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
When a shared library providing stubs is included in an APEX, only the
stub variant (i.e. version:"1", etc.) gets emitted to Android.mk. This
enforces that everything in the Make world to link to the stub providing
only the public APIs of the library. The non-stub variant (i.e.
version:"") isn't exposed, otherwise, others will be able to access
private part of the lib which isn't guaranteed to be stable.
This has been done by unhiding the stub variant when it is known that
the library is actually included in an APEX. Note that stub variants are
by default hidden.
The above mechanism however doesn't work when the shared library is
replaced by a prebuilt and when the APEX is also replaced by a prebuilt.
Then, the fact that the prebuilt library is actually in the APEX gets
lost. In that case, AnyVariantDirectlyInAnyApex() returns false for the
prebuilt library. As a result, the stub variant remains hidden and not
emitted to Android.mk.
This change fixes the problem by checking if the lib isn't available for
the platform at all. If not available for the platform (e.g.
apex_available doesn't have "//apex_available:platform"), the lib is
assumed to be included in an APEX even when it actually didn't go
through the apex mutator (... because it's a prebuilt).
Bug: 175166063
Test: m nothing
Change-Id: I41fabd5b368baecf4dc3c5a080b466f8bcd79d77
Generally ...DepsMutator methods add dependencies between modules but
the apexDepsMutator does not which can be confusing. This renames
apexDepsMutator to apexInfoMutator and adds some extra documentation to
clarify its function. It also renames the registered name for the
mutator and its associated providers from apex_deps to apex_info.
Test: m nothing
Bug: 171061220
Change-Id: Ic074a281215b23e982448ccff7ac075236123bee
Mostly documentation changes, but includes a few refactorings like
changing the variable names, reording functions, reordering statements
in logical order, etc.
Bug: 173472337
Test: m
Change-Id: I000c76e818722ed06bac03d9de87588b23552b08
The android.WriteFile rule takes careful escaping to produce the
right contents. Wrap it in an android.WriteFileRule that handles
the escaping.
Test: compare all android.WriteFile outputs
Change-Id: If71a5843af47a37ca61714e1a1ebb32d08536c31
Adds a singleton that traverses the module variants finding the ones
that are in the list (updatable and non-updatable) of boot jars and
add a ninja rule to ensure that they only contain packages from an
allowed list.
Replaces a hack that ignored any prebuilt boot jars supplied as dex
file with an equivalent one to ensure that they are still ignored.
A follow up change that switches to checking dex jars will allow the
hack to be removed.
The boot jars check can be strict or lax. If strict then all the boot
jars listed in the configuration must be found, otherwise it will only
check the ones it finds. It is strict by default unless
TARGET_BUILD_UNBUNDLED=true or ALLOW_MISSING_DEPENDENCIES=true.
Moves the script and data file from build/make.
Test: m check-boot-jars - for failing and passing cases
SKIP_BOOT_JARS_CHECK=true - no check-boot-jars target created
ALLOW_MISSING_DEPENDENCIES=true - not strict
TARGET_BUILD_UNBUNDLED=true - not strict
verified manually that apart from path differences the same
files (same check sum) were checked in both old make checks and
the new Soong ones
EMMA_INSTRUMENT=true EMMA_INSTRUMENT_FRAMEWORK=true m check-boot-jars
Bug: 171479578
Change-Id: I9d81d6650ba64fc0d48d2dab4ba5a3ba8dd03dec
Export information about static libraries, shared libraries and
exported flags through Providers instead of accessing the module
directly. Much more is left to be converted, but this significantly
simplifies the dependencies on libraries with stubs by making it easy
for a module to masquerade as another by simply exporting the
providers from the other module. Instead of depending on all the
versions of a library and then picking which one to use later, it
can depend only on the implementation variant and then select the
right SharedLibraryInfo from the variant.
