Because of a bug in metalava [1], the flag constants are incorrectly
considered not-hidden when metalava generates the API signature files.
This means the FLAG_ constant name will be used, instead of the FLAG_
constant value, which is what we want.
While waiting for b/300211291 to be fixed, explicitly mark the FLAG_
constants as hidden (and keep marking the classes as such).
This CL can be reverted when b/300211291 is fixed.
1. Item.isHiddenOrRemoved does not check if its parent item is hidden,
tracked in b/300211291
Bug: 297881670
Test: atest aconfig.test aconfig.test.java
Change-Id: I05de8344ccb424bee51c39f9609489ad312397ce
Support LOCAL_JAVA_LANGUAGE_VERSION := 11 or 17 and match the default value of 17 used
in Soong.
Bug: 298613711
Bug: 299937334
Test: m javac-check
Change-Id: I19aa24cc5197ab0e6cb03361124c36ce3e316810
Removing this flag as --apply-vbmeta should only be used for root vbmeta
partitions.
Test: m fastboot_info, fastboot flashall
Bug: 299440104
Change-Id: I4910008635b16c4f4194ebd45da26bd1a8875fc1
This reverts commit c0e8ffea41.
Reason for revert: The detected build breakage was unrelated to the aconfig change. Details in b/297881670#comment11.
Change-Id: Ia32bc2e796ed3de7fa049c2c2db340f4325a2853
This reverts commit 379e09d157.
Reason for revert: Seems to have broken the build. Reverting while investigating.
Change-Id: I37c9a065853ef07b1f1c18355205a2f49dd816c4
Improve performance slightly: remove unnecessary clone operations, or
use references where a new object is not needed.
Bug: 283910447
Test: atest aconfig.test
Change-Id: I75205ffa1723dd2654039baac882c225d2653c86
Flagged APIs are annotated using `@FlaggedApi(Flags.FLAG_NAME)`, where
Flags.FLAG_NAME was auto-generated by aconfig.
When metalava generates an API signature file, it can either
a) insert the value of the constant ("com.foo.bar.flag_name"), or
b) insert the name of the constant (FLAG.NAME)
In the case of @FlaggedApi, we want a). This requires that
x) metlava has access to the definition of the constant while generating the API signature file, and
y) the constant is not part of the API surface of the API signature file
x) is handled by the build system, y) is handled by the aconfig code
generation.
This CL @hide:s all generated Java code, to make sure it is accessible
within the platform, but never part of any API surface.
Bug: 297881670
Test: atest aconfig.test aconfig.test.java
Change-Id: I328ed1a652a4e5e293f2f4b11f916d29fc2fbcbd
This is to check if unknown file context labels are used in APEX. The
reason why this is done in Makefile is to cover vendor-specific SE
policy and prebuilt APEXes. Note that there's already a validation check
when building APEXes (in Soong), which only checks the generic rules,
not vendor-specific.
Bug: 299391194
Test: atest apex_sepolicy_tests_test
Test: m
Change-Id: I4378b7530e8433d2bf188db95cb5bf791e3bf778
build_image.py now accepts a --input-directory-filter-file argument
that contains a list of files to consider from the input directory.
A new, temporary input directory will be created, and only the
considered files will be copied to it.
This removes most of the need for `m installclean`, as old files or
files built manually will no longer end up on the partition.
Bug: 205632228
Test: Presubmits
Change-Id: I07e0bf8e840abf0b51a2fdf67b758905fb3c5b5b
Our team performs the mainline sdk releases and has to update the sdk
extensions version on every release.
Adding people who are involved in the sdk ext release process so that
changes related to sdk extensions version updates can be approved by
our team.
We will only be approving changes which updates sdk extensions version.
Bug: N/A
Test: N/A
Change-Id: I42755dbae7054161f56e789b57e064084eebc7c7
To avoid merge conflicts. These are buildifier errors that will be
fixed in a followup cl.
Test: presubmits
Change-Id: I22c91c796aee16ef16f05471c2ff5221fa41d7b3
Previously, when building the proguard dict zip, we got a lot of
warnings like these, and the files were always rebuilt:
```
[ 39% 9/23] Copy symbols with mapping: out/target/common/obj/APPS/HTMLViewer_intermediates/proguard_dictionary
ninja: Missing `restat`? An output file is older than the most recent input:
output: out/target/common/obj/APPS/HTMLViewer_intermediates/proguard_dictionary.textproto
input: out/soong/.intermediates/packages/apps/HTMLViewer/HTMLViewer/android_common/e10894d3da5db41239526c9048962355/proguard_dictionary
```
For some reason kati wasn't adding `restat=true` on these ninja rules.
Adding $(4) to KATI_RESTAT fixes it for some reason.
Test: m dist repeatedly doesn't give the restat warnings
Change-Id: Ie02526b3ab127f42bdc2db350d2dc468e7888fe2
Before this change, entry times before 1980 had the year changed to 1980
but the month/day/hour/minutes/seconds were all kept as-is.
Bug: 277978832
Test: atest zipalign_tests
Change-Id: I9f87e98a6b985002578490b87c654fee86c65d62
update_engine requiers per partition timestamp if this partition's build
prop includes build.date.utc.
Test: th
Bug: 293313353
Change-Id: I6bf3de8e855f0830b26bdbe2cbdc78ba891662a7
build_image.py now accepts a --input-directory-filter-file argument
that contains a list of files to consider from the input directory.
A new, temporary input directory will be created, and only the
considered files will be copied to it.
This removes most of the need for `m installclean`, as old files or
files built manually will no longer end up on the partition.
Bug: 205632228
Test: Presubmits
Change-Id: I4c82ac7f1ed49a37271755749c8f72afeeb7f295
This change adds annotation in the generated code. If the flag is read
only, annotations @com.android.aconfig.annotations.AssumeTrueForR8 or
@com.android.aconfig.annotations.AssumeFalseForR8 will be added in to
the corresponding methods in the interface FeatureFlags, and class
Flags.
Test: atest aconfig aconfig.test.java
Bug: 295328116
Change-Id: Ic1b62a86f0945a5e72ae2c4987cce77fe3f172e9
