Also disable dexpreopting for host. These are necessary to avoid adding
dependencies on dex2oat in various non-platform builds where they will
break.
Since we cannot assume at least one module enables dexpreopting now,
the two dexpreopt singletons are silently disabled if there has been no
call to dexpreopt.GetGlobalSoongConfig.
Bug: 145934348
Bug: 148312086
Bug: 148319588
Bug: 148690468
Test: m
Test: env OUT_DIR=out-tools prebuilts/build-tools/build-prebuilts.sh
on the aosp-build-tools branch
Test: build/soong/soong_ui.bash --make-mode static_sdk_tools dist DIST_DIR=out-dist BUILD_HOST_static=1
on internal (cf b/148312086#comment8)
Test: build/soong/soong_ui.bash --make-mode dist DIST_DIR=out-apps TARGET_BUILD_APPS=Launcher3 TARGET_BUILD_VARIANT=userdebug
on internal without art/ and external/vixl/ (cf b/148319588)
Change-Id: I240dade7204b87fc2d12181534ab23439eca8b46
Should cut down on a bit of copying, and also required for an upcoming
CL that'll change GetCachedGlobalSoongConfig.
Test: m nothing
Bug: 145934348
Change-Id: I6bed737d9b061b5239cc603ad881f4ccd4312e43
binary module.
This uses the Once cache for GlobalSoongConfig to propagate the dex2oat
path from a module dependency to the singletons (both the one that
writes out dexpreopt_soong.config and the one that creates the
dexpreopted boot images). Unless dexpreopting is disabled altogether
through DisablePreopt in dexpreopt.config, that means:
- We must ensure at least one module registers a dex2oat tool
dependency and resolves a GlobalSoongConfig using it, or else the
singletons fail. That means we litter dex2oat dependencies in java
modules even when they won't get dexpreopted and hence don't really
need them.
- We still assume there's at least one java_library or android_app in
the build.
This relands https://r.android.com/1205730 without changes - the
necessary fixes are in the child CLs.
Bug: 145934348
Test: m
(check that out/soong/dexpreopt_soong.config points to dex2oatd64)
Test: env USE_DEX2OAT_DEBUG=false m
(check that out/soong/dexpreopt_soong.config points to dex2oat64)
Test: env OUT_DIR=out-tools prebuilts/build-tools/build-prebuilts.sh
on the aosp-build-tools branch
Change-Id: I66661711b317d1e4ec434861982919bdde19b575
dexpreopt package.
Preparation for a future CL that will need to get the make-written
GlobalConfig from within dexpreopt.
Also rename the Load*Config functions to Parse*Config, since they don't
actually load the config files anymore.
This relands https://r.android.com/1211982.
Bug: 145934348
Test: m
Change-Id: Icb9332a93811d77d2d8b06e983b92501b180a358
caching of it.
Introduce a Once cache for GlobalSoongConfig to allow it to get binary
tool paths from ordinary module dependencies (coming in a future CL)
that are then reused in singletons.
This relands https://r.android.com/1205729.
Bug: 145934348
Test: m
Change-Id: I039d6e204bee5ddc16d8e2d85057fbec20e326fe
This patch excludes non-Soong system server jars from dexpreopting.
System server jars should be dexpreopted together for better
performance: each jar should have all preceding system server jars
in its class loader context (that is passed to dex2oat and recorded
in the resulting .oat file to be used by the loader).
This means that we need a total order on system server jars. The
default order provided by PRODUCT_SYSTEM_SERVER_JARS is not good, as
it does not always respect genuine dependencies between jars (counter-
examples are rare, but non-trivial to fix: b/148219586).
This patch adds a post deps mutator pass that creates additional
dependencies and enforces global order. These are later used to
generate precise class loader contexts and system server classpath.
Test: Class loader contexts in the oat files for system server jars
match expectations:
$ oatdump --oat-file=out/target/product/walleye/system/framework/oat/arm64/com.android.location.provider.odex | grep '^classpath'
classpath = PCL[]
$ oatdump --oat-file=out/target/product/walleye/system/framework/oat/arm64/services.odex | grep '^classpath'
classpath = PCL[/system/framework/com.android.location.provider.jar*1919890654]
$ oatdump --oat-file=out/target/product/walleye/system/framework/oat/arm64/wifi-service.odex | grep '^classpath'
classpath = PCL[/system/framework/com.android.location.provider.jar*1919890654:/system/framework/services.jar*4269704903:/system/framework/services.jar!classes2.dex*134345935]
...
