Age | Commit message (Collapse) | Author |
|
And clean up ClassTable.
Test: m test-art-host-gtest
Test: testrunner.py --host --interpreter
Bug: 119486698
Change-Id: Ie413bc5a56eb548352ae1fed0976b75c44e0c0d4
|
|
Customize the email sent by LUCI to include the status in the title
and a summary in the body.
Test: manual
Change-Id: I26deeb5a8a567369cdcbe159b636b2f328a7b61a
|
|
Test: N/A
Change-Id: I3f93b7b1072e6f0396f9e376882e76fc4fdec10c
|
|
Test: N/A
Change-Id: Iea49baf8c6a6812ad76d1c848b8b0345e1948fcb
|
|
Test: N/A
Change-Id: I8778a84c85be073249c996d58a4f0757228b23f9
|
|
Disable for investigation.
Bug: 132163727
Test: m
Change-Id: I7164747452598cfb333511651fb1058b48464f76
|
|
Symlinks aren't supported directly in the build graph and thus they must
be side effects of some other action. The symlinks in the symbols
directory is now created as a side effect of building
out/target/product/<product>/obj/FAKE/com.android.runtime_intermediates/symlink.timestamp
file.
For more details, please refer to
https://android-review.googlesource.com/c/platform/art/+/930848/5/Android.mk#392
Bug: 120846816
Test: m dist with marlin (flattened) and blueline (non-flattened)
examine symbols.zip file and check that unstripped shared libraries are
found under /apex/com.android.runtime directory
Change-Id: I408eea6b91d4daf5e873c479fe5f0a2faf8a859e
|
|
Bug: 132100102
Change-Id: I06440aa0e577176ac2280534ef446fc06d3f46a8
|
|
Bug: http://b/129068177
Test: treehugger
Change-Id: I6632e246bf76f535a14de5e521bcdf65481ef544
|
|
This reverts commit c8c7e0659a76a951f3ab110ed9b7f37777448d36.
Test: none.
Reason for revert: Causes art-tracing to flake/fail pop-frame tests.
Bug: 131865028
Change-Id: I16bda461da94405f372b6b0be54ac4b4efb08fda
|
|
We were incorrectly treating some JVMTI events as still requiring
global deopt even though that requirment had been lifted. This meant
that if one sets thread-specific event notification requests on
multiple threads only some of them will actually be deoptimized. This
could cause events to be missed if the relevant methods/threads had
been jitted.
Test: ./test.py --host --jit-on-first-use
Test: ./test.py
Bug: 131865028
Change-Id: Ib196faa635e61262eea7ccc576dc0d700b3e7eff
|
|
Adds a mode to track what methods are being interpreted. Good
for debugging interpreter usage during app startup.
Bug: 130185360
Test: test-art-host
Change-Id: I8119f79e534cdad7d6a72d4f01115a288e9c8d9a
|
|
We could incorrectly calling AddLocalReference with a null pointer
during the JVMTI GetLocalObject call. This would cause a check
failure because nulls are not allowed in the table.
Test: ./test.py --host
Test: Manual
Bug: 131856650
Change-Id: Iabc283061c8f3ca8b3f7421ef9de83bb67605402
|
|
Since the memory tool space calls Object::SizeOf, we need to free
classes after instances to prevent reading stale data.
Bug: 131542326
Test: art/test/testrunner/run_build_test_target.py -j50 art-gtest-heap-poisoning
Change-Id: I4b0583b2456e8db23ae3cd19ef6495bff955e7dc
|
|
Bug: 131709594
Change-Id: Id9110adfcb42a94701eccc421ca7d9cceacd4cd2
|
|
The (dir $@) is a shared directory that could collide with another
rule. So instead prefix temporary files with $@_.
Also remove verbose flag to the zipalign command.
Test: m
Change-Id: Ie93e2b486095ee7d8a007dc15dd393f26bab03e8
|
|
Temporary measure until we have a system server specific profile.
Test: m
Bug: 119800099
Change-Id: I58da2d96ec355e26f872b67b021c3b3dba26e2b4
|
|
Forward-declare code item accessors in ArtMethod.
Test: mmma art
Change-Id: Ib4b2f274130416ee07e873ec32cef98d10c95838
|
|
Follow-up to commit 4835d21a33b783869820c928bfc0a16f249d65f2. Clarify
template parameter.
Bug: 120022597
Test: m test-art-host
Change-Id: I0d7c5503bb0bb0bf87baf3d05ae284e85899fe1f
|
|
Make the domain a known char storage and forward-declare in DexFile
to significantly reduce the list of transitive includes. Fix up
missing includes.
Test: mmma art
Change-Id: I9d43113e982fc0b509fd9a8168c6788e9431500a
|
|
ArtField doesn't need dex_file_types. This in turn exposes a
missing include in modifiers.
Test: mmma art
Change-Id: I318e2919121f350a0447f6fc0814d81779fd9cc5
|
|
When writing to tombstoned fails, do not write stack traces to the
log. The failure is a likely symptom of a more fundamental issue.
