Previously, the hidden API encoding was done before resource merging.
However, hidden API modularization requires that the encoding be done
by the bootclasspath_fragment/platform_bootclasspath modules which will
be after the resource merging. Therefore, this change moves the hidden
API encoding after to match the future behavior.
It also moves the initHiddenAPI() method call after resource merging
too and passes it the result of the resource merging so it is available
for the bootclasspath modules via bootDexJar().
Although the resource merging was not always done when it was done it
would reorder the entries in the generated jar to match java ordering,
which puts the MANIFEST.MF first. This change preserves that behavior
by adding -j to the call to MergeZipCmds. This does mean that jars
which did not require resource merging now have a different order but
as both orders work that is not a significant change.
Bug: 179354495
Test: m com.android.art com.android.ipsec com.android.os.statsd com.android.conscrypt
- verify that apart from the ordering change in the jars that this does
not change the contents of the apex files
Change-Id: If74baad5659301ca6ca9c0f6484374420dda8c34
This change adds a test to verify the existing behavior to provide a
baseline against which following changes can be compared.
Bug: 179354495
Test: m nothing
Change-Id: Ib7dadf3e65151f4e925251f7b1a9c099e824ea63
* changes:
Don't fail if the target module is disabled in dex2oat tool dependencies.
Use oatdump rather than oatdumpd for boot jar boot.*.oatdump.txt files.
The AlwaysUsePrebuiltSdks() causes all java_sdk_library_import modules
to be preferred over the source, i.e. as if they had prefer: true set.
That interacts badly with the work that is being done to integrate the
bootclasspath_fragment/platform_bootclasspath modules into the build.
It would work fine once that integration has been completed but in the
interim it causes problems. e.g. it does not cause a problem in AOSP
because those java_sdk_library_import modules that are affected have
already been integrated into the build properly.
Unfortunately, internally that is not the case because there are
java_sdk_library/java_sdk_library_import modules that still need to
be updated.
Before the java_sdk_library_import can be safely preferred each
java_sdk_library/java_sdk_library_import module that contributes to the
bootclasspath must:
* Be in the contents of matching bootclasspath_fragment and
prebuilt_bootclasspath_fragment modules.
* Have an apex and one of a prebuilt_apex/apex_set that contains the
dex implementation jar and lists the prebuilt_bootclasspath_fragment
name in its exported_bootclasspath_fragments property.
Safely preferred in this context means that the whole build will
continue to work rather than the current situation which is that only
some of the build will work and some will fail if an attempt is
actually made to build it.
Unfortunately, many java_sdk_library_import modules are missing:
* The prebuilt_bootclasspath_fragment.
* The exported_bootclasspath_fragments property on the
prebuilt_apex/apex_set that contains them.
Together these cause the following symptoms:
1. The java_sdk_library_import does not have a dex implementation jar.
2. The java_sdk_library_import does not have a myapex variant.
These workarounds will avoid Soong reporting build failures. However,
the build will still fail if an attempt is made to build anything
produced by the platform-bootclasspath, e.g. hidden API processing or
a system image.
Bug: 188505921
Bug: 179354495
Test: m TARGET_BUILD_APPS=Calendar
Change-Id: I3226e21cd6a7f9e4d6bbe94e54129ac5e1d4c679
Merged-In: I3226e21cd6a7f9e4d6bbe94e54129ac5e1d4c679
At least until libc and libm builds, which are the common values in
system_shared_libs. We don't have a way to break the cycle of:
libstdc++.so (system_shared_libs) -> libc.so -> ...(whole_static_libs) -> libstdc++.a (system_shared_libs) -> []
because bp2build just sees it as:
libstdc++ (system_shared_libs) -> libc -> (whole_static_libs) -> libstdc++ (system_shared_libs) -> libc -> ...
therefore creating a cycle.
This CL punts the problem for now by ignoring system_shared_libs in the cc deps mutator, with a TODO.
Test: TH
Bug: b/186024507
Bug: b/186489250
Change-Id: Id4744593af39b7eb3b31418c8afaeeac99a0ab5f
dependencies.
dexpreopt.RegisterToolDeps runs late after prebuilt dependencies have
been resolved, and there's special code in dex2oatPathFromDep to
resolve the prebuilt from the source module. However, if the source
module is disabled then the dependencies check in validateAndroidModule
will complain, so we need to disable that check in this particular
situation.
Also add a comment to explain why dexpreopt.RegisterToolDeps needs to
run so late.
Test: m nothing
Bug: 145934348
Bug: 172480615
Change-Id: Ibc673303d0336768fa23261a2068e91a08f46a30
oatdumpd isn't available as a prebuilt.
Test: m SOONG_CONFIG_art_module_source_build=false droid
Bug: 172480615
Change-Id: I70317eb8f253272d629f23063cbe265d556caad3
The intention before was to use relative paths to a partition where a
config is defined. However, jars in /system_ext partition are planned to
be declared in /system's classpaths.proto config.
Bug: 180105615
Test: derive_classpath_test, CtsClasspathsTestCases
Change-Id: Icc3e1a903c34187cfcd67a3ae7bc3dd746445c03
This allows support of generated hdrs / srcs in mixed builds.
