Propagate anyhow errors to main.rs and commands.rs to improve the error
message. As an example. instead of just "bad flag declaration: exactly
one bug required", aconfig will now print the following.
---- 8< ----
Error: failed to create cache
Caused by:
0: failed to parse build/make/tools/aconfig/tests/test.aconfig
1: bad flag declaration: missing description
---- >8 ----
Error messages can be improved further by including additional
information in the protos.rs error cases. This will be handled in a
follow-up CL.
Bug: 290300657
Test: manual: introduce error in an aconfig file and run `m all_aconfig_declarations`
Test: atest aconfig.test
Change-Id: Id278f4877e5794b95913ae8ba0ca3ee211293f38
Some consumers of `aconfig dump` do not support multiple bugs. At the
same time, we want all flags to be associated with at least one bug.
Teach aconfig to require that the bug field in the flag_declaration and
parsed_flag proto messages appear exactly once.
This change could have been implemented as a change of `repeated` to
`optional` in the proto definition. However, the chosen approach, with a
runtime check, is easier to revert if we want to support multiple bugs
in the future.
Bug: 293156797
Test: m all_aconfig_declarations && printflags
Test: atest aconfig.test aconfig.test.java aconfig.test.cpp
Change-Id: Ib87dac68b392986a8daa64e56cd85477c92fbe83
Check BOARD_AVB_$(call to-upper,$(partition))_KEY_PATH to decide whether custom_image should sign AVB or not. If key path isn't set, the custom image will be excluded from AVB and copied to /IMAGES in target-files directly. This allows vendor to use custom_images flow packing unsigned image.
And to every non-avb custom partition, one image whose name is partition name must be added in its BOARD_<CUSTOM_PARTITION>_IMAGE_LIST.
BOARD_CUSTOMIMAGES_PARTITION_LIST := tvconfig
BOARD_TVCONFIG_IMAGE_LIST := \
device/xxxx/yyyy/tvconfig.img \
device/xxxx/yyyy/tvconfig_custom1.img
Test:
1) Build image, target-files, OTA package by m and m dist
2) Sign images by sign_target_files_apk.py
Fix: 285227850
Change-Id: I7477dafe023e4b168f0f08fb7aedd9e511a60e1b
Consolidate integration tests under tests/. Rename build targets to
follow the same pattern.
The important aconfig build targets are:
- aconfig : the aconfig binary
- aconfig.test : the aconfig unit tests
- aconfig.test.java : the Java integration tests
- aconfig.test.cpp : the C++ integration tests
- aconfig.test.rust : the Rust integration tests (not implemented yet)
Note: the Rust integration tests are blocked until the build system has
added support for aconfig Rust libraries.
Note: the flags used in the integration are not set correctly. A
follow-up CL will fix this.
Bug: 283911467
Test: atest aconfig.test.java aconfig.test.cpp
Change-Id: Idb8dad4c38a49231c2e89228775bacf065ee1f57
Fix syntax error in auto-generated Rust code that prevents the code from
compiling.
For reference, the error was:
---- 8< ----
error: expected `where`, `{`, `(`, or `;` after struct name, found keyword `impl`
--> .../src/lib.rs:6:1
|
6 | impl FlagProvider {
| ^^^^ expected `where`, `{`, `(`, or `;` after struct name
---- >8 ----
Bug: 292052163
Test: manual: create sample based on auto-generated Rust lib
Change-Id: I4c029c0ee16ea3849e7bbe381717a10402c9cb81
Add a new --format=verbose dump output format. This is identical to the
default 'text' format but includes the source files from which the flag
declaration and values were read, and is intended to help debug why a
flag isn't set to some expected value.
Bug: 283910447
Test: atest aconfig.test
Test: printflags --format=verbose # manually inspect output
Change-Id: I03a1c8d940e7a0c6f91e986c5bafa4aa6cd1125a
The --format=debug dump output format is largely superseded by the
textproto format now that aconfig doesn't use hand-written wrappers
around the auto-generated proto struct anymore. Remove the format to
reduce the risk of code rot.
Bug: 283910447
Test: atest aconfig.test
Change-Id: I6700fc4eafd3fa3a63952109c0105d34c7ffd98b
This change will sorted the parsed flags from the declarations. Without
this change the code will expect the passed in declarations sorted.
