As part of our evolving performance testing strategy, we are adding pixel6 devices to test on
One day these may replace the a51s but that is not known for certain
This patch adds the ability to run tests on bitbar pixels
Differential Revision: https://phabricator.services.mozilla.com/D191229
This patch addresses a few inefficiencies with the current testing setup
for CaR tests and builds. In particular we need to reduce the frequency
of the CaR toolchain builds for the OSX platform and not run on every
m-c push. Instead OSX will run once a day.
Also we don't need to schedule the chromium update cron anymore as that
is automatically fetched through the dependencies of the CaR-sp3 tests
and we don't run chromium tests anymore.
Additionally, we bump up the tier of the OSX arm64 CaR build and tests
from tier 3 to 2.
Differential Revision: https://phabricator.services.mozilla.com/D198305
Update the chrommium fetch cron to now include the ARM64 Mac OS relevant
files. This ensures we will have available the correct chromedriver for
CaR on Apple Silicon.
Differential Revision: https://phabricator.services.mozilla.com/D193865
This patch modifies the target tasks so that we don't run the sp3 tests
more than once a day for the Android custom-CaR application. This is for
consistency as this is already taken care of in the
`custom-car_perf_testing` which handles custom-car for all platforms.
Differential Revision: https://phabricator.services.mozilla.com/D192518
This patch updates the CaR cron with OSX builds.
Additionally, the time between building and running perftests is increased from 8 hours to 9 hours,
just to provide a bit more of a buffer in case of failures and in particular for OSX builds which take quite long.
Differential Revision: https://phabricator.services.mozilla.com/D180826
1. Mark 32-bit, `asan`, `tsan` tasks as `UNCOMMON_TRY_TASK_LABELS`,
which means they'll only be presented for selection when `--full` is
passed to Try.
2. Just run Mochitest on every `trunk` push, minus the platforms we disable.
Co-Authored-By: Kelsey Gilbert <jgilbert@mozilla.com>
Differential Revision: https://phabricator.services.mozilla.com/D179816
1. Mark 32-bit, `asan`, `tsan` tasks as `UNCOMMON_TRY_TASK_LABELS`,
which means they'll only be presented for selection when `--full` is
passed to Try.
2. Just run Mochitest on every `trunk` push, minus the platforms we disable.
Co-Authored-By: Kelsey Gilbert <jgilbert@mozilla.com>
Differential Revision: https://phabricator.services.mozilla.com/D179816
This patch adds a daily cron task to build the custom chromium-as-release (CaR) application and running
raptor browsertime tests with it.
Differential Revision: https://phabricator.services.mozilla.com/D177545
This patch should disable fenix/geckoview *fission* performance tests only.
The tests are still available to select through Try.
Differential Revision: https://phabricator.services.mozilla.com/D171133
There are no jobs remaining on 10.14, but we still had some lingering
supporting code.
Also change the treeherder label for the generate-profile task to match
the actual worker type.
Differential Revision: https://phabricator.services.mozilla.com/D170735
Instead of excluding android builds based on the `project` parameter, use
`release_type`, which is properly set to e.g. `esr102` by `mach try release`.
Differential Revision: https://phabricator.services.mozilla.com/D168563
This patch updates the S3 commit used, fixes an issue with local Windows benchmark runs, and enables the windows S3 tests in CI.
Differential Revision: https://phabricator.services.mozilla.com/D166323
This patch adds the speedometer 3 benchmark test to raptor enables it for CI. In CI, it'll start by running on Linux, and MacOSX on Firefox, Chrome, and Chromium. The patch also adds a new method for parsing benchmark tests that is more generic and works in conjunction with the `custom_data` option.
Differential Revision: https://phabricator.services.mozilla.com/D165988
Now that we can run performance tests on Safari, this patch adds these tests to the general_perf_testing cron.
This let's us run the existing Safari live and benchmark tests at the same frequency as Chrome tests.
Differential Revision: https://phabricator.services.mozilla.com/D163987
The `always_target` attribute is a feature which I believe I previously
mis-implemented. It doesn't belong in `generator.py` and should instead be
implemented somewhere in `target_tasks.py`. I chose to wrap the registration
decorator to guarantee backwards compatibility within Gecko. Though we may wish
to move it out to a filter in the future.
I'm making this change now to facilitate merging standalone and Gecko
taskgraphs. The `always_target` feature will be removed from standalone (as it
isn't being used and should consumers need something like it, they can roll
their own implementations).
With the feature removed from both Gecko and standalone in generator.py, this
file will now be identical across both Taskgraphs and can therefore be removed
from Gecko.
Differential Revision: https://phabricator.services.mozilla.com/D159181
