2.1.3 - April 18, 2019¶
- Fixed an issue where super accessors may be missing depending on the order of compilation.
- sbt-hydra smoothly falls-back to compile with vanilla Scala if the
hydra.licensefile is not present.
FileNotFoundExceptionwhen saving the output of
hydraBenchmarkreports the files where measurements are stored.
2.1.2 - April 5, 2019¶
- Fixed source generation race-condition in sbt-hydra that could result in "duplicate source" compilation errors.
- Improved heuristic to support installation of sbt-hydra in the global plugins.
- Also use the sources count to set the
HydraTagcompilation's task weight (allows better parallel execution in sbt)
2.1.1 - March 7, 2019¶
- Support for installing sbt-hydra in the global plugins
- Increased sbt-hydra parallel execution factor to use all available cores.
- Check for Hydra updates at startup
- Added hydraStartTrial command to request trial license from inside sbt
- OpenJDK9 support for activating Hydra licenses
- Issue a warning in hydra.log when non-standard phases are detected
- Unbalanced workers warning was sometimes noisy. Now it is reported in hydra.log
- Improved treatment of nested imports that was leading to cyclic reference errors
- Made package objects that inherit from traits more robust
- Fixed an issue where Hydra could hang or take a very long time to invalidate long HList types
- Fixed a spurious cache invalidation issue that could lead to longer compilation times involving class tags
- Fixed a Hydra crash involving partial functions with inferred return type
- Fixed cache invalidation issue with local definitions.
- Added license management to the Hydra settings panel
- Allow requesting a trial license from the Hydra settings panel
- Added cache invalidation button to Hydra settings panel
2.1.0 - Didn't happen¶
2.0.0 - December 6, 2018¶
- Fix error reported when certain classes are deleted
- Improve performance when compiling code with dependent types
2.0.0-RC03 - November 30, 2018¶
- Fix assertion failure "The startSource can be called only once per source file"
- Fix issues with spurious cache invalidations, leading to worker restarts
- hydraBenchmark saves complete benchmark results inside .hydra, including Hydra version, number of workers and Scala version
- Location of .hydra is configurable, see Sbt plugin user guide
- Fixed a StackOverflowError in certain kinds of code
2.0.0-RC01 - November 22, 2018¶
- Smarter cache between builds, leading to up to 2x speedups compared to the previous version
- Fixed a bug when multiple routes files with the same name were used in Play projects
- Fixed a resolution bug in dashboard client when its dependencies might get replaced by the current project dependencies
- Clearly report failures that are due to a stack overflow or a out of memory error.
- Add a command to invalidate caches
- Report the amount of memory used by Sbt on start
1.1.1 - September 3, 2018¶
- Performance improvements in compiling programs written in a pure FP style
1.1.0 - August 21, 2018¶
- Significant performance improvements in Slick-based projects, and generally type-level Scala programs
- New metrics client configuration, see Dashboard architecture
- Released scala-maven-plugin
3.2.2-hydra08 that correctly launches the new metrics service.
1.0.3 - August 9, 2018¶
- Fix crash when launching the metrics service in 2.11 projects
1.0.2 - July 17, 2018¶
See upgrade instructions for how to stop the metrics client
- Hardened assertion that sometimes could fail compilation if symbolic links are used for source directories
- Don't fail build if all floating licenses are in use, instead use a single worker.
- Limit metrics pushed to the server to only the slowest 20 source files
- Issue license check warnings at info, so
-Xfatal-warningsdoesn't fail the build
- Issue warnings about common performance problems
- Fixed high cpu consumption occurring when the dashboard server is not reachable.
- Capped metrics service logging to 100MB.
- Metrics service is no longer started on build load.
- Metrics service shutdown after 1 hour inactivity (previously, it would never shutdown)
- Metrics client uses exponential backoff if the server does not respond.
1.0.1 - June 12, 2018¶
- Fixed regression in the sbt-hydra plugin that was breaking tasks aggregation of the automatically generated root project (the issue was caused by this sbt bug).
