To be able to avoid registering dependencies on nonexistent host APEXes
in host builds.
Test: art/tools/buildbot-build.sh --host
with http://r.android.com/q/topic:libdexfile-noext applied
Bug: 183217299
Change-Id: Iaa6411b511b6f50da01827b49852607ae825bc83
building test_for modules.
This extends the current approach where test modules always depend on
the platform variants of the APEX libs, and only skips the stubs on
them. It still has the limitation that the internal libs must have the
exact same apex_available lists.
Also some improvement of the test accuracy in TestTestFor.
Test: m libartagent-target
with http://r.android.com/q/topic:libdexfile-noext applied
Bug: 183217299
Change-Id: I2118b8a22c887077867a3ddbbe73437b4a29a6ad
Bug: http://b/181740505
Explicitly adding dependency to profile runtime causes apex dependency
errors. We're already passing -fprofile-instr-generate to the linker so
depending on the profile runtimes is unnecessary.
Test: pgo-coral-config1 on internal release branch and run tests on
APCT.
Change-Id: Ie54785726c7ac044e5a5a2299073d02d45d5e0a7
Background: `min_sdk_version` of a crt object is the oldest SDK version
that the crt object supports. When it's set to for example 16, Soong
creates a number of versioned variants of the crt object starting from
16 to the latest version. The variant for version X is provided to NDK
clients having `min_sdk_version` set to X.
Problem: all versioned variants of a crt object were built with `-target
<arch>-linux-android16`. Therefore they all have been with `#define __ANDROID_API__
16`. This is because the mutated variants still have the same
min_sdk_version property and the clang triple follows min_sdk_version,
not sdk_version. This is too conservative and against our intention to
provide the latest crt object that matches with the min_sdk_version of
the client.
In the other hand, the platform(non-sdk) variant of the crt object
doesn't have such a problem. min_sdk_version is completely ignored.
However, this is a bug and will be fixed by aosp/1640364. As a side
effect of the fixing, the platform variant will begin to built with a
very old __ANDROID_API__ which unnecessarily turns off the new platform
features like the TLS segment over-alignment.
This change fixes the problems:
* For the versioned variants of crt objects, we set both
`min_sdk_version` and `sdk_versio` to the version that the variant is
created for.
* For the platform variant of crt objects, `min_sdk_version` is force
reset to "current".
Bug: 183191008
Test: m
Change-Id: I8c9d0fcea816de8cd1532dac4a47eee4f726c037
modules converted with bp2build_available are will also be available to
be used in mixed builds.
Test: build/bazel/scripts/milestone-2/demo.sh full
Test: go tests
Change-Id: I49f16ec3ba5bb11dfed8066af069c27eb04371fb
Previously when an APEX whose min_sdk_version is set is linked to an
external library providing multiple versions of stubs, the
maximum version that is less than or equal to the min_sdk_version was
chosen. For example, if the versions of a library stubs are 28, 29, 30,
and 31, then APEX with min_sdk_version: 29 linked to the version 29 of
the stub.
This was to ensure that the APEX doesn't use any new APIs whose
existence can't be guaranteed.
This however imposes a severe restriction that the APEX can never use
new APIs even when the APIs are actually available: i.e. when the
APEX is running on a newer platform.
With the recent work about unguarded availability, using the future APIs
became much safer. When you use an API that is newer than your
min_sdk_version, the API is automatically declared as a weak symbol
(thus no link error at runtime), while the call to API is guaranteed to
be guarded with the `__builtin_available(...)` macro.
So, there really is no reason to use the old version of the stub. We can
always use the latest version of stub safely.
Bug: N/A
Test: m
Change-Id: Iaac0d8761d8929154527dc2e861a51ae31e23d49
This allows more direct access to the bazel label and whether the module
is bp2build available.
Test: go test *
Change-Id: I23375835d20fa53d7d94127b2dc2d5bb20487bfb
This CL introduces a basic bp2build converter for cc_object modules.
cc_objects maps cleanly to cc_library targets, but with -fnoaddrsig.