Test: Manually verified that libc_bionic_ndk passes compilation (failed
previously due to missing generated heaer)
Test: bp2build and mixed_libc CI scripts
Test: New aquery test
Change-Id: I88e359a4bd9eba383c207d5cf812272725ff0a3d
See b/188503688 for more info on why libjemalloc5 is in the
bp2buildCcLibraryStaticOnlyList.
Bug: 188503688
Fixes: 187012490
Fixes: 186828626
Test: TH
Change-Id: I2b70cf99b2efa81f3f35ab0b61716eb0f6f9ebd6
Consider this case:
apex {
name: "com.android.foo",
native_libs: ["foo"],
}
override_apex {
name: "com.mycompany.android.foo",
base: "com.android.foo",
}
cc_library {
name: "foo",
}
There are two APEXes defined: "com.android.foo" and
"com.mycompany.android.foo" which is a copy of "com.android.foo" with
some properties overridden (e.g. signing keys).
The module "foo" is mutated into two variants by the apex mutator: the
platform variant and the apex variant. The former has the variation name
"" and the later has "apex<min_api_ver>" which usually is "apex10000".
Internally, the apex variant has an alias "com.android.foo".
ApexInfo.InApexVariants() returns only "com.android.foo" when called for
the module "foo".
We can see that the information that "foo" is also part of
"com.mycompany.android.foo" is completely lost. This is causing problem
when we compare the apex membership by their "soong module name", not
the "apex name". In the example above, the two modules have different
soone module names, but have the same apex name: "com.android.foo".
To fix that, this CL introduces a new field `InApexes` to the `ApexInfo`
struct. It has the actual name of the APEXes that the module is part of.
With the example above, `InApexes` is ["com.android.foo",
"com.mycompany.android.foo"].
Bug: 180325915
Test: m nothing
Test: m nothing on non-AOSP targets with ag/13740887 applied.
Change-Id: I4e7a7ac5495d2e622ba92a4358ed967e066c6c2e
.. in preparation for the upcoming change. This change doesn't alter any
behavior.
InApexes is a misleading name. People expects that it has the list of
soong module names of the APEXes that a module is part of. So, for
example, `core-oj` is a part of both `com.android.art` and
`com.google.android.art`. However, in reality, that's not true. The
field has `com.android.art` only. This is because the two APEXes
(android and Google) have the same apex name which is `com.android.art`.
That apex name is used in various places like the `apex_available` and
allows us to keep using the same name regardless of whether the APEX is
overridden or not.
However, this is causing problems in some cases where the exact list of
soong module names is required. The upcoming change will add a new field
to handle the case and the new field actually will get the name
'InApexes'. So, the existing field is renamed to a less misleading name
`InApexVariants`.
Bug: 180325915
Test: m nothing
Change-Id: I0c73361b452eddb812acd5ebef5dcedaab382436
* changes:
Add systemserverclasspath_fragments property to apex.
Add "contents" property to systemserverclasspath_fragment.
Move classpaths.proto related info into a separate provider.
- All contents of the fragment are added as java_lib dependencies.
- Generated classpaths.proto is added into etc as required.
Bug: 180105615
Test: m nothing
Change-Id: I8e8e8b019c4ca2909182f205a47deffa946de6da
Similar to bcp_fragment's contents, this property lists all java library
contributions made by this fragment.
Bug: 180105615
Test: m nothing
Change-Id: Ifb1f54d5db290fffaa31933d15207014bb72d2fb
The new info struct can be easily shared with systemserverclasspath
fragments.
Bug: 180105615
Test: m nothing
Change-Id: I9986e64fdf19f4168da63c156de3dc9bcafac8d8
This would allow to start introducing these modules for apexes that
contribute to SYSTEMSERVERCLASSPATH.
In follow up, it will be evolved:
- platform_systemserverclasspath would have "fragments" property to
list all individual systemserverclasspath_fragments;
- systemserverclasspath_fragment would have "contents" property to list
contibuting java libs;
- systemserverclasspath_fragment would generate non-empty
classpaths.proto config within individual apexes.
Bug: 180105615
Test: m nothing
Change-Id: Ibaaa3fae5f1eab9a41ceecc1214a53be6bbc8ba6
There are two pieces to make this work:
1. Local include paths must include $(BINDIR)-relative paths, to support
generated headers in those directories.
2. The srcs that bp2build outputs for BUILD targets must include labels
for targets given in generated_hdrs and generated_srcs.
Support for exported_generated_hdrs intentionally deferred.
This allows us to remove several targets from the bp2build denylist.
Some are moved to the mixed build denylist, because genreated headers are
still unsupported in mixed builds.
Test: bp2build.sh CI script
Change-Id: Ib4f9dac20f6445487b8dad53b91eac01f437a590
In order for the bootclasspath_fragment to perform dex encoding on its
contents it needs to know whether the dex file is uncompressed or not.
This change makes that information available by passing it to
initHiddenAPI, storing it in hiddenAPI struct and providing access
through the hiddenAPIModule.
