We are bringing up support for ARM Memory Tagging Extension in the platform,
but the required patches have not yet landed in the upstream kernel. This
change adds support for a product variable that products may set to true to
declare support for an experimental userspace interface to MTE. Setting the
variable to true causes a product configuration to be enabled which will
activate code in bionic and scudo that uses the interface.
This variable should not be set to true in shipping products because
the userspace interface has not been finalized and is subject to change.
Once the support patches have landed in the kernel, bionic and scudo will
be changed to use the official interface and the variable will be removed.
Bug: 135772972
Change-Id: I8ba5dea4de374eb6775edb6d4e3c0cf9cad8d760
Dexpreopt build commands should only be generated on targets that
include build/make/target/product/runtime_libart.mk, which sets the
necessary variables such as PRODUCT_SYSTEM_DEFAULT_PROPERTIES that
contain default values passed to dex2oat. This file also sets a
variable PRODUCT_USES_DEFAULT_ART_CONFIG that is used in
build/make/core/dex_preopt.mk to decide if boot images should be
installed.
On some targets build/make/target/product/runtime_libart.mk is not
included. Prior to this patch, on such targets invalid dexpreopt
commands were generated, but not used, so they did not cause any
visible build failures. The invalid commands can be grepped as:
lunch qemu_trusty_arm64-userdebug && m nothing \
&& fgrep -e '-Xms ' $ANDROID_BUILD_TOP/out/soong/build.ninja
In this case '-Xms ' is an ill-formed option passed to dex2oat (the
option expects one argument, but none is passed).
This patch makes the DisablePreopt variable passed from make to
soong more strict: it not only requires WITH_DEXPREOPT, but also
PRODUCT_USES_DEFAULT_ART_CONFIG. This means that dexpreopt commands
will not be generated on targets that do not include
build/make/target/product/runtime_libart.mk.
Test: lunch aosp_walleye-userdebug && m \
&& find $ANDROID_BUILD_TOP/out -name 'boot.art'
# expect to find files in /out/target/product/walleye/system/framework/$ARCH/
Test: lunch qemu_trusty_arm64-userdebug && m \
&& fgrep -e '-Xms ' $ANDROID_BUILD_TOP/out/soong/build.ninja
# expect empty output
Change-Id: If2d4fe2cdcb6a81c7c6d730d18c2b681a74fb0b7
Current routine to check whether BOARD_VNDK_VERSION variant is created
or not is not correct, because only non-VNDK modules are installed in
/vendor. Thus, BOARD_VNDK_VERSION variants should only be created with
non-VNDK modules.
Bug: N/A
Test: m nothing
Change-Id: I2ee7821fce407214ac6be4f4b6d62483c907971f
This is a follow-up CL of I9a24f6975bd4b226a94f61a13d43857dcdce6b88
Generated files are passed to make via MakeVars:
- SOONG_LLNDK_LIBRARIES_FILE
- SOONG_VNDKCORE_LIBRARIES_FILE
- SOONG_VNDKSP_LIBRARIES_FILE
- SOONG_VNDKPRIVATE_LIBRARIES_FILE
- SOONG_VNDKCOREVARIANT_LIBRARIES_FILE
- SOONG_VNDK_LIBRARIES_FILE
Previously filenames were "guessed" from module names. Now VndkMutator
stores filenames as well and generate those files with them.
Bug: 142963962
Bug: 141450808
Test: m && device boots && TH
Change-Id: I0c248b707188f904df60ead50059fefe90bfd17f
When use_vendor is used, native modules are built with
__ANDROID_VNDK__ and __ANDROID_APEX__, which may cause
compatibility issues. (e.g. libbinder)
Even though libbinder restricts its availability via
'apex_available' property and relies on yet another macro
__ANDROID_APEX_<NAME>__, we restrict usage of "use_vendor:" from
other APEX modules to avoid similar problems.
Bug: 142684427
Test: m -j
Change-Id: Ibc781de2efcd20cb6688a183b08e908a8a6e2593
Create variant of image, zip or flattened according to
TARGET_FLATTEN_APEX and payload type.
If payload type is zip, only zip variant is created because flattened
apex is not supported. And if payload type is image, image and flattened
variants are created.
Bug: 139053989
Test: m -j
Change-Id: Ibde18490d23ec602c4cca97cf97db90a562e014e
62c7829595c0df53e96addcd347c11ac01012eee introduced the new stem
property to java modules, but it wasn't propagated to Make. Fixing the
problem.