Test: The phone boots and logcat has no scary messages related to
class loader contexts:
$ lunch aosp_walleye-userdebug && m
$ adb reboot bootloader && fastboot flashall -w && adb wait-for-device
$ adb root
$ adb shell stop
$ adb logcat -c
$ adb shell setprop dalvik.vm.extra-opts -verbose:oat
$ adb shell start
$ adb logcat | egrep -i 'system_server: .*load.*/system/framework'
02-03 14:14:26.912 5016 5016 I system_server: Loading /system/framework/oat/arm64/com.android.location.provider.odex with executable: 1
02-03 14:14:26.914 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/com.android.location.provider.odex with executable: 1
02-03 14:14:26.914 5016 5016 I system_server: Loading /system/framework/oat/arm64/services.odex with executable: 1
02-03 14:14:26.916 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/services.odex with executable: 1
02-03 14:14:26.927 5016 5016 I system_server: Loading /system/framework/oat/arm64/wifi-service.odex with executable: 1
02-03 14:14:26.933 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/wifi-service.odex with executable: 1
02-03 14:14:26.933 5016 5016 I system_server: Loading /system/framework/oat/arm64/ethernet-service.odex with executable: 1
02-03 14:14:26.934 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/ethernet-service.odex with executable: 1
02-03 14:14:26.946 5016 5016 I system_server: Loading /system/framework/oat/arm64/com.android.location.provider.odex with executable: 0
02-03 14:14:26.947 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/com.android.location.provider.odex with executable: 0
02-03 14:14:26.947 5016 5016 I system_server: Loading /system/framework/oat/arm64/services.odex with executable: 0
02-03 14:14:26.948 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/services.odex with executable: 0
02-03 14:14:26.948 5016 5016 I system_server: Loading /system/framework/oat/arm64/wifi-service.odex with executable: 0
02-03 14:14:26.948 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/wifi-service.odex with executable: 0
02-03 14:14:26.949 5016 5016 I system_server: Loading /system/framework/oat/arm64/ethernet-service.odex with executable: 0
02-03 14:14:26.949 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/ethernet-service.odex with executable: 0
02-03 14:14:30.480 5016 5016 I system_server: Loading /system/framework/oat/arm64/com.android.location.provider.odex with executable: 1
02-03 14:14:30.481 5016 5016 I system_server: Successfully loaded /system/framework/oat/arm64/com.android.location.provider.odex with executable: 1
Bug: 141785760
Bug: 140451054
Bug: 148944771
Change-Id: Idac678dbd1f5fe0e381ce8e0e3561423f8a31389
This is to unify boot jars with system server jars:
PRODUCT_SYSTEM_SERVER_JARS and PRODUCT_UPDATABLE_SYSTEM_SERVER_JARS
do not overlap.
Test: aosp_walleye-userdebug boots
Exempt-From-Owner-Approval: cherry-pick.
Bug: 146363577
Change-Id: If001049a46f8e9d38921dabeedbd61e971ebfe73
Merged-In: If001049a46f8e9d38921dabeedbd61e971ebfe73
(cherry picked from commit ef4358e536)
Previously, the .docs suffix was used for droidstubs internal modules
that generate the stubs source. The name was confusing so this change
replaces the suffix with .stubs.source to clarify the purpose of those
modules.
Bug: 148080325
Test: m droid
Change-Id: I9d7a2ab19e4a331b515408912ac8f7e8b0e1b35d
Previously a java_sdk_library_import module did not replace the
corresponding java_sdk_library module, even when it was marked as
prefer=true. That is because the java_sdk_library_import had an empty
set of sources. However, the stubs modules that are created by the
java_sdk_library_import did override the stubs modules created by the
corresponding java_sdk_library module.
That created inconsistent behavior between the cases where only the
prebuilt is available and both source and prebuilt are available and
the prebuilt is preferred. e.g. assume a java_sdk_library/import module
called SDKLIB.
When both prebuilt and source modules are available for SDKLIB then
even if the prebuilt is preferred then any dependencies on the
SDKLIB module would use the source module.