(cherry picked from commit 6cca24104ae06c7cd948e466af127bf1bc620f81)
Test: make
Bug: 130025619
Change-Id: I3c7e895dff7d9df5f8d14c1b1235c1c943d32295
Merged-In: I3c7e895dff7d9df5f8d14c1b1235c1c943d32295
|
|
Test: art/test/testrunner/testrunner.py -b --host -t 005-annotations
Bug: 130028992
Change-Id: I624f8afbfe4c4c1cd3f326f6e28d5967cf4ce2c1
|
|
way around.
This relands http://r.android.com/952236:
There isn't a way to make a namespace properly an alias for another - just
adding a link with allow_all_shared_libs true doesn't quite do it since
links aren't transitive. That meant that "runtime" in the previous setup did
not pass on links to "default" when "classloader-namespace" in
libnativeloader added links to "runtime".
With this we link from "default" to "runtime" instead, and only APEX
binaries will start in "default" - nothing else should create links to that
namespace when this config is in use.
This also ensures no libraries are actually loaded in "default", which could
otherwise create duplicate instances with libraries loaded in "runtime".
Also make the "platform" namespace visible, to work with
android_link_namespace() in the classloader-namespace setup, and update the
reason why those namespaces need visibility.
Test: Flash and boot
Test: Run a target run test similar to art/test/run-test:
Push libarttest.so to the test directory, then
ANDROID_ADDITIONAL_PUBLIC_LIBRARIES=libc++.so:libbacktrace.so:libbase.so:libnativehelper.so:libart.so:libdexfile.so:libprofile.so:libartbase.so \
dalvikvm -Djava.library.path=. -cp 203-multi-checkpoint.jar Main arttest
Test: lunch aosp_cf_x86_phone-eng;
atest android.compilation.cts.AdbRootDependentCompilationTest#testCompile_curProfile \
com.android.cts.dexmetadata.InstallDexMetadataHostTest#testProfileSnapshotAfterInstall \
installd_dexopt_test
Bug: 130293232
Bug: 121117762
Change-Id: Ica4cea3103c061e01bdcbecce4dc1cb9690ee186
|
|
Workaround until we properly support an apex in go/art-buildbot and go/lem.
Test: m
Change-Id: Id9900e847223c0e900a211b7a4e551e93f82c81f
|
|
We should not be holding the mutator lock since that will maybe
cause DCHECK failures.
Bug: 131318971
Test: test-art-host
Change-Id: I14064b06adb9ec0e83af6afb70320f184fd2aed0
|
|
Have `tools/buildbot-build.sh --target` and Golem's target build
rule (`build-art-target-golem`) build targets `tzdata-art-test`,
`tzlookup.xml-art-test`, and `tz_version-art-test`, copying timezones
prebuilt files into /system/etc/tz, so that they can be found even if
the Runtime APEX is not available, by setting the environment variable
`ART_TEST_ANDROID_RUNTIME_ROOT` to "/system" on device.
This is a temporary change needed until the ART Buildbot fully
supports the Runtime APEX.
Test: Rely on the ART Buildbot and Golem.
Bug: 131667225
Bug: 121117762
Bug: 129332183
Bug: 119293618
Bug: 131239046
Change-Id: Ib9f748dea155b54eec0d000dd9e0f0dfb29ca626
|
|
Follow up to http://r.android.com/943476. This library is only available in
debug builds of the Runtime APEX.
Test: Flash and boot with eng build
Test: adb shell ls apex/com.android.runtime/lib\*/libdexfiled_external.so
Change-Id: Ib6bbfacc19416ac1221a962f111b3937e61193e7
|
|
Testrunner should drive this. Having individual tests do this
puts them in the situation of double loading debug and ndebug libraries.
Test: 018-stack-overflow
Change-Id: I4091f67080c08167dbe7a44605540e729cae040a
|
|
Test: set maxTargetSdk=Q on an API, compile, check hiddenapi-flags.csv
Change-Id: I05b3418d76bdb2ef34a717e199abdda73f73ffd7
|
|
Change-Id: I8d905bc1e2aeec7a5610fac8b60bbcd5c9b1e8aa
|
|
We need both versions and load the correct one.
Test: test.py -r --ndebug -t 115
Test: test.py -r --debug -t 115
Change-Id: I189d4d92a7eea63d6ebd4040d12926bc0c3f1154
|
|
And when compiling an apex image, discard dex files not present
in an apex.
Test: m
Bug: 119800099
Change-Id: Ie91c5b8d271783f04e4c1501f315a3ec59137475
|
|
http://r.android.com/939394 changed the API but forgot to update this file.
Test: make update-api
Change-Id: I774ca4bc94ef578b626a97bf9a79b96295ff0b5c
|
|
Because loop unrolling is part of a general loop optimization pass,
it needs to update induction ranges as it will invalidate its
instruction cache with new instructions.
Bug: 131174583
Test: 696-loop
Change-Id: Id3628efe316b58f69abbd9ebd43e891a8e42529f
|
|
We load both the debug and release build when running tests.