This change also fixes a problem that (module name) == (file name) is
assumed in dexpreopt_config.go, which no longer is the case. A mutator
runs to build a map from module name to its stem. The map is then used
when filling up the file paths in the bootImageConfig struct.
Bug: 139391334
Bug: 143494499
Test: m
Test: BootImageProfileTest
Change-Id: Idbc894f877692401471130de6cbfe5e0dd129da9
This is causing some build errors because the config.json file isn't
associated with a rule (because the rule looks for .txt).
Test: ran locally, checked output directory
Bug: 142551000
Change-Id: Idcdb226fa2d23055a8850b2a0545202921edc3b1
This reverts commit d5df949385.
Bug: 143594594
Bug: 143593500
Reason for revert: Some builds are failing.
Change-Id: I69986b472bce39266095e526fcd7ef5f48ece85e
Exempt-From-Owner-Approval: Going back to green.
* Rust tests are like binary files compiled with --test.
New test.go follows binary.go code patterns and reuses
some code in binary.go.
* Generate one test per source file as testPerSrc in cc/test.go.
The "all tests" variation feature of cc/test.go is not copied yet.
Fix some Stem and SubName settings to make testPerSrc work.
* Move cc.CheckDuplicate to android.CheckDuplicate,
which is now shared by cc and rust.
* Refactor tests in binary_test.go and add new test_test.go.
Bug: 140938178
Test: mm in rust projects, added rust_test and rust_test_host
Change-Id: Ia6fec8b4cf2572fd352ab1938a1f3c7b5cca2212
Unstripped binaries allow for online symbolization, and offline
symbolization using the debug info in the binary. Debug information
isn't stripped for host binaries, but is for target. Target should also
contain debugging info.
Bug: N/A
Test: m fuzz && # Check that fuzz targets in the fuzz-*.zip package have
debug info.
Change-Id: Ibd3002674d519e927340a50dfdfbf44f2d809d58
When TARGET_VNDK_USE_CORE_VARIANT is set, vndk apex should have only
- vndk sp libraries
- vndk core libraries which are marked as "must use vendor variant"
Previously, vndk sp libs are discarded when TARGET_VNDK_USE_CORE_VARIANT
is set.
Bug: 143374795
Bug: 139772411
Test: TARGET_VNDK_USE_CORE_VARIANT=true m com.android.vndk.current
see if vndk sp lib(e.g. libunwindstakc) is included.
Change-Id: Iae218fff5c354eb874839a1f71be30633fb8a9ab
Since apex_vndk deals with its own dependencies, it should not follow
dependencies while packaging.
Bug: 139772411
Test: TARGET_VNDK_USE_CORE_VARIANT=true m com.android.vndk.current
see if there are unexpected libs in the apex
for example, android.hardware.audio.common@2.0.so should not be
included in the apex since it is not listed as "must use vendor
variant".
Change-Id: Ibada600b3099fb19630d1e327a2e09cd26b2deb0
__ANDROID_APEX__ was defined with the name of apex module.
-D__ANDROID_APEX__=com.android.foo
But in this way, conditional compilation is not easy since comparing
macro's string value is not supported in C/C++.
(There's no usages of this value in source tree.)
In most cases, modules can check if __ANDROID_APEX__ is defined to see
if they are compiled for apex.
For modules which should behave differently according to which apex they
are included, they can check __ANDROID_APEX_<NAME>__.
Bug: 142582178
Test: m (soong tests run) && boot device && TH
Change-Id: I0f5e3e9463ccd96cbba333a8bdd648470c5c912d
The smartStatusOutput.done channel is read from a goroutine that
locks smartStatusOutput.lock. If the goroutine that writes to
smartStatusOutput.done is holding smartStatusOutput.lock it can
lead to an AB-BA deadlock. Call stopActionTableTick from outside
the lock.
Fixes: 143558785
Test: none
Change-Id: I93a10ef9ff16c3953a1c5ccb102b024158358fe4
* changes:
Use java language 1.9 for sdk_version: "current"
Remove special case for sdk_version: "none"
Use system modules for turbine
Make javaVersion an enum
Move TestConfig sdk versions forward
Split java 8 and 9 classpaths in TestClasspath
Java language was set for 1.8 for anything building against the
current SDK because the stubs were not built in Soong, so the
system modules could not be built. The stubs have been built
in Soong since Iabd32b30954b3f4a6d9a779fde52a032b684807e, but
I5e870c34dd0ebc8ae3f888ec627da590c846a76f missed updating
this TODO.
