Improved network failure handling in ew-cli. We now retry failed requests in a more robust way, and we also increased
some of the longer network call timeouts (like file uploads). This should help in cases where you are running many tests
in parallel from the same CI job.
Fixed an escaping issue where environment variables containing the backslash \ character were
not passed correctly to the test runner. You were most likely to encounter this when using
regexes in your test runner arguments.
Added a workaround for a case where a single test method was passed as a --test-target and
it expanded to multiple parametric test cases. This was previously causing the test run to fail
and works now without issue.
ew-cli 0.10.8
The CLI is now better at printing some unexpected errors, for instance when misconfiguring the API token.
All of our emulators are now in the Atlantic/Reykjavik timezone (same as UTC).
This should improve compatibility with tests that use the system timezone identifier
and don’t support the Etc/UTC timezone identifier.
If your test manages to crash the surfaceflinger process and the whole System UI with it,
we’ll trigger an automatic retry of the test run. This is a rare event that may happen with
tests making heavy use of surfaces like utilizing the camera preview. The retry does not add any
billable minutes to your account. NOTE: just like with flaky repeats, we do not attempt any retries
when using the side-effects flag.
The test results link now always displays the model/version combination, even if the test run was
run on a single combination.
gradle-plugin 0.16.0
The Gradle plugin now clears the output directory before running, fixing issues where you
would get mixed outputs from multiple test runs when running the task repeatedly.
gradle-plugin 0.16.1
Updated the transitive org.json library to a newer version.
Fixed an issue where sharding together with the useTestStorageService flag would run all the tests again in
every shard. NOTE: there is a known issue about AndroidJUnitRunner 1.5.2 or earlier, we will now
warn you if you happen to have this combination.
Also fixed a case where --test-targets would not work as expected when used together with balanced sharding.
We now display branch names and PR links in the test results list view. This makes it easier to understand which branch and PR a test run is associated with. The PR link is clickable and will take you to the PR on GitHub. This feature is available for test runs started after this release.
Emulators on API levels 33 and 34 can now benefit from GPU acceleration. Use the gpu=auto flag
(e.g. --device model=Pixel2,version=34,gpu=auto) to enable GPU acceleration. Read our blog post on
the benefits of GPU acceleration here.
Disabled GPU acceleration for Automated Test Device images. These devices do not benefit from
GPU acceleration, and it was causing issues.
When using the --json flag or the summary output format, the consumed test run minutes are now available in the JSON outputs of test runs. The consumed test run minutes are available under the billableMinutes key in the JSON output.
This is useful when estimating the cost of using emulator.wtf in the long-term. Read more on how
we calculate billableMinuteshere.
emulator.wtf web app
The Usage view in the emulator.wtf web app now defaults to the past 30 days instead of showing usage
from the beginning of the current calendar month.
Fixed a bug where canceled test runs during the apk upload phase would show up as perpetually running in
the emulator.wtf webapp test runs list view.
The audit log can now be exported either as a CSV or a JSON file. The file contains
a log of all the actions taken by users in the emulator.wtf web app. NOTE: audit log is an enterprise plan only feature.
Added validation of uploaded APK files to prevent corruption issues due to incomplete
uploads. This issue compounded with the test file cache where we would keep the corrupted
apks around and future test files would be corrupted as well.
Organization owners can now disable public test result links. This is useful
when you want to keep your test results private and only accessible to your team members.
NOTE: the setting does not apply retroactively, so any existing public test result links will remain accessible.
