If profile guided dexpreopt _and_ optimization/obfuscation is true, then
apps must use the rewritten profile created by r8/d8
Bug: 335418838
Test: m nothing --no-skip-soong-tests
Change-Id: Iac45b614f43e52ac40ad1d13df0338bcafa788c3
Currently, dexpreopt supports profile guided optimization. This does not
work well with r8/d8 optimization, since the checked-in profile will not
match the dex signatures after r8/d8 has optimized the dex code.
This CL introduces a new property `dex_preopt.enable_profile_rewrting`.
If set, the checked-in profile will passed as `input` to r8 via
`--art-profile <input> <output>`. The <output> from the previous command
will be used as the profile for dexpreopt.
Test: m nothing --no-skip-soong-tests
Test: m CredentialManager with https://ag.corp.google.com/27448930
and obfuscation turned on
Test: nm -U symbol.odex # contains obfuscated methods
Bug: 335418838
Change-Id: I53beed9ed76f013262f1c503de0f2b74997c2a7f
`UpdateDirectlyInAnyApex` is used to propagate the ApexProperties
`DirectlyInAnyApex` and `InAnyApex` to the direct dependencies of the
direct dependencies of an apex bundle. In other words, this will be
propagated only to two-levels max dependency from the apex bundle.
However, the implementation library of the sdk library can have longer
dependency chain from the apex bundle than two levels:
e.g. apex -> bcpf -> sdk_lib -> sdk_lib impl lib
Therefore, even if the implementation library of the sdk library is
registered as a dependency using the "CopyDirectlyInAnyApexTag"
dependency tag, the ApexProperties would not be propagated to the
implementation library. In order to resolve this issue and recognize
the implementation library to be directly in apex and allow
instrumentation for the implementation library, this change proposes
propagating `DirectlyInAnyApex` and `InAnyApex` to transitive
dependencies on top of direct dependencies.
Test: DIST_DIR=dist_dir TARGET_BUILD_VARIANT=userdebug PRODUCT=mainline_modules_x86_64 COVERAGE_MODULES="uwb" ./vendor/google/build/build_unbundled_coverage_mainline_module.sh && \
unzip -l out/target/product/module_x86_64/jacoco-report-classes-all.jar and ensure that framework-uwb is included
Bug: 341170242
Change-Id: I27d7a74f6e5bc3e0a044d13c619f4897b6b2eb57
https://r.android.com/3094444 has removed instrumentation from the
implementation library of the java sdk library. However, given that the
top level java sdk library was not being instrumented before removing
the compilation action, there is no way to get the coverage data for the
java sdk library included in the mainline modules with the
aforementioned change.
As seen in b/340174053, instrumenting all impl libraries by default
may lead to double instrumentation execution time error in some edge
cases (especially with make to soong dependency).
Given the above two conditions, this change allows instrumentation for
the implementation library of the java sdk library that belongs to
apex(es).
Test: lunch husky-trunk_staging-userdebug && EMMA_INSTRUMENT=true EMMA_INSTRUMENT_FRAMEWORK=true m droid dist
Test: DIST_DIR=dist_dir TARGET_BUILD_VARIANT=userdebug PRODUCT=mainline_modules_x86_64 COVERAGE_MODULES="uwb" ./vendor/google/build/build_unbundled_coverage_mainline_module.sh
Bug: 341170242
Change-Id: I175e02f61e139296f5a177e6fc1c3ded760ac028
This replaces the global override of min_sdk_version (via
`APEX_GLOBAL_MIN_SDK_VERSION_OVERRIDE`) with an min_sdk_version override
that can be set by each individual soong override_app.
The use case for this are go apps which are only installed in T and
above, even though the base AOSP apexes might be installable on < T
devices.
Test: added a unit test
Bug: 295311875
Change-Id: Ie2e738a6786bb24417c675617f7c78358017c96c
android.ModuleBase already calls aconfigUpdateAndroidBuildActions()
that is the same with CollectDependencyAconfigFiles(). Remove the
CollectDependencyAconfigFiles() to avoid duplication with
aconfigUpdateAndroidBuildActions().
To make the aconfig information available in
GenerateAndroidBuildActions() of all modules, call
aconfigUpdateAndroidBuildActions() before calling
GenerateAndroidBuildActions() of each module.