Test: m checkbuild
Test: only expected changes to build.ninja
Change-Id: I1fd9eb4d251cf96ed8398d586efc3e0817663c76
A global variant was used to store the global mapping between
modules and APEXes. Replace it with storing pointers to APEX
contents inside each module so that they can query the contents
of any APEXes they belong to.
Bug: 146393795
Test: all Soong tests
Test: single line change to build.ninja host install dependency ordering
Test: no Android-${TARGET_PRODUCT}.mk, make_vars-${TARGET_PRODUCT}.mk or late-${TARGET_PRODUCT}.mk
Change-Id: Id2d7b73ea27f8c3b41d30820bdd86b65c539bfa4
apex_test.go wasn't listed in the Android.bp file, which allowed
it to bitrot. Make the API level methods take a PathContext
so that they can be called from a test using configErrorWrapper.
Also fix an int that was converted to a string.
Test: apex_test.go
Change-Id: I1ff87134c837bd5d344d22550baabde10d1b0b2e
Instead of tracking per module and per module variant, track allowed
list of dependecies for all modules combined. This avoids issues with
different products and different downstream branches having different
build graphs.
To compare allowed_deps.txt vs head, run:
:; m -j out/soong/apex/depsinfo/new-allowed-deps.txt.check
To update source allowed_deps.txt, run:
:; build/soong/scripts/update-apex-allowed-deps.sh
Bug: 149622332
Test: m
Change-Id: Ic518fbd9ebfe1b46aaf9a58df731780a7e5a676b
Merged-In: Ic518fbd9ebfe1b46aaf9a58df731780a7e5a676b
(cherry picked from commit 453555083b)
(cherry picked from commit e5207cd9a6)
I2954bb21c1cfdeb305f25cfb6c8711c930f6ed50 switched normalizeVersions
to work on ApiLevels, which inadvertantly caused it to return "current"
instead of "10000" for libraries that specify "current" in their stubs
property. ChooseSdkVersion couldn't handle "current" because it was
manually converting the version to an int. Switch ChooseSdkVersion
to use ApiLevels instead so that it can handle "current".
Test: m checkbuild
Change-Id: Id412359e092483ba419118dd03bc206fae702a96
APEX variants that share the same SDK version and updatability
almost always use identical command line arguments to build but
with different intermediates directories. This causes unnecessary
build time and disk space for duplicated work.
Deduplicate APEX variants that would build identically. Create
aliases from the per-APEX variations to the new shared variations
so that the APEX modules can continue to depend on them via the
APEX name as the variation.
This has one significant change in behavior. Before this change,
if an APEX had two libraries in its direct dependencies and one
of those libraries depended on the other, and the second library
had stubs, then the first library would depend on the implementation
of the second library and not the stubs. After this change, if
the first library is also present in a second APEX but the second
library is not, then the common variant shared between the two
APEXes would use the stubs, not the implementation.
In a correctly configured set of build rules this change will
be irrelevant, because if the compilation worked for the second
APEX using stubs then it will work for the common variant using
stubs. However, if an incorrect change to the build rules is
made this could lead to confusing errors, as a previously-working
common variant could suddenly stop building when a module is added
to a new APEX without its dependencies that require implementation
APIs to compile.
This change reduces the number of modules in an AOSP arm64-userdebug
build by 3% (52242 to 50586), reduces the number of variants of the
libcutils module from 74 to 53, and reduces the number of variants
of the massive libart[d] modules from 44 to 32.
This relands I0529837476a253c32b3dfb98dcccf107427c742c with a fix
to always mark permissions XML files of java_sdk_library modules as
unique per apex since they contain the APEX filename, and a fix
to UpdateUniqueApexVariationsForDeps to check ApexInfo.InApexes
instead of DepIsInSameApex to check if two modules are in the same
apex to account for a module that depends on another in a way that
doesn't normally include the dependency in the APEX (e.g. a libs
property), but the dependency is directly included in the APEX.
Bug: 164216768
Test: go test ./build/soong/apex/...
