Link against the libunwind.a shipped with the NDK so it matches what's
linked into the NDK's libc++_shared.so.
For arm32, binary.mk and cxx_stl_setup.mk had been adding two different
versions of the LLVM unwinder to the same linker command-line. Remove
the duplicate one from cxx_stl_setup.mk.
-Wl,--exclude-libs is now unnecessary because the symbols are marked
hidden in the archive itself.
Bug: http://b/153025717
Test: device boots, manually inspect some linker command lines
Change-Id: Ie7182bd565014348b16448976dc6587d2885ba99
PlatformSepolicyVersion and BoardSepolicyVers haven't assigned correctly
so far. Below is the reason why it hasn't been discovered yet.
DeviceConfig.PlatformSepolicyVersion() was added to support mixed
sepolicy build (setting BOARD_SEPOLICY_VERS and building vendor sepolicy
with old plat policy files). Soong compares PlatformSepolicyVersion()
and BoardSepolicyVers(), and used old vendor sources if both are
different. Back then, the only place where such logic played a role was
selinux contexts. Test codes were running as intended: after setting
BOARD_SEPOLICY_VERS the context files were built against old policies.
But there were two mistakes:
1) PlatformSepolicyVersion() was not added to soong_config.mk, so it was
always empty.
2) BOARD_SEPOLICY_VERS was set to default in system/sepolicy/Android.mk,
which was processed after soong_config. So if BOARD_SEPOLICY_VERS was
not set in BoardConfig.mk, BoardSepolicyVers() was empty, not
PLATFORM_SEPOLICY_VERSION.
And there were no issues as Soong only checked equality.
To fix the issue correctly, this commit adds the variable
PlatformSepolicyVersion, and then handles BoardSepolicyVers correctly by
returning PlatformSepolicyVersion if it's empty.
Test: set BOARD_SEPOLICY_VERS and see vendor_file_contexts changes
Change-Id: I4e306ec1f5225094a39f0c8d94f5e7683d70d60e
This is needed for test mapping to have module-info.json built out for
device targets so that test harness can do some checks to ensure the
test is properly configured in TEST_MAPPING files.
module-info has alreay depended on droidcore, which will be
built out in the infrastructure.
Verified in a forrest run: https://android-build.googleplex.com/builds/forrest/run/L81000000840964024
Bug: 154931418
Test: m -j droidcore
m -j module-info
Change-Id: I76b97debe3fbd51c2dc033c91ed9c2eb73cfa8a2
These images are never used by GSI. Building them is a waste of resource
and a potential source of confusion, for they are packaged into the
*-img-*.zip of GSI. Skip building them.
Bug: 183068624
Test: "m dist" on GSI and check the build artifacts under OUT and DIST
directories, and check the contents of *-img-*.zip
Change-Id: Ic14cfdd10ed26d975b369ce128e4b284378219ef
If system and vendor are built separately, none of the two
builds contained kernel information. The process of extracting
kernel information shouldn't depend on system and vendor
builds, but on the existance of the kernel image.
With this change, one of system or vendor build may have
INSTALLED_KERNEL_IMAGE defined and the other has
PRODUCT_OTA_ENFORCE_VINTF_KERNEL_REQUIREMENTS defined. The one
with INSTALLED_KERNEL_IMAGE defined will contain kernel information
in target files.
After target files are merged, check_target_files_vintf will kick
in to do the checks properly.
Test: forrest
Bug: 180475190
Change-Id: I8d887dd94e1171ab277f02f25534caf098e3faae
libneuralnetworks_shim.so allows vendor service to run
a NNAPI Driver sAIDL service backed by a NNAPI SL Driver.
SL Driver is a shared library with a known stable C API,
that contains vendor driver implementation.
libneuralnetworks_shim.so is part of com.android.neuralnetworks
APEX package.
Test: run VTS driver sample backed by shim on CF
Bug: 172925288
Change-Id: Ibbb0af00377a2ade212740b61158274a6942be9d
Compatibility library "android.hidl.manager-V1.0-java" should go
before "android.hidl.base-V1.0-java" in class loader context for
dexpreopt, because this is the order used by PackageManager when it
constructs class loader context on device.
