Currently, emulator system image is the non-legasy GSI, which is in sparse
format. Several internal clients don't support sparse format, and other
internal/external clients usually support both. The reason to choose sparse
format for GSI may not be valid any more. See bug for detail.
Bug: 130341962
Test: Checked the header of system.img to see if it's unsparse.
Boot emulator aosp_x86 successfully.
Boot mixed GSI and CF x86 successfully.
Change-Id: I839f49809429227607b2e876a4a3628df43a97b4
It is now in the Runtime APEX and considered internal there.
Test: Flash and boot
Test: atest CtsCompilationTestCases CtsBionicTestCases
Bug: 118374951
Bug: 124293228
Change-Id: I33bb9c238d7db46795deb592c9d20fe6591c1654
Adds the following line into a device.mk in order to include
GSI keys into the first-stage ramdisk.
$(call inherit-product, $(SRC_TARGET_DIR)/product/gsi_keys.mk)
Bug: 112293933
Test: make checks $OUT/ramdisk/avb/.
Change-Id: I8e4a9a8dd607916a3f45a07e0ba14c58def5d8d3
The jar for the media command line tool is now installed as a
dependency of the media package.
Bug: 122332340
Test: lunch generic-eng && m nothing
Change-Id: I15ef7cb373ff249887f5641ca9c9722bfc3263a2
Merged-In: I15ef7cb373ff249887f5641ca9c9722bfc3263a2
The _a targets are build for Legacy GSI for non-system-as-root
partition layout devices. The Android build system is not able
to build a system image with non-system-as-root anymore.
Also, `init` supports mounting system-as-root system image at
all, all devices should no problem to mount it.
Bug: 124278972
Test: none
Change-Id: I9c56e7e5d8517a630854f030616c3b3ecb94b9a8
This flag is not allowed to be true when dynamic partitions are
enabled, which mainline devices are expected to do.
Bug: 80410283
Test: make
Change-Id: I50c9fe9332ea1490a84484f9311cfe001a083754
Building of host tools stopped when moving to APEX. This change
re-enables building of host tools.
Bug: 128990084
Test: m -j72 && find out -name 'oatdump'
Change-Id: Iad42b4e36561001709634174014192e56a64aafd
This setting moves the PRODUCT_PROPERTY_OVERRIDES to
the vendor partition.
Bug: 80410283
Test: tail system/build.prop | grep ro.carrier=
(empty)
Change-Id: I30d564fbd2dd3b56763ccff77979b223857f1c2f
This config was in a weird state where it sets the OUT
directory for these partitions but not the file system
type. Within the build system, both of these variables
are used interchangeably to check whether these partitions
are expected to be present.
A typical mainline device is expected to /vendor and
/product, so just add these.
Bug: 120974093
Test: make
Change-Id: I803d7afdcadc37e4384ae80f08bc1c8db535d2f4
Configure this product not to build any of the other partitions.
Note that the only partition that currently has artifacts produced
is system_other, and this change doesn't appear to prevent odex/vdex
files from being created there anyway. However, it does allow
configuring the target device (mainline_arm64) to have more partitions
in a followup change.
Bug: 111533333
Test: make
Change-Id: Ib115cb03bc99b9b61e0d6f1b622784e7ad5fcea0
The adb_debug.prop will be loaded by init when the file
/force_debuggable is present in the first-stage ramdisk, and the device
is unlocked. This file has the highest priority to override other system
properties, thus can override ro.debuggable to 1 to allow adb root.
Bug: 126493225
Test: boot a device with /force_debuggable in ramdisk, checks
adb_debug.prop is loaded
Change-Id: I75d661bb04178f584d8b6f5eba26ae35ccb01239
The file $(TARGET_OUT_ETC)/update_engine/update-payload-key.pub.pem
is installed conditionally based on this variable, and was found
to differ between mainline_system and the intended device targets.
Also move the variable from gsi_common.mk as GSI inherits the mainline
boardconfig. The variable is not a PRODUCT flag (and not inherited as
such), so fits better in the BoardConfig I think.
Bug: 80410283
Test: presubmit
Change-Id: Ida6fa923fd01b062ff3b93e7737e1e6edf7e4fa1
It used to be depended on by something on /product, but this
is no longer the case.