Change-Id: I2ae170601f764e5b88d0be2e0e6adc84e3a4d9cc
In preparation for reusing the same variation for multiple apexes,
rename ApexName to ApexVariationName.
Bug: 164216768
Test: all soong tests
Change-Id: I88f2c5b192ffa27acd38e01952d0cefd413222a0
unavailable-to-platform case.
This removes the special case added in https://r.android.com/1274763
from SkipInstall(), so that it doesn't cause conflicting AndroidMk
entries when a cc_prebuilt_library_static module has prefer:true and
the corresponding source module exists.
Test: `m` in a tree with a snapshot created from art-module-sdk where
the libartimagevalues module has prefer:true
Bug: 151303681
Change-Id: I651ae325753b707296892adb4cae80daaddb6af2
com.android.gki.* matches any APEX with the prefix com.android.gki.,
like com.android.gki.bar.
Test: use it
Bug: 162267963
Change-Id: Ie46fcb08b031611d26b2b6cde157253f51ba9bfc
apexDepsMutator marks all dependencies of apex modules. Previously, it
was converted from WalkDeps() to Top-down mutator to avoid the pitfall
of WalkDeps() bug. (It did't handle multiple visits via different
paths.)
Because WalkDeps() problem solved in aosp/1277516, apexDepsMutator can
be reverted to use WalkDeps().
Even though there's no observable difference between them, I revert this
for the up-coming change, which requires different pruning strategies
per apexes.
Bug: 159195575
Test: m
Change-Id: Ib09cbc7a3dfd143dd37b660b1aea6c71392ce2e3
Enforce min_sdk_version for every payload dependency of updatable
APEX/APKs.
android.CheckMinSdkVersion() calls ApexModule.ShouldSupportSdkVersion
for every transitive dependency from APEX/APK modules to see if it
meets the min_sdk_version requirements.
The common implementation for apex/android_app is provided in
android/apex.go.
Bug: 145796956
Bug: 152655956
Bug: 153333044
Test: m nothing
Change-Id: I4a947dc94026df7cebd552b6e8ccdb4cc1f67170
Introduce a singleton apex rule to merge contents of individual
deps-info results into a single output file.
Bug: 149622332
Test: m
Change-Id: I4ab7e1a3527fead97a81a5a2cb0e1e93a429117c
For java.AndroidApp app, app.MinSdkVersion() was selected from
ApexBundleDepsInfo, not from java.Module.
This caused app.MinSdkVersion() to return "" always.
Besides, having two embeded fields and relying on depth to choose one of
them makes it hard to read.
Bug: 156537348
Test: m
Change-Id: Ib525b2f5132c596db8e33492d0a565ea23e91d1c
Compared to full list, flat list drops dependency edges and simply
lists all transitive dependencies for a top-level apex bundle.
Bug: 149622332
Test: m, manually build flatlist
Change-Id: Ibd521c96b7aeab90b95965c1b524e0a0152aaf5a
Move depsInfo into android for easier sharing with APK code.
Bug: 149622332
Test: m, diff'ing outputs for conscrypt module.
Change-Id: If0ee967d37425540e69b4ce9304229d9f2cd86bd
This change adds 'test_for' property to cc_test_* types. The property is
used to mark a module as a test for one or more APEXes, in which case
the module has accecss to the private part of the listed APEXes. For
example, the module is linked with the actrual shared library in the
APEX instead of the stub of the shared library.
Exempt-From-Owner-Approval: already +2'ed by owner
Bug: 129539670
Bug: 153046163
Test: m
Change-Id: I45ed0d7a15540b0d69b2a3b8d9c4cb202adff6f2
A module is marked unavailable for platform when 1) it does not have
"//apex_available:platform" in its apex_available property, or 2)
it depends on another module that is unavailable for platform.
In that case, LOCAL_NOT_AVAILABLE_FOR_PLATFORM is set to true for the
module in the Make world. Later, that flag is used to ensure that there
is no module with the flag is installed to the device.
