| Age | Commit message (Collapse) | Author |
|---|
|
android14-tests-release
Change-Id: I5ba0279aa0a1ce6bd4947e4d49618be06d946bff
|
|
Original change: https://android-review.googlesource.com/c/platform/external/perfetto/+/2995838
Change-Id: I0b07460266573958737d2f8055df34c5e55f2106
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
Original change: https://android-review.googlesource.com/c/platform/external/perfetto/+/2995838
Change-Id: I0333874266277d590556036a6f5d691c66a805cd
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
These tests:
1A. Start an app.
1B. Start recording a heap profile on that app (in some tests the
operations are reversed).
(In some cases 1A and 1B are swapped)
2. The trace is recorded for 4 seconds.
After 4 seconds, the tests read the trace and check for recorded memory
allocations.
On some particularly slow devices we've observed that the test app takes
more than 4 seconds to start.
Increasing the trace recording duration makes the test slower, but makes
it work also on slow devices.
Bug: 323332122
Merged-In: Ib3bdde1449d0e482e5428c1751381eae100e3536
Change-Id: Ib3bdde1449d0e482e5428c1751381eae100e3536
(cherry picked from commit 8fc000d37cce32a26661745b2ddcc1f7ddabc268)
|
|
The proto field id for clone snapshot has been changed in
U QPR1 to deal with bugs in the feature (fixed only in QPR).
The test, however, is part of CTS which needs to work on all
U releases. This CL disables the test on the android14-tests-dev
branch and keeps it in main.
Bug: 302438506
Bug: 296227911
Change-Id: Iee267f52cf3b9e5da678592fc8f01521e8f7d7b0
|
|
While merging this from `android12L-test-dev` to `main`, the `data`
field got duplicated.
Merge the duplicated entries into a single one.
Change-Id: Ie97857cf1d4a41fd479e100616b533265c807de8
Merged-In: Ie97857cf1d4a41fd479e100616b533265c807de8
Bug: 285215697
(cherry picked from commit 19e99ad003838ace452fba806e60c818e67d3366)
(cherry picked from commit ee31a6e6c21bd44cc66a02a41116efc91341ad65)
|
|
CTS tests on user builds fail, because the `sched/sched_process_fork`
event in not accessible (unlike on userdebug builds).
Let's use sched/sched_switch instead. The exact event doesn't matter, we
just need an event that can be successfully enabled.
History of the test:
* On android 12, the `PerfettoFtraceIntegrationTest` suite was happily
included on CTS tests and was passing on user and userdebug build.
* We mistakenly removed suite from CTS on android 13 with
https://r.android.com/1952222, but it was kept in
`perfetto_integrationtests` (which were only run on **userdebug**).
* We added a new test to the suite in https://r.android.com/2027444.
* I readded the `PerfettoFtraceIntegrationTest` in
https://r.android.com/2612391
Tested by running:
```
atest CtsPerfettoTestCases:PerfettoFtraceIntegrationTest#ReportFtraceFailuresInStats -- --abi arm64-v8a
```
on a user master build.
Bug: 288274277
Bug: 291679694
Bug: 289783494
Bug: 291870885
(cherry picked from https://android-review.googlesource.com/q/commit:d04c69d715d45a75c9ccc44e527b8951570eed10)
Merged-In: Ifea07f718bbbd4917170604c3470475811e48b3b
Change-Id: Ifea07f718bbbd4917170604c3470475811e48b3b
(cherry picked from commit 11dc17dc6ad015caf8c3d208f318d5d6ec572a3e)
|
|
5c8d324994 -s ours
am skip reason: subject contains skip directive
Original change: https://android-review.googlesource.com/c/platform/external/perfetto/+/2665397
Change-Id: I56db9897a3f0089cf75b72f34ed16d86d782868a
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
While merging this from `android12L-test-dev` to `main`, the `data`
field got duplicated.
Merge the duplicated entries into a single one.