Also, we don't need SetAconfigFileMkEntries(), which is a duplicate
of aconfigUpdateAndroidMkData()
Bug: 335363964
Test: diff `adb shell printflags` before and after the change.
Change-Id: I52808e442e9fed7db1eae7b7c5ed0b1c5ba74f5d
This change modifies the build actions of java_sdk_library module type
so that it does not perform any compilation actions (i.e. does not
create the top level java_sdk_library jar file). Instead, it delegates
the build actions the top level jar file was performing to the
dynamically created ".impl"-suffixed java library module. The build
actions that are delegated to the impl library module include hiddenapi
processing, dexing, and dexpreopt.
This change relands https://r.android.com/3035972. Implementation
changes from the original change:
- "all_apex_contributions" is added as a dependecy to the implementation
library modules where the parent sdk_library module has a prebuilt
equivalent. This allows the source apex variant to be hidden from make
when the prebuilt is active.
Test: patch in internal main, lunch barbet-ap2a-userdebug && m nothing
Test: m nothing --no-skip-soong-tests
Bug: 332785297
Change-Id: I017938e5567aef82e428e7ceb557d9c9090e0257
Bug: 331528424
Test: m --no-skip-soong-tests
Ignore-AOSP-First: Depends on internal changes. Will cherry-pick once merged.
Merged-In: Ied2821f11b6a5056ecf577e1e25765bc6dd212c0
Change-Id: Ied2821f11b6a5056ecf577e1e25765bc6dd212c0
This is a no-op change for a majority of cases.
Before this change, the contruction of the manifest check inputs is
confusing. It mutates uses_libs properties in place just for the
manifest check, by replacing module names with library names for
direct dependencies and merging library names from CLC for both direct
denpendencies and transitive denpendencies, and then constructs manifest
check inputs from those mutated uses_libs properties. This is
error-prone and leads to insistency: the goal is to check that the CLC
matches the manifest, but the inputs to the check don't reflect the CLC.
After this change, we no longer mutate uses_libs properties in place.
Instead, we maintain a separate list of missing denpendencies, and then
construct manifest check inputs directly from the CLC for all existing
libraries, no matter they are direct or transtive, and from the separate
list of missing libraries. This change makes the logic more
consistent and straightforward, and it also allows us to easily do the
next change, which is to propagate transtive missing denpendencies.
In fact, this change revealed several bugs around library optionality
and order in CLC construction, and fixed them.
Bug: 331528424
Test: m --no-skip-soong-tests
Ignore-AOSP-First: Depends on internal changes. Will cherry-pick once merged.
Merged-In: I0de82e76c47995b54aba9efd41538d950256a95f
Change-Id: I0de82e76c47995b54aba9efd41538d950256a95f
As part of aosp/3022586 where we added the idea of "test-only" modules
and top_level_test_targets, this CL implements that for java modules.
We let users set "test-only" on java_library, but not on other modules
where the module kind is implicitly test-only, like java_test.
The implementation, not the user decides it is test-only.
We also exclude it from java_defaults.
% gqui from "flatten(~/aosp-main-with-phones/out/soong/ownership/all_teams.pb, teams)" proto team.proto:AllTeams 'select teams.kind, count(*) where teams.test_only = true and teams.kind not like "%cc_%" group by teams.kind'
+--------------------------+----------+
| teams.kind | count(*) |
+--------------------------+----------+
| android_test | 1382 |
| android_test_helper_app | 1680 |
| java_fuzz | 5 |
| java_test | 774 |
| java_test_helper_library | 29 |
+--------------------------+----------+
% gqui from "flatten(~/aosp-main-with-phones/out/soong/ownership/all_teams.pb, teams)" proto team.proto:AllTeams 'select teams.kind, count(*) where teams.top_level_target = true and teams.kind not like "%cc_%" group by teams.kind'
+--------------+----------+
| teams.kind | count(*) |
+--------------+----------+
| android_test | 1382 |
| java_fuzz | 5 |
| java_test | 774 |
+--------------+----------+
Test: m nothing --no-skip-soong-tests
Test: go test ./java
Test: m all_teams
Bug: b/327280661
Change-Id: I9c3ad947dc3d68d6427abada27449526d69daa6b
With this change, the jarjar repackage actions are only on
the local classes of each module instead of the combined jar
that contains the static libs. The static libs don't need
jarjar repackage action on this module level because it has
been repackaged when building itself.