1.0.0 - June 6, 2018¶
- A new "Deck" dashboard that provides a quick glance at how your project is performing.
- Fixed an issue with the order of Hydra compiler arguments in Sbt 0.13 when using
- Fixed an issue where sources were type-checked even when newer classes files were available
- Fixed an issue where warnings about unused variables may crash when no position information was available
- First release of the Gradle Hydra plugin.
0.11.0 - May 30, 2018¶
- New tips for tracking down compile time inefficiencies.
- Renamed system property
- Support for scalameta semanticdb.
- Better handling of clashing sources with same package/filename in classpath (relevant for Play projects).
- Better handling of empty package (relevant for Play projects).
- LoC metrics works with scalameta too.
- Various improvements to the charts displayed in the web Dashboard.
- Released new version of Maven Hydra that supports the Dashboard: 3.2.2-hydra07.
0.10.0 - May 8th, 2018¶
- Compile time monitoring with a slick web-based dashboard. Go through the Quick start to set it up!
- Released new version of Maven Hydra that supports the Dashboard: 3.2.2-hydra06.
0.9.12 - May 1st, 2018¶
- Support for Scala 2.12.6
- Consider a build to be a full build even if the output directory contains some non-class files
- Fixed a binary compatibility with some versions of semanticdb-enabled builds
- Support sources with the same name and package in different source roots
- Better support of the empty package (sometimes used by Play route files)
- Collect new metrics: lines of code (LoC), LoC/second, tree nodes (after parser, after typer) See Metrics
0.9.11 - April 4, 2018¶
- Enforce license checking, see docs
- Increased threshold for reporting the GC error to 5 seconds (it was previously 2 seconds).
- Threshold for GC error is now configurable via the system property
- Added a Reference page with all configuration options
- License activation can create the .triplequote/ directory if it doesn't exist
- Add the date to log messages in hydra.log
- Released new version of Maven Hydra support: 3.2.2-hydra05
- Fixed issue with GC errors not being properly reported when using the Zinc server
0.9.10 - Didn't happen¶
0.9.9 - March 2, 2018¶
- Fixed issue with compilation of package object that could result in an error
not found: type <Foo>.
- Fixed a common cause for
knownDirectSubclassescompilation error (you should no longer need this workaround).
- Monitor GC during compilation and report an error if too much GC is observed (read here for details).
- Released a new version of the Maven Hydra support: 3.2.2-hydra04.
- Released Zinc server for Hydra. See Maven guide
0.9.8 - November 9, 2017¶
- All files generated by Hydra during compilation are now under
0.9.7 - November 9, 2017¶
- All Hydra artifacts are now published in a Maven repository
- Fixed XML literal patterns leading to
not found:type errors
- Better support for annotation macros like Simulacrum's
- Fixed a couple of crashes involving incremental compilation
- More robust handling of errors and logging during late phases of compilation
- Fixed an error leading to cyclic reference errors in calls involving default arguments that match symbols already in scope
- All files generated by Hydra during compilation are now under
- hydra-bridge is now published in a Maven repository
hydraBenchmarkcommand works now also with sbt 1
- The Maven Hydra plugin is now published in a different repository. Check the Maven user guide for correctly upgrading
- Released a new version of the Maven Hydra support: 3.2.2-hydra02
- Feature parity with the sbt support (minus the
- Changed location where the zipped distribution is published (get the new link in the CLI - Download section)
0.9.6 - October 24, 2017¶
- Scala 2.12.4 support
- Correct logging of Scala version
0.9.5 - October 9, 2017¶
- Fixed a problem in the sbt plugin that caused some error messages to not be reported.
- Fixed a regression in handling classpath conflicts between source and class files.
- Compute compilation time metrics only on full builds.
0.9.4 - October 2, 2017¶
- Support for sbt 1.0 (the
hydraBenchmarkcommand is not yet available on sbt 1.0).