Change-Id: Ie97857cf1d4a41fd479e100616b533265c807de8
Merged-In: Ie97857cf1d4a41fd479e100616b533265c807de8
Bug: 285215697
(cherry picked from commit d4482a5a8d4b70722f8c573290661bfa6e85d368)
|
|
Original change: https://android-review.googlesource.com/c/platform/external/perfetto/+/2652679
Change-Id: I5446f4aae8906b74edb36d10d2713f5d23d9bcb6
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
Original change: https://android-review.googlesource.com/c/platform/external/perfetto/+/2652679
Change-Id: I423fa2fdb2483ad4539042f346c24a8f03552097
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
The CDD doesn't enforce traced to be enabled (`persist.traced.enable`)
by default in all device types. In fact, only handheld is required to
have do to so.
Originally, the test assumes that traced is enabled by default, which
may not be true for all device types.
Improve that by temporarily enabling traced if it's not enabled, run
the tests and disable traces again if enabled by the test case.
Bug: 285215697
Change-Id: I3e93e4b93a0349fe3d9f790525ae61a0f484bc97
|
|
This CL introduces support for filtering strings in traces (i.e by
redacting contents) using regexes. It also adds faster filtering
primitives for atrace which is expected to be where this code is used
the most.
(cherry-picked from aosp/2629209)
Bug: 283246642
Change-Id: Ie3624e4066bbd3d59fbde3225e5a368c18b46b18
Merged-In: Ie3624e4066bbd3d59fbde3225e5a368c18b46b18
|
|
This CL introduces a new bytecode v2 format which adds a new opcode
for filtering strings in protos. This CL does not implement the actual
filtering logic yet, instead making the field a noop at the moment.
(cherry-picked from aosp/2616329)
Bug: 283246642
Change-Id: I4e1dfbeafa7b7728faf3dc063f868c9a36f1672a
Merged-In: I4e1dfbeafa7b7728faf3dc063f868c9a36f1672a
|
|
This is a pure refactor CL which makes it easier to reason about the
introduction of string rewriting.
(cherry-picked from aosp/2615453)
Bug: 283246642
Change-Id: I405d2cff15c738e76c51291a58fdc95553315a22
Merged-In: I405d2cff15c738e76c51291a58fdc95553315a22
|
|
This could cause an invalid bytecode to silently be accepted as we
were not checking correctly for the end of message.
(cherry-picked from aosp/2615174)
Bug: 283246642
Change-Id: I3d03f92f5a3ee36c680c7eae6b5f020baa03eb90
Merged-In: I3d03f92f5a3ee36c680c7eae6b5f020baa03eb90
|
|
|
|
|
|
The LSB is used to correlate tracing sessions with statsd subscriptions.
Retain the LSB to avoid breaking downstream pipelines that depend on this.
Bug: b/260112703
Bug: 287071990
Test: TracingServiceImplTest.CloneSession
(cherry picked from https://android-review.googlesource.com/q/commit:8755774a3fbe4e8c664a6436eb4fd9ea5e34a297)
Merged-In: I4c512701143a21ac1b01a2f785ebe005e9983556
Change-Id: I4c512701143a21ac1b01a2f785ebe005e9983556
|
|
Prior changed introduced generalized support for CloneSession
and rewrote the --save-for-bugreport feature on top of that.
Unfortunately we didn't notice that the CloneSession feature
is very strict on UID checking, enforcing that source and target
consumer UID match. This is good in general but not desirable
(And a regression) for the case of bugreports, where the two
uids differ.
In this case bugreport_score acts as an "I am fine exporting
this trace to other uids".
Bug: 286908237
Bug: 260112703
(cherry picked from https://android-review.googlesource.com/q/commit:5457722190e8cad7dddbedc71b8957ab45402a6c)
Merged-In: I99bd22aef0bfa402cb24559b8515c5bdf0b48ba9
Change-Id: I99bd22aef0bfa402cb24559b8515c5bdf0b48ba9
|
|
There are a bunch of end to end tests in test:integrationtests
(android_integrationtest.cc, cmdline_integrationtest.cc,
ftrace_integrationtest.cc, traced_integrationtest.cc).
These tests are linked in perfetto_integrationtests, which is run in CI
on a couple of devices.
I think the intention in the code base was for these tests to be
included in CtsPerfettoTestCases. All devices have to pass CTS, so that
gives us a better guarantee that perfetto is working.
Unfortunately, our build files did not include these tests correctly
into the CtsPerfettoTestCases.