This change fixes that inconsisteny by making the array of sources
non-empty.
Bug: 148080325
Test: m droid && TARGET_BUILD_APPS=Camera2 m
Change-Id: I25395e020393921735ada20c5492f27f1260f6c5
Prior to this change droidstubs modules that set sdk_version
did not get framework.aidl added to its aidl includes.
Bug: 149138391
Test: patch CL in bug && m system_aidl_test-droidstubs
Change-Id: I92ab344c8a4311e10c1e5c8ebf525fa2dc704075
(cherry picked from commit f278ca60e06da86c67f6a3865c290f8451657ce9)
When building java.lang classes it is necessary to compile them using
patch_module: "java.base". This change causes patch_module to be passed
through to the java_library created to compile the stubs to allow this
to be used to generate stubs for java.lang.
Test: m droid
Change-Id: I7c27953a5d782eeedd7f25e849ab444d28e28228
Adds an SdkMemberType implementation for java_system_modules. It
specifies that java_system_modules can be used with sdk as well as
module_exports, and also that the libs property should be included
as transitive members in the sdk.
It also adds support for treating appropriate tagged properties in
the snapshot prebuilts module as references to sdk members so that
they are correctly transformed when creating the versioned modules.
Bug: 142940300
Test: m nothing
Change-Id: Ic10b5a6d5b92b6018334fe876f06feaf79cc55e9
Framework and other dex files are used without image.
Test: taimen-userdebug boots when built with
DEXPREOPT_USE_ART_IMAGE=true
Test: Check logcat for checksum verification failures.
(Build ART with extra logging in OatFileAssistant.)
Test: Check that bootclasspath-checksums from some prebuilt
oat files (say input.odex) contain only one image
checksum followed by dex file checksums with
grep -az -A1 -E '^bootclasspath-checksums$' <oat-file> | \
xargs -0 echo | gawk '{print $2}'
Bug: 119800099
Change-Id: I65c2f247656e41f2c37df1ecb9e06af7dabab76e
Previously, a java_sdk_library called "SDKLIB" would create a
prebuilt_etc module called "SDKLIB.xml" which installs the generated
XML permission file to /etc/permissions/SDKLIB.xml. That module
depended on the java_sdk_library "SDKLIB" to generate the XML file
as one of its outputs by specifying srcs: [":SDKLIB{.xml}"].
If the java_sdk_library is replaced by a prebuilt then the SDKLIB.xml
module expects the prebuilt to provide the XML permissions file which
it doesn't because that is an implementation detail and so the build
breaks.
A couple of alternative approaches were looked at to fix this. One was
to have the logic that replaced the source module with the prebuilt to
inform the source module that it was being replaced so it could disable
its created module. That lead to a dependency cycle where
SDKLIB -> SDKLIB.xml -> SDKLIB{.xml}
Another solution was to mark dependency tags in such a way that the
prebuilt could automatically identify and disable the SDKLIB.xml
module. Similar to how the visibility code will ignore dependencies
that are tagged with ExcludeFromVisibilityEnforcementTag. That became
very convoluted.
Instead the java_sdk_library was changed so that it was not responsible
for creating the XML permissions file. Instead it created a genrule
called "gen-SDKLIB.xml" to create it and then "SDKLIB.xml" depended on
that. The java_sdk_library also depended on the genrule to make the XML
permissions file available for APEX and testing.
Some refactoring of the APEX code and tests was necessary because they
had knowledge of the internal implementation of java_sdk_library. The
refactoring insulates them a little better from those details.
Bug: 148080325
Test: m droid && TARGET_BUILD_APPS=Camera2 m
Change-Id: I597bccbb177b6b6320c3a3edeff467243230d384
This change fixes a bug that apex_available is not enforced for static
dependencies. For example, a module with 'apex_available:
["//apex_available:platform"]' was able to be statically linked to any
APEX. This was happening because the check was done on the modules that
are actually installed to an APEX. Static dependencies of the modules
were not counted as they are not installed to the APEX as files.
Fixing this bug by doing the check by traversing the tree in the method
checkApexAvailability.
This change includes a few number of related changes:
1) DepIsInSameApex implementation for cc.Module was changed as well.
