Benchmarks on bullhead-userdebug show that there is a negligent
performance impact when using sparse entries on a 30% loaded
sparse type of 1000 resources.
Benchmark Time CPU Iterations
-----------------------------------------------------------------------------------
BM_SparseEntryGetResourceSparseLarge 255 ns 254 ns 2751408
BM_SparseEntryGetResourceNotSparseLarge 254 ns 254 ns 2756534
Bug: 27381711
Test: make libandroidfw_tests aapt2_tests
Change-Id: I051ea22f2f6b2bc3696e446adc9e2a34be18009f
- Use FileMaps to open Assets (prevents closing of ApkAssets underlying
zip)
- Implement OpenDir and List methods
- Fix issue where DynamicRefTable wasn't properly constructed
Test: make libandroidfw_tests
Change-Id: Ib21a84e1114d028120744aa3bc1c6eb9d9399fa8
The color mode lets an application request a wide color gamut for
a specific window. This will also be used in the future to request
HDR. The color mode is currently either default (sRGB) or an undefined
wide gamut color space chosen by the platform. These attributes could
later be used to choose a specific color space if we deem this important
or useful.
This change also renames the various "colorimetry" attributes and
constants to "color mode" for consistency. These symbols were
added in O and can be safely renamed.
Test: CtsColorModeTestCases
Bug: 32984164
Change-Id: I4d4691dd12dbe3f3aa6a5cf893cff39aa16c739e
libandroidfw needs to make use of StringPiece, so
move it to libandroidfw and update all code referencing
StringPiece in aapt2.
Test: make libandroidfw_tests libaapt2_tests
Change-Id: I68d7f0fc7c651b048d9d1f5e7971f10ef5349fa1
This adds support for shared resource libraries in the new
ResTable/AssetManager implementation.
The dynamic package map encoded in resources.arsc is parsed
and stored with LoadedArsc, and combined to form a resolved table
in AssetManager2.
Benchmarks show that this implementation is an order of magnitude
faster on angler-userdebug (make libandroidfw_benchmarks).
Test: libandroidfw_tests
Change-Id: I57c80248728b63b162bf8269ac9495b53c3e7fa0
Due to legacy reasons, Android translations of European Spanish were
kept under 'es', while Latin American Spanish translations were kept
under 'es-US'. The combination of this, and the new locale
preference rules in Nougat, resulted in 'es' winning over 'es-US' for
all Latin American locales, since 'es' was a direct ancestor, while
'es-US' was just a fallback.
The changes in Nougat had assumed that app developers would put Latin
American Spanish translations under 'es-419', but that could create a
backward-compatibility problem under older Android versions that did
not support three-digit region codes properly.
This CL keeps the Nougat logic and its locale parent tree, but
special-cases es-US and es-MX to be treated as equivalents of es-419
in cases where they are present and es-419 is not.
Bug: 31545805
Bug: 34126460
Test: unit tests are included
Change-Id: Iab26f41294587ee044685a5a6560520c7cbb06f7
The multiwindow model and Resources-per-activity
model that came in N puts greater demands on AssetManagers.
They are created whenever window dimensions change, which
can be frequently. There is a need to be able to cheaply
create a new AssetManager for each Activity, which shares
a lot of underlying state.
In order to make the creation of AssetManagers cheap,
we need a new implementation of the native AssetManager
and ResTable to support immutable representations of
APKs. This new data structure/class is ApkAssets.
ApkAssets have the same functionality of an AssetManager, except
that they operate on a single APK, and they do not do any caching.
Once loaded, they are immutable.
ApkAssets will be exposed as a Java object, with its implementation in
native code. The existing Java StringBlock will be owned by ApkAssets,
which means that Strings can be shared across AssetManagers.
ApkAssets can be cached by the ResourcesManager. Creating an AssetManager
requires only a list of ApkAssets and a configuration.
AssetManager2 (named with the suffix '2' for now while transitioning
to the new implementation) caches bags that are accessed.
Since ApkAssets are expected to be kept around longer, they do more validation
of the resource table, which cause slower load times. Measured on an angler-userdebug,
loading the framework assets takes 11ms with ApkAssets, and 2ms with the old
AssetManager implementation.
The tradeoff is that there does not need to be any security checks once an ApkAssets
is loaded, and regular resource retrieval is faster. Measured on an angler-userdebug,
accessing resource (android:string/ok) with many locales takes 18us with AssetManager2,
and 19us with AssetManager (this is per resource, so these add up).
