dexpreopt of apex system server from prebuilts involves three soong
modules
1. prebuilt_apex / apex_set
2. an internal deapexer module created by the prebuilt apex
3. java_import/java_sdk_library
(3) acts as a shim for the deapexer to set the dexjar extracted from the
prebuilt apex. This methodolody requires a 1:1 correspondence across the
three modules
This breaks down when we have multiple versions of the same prebuilt
apex in the tree. In preparation for this, move the dexpreopt
processing from (3) to (1). Each prebuilt_apex will create the necessary
rules for dexpreopting the jars deapexed from itself. In the future,
apex_contributions will be used to pick which service-foo.{odex|.vdex} to
install depending on which prebuilt apex is selected.
Implementation details
- Embed dexpreopter in prebuiltApex structs so that this module type can
register the dexpreopt rules. Since a single apex can have multiple
system server jars, this also requires creating an additional scope in
dexpreopt.go to prevent name collisions
- Add the dexpreopt modules as required in initApexFilesForAndroidMk
- Add the depreopt modules to androidMk in AndroidMkEntries. Drop the
equivalent from java_import and java_sdk_library_import
Bug: 308790457
Test: existing soong unit tests
Test: lunch cf_x86_64_phone-next-userdebug && m out/target/product/vsoc_x86_64/system/apex/com.google.android.adservices.apex
Test: Verified that the above command installs
/out/target/product/vsoc_x86_64/system/framework/oat/x86_64/apex@com.android.adservices@javalib@service-adservices.jar@classes.{odex|vdex} and the equivalent files of service-sdksandbox
Test: presubmits
Change-Id: I01cea8956d2857fb864b415e73d3d2686d069b5e
This change adds an overrideable property flags_packages to android_app,
which is used to list the aconfig_declarations module names that the app
depends on. The build action of android_app is modified to pass all
flags text file provided by the aconfig_declarations to aapt2 link as
--feature-flags arguments.
Test: m nothing --no-skip-soong-tests
Bug: 306024510
Change-Id: I4924f88b9954950cc1936a472cd7ac70f41add5d
These were the only 2 pre-singletons in soong. sdkPreSingleton is
totally unused. overlaySingleton can be done during
GenerateAndroidBuildActions instead.
Test: m nothing --no-skip-soong-tests
Change-Id: Ieb5ab92f18cb56be4049c0842f61df8aa02dc52c
This property is similar to `asset_dirs`, but for individual files. This
is especially useful when the user wants to pack generated asset files
into the APK.
Bug: 257532944
Test: m nothing
Change-Id: If3062124e9dedb41315dcaf5d379803e5419b8cd
When building with use_resource_processor: true R.jar files from
shared andoid_library dependencies need to be added to the classpath
so that the generated R classes can be referenced.
Bug: 294256649
Test: m DocumentsUIPerfTests
Change-Id: I30a6bddc3f378ecf58f142f94049e67ba33a47e3
Currently the product variable that is controlled by
APEX_GLOBAL_MIN_SDK_VERSION_OVERRIDE only overrides apexes, but it
doesn't apply to android_apps. This commit allows android_apps which set
updatable: true in their Android.bp to have their min_sdk_version be
overriden by this product variable.
Bug: 295311875
Test: go test
Change-Id: If3cb82a17cae4553b577dd1a4637ee13c3c95302
The R.Java files generated by aapt2 link --no-static-lib-packages
cause scaling problems by combining all resources into every package
listed in a dependencies' AndroidManifest.xml file. For SystemUI-core
this results in 74 R.java files, each with 76k lines, and takes 20
seconds to compile in javac.
Both AGP and Bazel have workarounds for this that avoid using the
R.java files generated by aapt2, instead generating more efficient
R.class files directly based on the R.txt file.
Bazel uses the ResourceProcessorBusyBox tool that is already present
in our tree to process the resources. Reuse the same tool in Soong
to create the R.jar.
The more efficient R.class files require modifiying source files
that use incorrect packages to refer to resources.
Bug: 284023594
Test: TestAndroidResourceProcessor
Change-Id: I026073b40dabcfdb10e5d7a52e9348205b0e9a66
Merged-In: I026073b40dabcfdb10e5d7a52e9348205b0e9a66
Instead of rolling assets and manifests up through each static lib,
provide them as a DepSet. This will help with the next patch, which
needs to pass all the transitive manifests and R.txt files together.