Previuosly, it returned false only when the dependency is actually a
stub variant of a lib. Now, it returns false when the dependency has one
or more stub variants. To understand why, we need to recall that when
there is a dependency to a lib having stubs, we actually create two
dependencies: to the non-stub variant and to the stub variant during the
DepsMutator phase. And later in the build action generation phase, we
choose one of them depending on the context. Also recall that an APEX
variant is created only when DepIsInSameApex returns true. Given these,
with the previous implementatin of DepIsInSameApex, we did create apex
variants of the non-stub variant of the dependency, while not creating
the apex variant for the stub variant. This is not right; we needlessly
created the apex variant. The extra apex variant has caused no harm so
far, but since the apex_available check became more correct, it actually
breaks the build. To fix the issue, we stop creating the APEX variant
both for non-stub and stub variants.
2) platform variant is created regardless of the apex_available value.
This is required for the case when a library X that provides stub is in
an APEX A and is configured to be available only for A. In that case,
libs in other APEX can't use the stub library since the stub library is
mutated only for apex A. By creating the platform variant for the stub
library, it can be used from outside as the default dependency variation
is set to the platform variant when creating the APEX variations.
3) The ApexAvailableWhitelist is added with the dependencies that were
revealed with this change.
Exempt-From-Owner-Approval: cherry-pick from internal
Bug: 147671264
Test: m
Merged-In: Iaedc05494085ff4e8af227a6392bdd0c338b8e6e
(cherry picked from commit fa89944c79)
Change-Id: Iaedc05494085ff4e8af227a6392bdd0c338b8e6e
apex module accepts PlatformCompatConfigIntf as prebuilt,
and places it in the etc folder of the apex.
Test: m
Test: flash device with dummy config in mediaprovider APEX -
the config is present
Change-Id: Ifc62cd262f6c6571c1bf6c2943879aa20877ecad
Previously, the java_sdk_library had special support for disabling the
stubs library when the build was configured to use prebuilts for sdks,
e.g. in an unbundled app build. That caused the prebuilt version of the
stubs library to be used instead. The disabling was done irrespective
of whether a prebuilt was available which prevents java_sdk_library
from being used in situations when prebuilts are not available, e.g.
when they have not been created yet.
This change moves the logic into tha java_sdk_library_import and
leverages the existing prebuilt/prefer mechanism to ensure that the
prebuilt version is used when required.
* Adds a ForcePrefer() method to Prebuilt to allow a module to forcibly
set the value of the prefer property.
* Sets prefer true on the java_sdk_library_import and the stubs modules
it creates when the prebuilt version should be used.
* Refactors PrebuiltJars for use by both java_sdk_library and
java_sdk_library_import as they both need to provide access to
prebuilts for previously released versions of the library.
* Removes disabling logic from java_sdk_library.
This will probably require some additional java_sdk_library_import
modules to be added to prebuilts/sdk/current/Android.bp.
Bug: 148080325
Test: m droid && TARGET_BUILD_APPS=Camera2 m
Change-Id: I0b5f751e82a2179a967ae64ca03dc9b9e7665c16
We add a compat config build rule to extract the merged config, and
then update the droiddoc build rule to consume that.
Test: m -j offline-sdk-docs
Bug: 144927670
Change-Id: Ib1e85f97895c89227882e665572bda9bfc2a8cba
Exempt-From-Owner-Approval: ag/10097965 approved internally, Colin requested to patch to aosp
Previously, when a library that used a numeric sdk version also
referenced a java_sdk_library it would use the current version of its
API. That was dangerous as there is an expectation that an app building
against a numbered version will also be targeted at that version and
so building against a later version of the API could hide runtime
incompatibilities.
This change will use prebuilt versions of the java_sdk_library's api
when being built for a numbered sdk version.
Bug: 148080325
Test: m droid
Change-Id: I3fd416553950785a443c1702e495a96debc33331
The legacy usages have all been updated so the legacy properties
can be safely removed. The Libs property is kept so it can be used to
specify properties common to all scopes.
Bug: 148080325
Test: m droid
TARGET_BUILD_APPS=Camera2 m
Change-Id: I252ebbedbb463db3c7346e86d86b5880eea76fe9
Previously, the java_sdk_library_import only exposed the public stubs.
This change adds support for exposing system and test stubs too by adding
separate structures for public, system and test scopes. The existing
properties are kept for legacy reasons (and because libs can be common
across the differents scopes).
