Previously, when Q-targeting apexes are bundled-built, they are built
against the latest stubs.
It was because unwinder is linked dynamically in R and APIs are provided
by libc while Q apexes should run on Q where libc doesn't provide those
APIs. To make Q apexes run on Q device, libc++ should be linked with
static unwinder. But, because libc++ with static unwinder may cause problem
on HWASAN build, Q apexes were built against the latest stubs for bundled
build.
However, Q apexes should be built against Q stubs.
Now, only for HWASAN builds, Q apexes are built against the latest stubs
(and native modules are not linked with static unwinder).
Bug: 151912436
Test: TARGET_SANITIZE=hwaddress m
=> Q apexes(media, resolv, ..) are linked with the latest stubs
m
=> Q apexes are linked with Q stubs,
and Q apexes' libc++ is linked with static unwinder
Merged-In: If32f1b547e6d93e3955c7521eec8aef5851f908c
Change-Id: If32f1b547e6d93e3955c7521eec8aef5851f908c
(cherry picked from commit 7406660685)
Exempt-From-Owner-Approval: cp from internal
Change-Id: If32f1b547e6d93e3955c7521eec8aef5851f908c
It looks like sometime in late February our Mac builds started taking
~10 minutes longer than before. On my local workstation the Soong tests
were taking >25 minutes before completing (likely because I don't have
the older SDKs installed, and we iterate from older to newer to find the
oldest installed SDK).
Most of this time was spend running the `xcrun` tools to interrogate the
system about which Mac SDKs are installed and where the tools are. This
will never change during any build or test, so wrap it all in a
sync.Once so that we only ever call them once.
And remove the macSdkPath variable, which has been unused for years and
no longer works (as we don't allow the use of xcode-select during the
build).
Bug: 153010389
Test: prebuilts/build-tools/build-prebuilts.sh on a Mac
Change-Id: I39b2d49739e628e4c11bec4805b25039115d2fd0
A library may have both VNDK variant and stubs. getRefAbiDumpFile should
differentiate its vendor and core variants.
Bug: 152277104
Test: Add dumps to platform/ and vndk/ ; make libselinux.vendor libselinux
Change-Id: Iad038cf4cd3eccc3dfbef13fab67da044498ce77
Necessary to get correct prebuilts for many Bionic libs.
Cleaned up numerious "system_shared_libs: []" from test fixtures, since
they otherwise would need correction in the expected results, and it is
better to have a single test focused on testing system_shared_libs
propagation.
Test: m nothing
Bug: 152255951
Change-Id: If2e8a5296223e6281d833312660e8e9e4cd184c0
This causes the compiler to emit some additional debug infomation that
will be used for sampling PGO. These debug infomation will get stripped
so it only affects intermediate files.
Test: build
Bug: 79161490
Change-Id: Ie4d1d5ffbd311ba6e268cb94a618f5272be246ef
Also remove a deprecated flag -fsanitize-hwaddress-abi=platform.
Switching to the new pass manager reduces size of the "lib64" directory
by almost 6%.
Bug: 135298400
Test: objdump -d path/to/symbols/system/lib64/*.so | grep __hwasan_check_
Test: adb shell sanitizer-status
Change-Id: I327ee982e20c02b3b2126d93189048d9f4bfb301
Previously passing additional information to the implementations of
AddPrebuiltModule() or the SdkMemberProperties interface would have
required making changes to the API. This change added an
SdkMemberContext object into which additional information can easily
be added without requiring changes to existing implementations.
The BuildSnapshot() method was not modified because it is deprecated
and will be removed in a follow up change.
It also switches the API from passing variants as android.SdkAware to
android.Module. That is for a couple of reasons:
1) SdkAware is designed for managing the relationship between the
module and the SDK, not for generating the output snapshot. As such
there is nothing in SdkAware that is needed for generating the
output snapshot.
2) Accepting android.Module instead makes it easier to use the
underlying code for generating the snapshot module as well as the
individual member modules.
This is in preparation for a number of improvements and bug fixes in
both the snapshot creation code and implementations to address found
while trying to built the platform against ART prebuilts.
Bug: 151937654
Test: m nothing
Change-Id: Iac10f1200c0f283aa35402167eec8f9aeb65a38e
The macro is required only for apex variants regardless of useVndk.
Before the enforcement of LLNDK sdk version, the macro was not passed to
vendor variants.
Bug: 151689896
Test: TARGET_BUILD_APPS=com.android.media.swcodec m
libbase in swcodec apex is linked with liblog#29
(compiled with __ANDROID_SDK_VERSIO__=29)
Change-Id: I57fa4afe027eb39b98bd94d534be9ebe11713f19
This reverts commit 380fc3615c.
Reason for revert: breaking one of internal targets
Change-Id: Ica96c44078e5a7f674410828af3ba851317775dd
Exempt-From-Owner-Approval: revert to fix build failure
Added link type support in module creation code as cc_prebuilt_library
has to provide both shared and static libraries.
Had to add some calls to FirstUniquePaths() around the include dirs
settings in library_sdk_member as otherwise the shared variant contains
duplicate include dirs (presumably because it gets one set from its
dependency on the static variant and one set of its own). That
difference in include dirs causes problems in the generated
cc_prebuilt_library.
Test: m nothing
Bug: 142918168
Change-Id: Ie7f23fc2341c83c7814cc98e3970df4f5d4c8423
The marked library(ies) were available to the APEXes via the hand-written
whitelist in build/soong/apex/apex.go. Trying to remove the whitelist
by adding apex_available property to the Android.bp of the libraries.