This change also removes the skip_jarjar_repackage property
since it's incompatible with this change. Actually skipping
jarjar repackage on a dep may result in incomplete repackage
on the module output.
Test: CI and observing the build time of SystemUIGoogle
Bug: 328067025
Ignore-AOSP-First: Will cp to aosp
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:78ffdd47a658dec4bf79e63d11a5f0f3b94876a4)
Merged-In: I476d959af025c46d2ba6d3f48ea378a086666a33
Change-Id: I476d959af025c46d2ba6d3f48ea378a086666a33
droidstubs module require aconfig_declarations modules to be specified
when the module depends on the java_aconfig_library module in order to
generate the "exportable" stubs. This adds burden to the droidstubs or
java_sdk_library module owners, as module should specify both the
java_aconfig_library and aconfig_declarations modules to genreate the
"exportable" stubs, although the necessary information from the
aconfig_declarations module can be provided from the
java_aconfig_library modules.
In order to resolve such burden, this change enables the intermediate
cache files from the associated aconfig_declarations module of a
java_aconfig_library module to be propagated to its reverse dependencies
without having the specify the aconfig_declarations modules in the
droidstubs or java_sdk_library modules definitions.
This does not mean that the intermediate cache files of every transitive
dependencies of the java_sdk_library or the droidstubs will be passed to
aconfig to retrieve the state of the aconfig flags.
Specifically, only the java_aconfig_library modules or the java_library
modules that have static dependency on java_aconfig_library modules that
are passed to droidstubs via `libs` or `srcs` (using
":module_name{.tag}" syntax) will provide the intermediate cache files
to generate the "exportable" stubs. For java_sdk_library, all modules
listed as `libs`, `static_libs`, `stub_only_libs`, and `apiScope.libs`
are passed as `libs` and all modules listed as `api_srcs` and `srcs` are
passed as `srcs` to the droidstubs modules dynamically generated in
java_sdk_library module per api scope, thus these properties will be
affected.
Note that the test is being added in the apex package. This is because
trying to register the codegen package build components in the java
package leads to circular dependency between the codegen and the java
package, as codegen package imports the java package.
Test: m nothing --no-skip-soong-tests
Bug: 329284345
Change-Id: I7953ab64776f6947808321ce8a3598154501bcfe
Previously JAVA dependency over partition was checked only when device
VNDK version is not empty. This change updates logic to enforce inter
partition dependency despite of VNDK version based on VNDK deprecation.
Bug: 330100430
Test: AOSP CF build succeeded
Change-Id: I7444ad612aa6c525e61cc5c2a89f5421ab4465b4
Add support for android_library_import to propagate the proguard specs
from its static dependencies.
Bug: 326265727
Test: TestExportedProguardFlagFiles
Change-Id: I174c1d7937a529958f8d240214b279062ef74868
This change propagates the intermediateCacheFiles generated by the
aconfig_declarations to the static rdeps that are java modules or the
rdeps that are filegroups.
Test: m nothing
Bug: 329284345
Change-Id: I02431336c1aa0378d03248f3bb6edf2f57ec3b7f
Bug: 329762127
Test: cd build/soong/java ; go test -run TestJavaLibHostWithStem
Test: cd build/soong/java ; go test -run TestAppStem
Test: cd build/soong/java ; go test -run TestAndroidAppImport_ArchVariants
Test: cd build/soong/java ; go test -run TestOverrideAndroidAppStem
Test: cd build/soong/java ; go test -run TestOverrideAndroidApp
Test: cd build/soong/java ; go test -run TestPackageNameOverride
Change-Id: I8fbdc82116f9a0c68121fbbeb1ef4f78f2b6fea3
extraCombinedJars contains the R.jar generated by
ResourceProcessorBusyBox for apps. It needs to go before any
static_libs dependencies so that the finalized R values take
precedence over any non-final R values that may have been included
in the dependencies. Moving after the javac and kotlinc compiled
jars helps minimize differences when turning on use_resource_processor
by default by using the MANIFEST.MF generated by soong_zip instead of
the one generated by ResourceProcessorBusyBox, which contains a
different Created-By field.
