Age | Commit message (Collapse) | Author |
|
Change-Id: I30c4990152921741b7ffd435a0864f63dcd74c3d
|
|
A previous CL caused perfetto_enable_git_rev_version_header to
become unexpectedly false in the amalgamated build.
That is undesirable as omits the GIT commit in the version.
The last version (v20.1) had it.
Test: manual (while following the SDK release docs)
Change-Id: If4e0e00f80bd5a16e52b3a3c0038c8c681ccfc33
(cherry picked from commit adb22cfb36ccc1a8c485d6f98b6cbc42481c3887)
|
|
Change-Id: Idd2cbdbb2fbd62e5fc01893b782bd2f863da3bff
(cherry picked from commit ea72787a8b9beb5b12a238e2ce39441fcc2b6d0f)
|
|
|
|
|
|
|
|
Compared to Catapult, Perfetto UI lacks immediately visible "?" button
showing basic controls. The corresponding menu item exists and available
when clicking "Support" group, but user has to experiment with the UI in
order to be able to find it, which has been a problem in practice (the
problem has been brought up by a user in response to request for
feedback).
This CL makes the "Support" section expanded by default and swaps it
with "Sample queries"; without the swap "Sample queries" is collapsed
in-between expanded section which makes the menu look a bit weird.
Bug: 204404247
Change-Id: I8aaad88ed13a504f67cda9226b83d78133fa04a4
|
|
If saved in local storage serialized recording configurations, instead
of just silently repairing them, the typed deserializer will now also
retain information about extra fields and fields with invalid values
(keys that are present but failed to validate, e.g. because of erroneous
type)
Bug: 204041135
Change-Id: Ic0765272c3b4bd734ecaf9121c364a8a33bcc667
|
|
|
|
|
|
This prepares the metric for backporting to API 23.
Change-Id: I4f96954ba2fcf0d66bc7d3fbdadc53e9cb599e85
Bug: 190219056
|
|
Bug: 202010519
Change-Id: I8d7c69edbe4e5850aea0095fb4b30336e1aa513a
|
|
post_message_handler."
|
|
|
|
|
|
This should not have any effect, it just fixes a warning reported by
clang-tidy.
Change-Id: I460424e33386935848e794b2e08bb31f28988020
|
|
http://android-review.googlesource.com/1824203 updated libunwindstack
for android in-tree builds and broke standalone builds.
This updates libunwindstack for standalone builds.
Change-Id: Ic3bc2eae577ea2ffc34605a64ffff1211a21cd7d
|
|
|
|
post_message_handler.
Without this, URLs that contain + have that sanitized to ' ', which
makes the URL invalid!
Bug: b/204125489
Change-Id: Ic6053d953e1b03261b5022c7a2eb4c343384776e
|
|
It is useful to distinguish renderers hosting an extension when
analyzing resource usage in traces.
Companion CL: https://chromium-review.googlesource.com/c/chromium/src/+/3252934
Change-Id: I78ff7801e7a4aa7bd6043a74c46b42d6e7815a06
|
|
Bug: 120606663
Test: Unit tests pass.
Change-Id: I27eb820546308048be2766bf1173bf5ba06d3d70
Merged-In: I27eb820546308048be2766bf1173bf5ba06d3d70
(cherry picked from commit 4a216d7154cfbc36b95f4cef84343273126ad35a)
|
|
Plumbs a WatchdogCrashReason crash key to disambiguate
the reason of a watchdog-related crash.
Bug: 202283508
Test: manual
Change-Id: I911fd606e83dd155102449bba1885ded8f8ab62c
|
|
perfetto_cc_protocpp_library"
|
|
This was misleading some users into thinking that lightweight mode
performed differently on-device than normal mode. This update
changes the name of that mode to "simple mode" and clarifies that
there is no behavior change compared to "normal mode".
Bug: n/a
Change-Id: Iff2dbc038a90b8d57278d73218ec3d48650483b0
|
|
The libperfetto static library might be used outside of its build
directory, therefore we set its install flag to true and avoid to
generate just a thin archive [0].
[0] https://github.com/mesonbuild/meson/commit/0a3a9fa0c3ebf45c94d9009a59cead571cbecf7b
Change-Id: If47189deee996a9682846e0f2016cc54d05d7a60
|
|
|
|
This hides the GPU and Android probes from the recording page if target
OS is Chrome OS or Linux.
Bug: 204405690
Change-Id: I004c7f7f9064163903648916f402d79896c772e7
|
|
Amending CL:1867790, the two macros need to support overriding more
rules in internal google builds as well, for other targets such as
traced and traced_probes, to propagate non_prod rules internally.