This allows to avoid "ClassLoaderContext classpath element mismatch"
errors on first boot for Calendar and messaging apps, which have old
enough targetSdkVersion to need HIDL compatibility libraries. Previously
the errors were masked because these apps used the deprecated
&-classpath hack.
Bug: 132357300
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 grep output, no errors
Change-Id: Ibde9a4578cd86b85a9e7f11d8752716b6567e51e
Java modules that are defined in makefiles are not processed in
topological order, so it is necessary to communicate information from
dependencies via dexpreopt.config files. This has already been done in
make/core/dex_preopt_config_merger.py, and now manifest_check.py also
needs to get library names from their dexpreopt.config files.
This is to accommodate Java libraries which name differs from their
modules name. Soong properties `uses_libs`/`optional_uses_libs` and
makefile vars `LOCAL_USES_LIBRARIES`/`LOCAL_OPTIONAL_USES_LIBRARIES`
contain module names, not library names, so it is necessary to translate
them when comparing against library names in the manifest.
Bug: 132357300
Test: lunch cf_x86_64_phone-userdebug && m
Change-Id: I769b508a927d87a5ffbabf9aa45eebfb954b8bd2
Move manifest_check (a.k.a. verify_uses_libraries check) from makefiles
for specific module types to common makefile dex_preopt_odex_install.mk,
which is included by all Java modules that may require dexpreopt. If a
modules locally disables dexpreopt, it still goes through manifest_check
(unless dexpreopt is globally disabled or the module has no Java code).
This CL allows to have manifest_check and dexpreopt in the same makefile
(which is needed for a follow-up CL that will reuse dexpreopt variables
for manifest_check).
Bug: 132357300
Test: lunch cf_x86_64_phone-userdebug && m
Change-Id: Ia217cfc247ae43d8fc716bfc1fe9dcce1d00aa7f
Commit I9967d06bde0e18a12b84b5b0b568db09765fe305 supports adding a
generic boot_signature into boot.img v4. This change allows replacing
the boot_signture signing key with a release key during the release
process.
The default GKI signing key can be specified in a BoardConfig.mk via:
BOARD_GKI_SIGNING_KEY_PATH := external/avb/test/data/testkey_rsa2048.pem
BOARD_GKI_SIGNING_ALGORITHM := SHA256_RSA2048
BOARD_GKI_SIGNING_SIGNATURE_ARGS := --prop foo:bar
The release signing key/algorithm can be specified by the following options
when invoking sign_target_files_apks:
--gki_signing_key=external/avb/test/data/testkey_rsa4096.pem
--gki_signing_algorithm=SHA256_RSA4096
Additional arguments for generating the GKI signature can be
specified as below:
--gki_signing_extra_args="--prop gki:prop1 --prop gki:prop2"
Bug: 177862434
Test: make dist
Test: sign_target_files_apks \
--gki_signing_key=external/avb/test/data/testkey_rsa4096.pem \
--gki_signing_algorithm=SHA256_RSA4096 \
--gki_signing_extra_args="--prop gki:prop1 --prop gki:prop2" \
./out/dist/*-target_files-eng.*.zip signed.zip
Test: Checks GKI boot_signature is expected after signing:
`unzip signed.zip IMAGES/boot.img`
`unpack_bootimg --boot_img IMAGES/boot.img --out unpack`
`avbtool info_image --image unpack/boot_signature`
Test: unit test: releasetools_test and releasetools_py3_test
Change-Id: I61dadbc242360e4cab3dc70295931b4a5b9422a9
Update the size limit for madvising (MADV_WILLNEED)
odex and vdex files to 100MB.
Art file continues to be madvised fully, with a UINT_MAX limit.
Test: presubmit
Bug: 178853586
Change-Id: Ie46a0658bf30a6a9908ab30b28ce097e434e7e2d
The application rbcrun executes Starlark scripts that define Android product configurations.
See README.md for details.
Test: go test
Fixes: 180529448
Change-Id: I7d728b47d3f381b7052a0d7d51c9e698e5c2e316
