* changes:
Keep the first and last snapshot of variables.
Generate GenericConfig objects from MakeConfig objects.
Emit and parse the product config variables from kati/make
Add class to fork and exec kati, based on the commandline option given.
Add a CSV parser to parse the output from kati.
Take the unwanted JNI lib removal action out of JNI uncompression
function so that they work independently when building prebuilt apks.
Fixes: 180475879
Test: Manually with a test apk.
Change-Id: I7392d0bf4ca7db76f27e883df08f7cb256162440
This to allow operating on the complete ART_APEX_JARS list from make in
a follow-up CL.
Test: m nothing
Test: m nothing EMMA_INSTRUMENT=true EMMA_INSTRUMENT_FRAMEWORK=true
Bug: 180325915
Change-Id: Ic08148edf25738f6f4769e5359a573237a38b0e9
* changes:
Set BUILD_BROKEN_VENDOR_PROPERTY_NAMESPACE for goldfish
Add BUILD_BROKEN_VENDOR_PROPERTY_NAMESPACE to BoardConfig
Add PRODUCT_SHIPPING_API_LEVEL to productVariables
TARGET_BUILD_APPS.
Test: Run build_mainline_modules.sh and check that
SOONG_CONFIG_art_module_source_build is true even if the fallback
default is false.
Bug: 172480615
Change-Id: Ie0e9566d4d09b1e54af00c06021aa355adcf0a1a
* Default value will be empty and not work in -header-filter.
Bug: 179530304
Test: make with WITH_TIDY=1 and various setting of DEFAULT_TIDY_HEADER_DIRS
Change-Id: I3f784f545acd12285a5bf6dcae1bab42c9185fe7
Library order is important because it is used to construct class loader
context, which is then written into OAT/ODEX files and chacked against
class loader context constructed by PackageManager on the device. If the
orders are different, dexpreopted code is rejected.
This CL fixes a few problems that caused reordering:
- 'filter' function arguments are swapped so that patterns list comes
first, and the library list second
- JSON representation of class loader context is changed to avoid
unmarshaling it to Go maps, which may reorder keys
- library list is no longer sorted (it's unclear why it was sorted)
Bug: 132357300
Test: lunch 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: Ie76996d497e60da0948f1879d6db589ff3e968a2
This adds the configuration necessary to allow the cts_root
test suite to be built.
Bug: 162933947
Test: m cts_root
Merged-In: I4732bf82d3cedc77e30d7fa5eff285d382992853
Change-Id: I4732bf82d3cedc77e30d7fa5eff285d382992853
Jarjar doesn't exit with a nonzero return code when there is a syntax
error in a rules file and doesn't write the output file. Without a
nonzero return code ninja prints a warning but continues, which leads
to stale results on incremental builds where the output file already
exists, or delayed errors on clean builds whre the output file didn't
exist.
Delete the output file before running jarjar, and then check if it
exists after running jarjar.
Fixes: 119516143
Test: m out/target/common/obj/JAVA_LIBRARIES/noto-emoji-compat-java_intermediates/classes-jarjar.jar
Change-Id: Ib99167403948ee11f039dd4d2d7e63052bb92136
Use_lmkd_stats_log used to control LMKD_LOG_STATS when compiling lmkd.
However, for newer Android releases, we mandate that lmkd informs statsd of process kills (change id I8cb123b9488fbc6e88863c2f0e75f1422bcd282e)
After aosp/1555318 we compile lmkd with statsd by default, so this setting is a noop.
Change-Id: I91bd08198fe2275fb8aca77ae5ea637df360d416
Test: build
Bug: 177985094
repack_bootimg is useful to add files into a ramdisk, then repack
a boot.img. Note that mkbootimg, unpack_bootimg, lz4, minigzip, etc.,
that repack_bootimg depends are already in otatools.zip.
Bug: 174443192
Test: `m otatools-package`, unzip $OUT/otatools.zip, export $PATH
with ./otatools/bin and checks unpack_bootimg can run
Change-Id: Id4fbf2076b50522b95ab46dfeb1e2b7115d6606d
BOARD_<partition>_KERNEL_MODULES_BLOCKLIST_FILE<kernel_version> checks
the syntax of and installs a prebuilt modules.blocklist file.
Example usage:
```
BOARD_VENDOR_KERNEL_MODULES_BLOCKLIST_FILE := \
$(wildcard <path to prebuilt modules.blocklist>)
```
Then the modules blocklist would be install to
(vendor|vendor_dlkm)/lib/modules/modules.blocklist.
Bug: 175165656
Test: Create a blocklist with incorrect syntax, verify build fails
Test: Verify the new macro indeed installs the blocklist file
Change-Id: Icb75aae1bdb45546792f28f4c4e8d06d9439391a
target_files.zip passes a list of files to soong_zip in a file list
file, which soong_zip interprets as a glob. Adding a file called "["
causes soong_zip to error with "syntax error in pattern". Use an RSP
file instead, which after I1de7829c668538204fc4a2ac793a4f88dbc886c9
does not interpret entries as globs.