- Fixed 'not found' compilation errors when a classfile (on the classpath) and a source file (in the project) had the same name and the same directory path.
- Fixed cast exception occurring when using scala-clippy.
- Fixed wrong warnings
"No classes or objects found ..."that could be reported if mixed Java and Scala sources were placed under the same source directory.
- Stop workers as soon as errors are reported by a worker (fixes occasional crashes).
- Fixed "unsupported Scala version" warning reported when adding sbt-hydra to the global plugins.
- Compilation unit timings are now sorted in decreasing order.
- Support for multiple output directories.
0.9.3 - August 3, 2017¶
- Performance improvements for code using macro annotations, such as Circe, Grafter, MacWire, etc.
- Support Scala 2.12.3
- Fixed a regression causing the auto-balancer to produce unbalanced workloads.
- Update compilation unit timings file only when there are no compilation errors.
0.9.2 - July 20, 2017¶
- Use weighted HydraTag for compilation. This allows easier interoperability with sbt parallel execution. See Parallel execution
- Fixed assertion failure occurring with the "auto" partition strategy.
- Fixed benign NullPointerException occurring when retrieving compilation metrics.
- Fixed printing of milliseconds in the times reported by the
- Use relativized path in the unit-timings.csv file.
- Report time in milliseconds instead of nanoseconds in the unit-timings.csv file.
- Warn when a package object is not following the Scala convention.
0.9.1 - June 30, 2017¶
- A new
"auto"source partition strategy that automatically balances workload based on compilation times of individual source files. This is now the default partition strategy assigned to
hydraSourcePartitioner. See Unbalanced workers.
- Incremental compilation uses the compilation times of individual source files to automatically balance workload (this works with any source partition strategy).
- Documentation about tuning your Continuous Integration builds.
- Don't do a clean build after changing the number of workers or source partition strategy.
- Added a
hydraBenchmarktask for measuring performance of Hydra vs the vanilla Scala compiler. See Benchmark.
- Hydra log and timings files are now placed inside a
.hydrafolder located in the project's root directory.
- Documentation about how to pinpoint compile time inefficiencies.
0.9.0 - June 9, 2017¶
- Significant performance improvements in projects with large classpaths, such as Play projects.
- Fixed execution of publishLocal/publish.
- Tagged compilation task with
HydraTagfor better interaction with
parallelExecution. See Concurrent Restrictions.
0.8.3 - May 19, 2017¶
- Fixed spurious unused variable warnings.
- Fixed spurious warnings on methods marked scala.js.native.
- Fixed support for Scala 2.12.2 (resolves compiler crash).
- Added a Troubleshooting page to the documentation.
- Added documentation about Hydra Memory Requirements.
0.8.2 - May 12, 2017¶
- Generate a separate CSV file recording compilation times for each run. See Timings file.
- Sources in the explicit partition file are now relative to the project's root directory. See Explicit Source Partitioning.
0.8.1 - May 5, 2017¶
- More robust handling of malformed input during initialization, sometimes seen with XML in case patterns.
- Fixed scoping of
hydraSourcePartitionersetting. See Explicit Source Partitioning.
- Explicit partitioning file allows line comments starting with
0.8.0 - April 28, 2017¶
- Less noisy 'unbalanced workers' message.
- Vastly improved support for the explicit partitioner. See Tuning.
- Added Tuning guide
- Fixed Hydra logging configuration (avoid intercepting application's logging).
- Only issue unbalanced warning when compilation succeeded.
- Fixed message displaying the number of needed workers (it was off by one).
- Fixed performance regression introduced by the flat classpath in 2.11 (via Scala patch
0.7.3 - April 20, 2017¶
- Add support for 2.11.11 and 2.12.2.
- Allow to set log level via environemnt variable. Set
hydra.logLevelto a valid log level.
- Bump to use hydra10 and removed loader setting.
- Added Hydra license and third-party licensing information.