Age | Commit message (Collapse) | Author |
|
This reverts commit 399f4f61132f363e9479e59c550a7bcf005922d8.
Reason for revert: b/187908823
Change-Id: I81f7cd3672e65ce1f7e1fa846b6c3e2f87147a26
|
|
Note: previously, dagger.android tests have existed in GitHub, but they were run as a noop in our GitHub Actions because robolectric tests were not enabled. This CL enables them.
RELNOTES=N/A
PiperOrigin-RevId: 358903809
|
|
This is done by splitting the SwitchingProviders into separate classes based on scoping:
1. Unscoped: SwitchingProvider<T> implements Provider<T>
2. SingleCheck: SingleCheckSwitchingProvider extends SingleCheck<T>
3. DoubleCheck: DoubleCheckSwitchingProvider extends DoubleCheck<T>
This allows SingleCheck and DoubleCheck to properly bypass memoization when it's not needed.
RELNOTES=Fix redundant DCL for scoped bindings requested as Lazy
PiperOrigin-RevId: 330015462
|
|
PiperOrigin-RevId: 329394738
|
|
RELNOTES=N/A
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=324646786
|
|
...by load()-ing java_library and other rules wherever we use them.
Compare to CL 297412705 for Flogger.
This CL includes updating to a new version of bazel_common to avoid --incompatible_load_java_rules_from_bzl errors in bazel_common. See https://github.com/google/bazel-common/pull/104.
Note that this CL also changes the way we get zlib() (a dependency of protobuf) from bind() to http_archive().
http_archive() seems to be the more recommended pattern:
- https://docs.bazel.build/versions/master/external.html#repository-rules
- https://github.com/bazelbuild/bazel/issues/1952
But my immediate motivation was that bind() wasn't working with the new version of protobuf. (The new version of protobuf is necessary to avoid --incompatible_load_java_rules_from_bzl errors inside protobuf. It comes with the new version of bazel_common.)
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=297608393
|
|
We haven't found this to be effective on any of our larger projects, and we are going to focus our efforts on other build performance improvements.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=253668388
|
|
This allows us to reconstruct ComponentImplementation instances from types that have already been compiled, saving time during component proccessing.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=242753707
|
|
this mode in near-midterm future. It's simple enough to revert if we ever want to bring it back.
I haven't gone ahead with merging InnerSwitchingProviders -> SwitchingProviders because I still want to get to doing Producers for fastInit, and SwitchingProviders may be benefitial for that.
This is a partial rollback of cc739fe17cb9bd666e14b7bce2bca78d1c497585
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=227772927
Signed-off-by: Ron Shapiro <ronshapiro@google.com>
|
|
In doing so, fix a bug where we had missed adding onProducerFutureCancelled(boolean) to the Hjar processing step.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=225548921
|
|
94c28e7ae1310964d4960fd9aa13392b3f9764ef
*** Reason for rollback ***
All functional tests now pass in ahead-of-time mode, so no more need for duplicated macros!
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=222078128
|
|
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=217333772
|
|
variants for GenJavaTests and GenRoboelectricTests. Duplicate each call to the above with the flag set. This will allow us to add exceptions (in the srcs exclude glob) as we work through bugs exposed by functional tests.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=215751328
|
|
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=211814342
|
|
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=204152825
|
|
to dagger.experimentalAheadOfTimeSubcomponents as that more accurately reflects
the change to generated code.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=201516630
|
|
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=199741327
|
|
RELNOTES=Rename experimentalAndroidMode to fastInit.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=199535162
|
|
androidExperimentalAndroidMode2 is much more similar to default mode than the original androidExperimentalAndroidMode. In this new mode, virtually everything is the same as default mode except that the initialization of providers is done with a inner static switching provider class instead of separate *_Factory classes.
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=193520283
|
|
We were only passing the variant javacopts to the libraries of those functional tests. However, that silently fails to apply the javacopts when the @Component is defined in the test class itself, since the javacopts were not applied to the test classes.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=193042437
|
|
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=189926134
|
|
RELNOTES=n/a
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=189676820
|
|
javac warnings
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=173539568
|
|
This adds private methods for requests other than INSTANCE (e.g. PROVIDER, LAZY, PROVIDER_OF_LAZY, etc.).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=171197859
|
|
together (or ignored together)
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=170376379
|
|
(If you're reading this CL description, please do not use this feature unless you've discussed with the Dagger team directly. It is very much experimental and not guaranteed yet.)
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=168758665
|
|
For the functional tests, apply Java7 source, target, and type inference for the code under test.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=153867150
|
|
maven structure (src/{main,test}/) to the {java,javatest}/ model typical to bazel
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=153174049
|
|
As we transition from Maven to Bazel, this moves our code into a more traditional Bazel structure and removes the need for duplicative src/{main,test,it}/java/ trees for each artifact that we have.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=146020680
|
|
The Maven infrastructure is not being removed in this CL. We will build and test with both Maven and Bazel until we have a successful release to Maven Central using bazel-built artifacts and automated tools to generate poms/ship to Central.
This CL does not address any of the scripts in util/ like publishing snapshots or generating javadoc. It also does not reorganize files into a more traditional bazel project structure.
Open questions:
- What should we do about shading?
- Right now, only auto-common is shaded. If we shade it, should we remove it from the deps of the compiler POM, since other users don't need it to build?
- How should we implement shading? One idea is to make a unified jar of the bazel-built compiler and auto-common, and then use jarjar to rename classes.
- What should we do with dagger-gwt? I'm not a GWT expert but it seems that we could recreate that jar with a genrule + jarjar
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=144448160
|