After this change the code will still guarantee the parsed flags in the
cache is sorted, but it won't expect the passed in declarations are
sorted
Test: atest aconfig.test
Bug: 291926035
Change-Id: I5f0637fe770003224b128591890e04277bc09345
Introduce rust codegen 2nd iteration with unit test support. The design
is described in detail in the design doc (go/aconfig_rust_codegen). The general idea is to generate
different code with the same signature for production and test build
targets, just like java/c/cpp codegen.
We will have a FlagProvider struct that has flag methods implementation.
This flag provider instance can then be used in injection pattern. In
additon, we also generate top level functions that wraps around flag
provider call so it can be used in static function style.
Things to be decided later:
should we just generate one set of code, and use cfg! as compile
time marco to compile the right code for the right targets.
Bug: b/279483360
Test: atest aconfig.test
Change-Id: Ic75cedbd0d27b5242014c3ac7fc80692d2ab4589
For the OTA build, target_files.zip requires setting fixed timestamp
for images. However, it caused regression on 'adb sync'.
Restoring image build process for target_files.zip instead of copying
them from the PRODUCT_OUT, only the images in target_files.zip set the
fixed timestamp.
By setting COPY_IMAGES_FOR_TARGET_FILES_ZIP flag, we still may copy
the images for target_files.zip for future use.
Bug: 287534409
Test: m; flashall; adb sync
Test: m dist; flash images from target_files.zip and check the timestamp
Change-Id: I4135eb721c33581f191621847e6595833263d707
1, Moved "#include <string>" from exported header to test flag provider
header file.
2, For production target and read only flags, the generated c
api should just return default value instead of calling into c++ api
3, Remove using namespace server_configurable_flags from header, instead
of just having the namespace to be spelled out in each api call. Having
using namespace xxx in header is not a c++ best practice.
4, Replace #ifdef #def #endif with #pragma once
Bug: b/279483801
Test: atest aconfig.test
Change-Id: I3e55a7b14301f3de419795467f33e2dc889d371e
Two major changes to c/c++ codegen
(1) explicit setter for each flag instead of a generic flag setter
void override_flag(std::string name, bool val) is replaced with
void <flag name>(bool val) for each flag name
This has several advantages:
(a) generated code is more c++ idomatic
(b) no longer need to create flag name string constants
(c) any typo in the code is caught early in the build time
(2) remove flag setter and flag override reset methods/functions when
generating code targets for production. If developers want to update
their main function to take command line arg for flag overrides, they
can use compile time macros to decide if the flag override code should
be included.
Bug: b/279483801
Test: atest aconfig.test
Change-Id: I6141f7f979b32fe0426154d578edeb997ae5ff7c
c codegen can be done at the same time with cpp codegen, the idea is to
create a c compatible header that defines the flag apis, including flag
override apis for test. then in the corresponding cpp file, the
implementation simply calls into cpp api.
c header supports static method interface, and unit test override, but
it does not support injection pattern compared to cpp header
Bug: b/279483801
Test: atest aconfig.test
Change-Id: Ie62b76d6524e443de5d3c2f9000f7f66623ab571
This change allows ARC to define ARC-specific system properties in
Android build without changing these property definitions for other
Android builds. Please see go/arc-sigprop-changes and
go/arc-android-sigprop-sync for additional details.
Bug: 195609932
Test: built bertha_x86_64 with forward declarations
Change-Id: I22bd9d60c2491506fe5c633dbbb9e7516f529b35
Python's zipfile doens't restore file permission by default, so we need
to manually restore permission.
Test: th
Bug: 290643514
Change-Id: I89c1e2ee178b534fa7e3f02afd04d170100d37e7
Move server_configurable_flags header include away from exported header,
instead put it in the flag provider headers. Otherwise, it can cause
compilation error if the source code who wants to use the generated c flag lib has
not added dependency on server_configurable_flags yet.
Bug: b/279483801
Test: atest aconfig.test
Change-Id: Ib75877ee88f995caf72b3fd2554c3da195032c14
Improve the error message returned when `aconfig dump` is fed multiple
declarations of the same flag: include the paths to the declaration
files.
In general all error messages from the protos::*::verify_* functions
should include paths to the offending files. This will be handled in a
follow-up CL.