Bug: 294256649
Test: m javac-check
Change-Id: I8d060594404121fd36ef650ac317f0c5d5f92b23
jarjar_prefix propagation logic currently depends on the type of the
providers of the dependencies, thus the logic needs to be updated when a
new provider is added. Since `aconfig_declarations_group` module type
utilizes its custom provider "CodegenInfoProvider", this change adds
jarjar_prefix propagation support for this provider.
Test: Set aconfig_declaration module's exportable property to true && m framework-minus-apex and inspect if "repackaged-jarjar" subdir exists in outdir
Bug: 310504781
Change-Id: I1689027d7810687a53ec34b9cfda479806f2ec87
Add more explicit checks to not propagate JarJarProvider when we will be
linking against stubs.
Bug: b/310504781
Test: treehugger
Merged-In: I3219d0bf7e84a386b16b248314f11b4e9edcd0fa
Change-Id: I3219d0bf7e84a386b16b248314f11b4e9edcd0fa
is_stubs_module property signifies whether the module compiles stubs or
not. The information is propagated to the reverse dependencies of the
java_library/java_sdk_library modules via JavaInfoProvider as
StubsLinkType, a ternary state that provides information about whether
the provided jars are compilation of stubs or implementation, or
unknown.
When java_sdk_library sets StubsLinkType, its state is Unknown as the
decision of providing stubs vs implementation jar differs by the reverse
dependency.
Note that this CL does NOT propagate the property up to its reverse
dependencies. This will be done in follow up CLs.
Test: m nothing
Bug: 310504781
Merged-In: Ic81488dbd1c9af1b5d31a33e5e9441d9f1416dbe
Change-Id: Ic81488dbd1c9af1b5d31a33e5e9441d9f1416dbe
For prebuilts, java.dexpreopt gets called twice
1. in the context of the top-level prebuilt apex
2. in the context of the java_library shim
Only the artifacts of (1) get installed. The artifacts generated by 2)
are unused. However the args used to generate the artifacts of (1) are
incorrect. It uses moduleName(ctx) to do special-handling of apex system
server jars. This special-handling does not happen in (1), so although
dexpreopt artficats get generated, they get generated with the wrong
args.
To fix this, add an additional parameter in java.dexpreopt to pass the
libraryName explicitly.
Details
- Delete moduleName function. This was used to determine the java
library name, which is no longer safe
- Add a libraryName parameter to java.dexpreopt
- Most module types will use j.Name() as libraryName
- prebuilt_apex and apex_set will iterate over its `contents` and use
each element as libraryName when invoking java.dexpreopt
With the correct args passed from (1), we can drop j.dexpreopt from (2)
completely. Dropping (2) also breaks profile guided dexpreopt tests.
These currently operate on (2). They will be moved to (1) in the next CL
of this stack.
Test: presubmits
Test: lunch cf_x86_64_auto-trunk_staging-userdebug && m nothing (this
was a postsubmit failure with aosp/2923733)
Test: art_standalone_dexpreopt_tests on next https://android-build.corp.google.com/builds/abtd/run/L86000030001579256
Test: art-gtest on git_master-art-host https://android-build.corp.google.com/builds/abtd/run/L07800030001550262
Bug: 308790457
Bug: 322255144
Change-Id: I8eb604c82f1fa5289d3cd1a20084d56e4d7485e3
(This was missed in aosp/2876756, which copied only the bcp jars)
This hardcoded location is used by dex2oat to compile the dexpreopt
artifacts. The copy rules are currently generated by java_(sdk)_library
for source builds, and their prebuilt counterparts in prebuilt builds.
After this change, the copy rule will be bifurcated between source and
prebuilt builds
1. For source builds, it will come from java_(sdk)_library
2. For prebuilt builds, it will come from the top-level prebuilt apex.
Since there can be multiple prebuilt apexes in trunk stable,
HideFromMake will be used to determine which deapexed jar to copy.