Bug: 179297188
Test: m target-files-package
soong_zip: don't glob RSP file entries
File arguments are considered globs, but RSP file entries should be
considered filenames without globbing. Escape RSP file entries so
they don't get treated as globs later.
Bug: 179297188
Test: zip_test.go
Change-Id: I0736f2d1802114217cba03ee32b67479c9d98da8
builds.
This CL is a nop since the default is still true, but it gives us a way
to toggle it without affecting coverage builds.
Test: env NATIVE_COVERAGE=true build/soong/soong_ui.bash \
--dumpvar-mode SOONG_CONFIG_art_module_source_build
env CLANG_COVERAGE=true build/soong/soong_ui.bash \
--dumpvar-mode SOONG_CONFIG_art_module_source_build
env NATIVE_COVERAGE=true CLANG_COVERAGE=true \
build/soong/soong_ui.bash \
--dumpvar-mode SOONG_CONFIG_art_module_source_build
check that all these print "true" when the default in the "else"
clause is "false".
Bug: 172480615
Change-Id: I648ed911773c59a5b2a117dbf3ddcb2ef338830d
Filter LOCAL_OPTIONAL_USES_LIBRARIES by PRODUCT_PACKAGES. This has the
drawback that some present libraries may be filtered out as well, as the
full list of product packages is unknown until all Android.mk files have
been read.
This CL unblocks <uses-library> fixes for individual apps, which will be
necessary to dexpreopt them in the nearest future when the &-classpath
hack is removed in ART. Apps that are not fixed by then will loose
dexpreopt completely. This fix, although not perfect, lowers the risk of
loosing dexpreopt for many apps. In the future it may be replaced with a
better solution.
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: I188cf240259f7a448623450e6a949e0784a21f87
If a module is installed to TARGET_RECOVERY_ROOT_OUT,
then its init rc file (declared in LOCAL_INIT_RC) is installed
to $OUT/recovery/root/system/etc/init instead of
$TARGET(partition)_OUT_ETC/init.
Fixes: 178859897
Test: m snapuserd.recovery
Change-Id: I82d0f40d6d7047e271612204391c5a027f84f52f
This empty file is required to be present.
Test: fvp_mini-eng* builds and boots on FVP
*: with TARGET_ARCH_VARIANT := armv8-a-branchprot
Change-Id: I689033f16663822e263f45ab5e46c9f120c9dca0
Since Make does not visit modules in topological order of their
<uses-library> dependencies, information from dependencies to the
dexpreopted module has to be passed via dexpreopt.config files. A build
rule for a dexpreopt.config file depends on dexpreopt.config files for
<uses-library> dependencies, and dex_preopt_config_merger.py script
extracts the necessary information from dependency configs and patches
the module's config.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: Id0b71170a4d2ab1d33059de0e9ad9d7e61f2345e
Support building cc modules with different library linking strategies
for apexes, to support different apex strategies pre/post S
Test: m APEX_BUILD_FOR_PRE_S_DEVICES
library_linking_strategy_sample_binary
Bug: 154663937
Change-Id: Ic95ddce5e2b0619adff5b6b8a4e2da0d44df7173
The new variable stores the name of the <uses-library>, if for some
reason it is different than the name of the module for that library
(normally they are the same, but sometimes differ). If the variable is
undefined or defined to an empty string, the name of the <uses-library>
is considered to be the same as the name of its module. It is recorded
in a module's dexpreopt.config file as "ProvidesUsesLibrary" field.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: Iac1be0fac79a944e4c26867cc58e274b76709da8
For now the subcontexts are always empty. A follow-up change will
actually populate that field from dependencies' dexpreopt.config files.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: I6cf0e8363dff4b406cb62ba3d76d566dbfec57e7
vendor-ramdisk-debug.cpio.lz4 will be packed into a
vendor_boot-debug.img and usually won't be flashed directly.
Moving it to the intermediate build directory.
Bug: None
Test: make vendorbootimage_debug
Change-Id: I86ea72c7a3484a90286139624408d07ce4bf994c
Bug: 176171716
Test: build_mainline_modules.sh and ensure that the ART boot image is
built: `find out -type f -name '*.art'` is nonempty for every arch.
Change-Id: I5163ebe1f8db3d4d921bab0f83cd388ba29419ec
Test: mm -j && mm -j dist, make sure virtual_ab_compression is set to
true in dynamic_partition_info.txt of target_File.zip
Change-Id: Id0d46bd8ed738b2b69a0c01481b7db0c7127f773
When building vendor_boot v4 with multiple vendor ramdisks, pass the
name of each ramdisk to mkbootimg.