Bug: 290300657
Test: atest aconfig.test
Test: manual: add duplicate flag and run `m all_aconfig_declarations`, inspect error message
Change-Id: I46dc23f7128dd5c68ced9f2e8518cfa89d81c2df
cpp codegen iteration 2, based on discussions with three internal teams
that use c++. Refer to the design doc "aconfig c++ codegen" for detailed
design. At a high level, we generate two sets of code artifacts with the
same signatured api: one for production that without any local flag
override capability, one for unit test that allows local flag overrides.
It supports static methods style interface as well as injection pattern.
Refer to the test points in the codegen_cpp.rs for examples of generated
code.
for production target codegen: aconfig create-cpp-lib --cache <cache> --out <out dir>
for test target codegen: aconfig create-cpp-lib --cache <cache> --out
<out dir> --mode test
Bug: b/279483801
Test: atest aconfig.test
Change-Id: I92fefb9623d5435525339a74f57bbd36d0afef08
Host side test infra needs the library to consume the dumped flags info
generated by aconfig.
Test: m libaconfig_java_proto_lite libaconfig_java_proto_full
Bug: 289906970
Change-Id: I9891cb8445b90f7269e0c679ceaae6bc102bbb0c
Add FlagAnnotationTests in presubmit to ensure that test filtering rule
always works as expected.
Test: atest --test-mapping
Bug: 277819423
Change-Id: I7cc3a70495f8d62ed7a5b26edc4fd240e73ccf61
Add resetAll method to test mode FeatureFlagsImpl. This method
is used to reset all the flags values to null. It provides convenient
way to the test tools to reset the flags values.
Bug: 280833463
Test: atest aconfig.test
Change-Id: I4bf1d3ba69ee106ef8d0c1cc62c00bbeca1b72aa
Change the java template to remove the extra new lines and spaces
in generated java code.
Bug: 280833463
Test: atest aconfig.test
Change-Id: I336050540f8a2b0966bf21cf6e034a8fa701458e
Introduce a new requirement on package fields: a package must contain at
least one dot character.
Bug: 289336036
Test: atest aconfig.test
Change-Id: Idadcd2a76783a484cc5c6d6e94778c0248fa475f
Use is_valid_name_ident to check the flag name, not
is_valid_package_ident.
Bug: 283910447
Test: atest aconfig.test
Change-Id: I3718e7de565b975a1696190e8effbcb077e5fde2
Rearrange the tests in test_is_valid_package_ident to make it more
apparent what constitutes a valid package pattern, and what doesn't.
Bug: 283910447
Test: atest aconfig.test
Change-Id: I3d2b87aed008d0ed3a0aa6e483e655178dda7277
Add java codegen test mode. The test mode will generate Flags.java and
FeatureFlagsImpl.java differently.
* Flags.java will have getter and setter function to switch the
FeatureFlagsImpl. Flags.java will not initialize the instance
of FeatureFlagsImpl during initialization, thus it will force the
user to set up the flag values for the tests.
* FeatureFlagsImpl removes the dependency on DeviceConfig, and
allows the caller to set the values of flags.
Command changes
This change adds a new parameter `mode` to `create-java-lib` subcommand.
The default value of `mode` is production, which will generate files for
production usage, and keeps the same behavior as before.
The new `mode` test is added to trigger the test mode. The command is
aconfig create-java-lib --cache=<path_to_cache> --out=<out_path>
--mode=test
Test: atest aconfig.test
Bug: 288632682
Change-Id: I7566464eb762f3107142fe787f56b17f5be631b7
Host-side simulation of APEX activation is done by a new tool
'apexd_host'. This simplies checkvintf invocation for local builds and
for target-files.
For local builds, checkvintf no londer depends on $OUT/apex, a flattened
view of APEXes. In fact, the build system doesn't need to install
$OUT/apex. They are installed for now only for its side-effect of
installing symbol files. We'd better not rely on $OUT/apex.
For target-files, scanning/activating apexes are extracted and moved to
the new tool. Now check_target_files_vintf is not more efficient because
it doesn't copy .apex files to a temporary directory.
Bug: 260358957
Bug: 288826922
Test: m (running checkvintf) # for local builds
Test: m target-files-package && check_garget_files_vintf target-files.zip
Test: atest releasetools_test
Change-Id: Iba23f429d96f9ec31814196aa14bdbb800649218
The flag enables list_files to output a list of .class files pattern
that will be generated when compiling the java files in the input
directory. The flag can be used to extract the appropriate class files
from stub jar file provided via dep_api_srcs in per api scope
sdk_library generated java_api_library modules.
