Clone Tools
  • last updated a few seconds ago
Constraints: committers
Constraints: files
Constraints: dates
Allow fixtures from `:ide` to be used in forked TAPI tests

Move fixtures from `:ide` into a separate package

Resolve releases from /libs-releases

Prior to this commit, the /libs-milestones virtual repository had been

configured to include /libs-releases, just as libs-snapshots had been

configured to include /libs-milestones. This "chaining" of virtual

repositories had the positive effect of providing builds with "one stop

shopping", i.e. the need to declare only one repository, but it also

came with the negative effect of effectively associating release

artifacts with non-release repositories on the client side. This may be

problematic for Gradle's own artifact caching, and is in any case

confusing to observe from the Gradle command line. One wonders, for

example, while a build is resolving dependencies "why is jackson-core

2.3.2 being downloaded from"

These repositories have been updated on the Artifactory side such that

they no longer do this "chaining". The resulting effect is that builds

must declare each of the repositories they wish to resolve from. If a

build depends only on GA release artifacts, resolve from /libs-releases

(which still proxies all remote release repositories, including

JCenter); if a build depends on milestone artifacts, resolve from

/libs-milestones as well; and likewise, snapshot artifacts should be

resolved from /libs-snapshots.

The performance impact on builds should be quite minimal, so long as the

repositories are declared in the correct order: /libs-releases, then

/libs-milestones, then /libs-snapshots, assuming this order corresponds

to where the bulk of the build's dependencies actually reside. Most

resolutions, then, will require only a single HTTP request against

/libs-releases; if the dependency is not found there, /libs-milestones

and /libs-snapshots will be checked (if indeed either or both are


Review items for Manifest API fix

- add @HasInternalProtocol to Manifest

- deprecate Manifest.writeTo(Reader)

- introduce CustomManifestInternalWrapper

- add coverage for custom Manifest type usage (e.g. shadow)

- add coverage for the OSGi plugin

- advertise added property on ManifestMergeSpec in release notes

After this commit, thanks to the CustomManifestInternalWrapper type

mentioned above, plugins providing their own implementation of Manifest

will benefit from the split multi-byte characters fix.

+review REVIEW-6007

    • -0
    • +2
Resolve from /libs-releases vs /libs-releases-local

Prior to this change, the affected build scripts resolved dependencies

from two repositories: /libs-releases-local and mavenCentral(), and in

that order. /libs-releases-local is, in Artifactory terms, a "local

repository", and it contains only the things we upload to it, e.g. our

gradle-core, gradle-logging, and other gradle-* modules. This means that

the vast majority of dependency resolution requests will fail to find

what they're looking for in /libs-release-local and will then fall back

to mavenCentral(). This is inefficient and slows down builds, because it

requires two client-side HTTP requests for almost every dependency.

This commit favors the use of the /libs-release "virtual repository",

which includes all the content from /libs-release-local, but also

proxies content from maven central (actually, it proxies content from

JFrog's JCenter, but JCenter itself is a proxy for Maven Central).

This means that just one repository declaration can be used instead of

two, and that the client build need only ever issue one HTTP request for

each dependency.

Note that one exception to the explanation above is

buildSrc/build.gradle, which is being changed from /libs to

/libs-release. This same change was recently made to the project's root

build.gradle as well, and the rationale for doing so can be found at


Remove unnecessary FileResolver injection in Ear

+review REVIEW-5941

Upgrade the AWS SDK for Java to v 1.11.4

- Removes the [dependency on the library](

- Upgrade jackson to 2.6.6

Restore lost access levels in codeQuality

+review REVIEW-5933

Restore lost access levels in Jacoco plugin

+review REVIEW-5933

Restore lost access level on EclipseClasspath.noExportConfiguration

+review REVIEW-5948

Update wrappers to first nightly with Kotlin support

This commit updates the gradle wrappers in this repository (main build

and samples) to use the first gradle-3.0 nightly distribution that

includes the gradle-script-kotlin module (at version 1.0.0-M1a).

We will likely return to working against custom gradle-script-kotlin

snapshot distributions after this, in order to track to our own snapshot

changes, but in the meantime this change validates that everything

works as expected when working against the main gradle distribution.

    • -1
    • +1
Remove timestamps from files

These files are updated by hand, meaning that the timestamps have become

out of date. And they're irrelevant anyway in the light of source


    • -1
    • +0
Fixed unit test to work on windows.

Fix formatting

+review REVIEW-6008

Changed `Jvm.forHome()`` to return the same instance as `Jvm.current()`` when the home directories are the same.

Removed mystery single quote.

Use the faster configuration closure DSL implementation in many more places. In particular:

- Configuring elements of containers, such as tasks, configurations, etc.

- Configuring extensions.

- Copy spec actions, and other file type actions.

  1. … 11 more files in changeset.
Changed configuration DSL to understand that Groovy treats maps specially.

Use some static factory methods rather than constructors for ClosureBackedAction.

Removed unit tests that only run on ibm jvm 6, removed preconditions to exclude ibm jvm 6.

Change the implementation for registering `DaemonExpirationListener`'s

+review REVIEW-6008

Add a spec to internally register `DaemonExpirationListener`'s


All Idea modules are generated when creating Idea project for composite

Removed unused import.

Merged release into master.

Include test coverage for continusou mode

+review REVIEW-6001

Address additional feedback on daemon message display +review REVIEW-5986

Tweaked preconditions for a couple of tests to run the right tests on the right versions of the ibm jvm.

Ignore test for now: can't use test fixtures in tooling-api integ-tests

Don't run this test on ibm jvm