Bug: 119619080
Test: build downstream users of mainline_system
Change-Id: Iabe1dbb45499b25295f741a62336cb86e4565d60
RescueParty reboot the device to recovery mode if there is any
priv app die continuously. The behavior blocks the testing.
Disable the feature temporary before all problem apps fixed.
Bug: 120679683
Test: Boot a device with GSI and a die app
Test: Unplug USB and does not auto reboot
Change-Id: I70e5b7222299c6d0173d87e41d7a68beeda77dd5
Some PRODUCT_PROPERTY_OVERRIDES has no effect, ex.
ro.dalvik.vm.native.bridge=0 in runtime_libart.mk,
because gsi_common.mk uses ':=', not '+='.
Bug: 125829345
Test: build aosp_arm64-userdebug
Test: check the properties in /vendor/build.prop
Change-Id: Ib5e026069e7919fed8a00f905f3dbaa15042107f
This makes the in-process configuration closer to the "standard" module
configuration, by using an APK in both cases, but making sure that
svelte builds use an in-process service to avoid the memory impact.
Bug: 127908503
Test: svelte build boots and has functional WiFi
Change-Id: I59f44df7c23a24429e9fe6be034260373e33a994
Adds icu-data_host_runtime_apex to fix unbundled builds which pull it in
via PRODUCT_PACKAGES, but are missing packages that would pull it in via
PRODUCT_HOST_PACKAGES.
Test: build/soong/build_test.bash
Test: in ub-timezonedata-master; tapas TimeZoneData; m
Change-Id: I1583c7582b386c3e8478711cb1df340518d763c1
Merged-In: I1583c7582b386c3e8478711cb1df340518d763c1
It used to be installed to recovery with the rsync from normal root to
recovery root.
Bug: 114042635
Test: `m -j bootimage` with aosp_walleye-userdebug.
Change-Id: If3451f21018f271b1df83a31d4ece382e0dfebb9
In BoardConfig.mk, ODM manifests can be specify as follows:
# installed to /odm/etc/vintf/manifest.xml
ODM_MANIFEST_FILES := device/foo/bar/odm_manifest.xml
# installed to /odm/etc/vintf/manifest_baz.xml
ODM_MANIFEST_SKUS := baz
ODM_MANIFEST_BAZ_FILES := device/foo/bar/odm_manifest_baz.xml
This is preferred over PRODUCT_COPY_FILES because it uses
assemble_vintf to check manifest at build time.
Test: builds
Change-Id: I18517b8ba09d3271627d6f21b8250b1bfc9e5db3
The major purpose of vbmeta.img built on GSI targets (e.g., aosp_arm,
aosp_arm64, etc) is to disable AVB. We should also set the rollback
index to zero, to prevent the device bootloader from updating the
last seen rollback index in the tamper-evident storage.
Bug: 122583908
Test: build aosp_arm64, then `avbtool info_image --image $OUT/vbmeta.img`
Change-Id: I48a49957f8dd3169003b9507fe80e519f301d5b5
It must be there for libjdwp to find it and load the debugger transport
correctly. This is a temporary fix while we are looking into loading libnpt
from the Runtime APEX.
Bug: 124498388
Test: adb forward tcp:5555 jdwp:`adb shell pidof com.android.email` && jdb -attach localhost:5555
Test: atest CtsJdwpTestCases
Change-Id: I644c092c2ad5f6c9669bdccc24c7953d3be6b328
GSI doesn't include gsi_system.prop on user build, and cause some
devices cannot boot GSI with user build. Some properies in
gsi_system.prop do not suitable for user build, the patch create
a new gsi_system_user.prop for GSI user build.
Bug: 124479721
Test: boot aosp_arm64-user on a blueline device.
Change-Id: Ide2bdc8e8733817afb832b50b28807768e22f011
This reverts commit 74f27716f3.
The key/cert pairs for APEXes are moved to the same directory where the
APEXes are defined.
Test: Check this for all APEXes
jarsigner -verify -verbose -certs out/target/product/blueline/system/apex/com.android.tzdata.apex shows
...
X.509, CN=com.android.tzdata, OU=Android, O=Android, L=Mountain View, ST=California, C=US
Change-Id: I8d867b434b93aa68cbc707fa20ed8dbd4cef755e