This CL also demonstrates generating include deps within a macro to
allow the cc_object compilation to depend on a relative-include header
within an include dir.
e.g. if "foo.cc" includes "android/log.h" and the latter is located at
"include/android/log.h", the autogenerated header deps would export
"android/log.h" correctly to the foo.cc upstream target.
Test: GENERATE_BAZEL_FILES=true m nothing && bp2build-sync write && bazel build //bionic/libc:crtbegin_so1
Change-Id: Ifd9e097051ec184ab0a1929d07918f0ff4f24d98
Similar to our vendor support, this adds support for linking rust static
libraries to vendor ramdisk cc modules.
A bug fix is also included where a restriction against setting rust_ffi
vendor-specific was not being enforced.
Bug: 179397942
Test: Example modules link, Soong tests pass.
Change-Id: I737cdf0c2f49ab349bcea2a0429e6298ebc1313e
This currently expands all globs, still need to support converting glob
syntax.
Test: go build_conversion_test
Test: GENERATE_BAZEL_FILES=true m nothing
Test: m nothing
Bug: 165114590
Change-Id: If7b26e8e663d17566fad9614ca87a8da1f095284
When a module has both 'product_specific: true' and
'vendor_available: true', the module does not have core variant but
have product and vendor variant modules. As the module generated
multiple variants, the names of the modules have the variant name
suffix. However, it did not create the module with the base module
name because it does not have core variant.
As the base module, in this case, is the product variant, use the
base module name for the product variant without adding the
'.product' suffix to it.
Bug: 161913599
Test: m hidl_test_product@2.0-adapter-helper
Change-Id: I904b43744579a6e89cecbc2e529122eaa8b14cae
The module names for vendor and product variants have the image
variant suffix to avoid conflict with the core modules. It requires
updating the module names in the dependency tree with the suffixes.
We had a hidden bug that used the original module properties to
update the names of its dependent modules.
Also, it must cover the product variants modules.
Test: updated cc_test.go and build
Change-Id: I6b4ea062d13c8fac1e699138d44376e52e0d7852
Propgate the Android.mk suffix from source modules into the snapshot so
that it can be used for the prebuilt modules.
Bug: 177098205
Test: vendor_snapshot_test.go
Change-Id: Iea151dc91395f714fbcad1df3a6fd0874e5455d9
Snapshots storead global sets of modules that should be replaced with
vendor snapshot modules. Move the data instead to a vendor_snapshot
or recovery_snapshot module type that depends on all the modules in
the snapshot, and then have modules that should use the snaphsot
depend on it to query for the set of modules that should be replaced.
Bug: 177098205
Test: vendor_snapshot_test.go
Change-Id: I2826adacfb473e9139b5ea93ba83b8a54cc1a56b
'vendor_available: true' creates a vendor variant from a system
module. The vendor variant of the module is installed to /vendor.
However, we may want to install the vendor variant to /odm, instead.
'device_specific: true' does not work for this purpose because
'vendor_available: true' is allowed only for the system or product
modules to create a vendor variant. But 'device_specific: true'
itself creates a vendor variant that may not work with
'vendor_available: true'.
To install the vendor variant to /odm, we define a new property
'odm_available'. 'odm_available' is exactly the same as the
'vendor_available' except the install path of the vendor variant.
By defining 'odm_available: true', the vendor variant of the module
will be installed to /odm or /vendor/odm instead of /vendor.
Bug: 176147321
Bug: 176079978
Test: check if a module with 'odm_available: true' is installed to
/vendor/odm
Change-Id: I2d16bd2c515796597b2fbd1eb66f7c2736434697
Windows prefer32 was implemented in cc, which caused a mismatch
between cc and other packages like python. This manifested as
missing variant when a python_test_host with compile_multilib: "first"
tried to depend on a cc_binary_host, which implicitly has multilib
"first".
