Clone Tools
  • last updated a few seconds ago
Constraints: committers
Constraints: files
Constraints: dates
issue 49

Publish 6.1-20200106000035+0000

Remove some minor hot spots in scheduling and executing Worker API actions.

- Don't create JVM options for a legacy worker action until needed.

- Don't create the parameters of a legacy worker action more than once.

- Improve inference of an isolated object's parameter type.

Also restructure to decouple the different kinds of worker action specs, to allow the different specs to carry different information and also to make clearer what state a given stage of the execution pipeline expects the spec and its parameters to be in.

  1. … 22 more files in changeset.
Publish 6.1-20200105000033+0000

Restructure project tree construction so that the containing `ProjectState` for a project is always available, to avoid throwing and then silently ignoring an `IllegalArgumentException` for each project.

Also reuse more project construction logic and remove a bunch of duplicate ad hoc code. This change means that the `ProjectState` object now owns the mutable `Project` object, as was always the intention, rather than existing as a parallel tree. This can continue to evolve so that all access to the mutable project object is mediated by the `ProjectState` that owns it.

  1. … 8 more files in changeset.
Publish 6.1-20200104000020+0000

Rebaseline several scenarios

Publish 6.1-20200103000024+0000

Publish 6.1-20200102124445+0000

Not use mirror for EC2 builds

Change the `JavaInstallationRegistry` API to represent the installation directory using `Directory` and `Provider<Directory>` instead of `File`.

Publish 6.1-20200102000035+0000

Serialise services marked with `@BuildTree` to the instant execution cache. Mark the `JavaInstallationRegistry` implementation with this annotation so that tasks and other types that reference this service work with instant execution.

Add a `@BuildTree` annotation that can be attached to a service implementation to indicate that it is intended to be present at build tree scope.

Other annotations will be added later, and used to provide some validation and other useful things, but for now this is intended to be just enough to mark up some services to work with instant execution.

Revert some changes to artifact transform execution, as these changes introduce a performance regression.

  1. … 16 more files in changeset.
Rebaseline RealLifeAndroidBuildPerformanceTest


Publish 6.1-20200101000030+0000

Push parallel execution of transforms down to the transformation step, so that this happens independently of whether the tranform is executed for a scheduled node or when visiting an `ArtifactCollection` or `FileCollection` contents.

Shuffle some transform logic around.

Publish 6.1-20191231000021+0000

Remove Maven API usage from javadoc (#11742)

Previously we use -link in javadoc,

which failed the build upon network issues. This commit removes such usages.

Add support for key fingerprints

In order to provide maximum security, it's now possible to use full key

fingerprints, in addition to long (64-bit) key ids, in trusted or ignored


It doesn't matter what format is used: if a trusted key uses a long id,

then it's possible that if there's a key collision, an artifact would be

trusted even if it shouldn't. If a fingerprint is used instead, then we

would use the full fingerprint for verification.

It's worth nothing that PGP doesn't provide the full fingerprint in signatures

for the key issuer. This means that when we're going to download keys, we

will still use the long ids.

Fixes #11770

  1. … 5 more files in changeset.
Attach gradleApi() sources to IDE models (#11751)

Attach sources to gradleApi() and gradleTestKit() generated jars in the IDE models when -all distribution is used and sources are available.

  1. … 3 more files in changeset.
Publish 6.1-20191230002137+0000

Merge some logic used for executing a chained scheduled artifact transform node and the other places artifact transforms are executed.

An implication of this change is that when a scheduled transform produces multiple output files, then a consuming scheduled transform will transform those output files in parallel.

  1. … 9 more files in changeset.
Merge some logic used for executing an initial scheduled artifact transform node and for visiting transform outputs included in the contents of an `ArtifactCollection` or `FileCollection`.

Serialize the artifact metadata for an `ArtifactCollection` instance to the instant execution cache.

Rebaseline GradleInceptionPerformanceTest to make new feature work

Publish 6.1-20191229000050+0000

Fix the serialization of `ArtifactCollection` instances that contain the output of artifact transforms to the instant execution cache.

Use a similar strategy to that used to capture the contents of a `FileCollection` that contains the output of artifact transforms.