R=lalitm@google.com, primiano@google.com
CC=chinglinyu@google.com, eseckler@google.com, skyostil@google.com
Change-Id: Icfbc746048d9addb4080601104eccfac290bb1b4
|
|
There are no images for Chrome categories presets, so there are a lot of
blank space in blocks with probe selectors at the moment. This CL fixes
that issue by introducing a more compact way to display a probe and
using it in Chrome categories page.
Bug: 202010687
Change-Id: I5bc0436efe2656756d449eaf091b95bd217d1e37
|
|
|
|
|
|
|
|
This CL aims to resolve a frustrating scenario with having to
re-configure recording settings from scratch in case a tab is reloaded.
In order to remedy that, two features are implemented:
1. Starting recording would automatically save recording profile,
available to load in the recording profiles list when the appropriate
flag is enabled.
2. Introduce a separate flag to automatically load the profile from
previous item. I've decided to put it behind a flag to avoid changing
the default behaviour; I plan to change this later according to
feedback.
Bug: 160016057
Change-Id: I2549b8ceb6ee23749c52079b47854694c9bc494b
|
|
|
|
Adds support for overriding more rules in internal google builds.
This is to propagate non_prod rules internally to all sub-targets
that come from AOSP.
R=lalitm@google.com, primiano@google.com
CC=chinglinyu@google.com, eseckler@google.com, skyostil@google.com
Change-Id: Iec046b8813b04b2cfe798f748470ce7a2293cbf7
|
|
- Disable the watchdog in bazel: the watchdog should not be enabled
for SDK users, only for services. There doesn't seem to be a good
reason for having it enabled in bazel, and it causes problems
in Google-internal builds (See bug)
- Disable the crash-log feature on Android versions < API 19.
We target minsdk API 21 in GN builds. However one internal
blaze user today targets API19. For now disable the crash
log feature doing minsdk detection until the root problem
is solved.
Bug: 203795298
Change-Id: Ief5f4249d90620f136a03929b9d456d07ee05e4f
|
|
Increase core count limit for CtsPerfettoTestCases
PerfettoDeviceFeatureTest#TestMaxCpusForAtraceChmod to support
future devices with more cores
Bug: b/203651019
Test: atest CtsPerfettoTestCases
CQ-DEPEND: CL:1867150
Change-Id: I2246d9ef10e553df3f9d12ccde78d8009f3fd7cd
|
|
|
|
Change-Id: I58fe5ca324da03ae2d6e3422cdb21f074e23c2b8
|
|
The values in the trace which crashed are the following:
https://screenshot.googleplex.com/7jqJw22JGrqw54D
Bug: 203613535
Change-Id: I3ed1975a17390097950031446af7a1bd243717a3
|
|
|
|
|
|
|
|
|
|
When TracingMuxerImpl gets reinitialized by ResetForTesting(), we
shouldn't try to use any TraceWriters that were active for the
previous instance, because their data would just get dropped on the
floor. To fix this, keep track of which muxer instance TraceWriters
were associated with, and recreate them if there is a mismatch.
Bug: 203421836
Test: perfetto_integrationtests --gtest_filter='PerfettoApiTest/PerfettoApiTest.TrackEventArgs_*'
Change-Id: Ic4802f759f83fc074ffc5497ff07aa7a1b3bbbdd
|
|
Avoid registering the built-in producers on other builds.
That is particularly misleading on Linux / CrOS, where the real
producers don't exist and perfetto --query just gives false hopes.
These builtin producers exist only to dynamically spawn the real
services backing them on-demand when a tracing config lists them
(this is to avoid the memory cost in Android low-end devices of
heapprofd / traced_perf).
Change-Id: I3e35ad69380461bb3c923b2a3b7caf6a80a20510
|
|
Before this CL TypedProtoDecoder ensured that the on-stack
storage is enough to cover the highest field id parsable by
protozero decoders. That lead to large stack frames of
1000 * 16 = 16KB per decoder, which in turn caused problems
in various internal usages (b/203389087, b/194510398).
This CL adds a little bit of extra complexity, allowing the
on-stack storage to be < than the highest known field id,
allowing to reduce the stack size by 10x, still reaping the
benefits of direct-on-stack decoding for most of the trace
protos.
Bug: 203389087
Test: manual:
$ parallel -j10 -N0 time trace_processor_shell-before -q /dev/null mpts_trace ::: {1..10}
$ parallel -j10 -N0 time trace_processor_shell-after -q /dev/null mpts_trace ::: {1..10}
On a 266MB MPTS trace:
Average: before: 14.6191, after: 14.6866
Median: before: 14.6385, after: 14.5365
Change-Id: I8a51e11f1a1deb31df4bb02412d7fa0efb7277af
|
|
|
|
Also switch some usages of sprintf to base::StackString<N>
Bug: 189749310
Change-Id: Iaceea6dc063f1257b286c9e244c51350bfe45041
|
|
|