Before this commit
GN:
```
//test/cts:
static_library("perfetto_cts_deps") {
deps = [ "test:integrationtests" ],
}
//test:
source_set("integrationtests") {
sources = [
"android_integrationtest.cc",
"cmdline_integrationtest.cc",
"ftrace_integrationtest.cc",
"traced_integrationtest.cc",
]
}
```
BP:
```
//test/cts/Android.bp (NOT autogenerated)
cc_test {
name: "CtsPerfettoTestCases"
static_libs: ["perfetto_cts_deps"]
}
//Android.bp (autogenerated from GN)
cc_library_static {
name: "perfetto_cts_deps"
srcs: [ "perfetto_test_integrationtests" ]
}
filegroup {
name: "perfetto_test_integrationtests"
srcs: [
"android_integrationtest.cc",
"cmdline_integrationtest.cc",
"ftrace_integrationtest.cc",
"traced_integrationtest.cc",
]
}
```
The files in the filegroup `perfetto_test_integrationtests` were
included into the cc_library_static `perfetto_cts_deps`. When
CtsPerfettoTestCases links `perfetto_cts_deps`, it doesn't include all
the objects, it includes only the one that strictly needs and it
excludes the one that only register initializers (i.e. there's no
--whole-archive).
This commit fixes the problem by listing the _integrationtest.cc
explicitly in `CtsPerfettoTestCases`.
In order to do that, this commit:
* Renames end_to_end_integrationtest_cts.cc to
producer_to_consumer_integrationtest_cts.cc so that the name
end_to_end can be reused later.
* Renames the source_set //test:integrationtests to
perfetto_end_to_end_integrationtests, so that it can be exported to
Android.bp (exported targets are not prefixed with their path).
Before this change CtsPerfettoTestCases was taking ~2 minutes and 30
seconds on my redfin. After this change CtsPerfettoTestCases is taking
2 minutex and 40 seconds, so this commit also updates the timeout in the
XML.
This is the diff between the old CTS test list and the new one:
```
HeapprofdJavaCtsTest.
DebuggableAppRuntime
ProfileableAppRuntime
ReleaseAppRuntime
DebuggableAppRuntimeByPid
DebuggableAppOom
ProfileableAppOom
ReleaseAppOom
DebuggableAppOomNotSelected
HeapprofdCtsTest.
DebuggableAppRuntime
DebuggableAppStartup
ProfileableAppRuntime
ProfileableAppStartup
ReleaseAppRuntime
ReleaseAppStartup
NonProfileableAppRuntime
NonProfileableAppStartup
JavaHeapRuntime
JavaHeapStartup
ProfilePlatformProcess
TracedPerfCtsTest.
SystemWideDebuggableApp
SystemWideProfileableApp
SystemWideNonProfileableApp
SystemWideReleaseApp
ProfilePlatformProcess
PerfettoDeviceFeatureTest.
TestMaxCpusForAtraceChmod
PerfettoCtsTest.
TestProducerActivity
TestProducerService
TestProducerIsolatedService
+PerfettoAndroidIntegrationTest.
+ TestKmemActivity
+ TestBatteryTracing
+PerfettoCmdlineTest.
+ InvalidCases
+ Version
+ TxtConfig
+ SimpleConfig
+ DetachAndAttach
+ StartTracingTrigger
+ StopTracingTrigger
+ NoDataNoFileWithoutTrigger
+ StopTracingTriggerFromConfig
+ TriggerFromConfigStopsFileOpening
+ Query
+ CmdTriggerWithUploadFlag
+ TriggerCloneSnapshot
+ SaveForBugreport
+ SaveForBugreport_WriteIntoFile
+ Clone
+ UnavailableBugreportLeavesNoEmptyFiles
+ DISABLED_SaveForBugreport_Triggers
+PerfettoFtraceIntegrationTest.
+ TestFtraceProducer
+ TestFtraceFlush
+ KernelAddressSymbolization
+ ReportFtraceFailuresInStats
+PerfettoTracedIntegrationTest.