The bifurfaction is expected to be temporary. It is needed for now since
the `apex_contributions` which will be used for source vs prebuilt
selection have not been populated completely.
Test: Added a unit test
Test: Presubmits
Test: git_master-art-host:
art-gtest (https://android-build.corp.google.com/builds/abtd/run/L40800030001459791)
Test: git_main:art_standalone_tests
(https://android-build.corp.google.com/builds/abtd/run/L09000030001463855)
Bug: 308790457
Change-Id: I3105d3b3a7e5c41cb601d07806f4ea483a61b50a
This is a followup cleanup for aosp/2876754 and replaces panic with
ctx.ModuleErrorf. The latter creates a more expressive build error.
Implementation details
- export moduleErrorf interface from build/soong/android. This minimal
interface will be used as a parameter for `DexJarBuildPath`
- Add ModuleErrorf to the function signature of DexJarBuildPath. This
parameter only gets used for Import and SdkLibraryImport structs.
These two can have duplicate deapexer definitions, and ModuleErrorf
will be used to report that error
- Create a minimal implementation of `ModuleErrorf` in tests of java and
apex
Test: m nothing --no-skip-soong-tests
Change-Id: I0febec651f40c3f04deb957e64133c94b80fbd78
Convert all of the callers of OtherModuleProvider/OtherModuleHasProvider
to use the type-safe android.OtherModuleProvider API.
Bug: 316410648
Test: builds
Change-Id: Id77f514d68761a262d9ea830a601dbed804bbbe5
Convert all of the callers of Provider/HasProvider to use the type-safe
android.ModuleProvider API.
Bug: 316410648
Test: builds
Change-Id: I73479de1625fa2865b6c73444cd477e50d56dc5a
Convert all of the callers of SetProvider to use the type-safe
android.SetProvider API.
Bug: 316410648
Test: builds
Change-Id: If58f4b5355264ddab2045bc3591a4eac19cd58fc
art java_test modules need to specify compile_data, make it common
to all java modules.
Bug: 307824623
Test: m lint-check
Change-Id: I68640f14137f9cadaf0c454d3b5abc9e2d1d9b4b
Remove the ConvertWithBp2build implementations from all the module
types, along with the related code.
Bug: 315353489
Test: m blueprint_tests
Change-Id: I212672286686a318893bc7348ddd5a5ec51e77a7
Passing the list of all transitive aconfig files to Make causes extra
Kati analysis runs when dependencies are changed in Android.bp files.
Since Make is going to merge them anyways, merge them per-module and
pass a single aconfig file to Make for each module.
Fixes: 313698230
Test: m out/target/product/vsoc_x86_64/system/etc/aconfig_flags.pb
Change-Id: Ifde4826bc93bc06e40338f72b4cb39eed26ca08d
Java libraries were writing lists of files that changed whenever
transitive dependencies changed to Android-${TARGET_PRODUCT}.mk, causing
Kati analysis to rerun whenever a dependency was changed in Soong.
In both cases, Make would immediately use the list to write a single
output file. Write the files in Soong and pass the path to the file
to Make instead, which will both reduce the size of
Android-${TARGET_PRODUCT}.mk and skip Kati analysis more often.
Bug: 309006256
Test: m checkbuild
Change-Id: I5dff16c6fb7cca8c6da927b37c612c7b1d0954e6
The jdepsGeneratorSingleton can get the module path directly, it doesn't
need to be collected by each module type that implements IDEInfo. Fixes
module types (like android_library) that didn't reach the code that
collected the path.
Bug: 309835196
Test: out/soong/module_bp_java_deps.json contains path for ExtServices.core
Change-Id: If8cb81b4f708e0367f156ade164bee253bf53492
Only propagate transitive libs by way of library-like references. This
avoids inclusion of transitive deps from tools like lint modules that
are unncessary for R8.
This yields build speedups for a number of targets, including:
* services.jar: -22% (66s -> 51s)
* telephony-common.jar: -27% (29s -> 21s)
* updatable-media.jar: -41% (17s -> 10s)
* framework-appsearch.jar: -46% (13s -> 7s)
Bug: 302383328
Bug: 307273642
Test: m
Change-Id: I60bb30e84dabe522ea9ac9333f00e739962ea91d