Test: make libandroidfw_tests
Change-Id: Id0e57ee828f17008891fe3741935a9be8830b01d
The out parameter `out_indices` is expected to be non-null
and so the extra null check adds a cost to performance
and opens the door to misusing the API by not supplying
`out_indices`.
Test: make libandroidfw_tests
Change-Id: Ie66fd837c5e24ec2838156e7b67f54c15cd27933
Changing any of the test APKs could change resource IDs.
The resource IDs are all made public and assigned an ID
to avoid this issue.
Test: make libandroidfw_tests
Change-Id: Idd17c25f4ac86a0ad5b2b8da6f968e5d9b2346c1
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
Bug: 32573798
Mark input uint32_t[] as const. Use Read-only JNI
array access for input as it's faster than critical access.
Use non-movable arrays for TypedArray so that the address can
be resolved and stored, avoiding the need to do JNI array
access for the output.
Indicies is always non-null, so remove the optional checks.
Eliminate unused return value.
Benchmark results:
twelveKeyInflate 4963us -> 4713us
simpleViewInflate 73us -> 60us
Test: Device boots, benchmarks show faster
Change-Id: Ic3bde5aee31407d8903913f97f2218daf074499a
Instead of hardcoding or assuming a path for testdata,
allow the testdata path to be specified via the command
line.
Test: make libandroidfw_tests
Change-Id: Ideae880b21c157b70a11bb5a90a94556771aead6
- Adds unit tests for attribute resolution. These include
some test data resource tables and compiled XML files.
- Convert touched files to Google style guide.
Test: make libandroidfw_tests
Change-Id: Ib3a36061dc874de5f6a266b4e82c0a12ef435f23
A race could occur when printing the list of Asset allocations for
debugging purposes.
Each Asset object would insert themselves into a global linked list
on construction and remove themselves on destruction. Iterating the list
and the insertion/remove operations all acquire a global lock.
The race occurs after the Asset subclass destructor runs but before the Asset
base class destructor runs, which performs the actual removal from the list.
The vtable of the object being destroyed ends up pointing at the base Asset class'
vtable, and during the iteration of the global list, a pure virtual method is called
leading to an abort, since the wrong vtable is dereferenced.
This change moves the insertion/removal of the Asset object into the global list
to the concrete class, which adds some maintenance overhead but solves the problem.
Bug:31113965
Test: make libandroidfw_tests
Change-Id: I1a620897e5e04a8519ee247883bba0719b1fa6f3
Previously, if a mix of "fil" and "tl" resources existed in Resources
(from mixing resources from libraries for example), only resources
from one or the other would be chosen, resulting in default resources
getting surprisingly used. Now, we resolve the equivalent languages
at a per-resource levels (breaking ties for the identical code).
Also, previously if both "tl" and "fil" resources were present in
assets, getLocales() could return a list with duplicate locales.
This change removes Filipino duplicates in the return value of
AssetManager::getLocales().
Finally, there was a bug in the replacement of "tl" with "fil" that
considered any locale starting with the letter "tl" to be Tagalog.
This failed in case of various languages, including Klingon ("tlh")
and Tlingit ("tli"). It's now fixed.
Bug: 29073000
Change-Id: I0e8b9ae337ced2e640a2575897948c4c5ca307d3
Theme values that would reference other theme values would not work
if they were declared in a shared library.
We now introduce a parallel resource type to TYPE_DYNAMIC_REFERENCE,
TYPE_DYNAMIC_ATTRIBUTE, which allows us to lookup and resolve
theme value references from shared libraries.
Bug:28687378
Change-Id: I4f2364e3e8b567679f90784fcaaea12b6b05e926
"tm_mon" format should align with "ZipEntry::setModWhen" in aapt.
"tm_isdst" should be initialized, or it will because random value
and cause error in function mktime().
BUG:28021145
(cherry picked from commit bb0eb3f4cb33c077b296a18555a9a1a966ff696b)
Change-Id: Ia39cf63fcd4bb39c24af080562960279106f526c
When we set the parameters for a ResTable, we can pre-filter which
resources match and only look at that smaller list when getting entries.
This helps A LOT with types that have many configurations, like strings
and all their various locales.