Bug: 177882072
Test: Modify a device BoardConfig and verify the built vendor_boot.img
with unpack_bootimg
Change-Id: I679d277aa88f1e10678d2f74b579e4140536e267
Commit I56dda56bab7def1540f4fb506323e3e605620cd4 adds
BOARD_MOVE_RECOVERY_RESOURCES_TO_VENDOR_BOOT and includes
$(TARGET_RECOVERY_ROOT_OUT) directory when building vendor
ramdisk if BOARD_MOVE_RECOVERY_RESOURCES_TO_VENDOR_BOOT is set.
We should do the same thing when building the vendor debug
ramdisk as well. Otherwise, lacking some recovery files
makes the device unable to boot into fastbootd/recovery mode.
Bug: 177712689
Test: `make vendorbootimage_debug`, then unpack
./obj/PACKAGING/vendor-boot-debug_intermediates/vendor-ramdisk-debug.cpio.lz4
to check it includes recovery resources via:
`lz4 -d -c ../vendor-ramdisk-debug.cpio.lz4 | cpio -idm`
Test: device can boot into fastbootd after flashing vendor_boot-debug.img
Change-Id: I4744c6600866436ed7d5a321e67e20b8f3da8a81
By enabling these hiddenapi CSV files to be prebuilt, it
becomes possible to create a split build that supports
the hiddenapi encode dex step, but doesn't contain all
of the java sources needed to generate the CSV files.
Bug: 175048716
Test: m nothing
Test: new TestHiddenAPISingletonWithPrebuiltCsvFile
Test: local build without prebuilt hiddenapi
Test: local build with prebuilt hiddenapi
Change-Id: Ia38c5016d2aeba54aa537a5ce601898d46330730
New clang compiler enables unsigned-shift-base as part of
integer sanitizers. But it makes some daemons crash at
libc++.
Bug: 177566116
Test: build.
Change-Id: I6d64fa5002b6035be4d960441eb5176c97152af9
As we add more partitions, INTERNAL_*IMAGE_FILES variables has grown
and the places where the variables are referenced had to be touched
everytime. This change introduces a new variable
INTERNAL_ALLIMAGES_FILES which concatnates INTERNAL_*IMAGE_FILES, and
replaces the uses of the variables with the new one.
Bug: N/A
Test: m
Change-Id: I62a0f725a0f6999421d04a6261370be9124e62d7
The reason that the jar file depended on modules_to_install was to
collect files from other partitions, not just system. It isn't the
intention to gather fake modules like phony modules and more importantly
their dependencies (LOCAL_REQUIRE_MODULES). This change explicitly
filters out the fake modules to make the intent clearer.
In fact, the jar file depending on fake modules can have a side effect
that overridden modules are installed. It can happen as follows:
PRODUCT_PACKAGES += A B C
LOCAL_MODULE := A
LOCAL_REQUIRED_MODULES := B
include $(BUILD_PHONY_PACKAGE)
LOCAL_MODULE := B
include $(BUILD_XXX)
...
LOCAL_MODULE := C
LOCAL_OVERRIDES_MODULES := B
include $(BUILD_XXX)
modules_to_install is set to
out/target/product/xxx/fake_packages/A-timestamp
out/target/product/xxx/system/.../C
; B is correctly overridden.
But there still is an order-only dependency from A to B which is created
by add-all-target-to-target-required-modules-deps. Unfortunately,
overridden isn't take into account here.
So, if the jacoco jar file depends on modules_to_install, both B and C
are installed. And if the creation of system.img happens later, the two
are included together, which is an error.
A preferred way of fixing this is to amend
add-all-target-to-target-required-modules-deps so that override is take
into account. However, it will likely to require more Make tricks. This
change takes a much easier path; pruning the fake packages from the
dependencies so that its dependencies which might be wrong can't be
installed.
Bug: 175400393
Test: build/soong/soong_ui.bash --make-mode SKIP_ABI_CHECKS=true
SOONG_COLLECT_JAVA_DEPS=true TARGET_PRODUCT=cf_x86_phone
TARGET_BUILD_VARIANT=userdebug
TARGET_USERDATAIMAGE_PARTITION_SIZE=17179869184 droid dist
DIST_DIR=out/dist EMMA_INSTRUMENT=true EMMA_INSTRUMENT_FRAMEWORK=true
CLANG_COVERAGE=true NATIVE_COVERAGE_PATHS="*"
NATIVE_COVERAGE_EXCLUDE_PATHS="art bionic/libc device"
continuous_instrumentation_tests continuous_native_tests device-tests
platform_tests
No com.android.art.apex in system.img
Change-Id: Ie8a917b7c3985c3a4f19372a4430accd84124e83
When an APK is packaged without any classes.dex files the resources
are packaged directly, and so the packaging rule must depend on the
resources.
Fixes: 177295654
Test: prebuilts/build-tools/linux-x86/bin/ninja -f out/combined-aosp_cf_x86_64_phone.ninja -t query out/target/product/vsoc_x86_64/obj/APPS/CtsSplitApp_x86_64_intermediates/package.apk
Change-Id: I786e71439f11e7bd5e20e010b66fbf529eb57a5a