This relands Id8b3aa2bed3771e82ab6bde192c9b43baa38b54c with a fix
to include additional manifests from dependencies in the final
manifest merger.
Test: app_test.go
Test: TestManifestMerger
Change-Id: Ied0a6cfee2f18e87188db145b9411d4a903ab6c9
Instead of rolling assets and manifests up through each static lib,
provide them as a DepSet. This will help with the next patch, which
needs to pass all the transitive manifests and R.txt files together.
Test: app_test.go
Change-Id: Id8b3aa2bed3771e82ab6bde192c9b43baa38b54c
Add a test for the aapt2 resource processing code that will be affected
by the resource processor feature in a subsequent CL.
Test: TestAndroidResourceProcessor
Change-Id: I39712853f7bbfb7963f1b71e234d039d5ccaae53
Starting from aosp/2594905, dexpreopt depends on
`$PRODUCT_OUT/product_packages.txt`. When PRODUCT_PACKAGES changes,
dexpreopt has to rerun for all apps. This is not ideal.
After this change, dexpreopt uses a per-app product_packages.txt that is
filtered by the app's dependencies, and it uses `rsync --checksum` to
prevent the file's mtime from being changed if the contents don't change.
This avoids unnecessary dexpreopt reruns.
Bug: 288218403
Test: m
Test: Change PRODUCT_PACKAGES and see no dexpreopt reruns.
Change-Id: I5788a9ee987dfd0abfd7d91cbcef748452290004
Use the APK's filename as the on-device filename for the allowlist.
Previously the filename of the source xml file was used, which leads to
a problem when the same source file is used for different apps (even
if only one of those modules ends up being installed).
Bug: 284500052
Test: app_test.go & presubmit boot tests
Change-Id: I0b43a35e3a310464d79d7f37e0900d639f4a479c
Before this change, dexpreopt was often broken with optional libraries.
This was because the CLC construction was done in Soong at an early
stage, where we don't have sufficient information to determine whether
an optional library is installed or not.
For example, the "Settings" package uses an optional library called
"androidx.window.extensions". On some devices, the library is installed,
but on some other devices, it's not. Soong always adds the library to
the CLC, meaning the CLC is wrong for devices which don't have the
library. This change fixes the problem. See the tests below.
After this change, the CLC construction is done by a Python script
invoked at a very late stage. It uses product_packages.txt, which is
generated by Make, to determine whether an optional library is
installed or not, and filter out libraries that are not installed.
Note that optional libraries are still added as dependencies by Soong.
This is because dependencies have to be added at an early stage. This
means what dex2oat eventually uses will be a subset of the dependencies,
which is fine.
Bug: 282877248
Test: m
Test: atest construct_context_test
Test: -
1. lunch aosp_cf_x86_64_phone-userdebug && m
2. Check the .invocation file of the "Settings" package (defined in
.bp file)
3. See androidx.window.extensions
Test: -
1. lunch aosp_redfin-userdebug && m
2. Check the .invocation file of the "Settings" package (defined in
.bp file)
3. Don't see androidx.window.extensions
Test: Check the .invocation file of the "Dialer" package (defined in
.mk file)
Test: -
1. Build a Pixel 5 system image and flash it to a Pixel 5 device.
2. adb shell pm art dump
3. See "reason=prebuilt" instead of "reason=vdex".
(https://diff.googleplex.com/#key=fB6Ls9q2QGSN, before: left,
after: right)
Change-Id: Ia112bd7c2328373e68db6bffb74bf34030f683d8
AndroidMk assumes that the app is the last file installed, and it uses
this assumption to populate the LOCAL_SOONG_INSTALLED_MODULE entry. This
CL moves the privapp_allowlist installation to before the app
installation to respect this assumption.
Bug: 242509786
Test: go test
Test: OUT_DIR=out.ref m nothing &&
cp aosp/2562351 && OUT_DIR=out.change m nothing &&
GOWORK=$PWD/build/bazel/mkcompare/go.work \
go run android/bazel/mkcompare/cmd -json \
<(sed -e "s/out\.ref/out/g" out.ref/soong/Android-aosp_cheetah.mk) \
<(sed -e "s/out\.change/out/g" out.change/soong/Android-aosp_cheetah.mk)
&& verify manually that the only diffs are related to the removal of
the prebuilt_etc module.
Change-Id: I95ec27070f575e79fb976de68493a219717ed89a
For a non-override android_app, we can assume that the privapp_allowlist
already contains the correct package_name, and so we don't need to
overwrite it in this case.