Test: m --build-from-text-stub in same topic
Bug: 275570206
Change-Id: I134fcd9781024dcb6781297d02645f5fb9befb18
Flattened apexes are deprecated. mege_dexopt does not need to care about
flattened apexes.
Bug: 278826656
Test: m otapackage
Change-Id: I527eed64c5f1a7466e2c48d73360f47f6cfa0911
Teach `aconfig dump` to print its as text protobuf.
Bug: 288713733
Test: atest aconfig.test aconfig.test.java
Change-Id: Id31ce53226162a563fbe2d4841a8f6cf96c22436
Add a `bug` field on the flag_declaration and parsed_flag proto
messages. This field is optional in the sense that it can occur zero or
more times, and aconfig will simply pass any value through.
Bug fields are included in the aconfig dump format, which can be
processed by other tools.
Also unify how protos.rs checks that fields marked 'optional' in the
proto file, but in practice are 'required', are actually set.
Test: atest aconfig.test aconfig.test.java
Bug: 288261336
Change-Id: I93de0005674822c6ff4d699bdc2c6509763a7f7f
Bug: 285303012
Test: for x in next trunk trunk_food trunk_staging ; do lunch aosp_panther-$x-eng ; m nothing ; done
Change-Id: I174ce3e609fa9077ea6bc61679ddb83e8efe57c8
python3.11's zipfile implementation does not handle symlinks. This
causes important symlinks in ramdisk to be broken, and later causing a
boo failure.
Test: unzip a target files with symlinks, make sure symlinks are created
Bug: 287896098
Change-Id: Ia7d6ac8ffb03807680a36ff648aa11afafb7f481
Change all required proto fields to optional. While the proto file is
supposed to be a backwards compatible API, and fields are not supposed
to be deprecated, this commit will allow for that option if needed.
Implementation wise this change doesn't matter much: any parsed data
needs additional verification outside what the protobuf crate's parser
provides anyway, so adding checks to verify that all required fields,
even though marked optional in the proto file, were found is a minor
increase in code complexity.
If in the future a proto field should no longer be used:
- keep the field in the proto, still marked optional and clearly
document that it is no longer in use
- change protos.rs from checking struct.has_field() to explicitly
dropping any value via struct.clear_field()
Bug: 286337317
Test: atest aconfig.test
Change-Id: Iad1ccfe50ecac286ff7a796aec909bec70b9520d
Generate Java constants for use in @FlaggedApi(flag = ...).
Also update the generated Java code to use the constants when reaching
out to DeviceConfig instead of hard-coding (duplicate) strings.
Bug: 285288440
Test: atest aconfig.test aconfig.test.java
Change-Id: I1127cacba650cc7a7896b1533e03631d7f5ec71b
Remove the hand-crafted wrappers around the structures auto-generated
from protos/aconfig.proto, and use the auto-generated structs directly
intead. This gets rid of a lot of manual repetition, and its inherent
risk.
Also unify how individual fields read from text proto are verified (e.g.
is the flag.name field a valid identifier).
Also change the intermediate cache format from JSON to binary protobuf.
The concept of a 'cache' as an intermediate internal format to represent
parsed input stays. The command line interface still refers to caches.
At the moment a cache file is identical to a parsed_file protbuf, and
the code exploits this internally.
A couple of points regarding the auto-generated structs:
- Vectors are named in the singular (e.g. parsed_flags.parsed_flag is
a Vec<ProtoParsedFlag>) because this improves ergonomics for all
devs working with aconfig input files
- The auto-generated structs have fields that are of type Option<T>
and convenience methods (named the same as the fields) to access T
Test: atest aconfig.test aconfig.test.java
Bug: 283910447
Change-Id: I512820cc4bc6c543dea9f6a4356f863120a10be3
Before java code will directly use the flag name as the method name.
This change adds funciton to try the best to convert flag name to
camelCase, and then use the camelCase string as the method name in the
generated code.
Bug: 279483816
Test: atest aconfig.test aconfig.test.java
Change-Id: I45fc6df46c9d535cd38a657a41313202f9b660af
Group statements with identical #[cfg(feature = "cargo")] attributes in
the same block. This reduces repetition and makes the code easier to
read and less error prone.
Bug: 284779868
Test: atest aconfig.test
Change-Id: Iebdcd20e7cd22cb641424a1af594f5c9ac57b623
