Age | Commit message (Collapse) | Author |
|
Just update the PERFETTO_VERSION_XXX macros.
Change-Id: I1d34c14b74449f72428814ba170e507ebbd8a961
|
|
The previous v12.0 branch cut missed the CL with the
updated CHANGELOG file.
Change-Id: I37a0d2d3b67d4e8992dc4715aaf2f882d93cd926
(cherry picked from commit 71a92399dcf1a54dad73b2ad476ba6fd3175faf3)
|
|
Change-Id: I17c8a5bcdd8224bec012d47836898b3edfa50e85
(cherry picked from commit 9c294a40a5d0470db2978f3103e2b7f2e1e50a2e)
|
|
Change-Id: I23c4e646e88733c8ab67b1621798cbbe348a5386
|
|
Bug: 178826286
Change-Id: I16552ed75353c6d3a8ac3536d10a5eaad1e93552
|
|
|
|
|
|
|
|
|
|
|
|
Change-Id: Iae526fbd02787cfc48d51021f9cd7940eb35391f
|
|
Change-Id: I8d39460e33f513b07c98bbbcf39c74a616957915
|
|
Change-Id: If9a165a4f41b055738ce4586fd75af6851a96799
|
|
|
|
|
|
|
|
Change-Id: I34518ddc604f1d8925b64a057d1dbb65a749121e
|
|
Change-Id: I0342a39bb702488bd3060816281528c87d5572b9
|
|
We count the time between the intent start and the reportFullyDrawn
slice starting to compute the time to report fully drawn.
Change-Id: I51dca019d6f7b6f0616d32cee4188efba19aed63
Bug: 174880603
|
|
Bug: 178467684
Change-Id: I69fd300ea3bb4248d93680456bd258e56ca5bec9
|
|
This way, we can tell if adapting sampling was applied.
Bug: 178467684
Change-Id: I7763d532fd5a4ab168cc6b2fbe11c9630446e4f3
|
|
|
|
|
|
Bug: 178378779
Change-Id: Ia5ea5f3bd28087ff5e3741772a21c7e0604dc65e
|
|
Before: [Thread 1, Thread 10, Thread 11, Thread 2, Thread 20]
After: [Thread 1, Thread 2, Thread 10, Thread 11]
Bug: 178753925
Change-Id: Ifac7b32f70547de6751665f507a3049c2de6eb49
|
|
|
|
|
|
This field will record IO handler locations for MessagePumpForIO.
Change-Id: I69f7391286ba523547bec2ac80fa24854adea0e8
|
|
Before 8a689afd048a60cc76e7543d9d8c4193df351f96 this was a no-op because
we forgot to set the TracePacket size. After that we would crash because
most data would fail to parse as a packet.
Change-Id: I37a95f0062fc87d2bdaa78eaac1284c662d88f37
|
|
|
|
|
|
|
|
Also add a diff test to prevent regressions.
Change-Id: Ic22560dce59c38764c373f0fed8753a1c13ca48a
|
|
Before, we would not set some necessarily cflags.
Change-Id: I40decf4a2bf7aa019961d3b3b08e86e3e5880fe7
|
|
This is a port of the TBMv2 frame_time metric [1] to TBMv3.
Some features required by this metric to work on all platforms are still
missing (most notably the SurfaceFlinger events on Android), so adding
this to experimental folder for the time being.
[1] https://source.chromium.org/chromium/chromium/src/+/master:third_party/catapult/tracing/tracing/metrics/rendering/frame_time.html
Change-Id: I9562d93256617357bc4dc7413ca7685a4f496ed3
|
|
|
|
|
|
This allows to use ubsan as well.
Test: python infra/helper.py build_fuzzers --sanitizer {address,undefined} --engine {afl,libfuzzer,honggfuzz} perfetto
Test: for f in buffered_frame_deserializer_fuzzer
cpu_reader_fuzzer
end_to_end_shared_memory_fuzzer
packet_stream_validator_fuzzer
producer_socket_fuzzer
protozero_decoder_fuzzer
shared_ring_buffer_fuzzer
shared_ring_buffer_write_fuzzer
trace_processor_fuzzer
unwinding_fuzzer; do
python infra/helper.py check_build --sanitizer {address,undefined} --engine {afl,libfuzzer,honggfuzz} perfetto $f;
done
Change-Id: I5adb3ff84a4b61806e0c41ffe73fae79db55521d
|
|
Fixes from CL cl/353893061
Change-Id: Ib5b6c1706363adf37db38d6cb32d35930981cb11
|
|
Change-Id: Iba423ba8b5c71089939bc4fc82ca9177d1236017
|
|
Change-Id: Ib4578491d455eb229eaf05c5f204576b9b547cc1
|
|
Test: python infra/helper.py build_fuzzers --sanitizer address perfetto
in oss-fuzz repo with patch to perfetto applied.
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=29820
Change-Id: I6c814014ef841cc873c444190bea69ad553d9f9b
|
|
|
|
|
|
Change-Id: Iccf59a5bed78629dfbdbf2a1e6f36d7fd70906c0
|
|
|
|
This is a follow-up to aosp/1515497. Previously we didn't
allow the --save-for-bugreport feature to steal traces
marked as write_into_file. That was because a trace
might have been partially written on disk and that would
make the bugreport re-routing more difficult (also
undesitable given those traces can be ~GB).
This CL softens the check allowing to steal also those
traces, but only if no data has been written yet.
The problem is that Traceur (System Tracing App), uses
write_into_file && --detach both for long and ring-buffer
traces (to keep the code path more uniform).
In ring-buffer mode, traceur sets the file_write_period_ms
to 7days (which really means, only save at the end), to
maintain the ring-buffer behavior.
The alternative would be a more invasive change to traceur
AND making our cmdline more awkward, as we would have to
support both "--detach + path, but only if write_into_file"
and "--attach + path, but only if !write_into_file"
Bug: 175591887
Bug: 170334305
Test: perfetto_integrationtests --gtest_filter=PerfettoTest.SaveForBugreport*
Change-Id: I9b3dbb2c1c27e80af3c25eba35ac8fc0d3992a6a
|
|
Bug: 174656725
Change-Id: I0fc63f5068127d7de7d3ce3e74ae99c0a6823bcd
|
|
Change-Id: I6bcc0b2b967ec0f2d244d10a13e52afd7776f5c9
|
|
|