Bug: 242509786
Test: go test
Merged-In: I0f137e34cae3478dc8b9178d138121ff1d936f07
Change-Id: I0f137e34cae3478dc8b9178d138121ff1d936f07
Getting the first certificate will panic if there are no certificates,
which can happen when AllowMissingDependencies is set and the
certificate property is a module reference to a missing module.
Only get the first certificate if the list is not nil.
Use mainCertificate since it handles the logic already.
Test: TestAppMissingCertificateAllowMissingDependencies
Bug: 283102635
Change-Id: I8b27f65aa7d071041171ad45ac52bf47fa31bf2b
This change allows override_android_app to use the same
privapp_allowlist as the non-override module so that they will always
remain in sync.
Test: go test ./java -v -run TestPrivappAllowlist
Test: go test ./apex -v -run TestApexWithApps
Test: m com.android.permission com.google.android.permission and verify
manually that apex_payload.img contains correct privapp_allowlist
Test: m com.android.permission before and after change &&
`diffoscope apex_payload_reference.img apex_payload_with_change.img`
&& verify that there are no semantic changes
Bug: 242509786
Change-Id: Ifdcb28af40763aed7a4aac9a7f681153554bc256
Currently, target_sdk_version of MTS tests were being set to 10000 even
in release branches. This should only happen in active branches.
Test: Unit test
Test: aapt2 on CtsContentTestCases in udc-dev and verified
targetSdkVersion is 34
Bug: 279507428
Change-Id: Ib79318febc0f6e771b3679c0281a97357930ec56
ld -r reorders symbols and invalidates the .llvm_addrsig section, which
then causes warnings if the resulting object is used with ld --icf=safe.
The warning is especially common when building with musl, as the
clang_rt.crt* objects have .llvm_addrsig sections, are linked into
libc_musl_crt* using ld -r, and are then linked into every other binary
and shared library with --icf=safe.
Strip the .llvm_addrsig section after ld -r to prevent the warnings.
Test: m USE_HOST_MUSL=true host-native -k
Change-Id: Ia52a4756b9ebbb62115898d0de9f8641e6fea705
target_sdk_version signifies device version and does not need an sdkKind to
describe it fully. Update the type and cleanup existing usages. As a
side benefit, we also get better error handling since users can no
longer enter something like `public_30` as a valid target_sdk_version in bp
files
Test: m nothing
Test: no change in ninja files (this should be a no-op)
Bug: 208456999
Change-Id: I3c19245e29184bd9e5660ad8981966f64dfa9424
Currently it would return the default one even if the requested one is
an active sdk.
Bug: 270609292
Test: go test ./java
Test: built `rkpdapp` locally in internal and verified that its
targetSdkVersion is U and V
Test: TH
Change-Id: Idb03ff4786ff87fb7911bf31205941618a662404
Tests that run against mainline modules generally need to be built
twice: once for aosp (unsigned) and once for internal (signed).
The override_android_test rule is really useful for generating a
signed version of the unsigned test.
However, mainline tests need to use a MainlineTestModuleController
to only run tests on targets with the expected module. Since unsigned
and signed mainline modules have different package names, this means
the test config needs to be updated when overriding a test.
Add a new "mainline_package_name" argument so that android test rules
can update the targeted mainline package. Without this, the only way
to create a mainline package test is to instead use android_test, which
results in a lot of copy-pasta.
Test: added tests to soong
Test: manual, using a locally-modified RkpdAppUnitTest
Test: test_config_fixer_test.py
Change-Id: Idaffd63f225719a2bfda41018fda630b17db0080
Enforce target sdk version flag could only be used by `android_app`,
moving the check into a common function `generateAndroidBuildActions`.
This would ensure that the `enforce_target_sdk_version` flag
can also be set by `android_test` and `android_test_helper_app`.
Bug: b/227460469
Test: m nothing
Change-Id: I86e0bf684a5083221dae53907d9f548a0390b673
Add a test that verifies that that missing certificate path inserted
when AllowMissingDependencies is set uses a well formed extension
that won't trip up sign_target_files_apks.
Fixes: 259861670
Test: TestCertificates
Change-Id: I00656809698aed5f9d3aaadaecda9848d2882223
Due to GMS target SDK requirements (https://docs.partner.android.com/gms/policies/preview/mba#min-target-sdk)
we need to ensure all the mainline packages to target latest API
level. Currently, the team chases each module to bump their target
API level.
