Using a lot of boolean variables can become very verbose without adding
really any new information:
variables: ["a", "b", "c"],
}
soong_config_bool_variable {
name: "a",
}
soong_config_bool_variable {
name: "b",
}
soong_config_bool_variable {
name: "c",
}
Now turns into:
bool_variables: ["a", "b", "c"],
}
Test: built-in tests
Change-Id: If5455a38433431c7ecbce1e5b32cfbb47f42602a
These were needed to create/track dependencies, but since we use src
jars now, they are no longer needed.
Test: bpfix_test
Change-Id: Ife7942f937cf94c66547af92821a4c11bd747e3f
Using a gomod-aware editor with build/soong requires that files
in build/soong can be mapped to the android/soong package path.
Move the partner androidmk and bpfix files such that their path
matches the package path when the android/soong package prefix is
replaced with the build/soong path prefix.
Test: go test ./...
Test: m bpfix androidmk partner_bpfix partner_androidmk
Change-Id: Ic7f7aad9e5eb9178eef0383f0b37e4fb93ce8314
libhidltransport/libhwbinder are empty and disallowed in Android.bp
(motivation is ~4kb per empty library per process overhead).
Bug: 135686713
Test: bpfix
Change-Id: I964215ad35068465217af74c5ef1322b43476428
Allows our partners to maintain partner-specific fixes in their repos. Converts most of androidmk into a library similar to bpfix. Makes some methods and types public for use by extended versions.
Bug:140727544
Test: Ran unit test cases &&
ran test conversions with sample
Change-Id: I7e1fbf3a6a7a8bd47334f43fe3eb68cbbd3426c1
There are a couple of makefiles that have etc with $(TARGET_OUT)/etc/firmware,
$(TARGET_OUT_VENDOR)/firmare as the local module path. Added the conversion in
androidmk for target.
Bug: b/133711895
Test: Unit test cases.
Change-Id: I74aca18a7eb8aa4dd692cde680fa4e112edcf6d1
There are a couple of makefiles that have etc with $(TARGET_OUT)/fonts
and $(TARGET_OUT_PRODUCT)/fonts as the local module path. Added the
conversion in androidmk for target.
bug: b/123403785
Test: Wrote and ran unit test cases, did a test conversion on
* platform/external/google-fonts/dancing-script/Android.mk
* internal/external/google-fonts/arbutus-fonts/Android.mk
Change-Id: Id67e10f80b32b27c5027cd274a7944801dad0880
There are a couple of makefiles that have etc with $(TARGET_OUT)/usr/share
as the local module path. Added the conversion in androidmk for target
and host.
Bug: b/132123818
Test: Wrote and ran unit test cases, did a test conversion on
external/neven/Android.mk makefile.
Change-Id: Iafed89f4cee499f561a1235f8870b1f1329e99bc
The value of a property to be merged may be a reference to a variable.
When we first create an attribute for a makefile variable (e.g. we
create 'android_static_libs' from LOCAL_STATIC_ANDROID_LIBRARIES), we
set its type correctly. However, reparse in bpfix erases this
information, so by the time 'android_static_libs' attribute is to be
merged with 'static_libs', it has no type and merge occurs. It isn't
easy to fix properly, so just don't merge in such case, Soong will
complain and it will be fixed manually.
Fixes: 125519127
Test: unit tests in androidmk_test.go
Change-Id: Ic66d7ab686a2fdde987f32e2b204c72d9bf0d026
Many Android.mk files for the CTS tests have
LOCAL_MODULE_PATH := $(TARGET_OUT_DATA_APPS)
statement. This can be dropped during the conversion to blueprint
files.
Also, ignore the assignments to obsolete LOCAL_CTS_TEST_PACKAGE
variable.
Fixes: 125405331
Test: Internal tests, selectively run androidmk on Android.mk's in cts/
directory
Change-Id: I2ed88acd3c8837f96b84be6eb7c7b9b0b5405f57
The conversion is a two-step process: first, when processing
BUILT_PREBUILT, convert LOCAL_SOURCE_PATH to a variable reference+fixed
subpath path in the blueprint AST. Then, set various boolean flags
depending on variable being referenced. androidmk_test.go has a test for
each handled case.
Change-Id: Iabd18d5f8645ca7077536863cd6640df5b24d24a
Fixes: 122906526
Test: treehugger
Now java_import with a host flag set to true is converted to
java_import_host.
Bug: 122372359
Test: bpfix_test.go + manual test with real examples.
