Change idmap to iterate over the resources in the overlay package
instead of the target package when scanning for resources defined in
both packages. This cuts down the runtime cost of creating an idmap
considerably since the algorithm now scales with the number of resources
in the overlay package (a handful) and not the number of resources in
the target package (android: 10k, SystemUI: 8k) at a minor cost to code
complexity.
Improvements on the runtime of ResTable::createIdmap (systrace on an
emulator running aosp_x86_64-eng):
- target=android: 12.5 ms -> 3.0 ms
- target=SystemUI: 8.6 ms -> 1.0 ms
The bulk of the cost of creating an idmap from installd is now the fork
and execl to call "idmap --fd ..." which weigh in at 16 ms.
Bug: 80150169
Test: make libandroidfw_tests
Test: atest OverlayHostTests OverlayDeviceTests
Change-Id: I98e18d5958c0cd835a73055b714f5bf0f4f95a09
The entry pool is encoded in UTF-8 not UTF-16. This would cause stringAt
to return a pointer to a cache that holds the UTF-8 string converted to
UTF-16. If the cache becomes invalidated, the UTF-16 pointer is no
longer valid and causes seg faults when looked up.
Test: libandroidfw_tests
Change-Id: I0f13dbeadd0d148de2805efd6cadb8e220368e3b
An overlay was incorrectly leaking its own resources into the
framework resource package, which caused warnings for every app
that tried to access framework resources (all of them).
This change skips including any resources that are not overlaying
anything (not present in IDMAP).
Bug: 36256974
Test: make libandroidfw_tests
Change-Id: I8c710af6849bb848938825aacca02799ee96c003
Tests would expect parts of the APK to be unzipped and
maintained. Instead, we now decompress the required files
from the test APKs on test setup. This simplifies
test maintenance substantially.
Test: make libandroidfw_tests && libandroidfw_tests --testdata=frameworks/base/libs/androidfw/tests/data
Change-Id: I3d2100af22df913e02401dedcf9842cdb32b2a3b
A ResTable_map entry has a name attribute, which
could be a dynamic reference if it comes from
a shared library. It was not being patched with
the correct package id.
Bug:16795890
Change-Id: Ia8df6a943269b2fefb2132c3ed74eb1997d7701b
This change allows the developer to add a base package for
which to build a feature split. The generated resource types
will begin after the base APK's defined types so as not
to collide or override resources.
Multiple features can be generated by first choosing an
arbitrary order for the features. Then for each feature,
the base APK and any preceding features are specified
with the --feature-of flags.
So with a base APK 'A' and features, 'B', and 'C',
'B' would be built with
aapt package [...] --feature-of A [...]
and 'C' would be built with
aapt package [...] --feature-of A --feature-of B [...]
Change-Id: I1be66e3f8df9a737b21c71f8a93685376c7e6780
In order to support APK split features, the resource
table needs to support loading multiple resource
tables with the same package but potentially new set
of type IDs.
This adds some complexity as the type ID space changes
from dense and ordered to potentially sparse.
A ByteBucketArray is used to store the type IDs in
a memory efficient way that allows for fast retrieval.
In addition, the IDMAP format has changed. We no longer
need random access to the type data, since we store the
types differently. However, random access to entries of
a given type is still required.
Change-Id: If6f5be680b405b368941d9c1f2b5d2ddca964160