Updating the build code to make sure that mainline modules
(i.e updatable modules) target the latest sdk version by default.
It would be by default set to 10000 before SDK finalization and updated to the new API level after finalization.
Effectively it would mean:
1. '10000' in aosp and internal master
2. Finalized number in development branches like "33" in "tm-dev"
3. As sdk hasn't been finalised in "udc-dev", it would be "10000"
which would be automatically updated to finalized version after sdk finalization.
For local development if the target sdk version is required to be set,
users would need to set Updatable flag to false.
Enforce_default_target_sdk_version flag needs to be used in bp file,
if this feature needs to be tested locally when updatable: false.
Bug: b/227460469
Test: m nothing
Change-Id: I05e0ae74ae44fd73df75e91b926bcc924446253f
MTS tests built on unstable branches (e.g. git_master) should be testable on
old system images (e.g. S). However, they run into an error during installation
on older images:
`Requires development platform $<current_codename>, but this is a
release platform`
This CL fixes this issue by updating the target_sdk_version of MTS test apps
targeting current to the magical sdk_version 10000
Bug: 227473065
Test: go build ./java
Test: TH
Change-Id: Ic0358a48a19dc239defbb4ee8ec99225cce75584
JNI libs present in AAR files are ignored in the build. If those
libraries are necessary, they must be manually extracted and a
prebuilt_cc_library module must be created for them. Instead, we can
extract these libraries and copy them to an output APK as part of the build.
Bug: 228848794
Test: created android_app that depends on an android_library_import with
JNI embedded, and verified that .so libs are present in APK
Test: created android_app that depends on an android_library which
depends on android_library_import with JNI embedded, and verified
that .so libs are present in APK
Test: verified that multiple .so's from a single architecture are copied
to APK
Test: verified that the absence of a jni/<arch> folder or files in the
folder will cause the Ninja action to error
Change-Id: I242a862d7bd881f7a4a8c23493924d8fe441ea25
AndroidApp had its own HideFromMake method and flag that shadowed
the one in ModuleBase. This caused performOverrideMutator to set the
AndroidApp flag, but ModuleBase.skipInstall to read the ModuleBase
flag, resulting in a conflicting install rule being created. Remove
AndroidApp's HideFromMake in favor of the ModuleBase one.
Bug: 232788722
Test: TestOverrideAndroidAppWithPrebuilt
Change-Id: I8c0dfcb50ff4dc1e4d0574f150b10d79908f46aa
Move overrides attribute from appProperties to overridableAppProperties
Bug: 220029162
Test: m
Change-Id: I6f527df3173f142311734333ad37018c83d5e279
Merged-In: I6f527df3173f142311734333ad37018c83d5e279
(cherry picked from commit a2ce78f80d)
With aosp/1640364, all variants of a cc_* module use min_sdk_version as
the version part of the clang triple. Therefore, checking
min_sdk_version of jni_libs should be sufficient to ensure that there is
no unintended access to symbols in newer Android versions
Test: go test ./java
Test: TH
Bug: 155209650
Bug: 209409604
Change-Id: I6c064f8a6ea12c8aa40165a9063380306a180c9b
This reverts commit 0b1c70efbc.
The reverted commit was based on the idea that uses-libraries that are
explicitly specified in build files should not be implicitly added to
the manifest, as that would mean that anything added to the build files
will flow to the manifest.
Although this logic is correct, it prevents propagation of
uses-libraries from dependencies, which is wrong: if a library has an
explicit uses-library property in Android.bp, this property is expected
to be propagated to the library's dependencies. Failing to do so would
mean that every user of that library has to add uses-library property to
their build files, which doesn't scale (see b/214255490 for example).
Bug: 214255490
Test: lunch aosp_cf_x86_64_phone-userdebug && m && launch_cvd \
&& adb wait-for-device && adb root \
&& adb logcat | grep -E 'ClassLoaderContext [a-z ]+ mismatch'
# empty output, no errors at boot
Change-Id: I6f420e76a89aa2f37be99f877711736640f2c361
Getting the first certificate will panic if there are no certificates,
which can happen when AllowMissingDependencies is set and the
certificate property is a module reference to a missing module.
Only get the first certificate if the list is not nil.
Bug: 228379411
Test: TestAppMissingCertificateAllowMissingDependencies
Change-Id: I046d75dbbd4f21f4a2b6851f558e430e9879fcff