It extracts some code that is common to both sdk library and sdk
library import.
The legacy support will be removed in a future change once all existing
usages have been switched over.
Bug: 148080325
Test: m droid
TARGET_BUILD_APPS=Camera2 m
Change-Id: I0b26cc8af9ee044437ff3b80c1eca611816b9386
module_* is a new API surface for OS modules (e.g. APEXes). It has
slightly bigger API surface than the system_* SDK. Specifically, APIs
with @SystemApi(client=MODULE_LIBRARIES) are added there.
Bug: 146757305
Test: m
Change-Id: I8980e50c0e3a4cd843048e0de1f638e854384f46
When building a library against sdk_version: system_current it uses the
system stubs of any java_sdk_library that it references. Previously,
when building against sdk_version: test_current this used the public
stubs of any java_sdk_library. This change causes it to use the
test stubs instead to be consistent with the handling of system and
public.
Bug: 148080325
Test: updated the test
ran m nothing which failed as expected
fixed the code
ran m nothing again which succeeded
Change-Id: I58ec5bd243701c5a5c75664e2bb615ce7b2f2441
The java_sdk_library duplicated a lot of code related to api scopes and
also did a lot of switching on apiScope. This change eliminates that
duplication and switching by converting apiScope from an enum to a
structure containing api scope specific information.
It associates dependencies with corresponding scope by using a
scopeDependencyTag that contains a reference to the associated
apiScope.
Tested by:
1) Running the command below without the changes.
2) 'mv out/dist/apistubs $ANDROID_BUILD_TOP'
3) Applying these changes.
4) Running the command again.
5) Verifying that the out/dist/apistubs was identical to before the
change using 'meld out/dist/apistubs $ANDROID_BUILD_TOP/apistubs'
Bug: 148080325
Test: m TARGET_PRODUCT=sdk TARGET_BUILD_VARIANT=userdebug dist sdk
m droid
Change-Id: I0de3268b12254122e94ca83d09309b06ca2f9dbe
Leaving out LOCAL_CERTIFICATE for flattened apex APKs causes
the apkcerts.txt to have empty keys for those APKs, which
confuses the signing tools. Set LOCAL_CERTIFICATE for them.
Also refactor the Certificate support to avoid introducing
duplicated handling for presigned certificates.
Bug: 147765187
Test: m apkcerts-list
Change-Id: Ife07661761cd5a89c9f009b8ce041db4dff9ec54
This CL adds RBE support to javac, r8, and d8 rules which is only
enabled if respective environment variables are set.
Test: an aosp_crosshatch build with and without the new variables.
Change-Id: Ic82f3627944f6a5ee7b9f3228170c2709b1bfcb8
The primary image name should be "boot.art" in case of
"boot-framework.art", and "apex.art" in case of "apex-framework.art".
Although "boot.art" and "apex.art" are identical binaries, dex2oat
has hard-coded logic based on the image name.
This has been broken since boot image extension has been enabled for
JIT-zygote config in CL I5493e575ebf90bad1d5ad2850004d54590bbc079.
Test: compare boot-framework.art and apex-framework.art,
they were identical before this CL, and differ after this CL.
Test: JIT-zygote config boots, steps 1-2:
1. Temporarily enable Jit zygote in the product device config (in
this case device/google/muskie/aosp_walleye.mk):
+# System server should not contain compiled code.
+PRODUCT_SYSTEM_SERVER_COMPILER_FILTER := verify
+
+# Use the apex image for preopting.
+DEXPREOPT_USE_APEX_IMAGE := true
+
+# Have the runtime pick up the apex image.
+PRODUCT_PROPERTY_OVERRIDES += \
+ dalvik.vm.boot-image=/apex/com.android.art/javalib/apex.art:/system/framework/apex-framework.art
2. Build and flash:
$ lunch aosp_walleye-userdebug && m \
&& adb reboot bootloader && fastboot flashall -w
Change-Id: I98de271852ecc33feb9fd4c9b0addf0feba01856
* changes:
Revert "Separate dexpreopt.GlobalSoongConfig to allow independen..."
Revert "Move the Once cache for dexpreopt.GlobalConfig into the ..."
Revert "Get the dex2oat host tool path from module dependency on..."