Fixes: 177904375
Test: m checkbuild
Change-Id: Iaaac16ae171c06d90d04d7cac11789d3f39b8d99
Native bridge modules never need to build against NDK stubs, so
don't create SDK variants. Also clear the sdk_version property
for modules that don't have SDK variants so that later code doesn't
use it to trigger SDK behaviors.
Test: m checkbuild
Change-Id: I1920fa82e9fab06235f01a62624382efa16cc6e3
The change will detect if a module is compiled against cfi support, and
will usebuild the coverage build against the libprofile that supports cfi. This is to resolve compilation errors when
building against modules with cfi support.
Bug: 177098919
Test: forrest build for cf_x86_phone-userdebug_coverage http://go/forrest-run/L81700000786828933
Change-Id: I8e0421cdf1c6e499292cfa3457cefd3c42f13155
Merged-In: I8e0421cdf1c6e499292cfa3457cefd3c42f13155
The check no longer tracks private dependencies like stubImplDepTag and
staticVariantTag. It also doesn't do the check for double_loadable
libraries that are not depended on by a LLNDK library.
Bug: N/A
Test: m
Test: add 'double_loadable: true, gen_trace: true` to an aidl_interface
module. The build doesn't break.
Change-Id: Iccd1a9d445a48d03c373708ba1bdd34b9a7f152d
Move the global lists of VNDK modules into boolean properties on the
modules themselves, and use the new SingletonModule functionality
to visit all modules and collect the list of modules with the properties
set.
Bug: 176904285
Test: all soong tests
Test: m checkbuild
Test: compare Soong outputs
Change-Id: Icf7e2f8f190a517d30f0780e98762bc0084ddb24
Previously, for cc_* modules, __ANDROID_API__ tracked the sdk_version
property. This however has caused a few number of problems:
1. It's confusing. __ANDROID_API__ has meant minSdkVersion. Therefore
the sdk_version property should mean minSdkVersion (since the macro
tracks the property). However, the introduction of the new
min_sdk_version property (which is currently for APEX) made this very
confusing. Also, this is not consistent with the java_* modules where
sdk_version means compileSdkVersion.
2. This is preventing go/android-future-symbols. The plan is to make the
APIs that are above the minSdkVersion available as weak symbols.
Previously those APIs had to be accessed via dlsym because they are
hidden behind the __ANDROID_API__ macro at build-time. To use make the
APIs visible at build-time, the module authors had to __ANDROID_API__
beyond their minSdkVersion. This is against the definition of
__ANDROID_API__.
To solve above problems, __ANDROID_API__ now correctly tracks
min_sdk_version. In addition, min_sdk_version now defaults to
sdk_version. Therefore, most of the modules that don't set
min_sdk_version aren't affected by this change.
Bug: 163288375
Test: m
Change-Id: I645e6bb1234c27ae0a69b7b87a59206cfd350744
Memtag_heap adds an ELF note that enables MTE heap tagging in
bionic/scudo. Ignored on non-executables. With diagnostic
(diag:{memtag_heap:true}) enables the SYNC mode, otherwise - ASYNC mode.
Memtag_heap defaults to set (with diag) on cc_test targets, unset
otherwise. Ignored on non MTE-compatible hardware.
Bug: b/135772972
Test: soong tests
Change-Id: I88fd0f159e609e17bd13487749980a1ba02cb91c
Adds a PlatformSanitizable interface which both CC and Rust can
implement so that the sanitizer mutators in CC can sanitize Rust
shared/static libraries appropriately.
Bug: 147140513
Test: m nothing
Change-Id: Ib31103b6c4902a4d5df2565c0d7c981298d100a3
As we replaced '(vendor|product)_available: false' with
'vndk.private: true', update the soong comments with it.
Bug: 175768895
Test: na
Change-Id: Iaa08d0caf2995d2327443e156abe8448c4e500c5
In case of VNDK, 'vendor_available: false' had a special meaning that
defines VNDK-private libraries. It is not trivial because not
defining a boolean property means 'false' normally. To avoid the
confusion replace it with the 'vndk.private: true' for VNDK-private
libraries and 'private: true' for LLNDK-private libraries.
