git@github.com:flutter/engine.git/compare/21ae92651a38...3e10d92b2bbb git log 21ae92651a38..3e10d92b2bbb --no-merges --oneline 2019-08-19 bkonyi@google.com Roll src/third_party/dart 0e201edeeb..9f13d07670 (27 commits) 2019-08-19 matthew-carroll@users.noreply.github.com Notify framework to clear input connection when app is backgrounded (#35054) (flutter/engine#9498) 2019-08-19 dnfield@google.com Re-enable firebase test and don't use google login (flutter/engine#11228) 2019-08-19 ariaye@google.com Update tflite_native and language_model revisions to match the Dart SDK (flutter/engine#11230) 2019-08-19 skia-flutter-autoroll@skia.org Roll src/third_party/skia f8221786d088..8566dda51b42 (4 commits) (flutter/engine#11220) 2019-08-19 chinmaygarde@google.com Update metal layer drawable size on relayout. (flutter/engine#11224) 2019-08-19 dnfield@google.com Make firebase testlab always pass (flutter/engine#11226) The AutoRoll server is located here: https://autoroll.skia.org/r/flutter-engine-flutter-autoroll Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+/master/autoroll/README.md If the roll is causing failures, please contact the current sheriff (stuartmorgan@google.com), and stop the roller if necessary.
Dart SDK dependency
The bin/internal/engine.version file controls which version of the Flutter engine to use.
The file contains the commit hash of a commit in the https://github.com/flutter/engine repository.
That hash must have successfully been compiled on https://build.chromium.org/p/client.flutter/ and had its artifacts (the binaries that run on Android and iOS, the compiler, etc) successfully uploaded to Google Cloud Storage.
The /bin/internal/engine.merge_method file controls how we merge a pull
request created by the engine auto-roller. If it's squash, there's only one
commit for a pull request no matter how many engine commits there are inside
that pull request. If it's rebase, the number of commits in the framework is
equal to the number of engine commits in the pull request. The latter method
makes it easier to detect regressions but costs more test resources.