The reason why this isn't entirely done in Soong is because Soong
doesn't know if a module will be installed to the device or not. To
explain this, let's have an example.
cc_test { name: "mytest", static_libs: ["libfoo"]}
cc_library_static { name: "libfoo", static_libs: ["libbar"]}
cc_library { name: "libbar", apex_available: ["com.android.xxx"]}
Here, libbar is not available for platform, but is used by libfoo which
is available for platform (apex_available defaults to
"//apex_available:platform"). libfoo is again depended on by mytest
which again is available for platform. The use of libbar should be
allowed in the context of test; we don't want to make libbar available
to platform just for the dependency from test because it will allow
non-test uses of the library as well.
Soong by itself can't tell whether libfoo and libbar are used only in the
context of a test. There could be another module depending them, e.g.,
cc_library_shared { name: "mylib", static_libs: ["libfoo"] }
can exist and it might be installed to the device, in which case
we really should trigger an error.
Since Make has the knowledge of what's installed and what's not,
the check should be done there.
Bug: 153073816
Test: m
Test: remove "//apex_available:platform" from libmdnssd (it is currently
installed to /system/lib), and check that `m system_image` fails
Change-Id: Ia304cc5f41f173229e8a154e90cea4dce46dcebe
For a given variant of a module that implements ApexModule interface,
the "updatable" property tests if this variant comes from an updatable
apex. For platform variants it is always false.
Test: lunch aosp_walleye-userdebug && m nothing
Bug: 138994281
Change-Id: I2d4c54fb397e29dc9b3203be7fb17be4536529f7
When a source and a prebuilt module are present in the same build a
dependency is added from the source module to the prebuilt module.
Previously, the code for generating the APEX did not recognize that
tag and in some cases (e.g. for cc_(prebuilt_)library_shared) will
fail the build.
This change:
1) Adds a test to reproduce the problem.
2) Improves the debug message by pretty printing the tag.
3) Adds a new ExcludeFromApexContents interface that can be implemented
by a tag to declare that it should be excluded from the APEX
contents.
4) Ignores tags that implement that interface when generating APEX
contents.
5) Implements that interface on prebuiltDependencyTag to fix the
test.
Bug: 153326844
Test: m nothing
Change-Id: I9dd4312c4f995c816c0a31d8d733eb5d7f56e1ea
The DepIsInSameApex() and RequiredSdks() methods were defined in a few
places to avoid having to depend on the whole ApexModule/SdkAware
interfaces directly. However, that has a couple of issues:
1) It duplicates functionality making it difficult to change, changes
to the definitions outside the main interfaces do not cause compile
time failures, instead they result in a runtime change in behavior
which can be difficult to debug.
2) IDE navigation (specifically in Intellij) does not detect that the
duplicate definitions can resolve to the definitions in the main
interface.
This change extracts the methods into their own interfaces and reuses
those interfaces instead of duplicating the methods to fix both of
these issues.
Bug: 152878661
Test: m nothing
Change-Id: I0cfdf342a14eb0bfb82b1bd17e0633d81c7facfb
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
Makes sure that the module snapshots do not rely on the white list
of apex available settings so that when those lists are removed it is
not necessary to update any snapshots.
Bug: 142935992
Test: m nothing
Change-Id: Iedcff7dfc2646a4da77258d16e06657dd2f411f9
Native modules within APEX should be linked with proper stub version
according to its min_sdk_version.
For example, when min_sdk_version is set to "29", libfoo in the apex
would be linked to libbar of version 29 from platform, even if it has
a newer version like 30.
Bug: 145796956
Test: m nothing (soong tests)
Change-Id: I4a0b2002587bc24b7deeb5d59b6eeba5e1db5b1f
Bug: 149075752
Test: tapas com.android.conscrypt com.android.tethering arm64
Test: m out/target/product/generic_arm64/{,symbols/}apex/com.android.{tethering,conscrypt}/lib64/libc++.so
Test: Verified that unwinder was dynamically linked to tethering's
Test: libc++ and statically linked to conscrypt's.