We must store the cached entries in a parallel data structure because parts
of the main Type object are shared with other ResTables, causing data races.
Bug:25499111
Change-Id: I63e37dcbd683fc9f1e7d0f3a6ed4c1c01e0fc575
Previously, a bit was kept to find if the script of a locale was
explicitly "provided" in a resource. This was not backward
compatible, and failed in some edge cases when the package was
created with older versions of AAPT that did not set the bit.
The cases would happen when the old resource had an explicit script
specified in its locale, but since the "provided" bit was not set in
the package, we would assume that the script was computed by us.
This CL replaces the "provided" bit with a "computed" bit, so the
default value of the bit (set to "false" for old packages) would be
correct.
Bug: 27156990
Change-Id: I99e7f1ad8f70c90e25ab3640ed34cc1a6f8d1d64
We get ResTables two different ways: one is from AAPT, another from
settings-based requests from the Java side. In the settings-based
requests, localeScript will be autocomputed, but for AAPT-filled
tables (especially if they come from older versions of AAPT), we need
to compute the script.
Previously, locales that came from packages were incorrectly assumed
to have "undeterminable" scripts, rather than "undetermined" scripts.
This led to us mistakenly falling back to the old logic of requiring
the locales' countries to match, rather than just looking at computed
scripts.
Bug: 27157452
Change-Id: Id7e346d3ecfb17273ffb63de5bcb4849a6eafbbd
When locale fallback landed, resources which specified an 'English'
locale started to be considered a better match for en-US, even though
traditionally, apps tend to ship US English resources under their
default locale.
This fixes that, and makes en-US requests match default locales.
Bug: 26756573
Bug: 26789680
Bug: 26803868
Change-Id: I460c276bfc6ddba0439dcdf87497a0aece0fa05d
* Add support for determining script from language and region.
* Add support for determining special parents of locales.
* Add support for smart comparison of locales with only a difference
in region, using the locale parentage tree.
* Fix LocaleData.matchScore() to not fallback to old locale matching
behavior if we can't determine a script.
* Allow four-character variant codes. (Previously, only five- to
eight-character variant codes were allowed.)
Bug: 7296673
Bug: 26589793
Change-Id: Ibde0a48c0564ff383b41068095a5cbacfe7b94bc
- Added aapt command line flag --app-as-shared-lib to build app resources
that could be loaded as shared lib at runtime.
- Added new method AssetManager.addAssetPathAsSharedLibrary() to load an
app resource as shared library.
Bug 22487604
Change-Id: Ib9b33c35f9c2b7129f3ba205de03d4564623ea39
The round qualifier denotes a device with a screen shape that
is round. The qualifier shows up after the 'long/notlong' qualifier
and before the orientation 'port/land/square' qualifiers.
Change-Id: I3044258b2703a9165694b79725bade770fa6cea1
Was here because UBsan found integer overflow in the parsing for hex
numbers, since hex numbers here are actually unsigned but assigned to
a signed integer.
Also fixes a number of missing error conditions.
Change-Id: Iaea576daedfc6c75521cde02de3fe9dd0198a3b7
A sentinal value of 0x00000000 was used to mark the first time an AttributeFinder
was used. If the resource ID of an attribute was also 0x00000000 (which occurs with
non-resource attributes, like 'style'), then it would be mistaken as the sentinel
start value.
Bug:18421787
Change-Id: I4be353e0f8c940cb6f262d155129f048dcc444ae
Shared libraries have their package ID assigned at run-time, so some
of the guarantees we used to have about sort order of attributes in
bags or XML elements no longer hold.
This CL adds back-tracking and can jump to the nearest attribute with the
same package ID and continue searching.
This means that attributes with the same package ID must be sorted by increasing
resource ID, as was the case before.
Attributes with the same package ID must be grouped together, but the groups can
be in any order. Ex: 0x02010001, 0x02010002, 0x01010000, 0x01010010, 0x7f010032
Bug:17666947
Change-Id: I9c198bbb6ca788849aac85b6323606ea5d9550d6
- char16_t is a distinct type, so stay consistent
with it throughout the code base.
- char16_t is defined as minimum size of 16 bits.
Since we mmap and cast data structures onto raw memory,
we need a precise definition (uint16_t), so we cast between
that (and static_assert that they are the same size).
Change-Id: I869c32637543bbcfb39d2643e7d9df10d33acd3c