Bug: 150999716
Test: m
Change-Id: I91d1b1076733a949ca2a959ba640ce34d0233492
Test: m nothing
Test: Add
sdk {
name: "runtime-module-sdk",
native_shared_libs: [
"libc",
"libdl",
"libm",
"ld-android",
],
native_objects: [
"crtbegin_dynamic",
"crtbegin_static",
"crtend_android",
],
}
to bionic/apex/Android.bp. Then:
build/soong/scripts/build-aml-prebuilts.sh runtime-module-sdk
Take the generated runtime-module-sdk-current.zip and unzip into a
master-art tree without bionic/, edit the generated Android.bp to
extend cc_prebuilt_* modules with:
nocrt: true,
stl: "none",
system_shared_libs: [],
apex_available: ["//apex_available:anyapex"],
recovery_available: true,
vendor_available: true,
ramdisk_available: true,
Then "m com.android.art.debug". This passes Soong but fails in the
build step because more members are required.
Bug: 148934017
Change-Id: I2ab8f6aadb1440b325697cae4a8ed761c62d15d2
When exporting Soong modules to Makefile, each VNDK snapshot module
whose version is BOARD_VNDK_VERSION has ".vendor" suffix. So makeLibName
should also return ".vendor", not the full vndk snapshot suffix like
".vndk.29.arm64.vendor".
Exempt-From-Owner-Approval: cherry-pick
Bug: 151274994
Test: manual build with snapshot
Change-Id: I48f606f90f6c5ca55d0a5bc29bad491c38a44f47
Merged-In: I48f606f90f6c5ca55d0a5bc29bad491c38a44f47
(cherry picked from commit 7ac1fa7823)
Even though use_vendor:true is prohibited, there is media.swcodec apex
which is still use_vendor: true and also needs to support Android10.
(min_sdk_version: 29)
Because LLNDK stubs were provided only for the current VNDK version,
media.swcodec couldn't be built against min_sdk_version: 29.
This change introduces additional versions for LLNDK stubs which are
enforced when an apex with use_vendor: true sets min_sdk_version.
To make things easier, the versions of LLNDK stubs are borrowed from its
implementation libraries.
Bug: 147450930
Bug: 149591522
Test: TARGET_BUILD_APPS=com.android.media.swcodec m
(with min_sdk_version: 29 set)
check if liblog/libc/libm/libdl stubs are 29
check if 29 stubs don't have new symbols.
Change-Id: I79946cbb4da6617138a96d2b254349d3a298e77b
Use the result of regex.ReplaceAllString()
Bug: 149591522 (not directly related to this)
Test: m (with new unit test)
Change-Id: Ifce07547ccc067f1ee5bd8467c2fb7d7f8387b8e
`make haiku` shouldn't build fuzz targets that are disabled on the
continuous fuzzing infra. This should reduce pressure on the repackager.
Bug: N/A
Test: make haiku && verify that example_fuzzer isn't there.
Change-Id: I6a687a3ab8cbb1eb75113e96aba2c4f9f8058010
This was only being used to set the "stl" property for cc library sdk
member type and so that functionality was moved to AddPrebuiltModule()
and FinalizeModule was removed.
Required a few test changes to move the property to the correct
position in the generated module.
Bug: 142918168
Test: m nothing
Change-Id: If6400189833d4ff3285e7a7adf63a9b509e2a03b
This change ensures that the runtime dependencies between a
binary/shared library are correctly specified in the snapshot so that
the build can ensure that shared libraries are built before the targets
that use them.
It adds support for differentiating between references that are
required to refer to another sdk member (required) and those that may
refer to either an sdk member or a non-sdk member (optional). The
latter is used for shared library references as the libraries used by
an sdk member may be provided from outside the sdk. e.g. liblog is not
part of the ART module but is used by some members of the ART sdk.
Bug: 142935992
Test: m nothing
Change-Id: Ia8509ffe79b208c23beba1880fe9c8a92b732685
Host binaries by default use the UBSan runtime whether or not
diagnostics are requested in the module definition. The rest of the code
assumes the runtime is only needed when diag properties are set.
This makes sure to add the runtime to the dependency graph of host
binaries that are UBSanitized.
Bug: 150336284
Test: ./art/tools/dist_linux_bionic.sh com.android.art.host
Test: runtime appropriately identified as a dependency of libziparchive
Change-Id: I2704aaba4abc7f4e6d96747917c77b5147fd1d56
The sdk snapshot creates two prebuilts for each member one that is
versioned and one that is not. If they are both installed then they
lead to duplicate rules in make for creating the same installed file.
This change adds an installable property to cc modules that will
prevent the installation of the file and then adds installable: false
on the versioned prebuilt for cc modules.
Bug: 142935992
Test: m nothing
Change-Id: I4cb294c2b0c8a3f411eea569775835d9e41726d6
Only disable CFI on incompatible variants rather than the base module.
Bug: 142157676
Test: Manual llvm-cfi-verify checks
Change-Id: I999e1dbcccc1ffb102b2fe01c47f1bda4df5e601
Updates the member snapshot creation code to support multiple os types.
It basically sorts the variants by os type, then applies the code to
optimize the arch properties and then it optimizes the properties that
are common across architectures and extracts any properties that are
common across os types.
The java and cc member types needed to be modified to make the location
of the generated files within the snapshot os type dependent when there
is more than one os type. That was done by adding an OsPrefix() method
to the SdkMemberPropertiesBase which returns the os prefix to use when
there is > 1 os type and otherwise returns an empty string.
Added three tests, one for cc shared libraries, one for cc binary and
one for java header libraries.
Bug: 150451422
Test: m nothing
Change-Id: I08f5fbdd7852b06c9a9a2f1cfdc364338a3d5bac