Change-Id: Ie38fa51a44c83c434927db4b742d6839c64b3d63
In androidmk, translate BUILD_JAVA_LIBRARY to java_library plus
installable: true, and BUILD_STATIC_JAVA_LIBRARY to java_library.
In bpfix, rewrite java_library_static to java_library.
Bug: 110885583
Test: androidmk_test.go, bpfix_test.go
Change-Id: I63c2f759ae9c62a43f3439526552d2cd8e8cedc3
android_test module are APKs that can be run as tests, either
as standalone unit tests or as instrumentation tests for another
APK.
Test: m checkbuild
Change-Id: I16661701637e4048fd99442029c3e195ebf373a4
And fix up androidmk / bpfix to provide warnings about what to do
instead.
Test: m blueprint_tools (runs the tests, ensures there aren't any tags left)
Change-Id: I1a3ad8600211050420041740207d6957f44463c8
This way we can remove the line the property was on, not just the
property itself.
Test: `m blueprint_tools` to run the unit tests
Test: diff bpfix results on all of AOSP before/after this change
Change-Id: I61fdd945e6ee711c620b79683dfee7b7c751b3c4
This will move the common properties:
name
defaults
device_supported
host_supported
To be listed first (and in that order) for every module. Other
properties are untouched.
This also adds a helper to test individual passes in an end-to-end
manner, and a helper to run passes that use PatchLists.
Test: `m blueprint_tools` to add the new tests
Test: Diff bpfix results over all of aosp before/after this change
Change-Id: I746a00a3731cb7597d2613ef2dc45a99654cd122
According to the documentation: "NewBuffer creates and initializes a new
Buffer using buf as its initial contents. The new Buffer takes ownership
of buf, and the caller should not use buf after this call."
Test: Run bpfix twice, only wrote the first time.
Change-Id: I52f88bfd9247240436b46f396c9196157774615b
This still leaves a blank line in it's place, but that's a more general
problem.
Test: m blueprint_tools
Test: Run bpfix over AOSP, diff.
Change-Id: I55c1d1c183cfa49beeca07abd539348bc2524c5a
The java library rewrites should only happen for java modules, not cc
modules.
Test: Ran androidmk on a number of Android.mk files
Change-Id: Ife2cfb5a69d7db37216671f08317033b99fcd3a1
Add support for translating LOCAL_MANIFEST_FILE, LOCAL_RESOURCE_DIR
LOCAL_SHARED_ANDROID_LIBRARIES, and LOCAL_STATIC_ANDROID_LIBRARIES.
Use the presence of non-empty LOCAL_RESOURCE_DIR,
LOCAL_SHARED_ANDROID_LIBRARIES or LOCAL_STATIC_ANDROID_LIBRARIES
to convert a java_library_static into an android_library module,
and then squash LOCAL_SHARED_ANDROID_LIBRARIES into
LOCAL_SHARED_LIBRARIES and LOCAL_STATIC_ANDROID_LIBRARIES into
LOCAL_STATIC_LIBRARIES.
Test: androidmk_test.go
Change-Id: I3ad2a3561f69ebd097eca97cb170754d64e06123
androidmk will start to generate multiple static_libs: properties
in a single module, add a pass in bpfix to fix them up into a
single property.
Test: bpfix_test.go
Change-Id: I30955b6efbb767c02ba77f2f18d44951ef094bad
Make a new object called Fixer to hold the state of the tree, and
make a copy of the input tree so the original doesn't get modified.
Test: bpfix_test.go
Change-Id: I1dc6fd99158c8b0e1db029df99e6cf72699a5e63
java_import and android_library_import modules can't be handled
directly in androidmk because the results may depend on properties
that haven't been parsed yet. Add a bpfix pass (which is
automatically included at the end of androidmk) to select
android_library_import vs. java_import based on the extension
of the prebuilt file, and convert the srcs property to jars or aars
as appropriate.
Bug: 73724997
Test: androidmk_test.go
Change-Id: I1024742e9e96d5e1e88c3cc139eeb0d5a2f6849b
The functions in bpfix that take a *parser.List and return a
modified *parser.List are always returning the same pointer
and mutating the target of the pointer. Remove the extra
unnecessary return value.
Also extract the getLiteralListProperty function.
Test: androidmk_test.go
Change-Id: I08d8aff955c72b7916741cda8083974a49af4d6f
Since in some cases they're not the default value.
Test: echo "cc_defaults{ system_shared_libs:[] }" | bpfix | grep system_shared_libs > /dev/null && echo ok
Bug: 66979076
Change-Id: I760b34f980281b955972819676bd62154a6c73f5