0e6a005767b
(cherry picked from commit ece15b4fa7cc4c622e41cc9a4cd28187ae584e9e)
|
|
This makes sure the mapping between af state and af mode are aligned
with spec. In other words, we make sure af state change aligns with af
mode change.
Bug: 30100921
Change-Id: Ie115eff5fa6fb7602e4218fd92510fe7062269ff
|
|
When the phone powers up for several days,
the timestamp (in ns, nano sec) of frames will exceed the
precision of float data type.
It will lead to incorrect timestamp calculation in
HFR batch mode. Use double instead
Ack-by: Gaoxiang Chen
BUG=24842044
Change-Id: I63598e56178b980fa817c1c77f77f0e6a005767b
|
|
In batch mode, only one metadata is sent from back-end per
batch. It can be *copied* as multiple metadata for rest of
the frames in the batch instead of parsing the same
metadata buffer again
BUG=28312446
CRs-Fixed: 1038197
Change-Id: I5be580e69b8f14e8a381111a787c38081fb04f71
|
|
Fix CTS issue: testDynamicBlackWhiteLevel.
Bug: 27224930
Change-Id: Ibe529d10f95c7ab48389eab59f7d2cf13383068f
|
|
Print out stream size, format, and last frame number for buffer loss.
Bug: 28427118
Change-Id: I5568d1a38a54243f094a679d80872a4c9d1b6523
|
|
|
|
Device open may fail when cameraserver and mm-qcamera-daemon are
killed at the same time.
Bug: 27869821
Change-Id: I03e006246c2df68b7ef176afd269fabb280b2534
|