Bug: 179354495
Test: m droid
Change-Id: I913416b4836766de194203fd8ed5124b61dfa3dd
Now that the individual modules no longer participate in the generation
of the monolithic files it is no longer necessary to select a single
primary module to provide the information they need.
Bug: 179354495
Test: m droid
Change-Id: If09796de710927e3e3f2ccecad0b57ca5fce5dc9
These two methods did very similar jobs and merging them together
simplifies the behavior.
Bug: 179354495
Test: m droid
Change-Id: Ibe1a23d54105e6a0e5693079cd8743679301fc85
The configurationName was intended to separate the name of the module
from the name used in configuration (such as BootJars) so that the
child implementation library of a java_sdk_library on the bootclasspath
would have hidden API encoding performed on it just as for the main
java_library embedded within the java_sdk_library.
While that did use to work it no longer does as the test added in the
preceding change proves. It is not surprising that this regression does
not appear to have caused any issues as the the child implementation
library is only a build time artifact and not used at runtime.
In future the only modules that will require hidden API encoding are
those that are part of a bootclasspath module so there is no point in
maintaining this capability.
Bug: 179354495
Test: m droid
Change-Id: Ief8136fa9e98600cdd8d36108ec22edc2ebd7c69
Fill a hole in the testing of hidden API encoding. Some comments in the
code indicate that the child implementation java_library of a
java_sdk_library should have hidden API flags encoded in its dex jar
just like the embedded java_library. However, this test proves that it
is not working. This will be fixed in a follow up change.
Bug: 179354495
Test: m nothing
Change-Id: Ia581a17f1e48dff252d17f16bf76adf039f46b60
This is so that we know whether the attached bug is caused by the disk
being full.
Bug: 187870338
Test: Presubmits.
Change-Id: Ib09f51d971de556e5737ad2754836f2a12f60281
This is so that its implementation is more amenable to implementing
`target: { android_arm: { ... } }` and the like.
In addition, add a Boolean return value to getChildPropertyStruct() and
use it at every call site so that we only merge property structs when it
makes sense.
Bug: 187530594
Test: Presubmits.
Change-Id: I1b441c14b3d90a12b84dc89f82dd053978e89c7e
Hidden API processing of a bootclasspath_fragment requires the fragment
provides information, such as dependencies on other fragments and flag
files to override the default flags. Failing to do so will cause hidden
API generation to either fail or to generate different flags to that
generated by the hidden API processing done by platform_bootclasspath
which will cause the build to fail.
Previously, this was handled by only performing hidden API processing
for those modules that provide stub libs and relied on there only being
bootclasspath_fragments defined for ART (which already supports hidden
API processing), and Conscrypt and I18n neither of which provide stubs.
Unfortunately, that can no longer be relied upon due to a couple of
recent changes:
1. A java_sdk_library in a bootclasspath_fragment's content property is
automatically treated a stub library. That avoids duplication for
most bootclasspath_fragments that provide the implementation and
stub libraries through the same java_sdk_library. It does not affect
either ART, conscrypt or i18n as they all define the implementation
separately to the stubs.
2. bootclasspath_fragment modules have been defined for a number of
android modules as they are needed for reasons other than hidden API
processing.
In combination this meant that rules to perform hidden API processing
were being created but they were not currently being used which is
good because they fail.
However, adding the fragment to the platform-bootclasspath will cause
those rules to be used as the platform_bootclasspath module performs a
consistency check on the hidden API flags generated by each of the
fragments to ensure that they are consistent with those it generates
itself.
The dynamic bootclasspath work will need to add fragments to the
platform_bootclasspath and without this change that would be blocked
until all fragments had been switched to generating hidden API flags
properly.
This change adds a new fragments property to the bootclasspath and
disables hidden API processing for everything other than ART and tests
if the fragments property or stub libs is empty.
Bug: 179354495
Test: - Before making this change.
- Add com.android.os.statds-bootclasspath-fragment to platform-bootclasspath.fragments
m out/soong/hiddenapi/hiddenapi-flags.csv
- hidden API processing fails in statsd as it cannot find java.lang.Object.
- After making this change, run the above command again and it should pass.
Change-Id: Ifbb362f8fcfb2c06595fbd5ae39421b536e329ef
rust modules that return false on EverInstallable don't get installed.
Currently, proc_macro modules are such ones.
Bug: 187810017
Test: m microdroid on aosp_arm64 and inspect that *.dylib.so files
from host x86_64 are not there.
Change-Id: Ia57289a04dbff6e4f2ce2733c60de5338305904c
The destination path, i.e. the path the file will occupy in the APEX is
more stable and less likely to change than the source path, i.e. the
path of the file in the out or source directories. This makes it easier
to diff changes in the copy commands.
Bug: 179354495
Test: m com.android.art com.android.ipsec com.android.os.statsd com.android.conscrypt
- verify that this does not change the contents of the apex files
- verify that this does change the order of copy_commands for the above apexes
Change-Id: Ia9ebe8418c59903cd738ee0fb9262b3f32a4ddbf
