Gradle Enterprise is now Develocity. Learn more about our rebrand here.
Gradle Enterprise 2019.5
Develocity 2019.5 helps deal with the problem of flaky tests by identifying and analyzing flaky tests in individual builds and across many builds, and processes incoming build scans much more efficiently.
Read on for more details.
Gradle and Maven build scans now show a test as FLAKY when it was executed multiple times and reported at least 1 FAILED and 1 PASSED result, such as when using the new Test Retry Gradle Plugin or specifying
rerunFailingTestsCount when using the Maven Surefire or Maven Failsafe plugin.
In addition, the top tests report in the Tests Dashboard now shows the tests that were the most flaky for your selected set of builds, along with an overall flaky tests trend chart.
As you’d expect, you can drill down to see which methods are flakiest and list recent builds where a single test was flaky or failed.
Learn more in this blog post.
When a Develocity enabled build is run, it publishes a record of what happened to the Develocity server. The server then processes the raw data and transforms it into a form suitable for long term storage. This process involves aggressive deduplication of data and other transformations to minimize the storage space requirements.
Organizations that perform many builds each day and/or have large builds that do a lot of work can generate significant amounts of data to be processed, which can strain the Develocity server and lead to delays in the availability of build scans after upload.
This data ingestion process has been improved to be more efficient in Develocity 2019.5. This reduces the processing latency and reduces the CPU and memory usage of the server when processing large volumes of data.
Temporarily degraded performance due to data reindexing
Upon upgrading, a data reindexing process will be initiated in the background with Develocity being usable for its duration. CPU usage will be increased and performance may be slightly degraded. For large installations storing many build scans, the reindexing process may take several hours. During this time some builds may be omitted from the Tests dashboard.
- [FIX] Replicated CVE-2020-10590
- [FIX] Phase filter in Maven build scan timeline section may contain empty value
- [FIX] Open/close stacktrace toggle is shown in build scan when exception has no stacktrace
- [FIX] Build deletion is too slow for installations with many recorded build scans
- [FIX] Focussing on deprecations in Gradle build scans is unreliable
- [FIX] Execution totals in Gradle build scans performance section are omitted when value is 0s
- [FIX] Task input comparison does not work with FireFox or Safari
- [FIX] Task input comparison incorrectly visualizes file name differences of classpath inputs
- [FIX] Tests dashboard does not include failed/flaky sort column choice when sharing page
- [FIX] Dependency comparison section is rendered incorrectly
- [FIX] Internal HTTP router logs are not garbage collected
- [FIX] Absolute task/goal inputs only differing in content are incorrectly denoted to have different paths
- [FIX] Task/goal input comparisons involving non existing files in one build do not render
- [FIX] Task/goal input comparisons involving duplicated file trees do not render
- [FIX] The first line of console log output cannot be linked to
- [FIX] Test execution summary incorrectly disregards class name of test
- [FIX] Build scan list sort order is incorrectly reset when search criteria is changed
- [FIX] SAML descriptor used to configure external IdP contains incorrect system address
- [FIX] Platform dependencies are properly captured in Gradle 6.2
- [FIX] Test duration in build scans are inaccurate
- [FIX] Test names containing underscores cannot be searched for in the tests dashboard
- [FIX] Builds occurring in an unknown timezone may fail to be indexed
- [FIX] Composite builds where an included build has a buildSrc project may fail to load
- [FEATURE] Tests dashboard visualizes flaky tests
- [FEATURE] Flaky tests are shown in build scans
- [FIX] Build scan ingest processing can be more efficient
- [FIX] Tests dashboard help does not convey that build count is constrained to builds that ran tests
- [FIX] Tests dashboard allows analyzing not only failed builds
- [FIX] Tests dashboard does not forward outcome criterion to data endpoints
- [FIX] Tests dashboard database indexes can require less storage space
- [FIX] Failures dashboard corrupts state of the outcome search filter
- [FIX] After encountering an error in build scans, going back doesn't restore a working state
- [FIX] Misleading zero console log line count is shown when console log is too big
- [FIX] Font rendering performance can be improved
- [FIX] Error is thrown when pressing enter on timeline search
- [FIX] Gradle artifact transforms create excessing logging in build scans