+ TestFakeProducer
+ VeryLargePackets
+ UnresponsiveProducer
+ DetachAndReattach
+ ReattachFailsAfterTimeout
+ TestProducerProvidedSMB
+ QueryServiceStateLargeResponse
+ TraceFilterLargePackets
```
Bug:286867751
(cherry picked from https://android-review.googlesource.com/q/commit:5567443fceff3fe08c16c3241cdd7ec96e3c4297)
Merged-In: Idcd04d6239f276b95abeca09ebb0891d5cf98a3f
Change-Id: Idcd04d6239f276b95abeca09ebb0891d5cf98a3f
|
|
(cherry picked from https://android-review.googlesource.com/q/commit:011149e55d3e2184d1c5256493ba422f7cce3282)
Merged-In: Ia692949743526f36182a0bd272dd651bb5dd8ac4
Change-Id: Ia692949743526f36182a0bd272dd651bb5dd8ac4
Bug: 260112703
|
|
The reporter test only works with a trace that's smaller than 1Kb.
Recent additions to the tracing service (chunk usage stats) caused the
final trace size to be bigger than that on some devices or builds (the
device and build names are included in the trace).
This commit fixes the problem by disabling a bunch of tracing service
feature. This reduces the tracing size on redfin from 1020 bytes to 400
bytes.
Bug: 282508742
(cherry picked from https://android-review.googlesource.com/q/commit:f3935233188da29f7f772f7685cbc4d1e014f06b)
Merged-In: I6c17948e7983828eb4f11bf760f71d8055b777ba
Change-Id: I6c17948e7983828eb4f11bf760f71d8055b777ba
|
|
expectations am: 8c5b81d9ac -s ours
am skip reason: Merged-In I20769f4ce47a2af4c90b87e4bfb709468b4568cf with SHA-1 d0b6ef14f3 is already in history
Original change: https://googleplex-android-review.googlesource.com/c/platform/external/perfetto/+/23410770
Change-Id: I4bcb8f7512cad7ec1a007c04f343d01bbb61e9d4
Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
|
|
Kernel commit 0b04d4c0542e("f2fs: Fix f2fs_truncate_partial_nodes ftrace
event") changed the format of an event in a way that causes perfetto to
hit a PERFETTO_FATAL.
The PERFETTO_FATAL was probably not intentional there, it was added
(probably by mistake) by 6db99a9ca272("Add PERFETTO_DFATAL.").
In any case, perfetto shouldn't crash for an event it doesn't
understand.
Bug: 281660544
(cherry picked from https://android-review.googlesource.com/q/commit:f1ee130d2d4fa37f3c57b4f05019d664ae6523a5)
Merged-In: I20769f4ce47a2af4c90b87e4bfb709468b4568cf
Change-Id: I20769f4ce47a2af4c90b87e4bfb709468b4568cf
|
|
Bug: 260112703
Bug: 278884624
Change-Id: I0d7fdad29902e683a40bdade9395da8b0a759b88
|
|
|
|
|
|
Change-Id: Idd713676f7beda9c3fb971c07f35a86d1ff0d8a0
|
|
Follows ag/23263017.
This brings internal<>AOSP in a state where the only difference
is statsd.cc, empty in AOSP, full content in udc-dev and int-master.
Bug: 271926639
Change-Id: I181f3041454de5dd43ad8666a4fe1c6f6d2adcdd
|
|
|
|
|
|
Otherwise test retries will hit the guardrails and fail.
Bug: 282508742
Change-Id: I76b806e119267c5e9c133775b9ac4576c3975ae1
|
|
Bug: 271926639
Change-Id: I800955331824d55d25de5ffae720c03cc93e12f4
|
|
This will add more context to the monitor contention metric and
allow joining with other CUJ metrics to understand the impact of
monitor contention on various CUJs
Test: tools/diff_test_trace_processor.py out/android/trace_processor_shell
--name-filter '.*monitor.*'
Change-Id: I83dabd7aa3f2cb3cce26db16c0bcd148d4a471a6
|
|
|
|
|
|
|
|
|
|
* changes:
Add a template metric for network activity query.
Fix battery stats function name and consistency
|
|
Change-Id: Ia3a99f226d018aad35ba3a6db103c97188062d5b
|
|
integration "
|
|
Change-Id: I8d3f361eb560a68abb622c47e17e7899b18ff891
Fixes: 282723377
|
|
|
|
|
|
|
|
integration
Change-Id: I890d42cff356df2b54c1ecd8d495f85beeba1cb1
|
|
|
|
Bug: 281694530
Change-Id: Ieb5d4fd580f51ad4241dfba6a5d54ca90b5c64aa
|
|
Change-Id: Ieedd86509a1cca6e4b4f7041435ec97e4e796147
|