All VNDK libraries must define 'vendor_available: true' and may have
'vndk.private: true' if they are VNDK-private.
With this change '(vendor|product)_available: false' is the same as
not defining the property.
LLNDK-private must define 'private: true' instead of
'vendor_available: false'.
Bug: 175768895
Test: build
Change-Id: I57fbca351be317257d95027f3cdcdbbe537eab23
The shared variant of a cc_library module depends on the static variant
of the same module. The dependency is just to optimize the build time by
not compiling the same source files twice. It doesn't mean that the
"static:" dependencies of the static variant are used to build the
shared variant. Therefore, the inter-variant dependency is considered
as NOT being in the same APEX.
Bug: N/A
Test: m
Change-Id: I7b8ff8e5bd5c9eb31fc4d8d18bd93dd80296fb81
To avoid adding ubsan to the apex allowed_dep list, this commit adds a
check on depedency tags to see if apex check should be skipped.
The check is only used on sharedLib dependencies when diag mode are enabled
for sanitizers.
Bug: 158010610
Test: make build for aosp-sargo and aosp_cf_x86_phone-userdebug
Change-Id: I3d7dbb70d8c80ffae1854819cf8cf9e6b0b15c00
To define VNDK-private libraries, we used `vendor_available: false`.
Because of it, `vendor_available == nil` had different meaning from
`vendor_available: false` for the VNDK libraries.
To clarify this, we change the logic for defining VNDK-private
libraries which was:
cc_library {
name: "vndk_private",
vendor_available: false,
product_available: false,
vndk: {
enabled: true,
},
}
It must be replaced with
cc_library {
name: "vndk_private",
vendor_available: true,
product_available: true,
vndk: {
enabled: true,
private: true,
},
}
Bug: 175768895
Test: m nothing
Change-Id: I81769f57c2231e54b682a28e4b82631ab9f3d390
With this patch, `vendor_available: true` will no longer creates
product variant. Instead, modules need to set `product_available:
true` if they have to be available to product vanriant.
If both properties are defined for VNDKs, they must have the same
values.
Bug: 150902910
Test: m nothing
Change-Id: I28fb6886e6114583227a31151136627c8516ac9c
Rewriting LLNDK dependencies with .llndk suffix requries referencing
a global data structure to determine if a given library is an LLNDK
library and therefore needs the .llndk suffix. References to
global data structures from mutators must be removed to support
incremental Soong analysis. Instead, move the LLNDK stubs rules
into the vendor variant of the implementing cc_library so that
the original name can be used.
As an incremental step, the llndk_library modules are left in
place, and the properties are copied into the cc_library via
the dependency specified by the llndk_stub property. A followup
will move the LLNDK properties directly into the cc_library and
delete the llndk_library modules.
The global list of LLNDK libraries is kept for now as it is used
to generate the vndk.libraries.txt file.
Bug: 170784825
Test: m checkbuild
Test: compare Soong outputs
Test: all Soong tests
Change-Id: I2a942b21c162541a49e27b2e5833c9aebccff1d0
SkipInstall is actually primarily used to prevent making a module
visible to Make, rename it and add new SkipInstall that actually
skips installation without affecting Make.
Call c.SkipInstall() for uninstallable cc modules to allow calling
c.installer.install, which will collect PackagingSpecs for
uninstallable cc modules, allowing them to be used by genrules.
Bug: 124313442
Test: m checkbuild
Change-Id: I8038ed5c6f05c989ac21ec06c4552fb3136b9a7a
After the next patch libraryDecorator.install will be called for
uninstallable variants of modules, manually filter them out when
copying to the NDK sysroot.
Bug: 124313442
Test: m checkbuild
Change-Id: I28b538d4ae271dc5e27c386d7cfa538ac0ed841b
Martin Stjernholm