Test: lunch flame-userdebug && m
Test: Verified that unwinder was dynamically linked to /system/lib64/libc++.so
Change-Id: I98eed7cb4316962b19b5c12e150c224c25d0e91d
apex { name: ["myapex"], native_shared_libs: ["libX", "libY"] }
cc_library { name: "libX", shared_libs: ["libY"] }
cc_library { name: "libY", shared_libs: ["libZ"], stubs: {...} }
apexDepsMutator was a bottom up mutator and it uses WalkDeps to traverse
the dependency tree rooted at myapex in a depth-first order. While
traversing the tree, if calls BuildForApex for a module that will be
part of the APEX.
libY is visited twice. Once via libX and once via myapex. If the visit
from libX was before the visit from myapex (since this is a depth-first
traversing), BuildForApex is not called for libY and its dependency
libZ, because libY provides a stub. And then when libY is again visited
via myapex, BuildForApex is correctly called for the module, but not for
its dependencies libZ because the paths from libY to libZ was already
visited.
As a result, the apex variant of libY has a dependency to the non-apex
variant of libZ.
Fixing the problem by changing the mutator a top-down one.
Bug: 148645937
Test: m
Change-Id: Ib2cb28852087c63a568b3fd036504e9261cf0782
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
This means everything that goes into apexes need to be
explicitly labeled to be available for apex.
Whitelist the current offenders. This list should be
trimmed down.
Bug: 147364041
Test: m
Test: multiproduct_kati -only-soong
Exempt-From-Owner-Approval: cherry-pick from internal
Change-Id: I837299c6a15d46f8a5ba544b613776b1cc27d7b8
Merged-In: I837299c6a15d46f8a5ba544b613776b1cc27d7b8
(cherry picked from commit 93488cbb10d4882845abb732f8e53714f0982031)
Create an APEX variation of a module only when it is either directly
included in an APEX or the dependency to the module demands that the
module should be included. For example, a non-static dependency to a
java library no longer creates an APEX variant of the library.
Bug: 146907857
Test: m (apex_test amended)
Test: cat out/soong/build.ninja | grep "Module:" | wc -l
reduced from 47132 to 45881
Change-Id: Icd79fb4b60668d852b9f018343b934b5e42a8eda
In case that an apex module depends on a module with stubs directly
*and* indirectly, the build system should follow the deps graph further.
Note that WalkDeps() visits deps in DFS and it won't visit again visited
modules.
Bug: n/a
Test: m (apex_test.go amended)
Change-Id: I1a6f135dbda6d1eb641575a3fbbc9bbee0622076
Java libraries with hostdex: true are available for the platform even if
it doesn't have "//apex_available:platform" in the apex_available
property. Note that the java libraries are still prevented from being
installed to the device.
Bug: 128708192
Test: m
Change-Id: I6463ebc59cf7fd861b812999d7a79c387bbb3335
The main purpose of apex_available is to prevent a module from being
installed to the system partition (or being statically linked to
something that is installed to the system partition). It's not the goal
to prevent host tools from using the modules.
However, since both host and device variants are not mutated for
platform when the module is not available to platform, the host tools
that depends on those modules couldn't be built. To solve the problem,
the platform variation is NOT skipped for the host variants.
Bug: 128708192
Test: m
Change-Id: I1d662cd6d165581f344138d872329a15bfc43d17
When an APEX is built with uses_sdks, any depedndency from the APEX to
the outside of the APEX should be from the SDKs that the APEX is built
against.
Bug: 138182343
Test: m
Change-Id: I1c2ffe8d28ccf648d928ea59652c2d0070bf10eb
apex_available property can be appended differently per the linkage
type. This will be used to restrict certain libs (e.g.
libc_malloc_debug) to an APEX while allowing them to be statically
linkable from platform for testing purpose.
Test: m (apex_test amended)
Change-Id: I6dec23129c5ac93a3ef06fea28f26f240c0ba410
Cindy Zhou