Use 1.9 when building against the stubs, but continue using
1.8 for unbundled builds until we have prebuilt system modules.
Always use the core-current-stubs-system-modules to avoid
splitting android.* packages between the system modules and
the classpath, which would cause new classes in android.*
packages in classpath jars to be ignored. Add a new
java9Classpath field to sdkDep that will contain the stubs jar
when targeting Java language level 1.9, and plumb it through
to javac and turbine. Rename the modules field to bootclasspath.
Bug: 142896162
Test: m checkbuild
Change-Id: Icfd32d0a863b2303a997c7cf03cb3708aade4724
The dependency handling for sdk_version: "none" (which propagated to
!sdkDep.hasStandardLibs()) was very similar to the normal useModule
case. Combine the cases by making decodeSdkDep set useModule: true
and put the system modules in modules for the sdk_version: "none"
case.
Test: TestClasspath
Change-Id: Icc9ff4d43a38da25cc0e3628be95951d61773ad5
turbine supports taking system modules on the command line,
now that we target Java language level 1.9 and use system modules
by default switch turbine to match javac.
Test: m javac-check
Change-Id: Ieee07502151da0d5693bb8929213d495c039106b
Adds the ability for rust modules to be compiled as C libraries, and
allows cc modules to depend on these rust-generated modules. This also
means that soong-rust should not have any dependencies on soong-cc aside
from what's required for testing.
There's a couple small fixes included as well:
- A bug in libNameFromFilePath that caused issues when library's had
"lib" in their name.
- VariantName is removed from rust library MutatedProperties since this
was unused.
Bug: 140726209
Test: Soong tests pass.
Test: Example cc_binary can include a rust shared library as a dep.
Test: m crosvm.experimental
Change-Id: Ia7deed1345d2423001089014cc65ce7934123da4
The snapshot script can now handle native shared libs in an SDK.
Bug: 138182343
Test: create following sdk module:
sdk {
name: "mysdk",
native_shared_libs: ["libc", "libdl"],
}
, then execute `m mysdk` and execute the update_prebuilt-1.sh as
prompted. Following directories are generated under the directory where
mysdk is defined at:
1
├── aidl
├── Android.bp
├── arm64
│ ├── include
│ ├── include_gen
│ └── lib
│ ├── libc.so
│ └── libdl.so
├── include
│ └── bionic
│ └── libc
│ └── include
│ ├── alloca.h
│ ├── android
│ │ ├── api-level.h
<omitted>
Change-Id: Ia1dcc5564c1cd17c6ccf441d06d5995af55db9ee
Specifically, this adds:
* Owner
* Disable (stop fuzzer from running in Haiku)
* Bug Component
* Bug Hotlist
The fields are all inside a new 'options' struct.
The values from these fields (if any) are written into a config file as
json.
Bug: 142551000
Test: ran locally with a modified build file and verified output in .zip
Change-Id: I86edf74c2cebe9912ac0ad203f99028be4062c8b
Remove the hardcoded checks against "1.9" by making javaVersion
an enum and implementing javaVersion.usesJavaModules().
Test: TestClasspath
Change-Id: I559eeb1f45880bb8177269c6d977ee4dfbadce57
The SDK version in TestConfig is very old, move it forward to trigger
the newer javaVersion code paths.
Bug: 142896162
Test: all Soong tests
Change-Id: Iaf2a6f6dfef43b006b0a09f193fbc2d2e103d90d
In preparation for adding the system modules jars to the classpath
for java 9, split the expected java 8 and java 9 classpaths into
separate fields. Also replace the system: "bootclasspath" special
case with a bool that specifies that the module will always be
built using a java 8 style command line.
Test: TestClasspath
Change-Id: I661ec50d77a039c435c03bc8e6443cc5a57c206e
Adds an interface, CcLinkableInterface, for cc linkable dependencies
which come from other toolchains such as Rust.
Bug: 140726209
Test: Soong tests pass, rust modules still compile.
Change-Id: I7378a46fad94fd0b735746aaf4e265fd2c2c04d8
Treehugger Robot