The different builds have different layout of some classes,
which can currently result in random native heap corruptions.
Fix the build dependencies to avoid the double loading,
and add runtime check to avoid it in the future again.
Test: test.py -b --host --64
Test: test.py -r --target -t 001-HelloWorld
Change-Id: Ie62f91dc06209c91e25ba5f11c9d61243ac7579d
|
|
Trying to diagnose a rare flake.
Bug: 111411286
Test: test.py --host -b -r -t 137-cfi
Change-Id: Icfd106f0d1fc03b22f3f4e85c6f2d2ee152e1be6
|
|
This reverts commit 5fe1b23d54303a1adf91b644f26f6df066e175ed.
Reason for revert: Breaks android.compilation.cts.AdbRootDependentCompilationTest#testCompile_curProfile and others: http://g/art-team/R2nEDA1Ka_s
Change-Id: I6bddceda13239fc6e27748128352c6954eaab660
|
|
Same as 570-checker-osr
Bug: 130313339
Test: test.py -r --host --trace -t 570
Change-Id: I1be08d72b569710c387e623d3b0df9a6a3a08465
|
|
Don't strip, repack or compress debug-info if explicitly
requested by the developer (using the -g compiler flag).
If enabled, the DWARF debug info has about 1:1 size
overhead relative to JIT code + data.
Bug: 131422204
Test: Check that gdb shows line numbers for JITed method.
Test: Hard-code always-enable generation and run maps.
Change-Id: If06de8ae2317af4d57d84e8a8bfae86a597dd4e4
|
|
Return an error message for it. Callers expect it.
Test: dex2oat_test
Bug: 131175467
Change-Id: I55c05f0a9c48cea252869f88740c6338b5b28b1f
|
|
around.
There isn't a way to make a namespace properly an alias for another - just
adding a link with allow_all_shared_libs true doesn't quite do it since
links aren't transitive. That meant that "runtime" in the previous setup did
not pass on links to "default" when "classloader-namespace" in
libnativeloader added links to "runtime".
With this we link from "default" to "runtime" instead, and only APEX
binaries will start in "default" - nothing else should create links to that
namespace when this config is in use.
This also ensures no libraries are actually loaded in "default", which could
otherwise create duplicate instances with libraries loaded in "runtime".
Also make the "platform" namespace visible, to work with
android_link_namespace() in the classloader-namespace setup, and update the
reason why those namespaces need visibility.
Test: Flash and boot
Test: Run a target run test similar to art/test/run-test:
Push libarttest.so to the test directory, then
ANDROID_ADDITIONAL_PUBLIC_LIBRARIES=libc++.so:libbacktrace.so:libbase.so:libnativehelper.so:libart.so:libdexfile.so:libprofile.so:libartbase.so \
dalvikvm -Djava.library.path=. -cp 203-multi-checkpoint.jar Main arttest
Bug: 130293232
Bug: 121117762
Change-Id: I0f0b58aac3ec869b42510f9ef6af7cb26667de96
|
|
Infra issues have been resolved.
Change-Id: Ia5f0a9dd138f1b42f6ef0c3a895d638697439ffe
|
|
We need to save callee save floating-point registers on ARM/ARM64.
No x86/x86-64 floating point registers are callee save.
Bug: 130313339
Test: test.py -r --target -t 570-checker-osr
Change-Id: I53dfd60edb136bd305389e3b4bd51b636875d429
|
|
These tests all use a common set of suspension triggering event code.
This will be useful in other tests later. To make it easier to use in
the future this code is separated into common ti-agent files.
Test: ./test.py --host -j80
Change-Id: I75f3fe8862c99bd32c5eabe0992fef911d7d9c56
|
|
Logic for constructing paths in IsLocationOnModule assumes that Android
root is provided with a starting slash and without one at the end. Even
though our default settings match that, some users have reported crashes
with a custom ANDROID_ROOT env variable. Change logic to support both.
Bug: 130817106
Test: set ANDROID_ROOT and compile
Change-Id: Id83ca985b05bd4e8f85d7ca0b4c8aeea98aa21ee
|
|
It can be called in a situation where dead blocks can have phis:
the DCE pass would re-build the dominator tree and RemoveInstructionsAsUsersFromDeadBlocks
used to assume there cannot be phis.
Test: 695-simplify-throws
Bug: 131174581
Change-Id: I853956482487a8c7c8fb99499aef318d099754e4
|
|
Compile JNI stub on first use to avoid the expensive generic stub.
Test: profile calendar
Bug: 119800099
Change-Id: Iaef9d0d528ff34c1636237cdcce6e8639c47c8ed
|
|
Remember the actual violation and use it as the abort message.
Bug: 131324618
Test: m test-art-host
Change-Id: I4c3edb328501917b838ec1b3443e454c6ac1a13f
|
|
Use better types and fix some typing issues exposed by clang-tidy's
readability-implicit-bool-conversion.
Bug: 32619234
Test: WITH_TIDY=1 mmma art
Change-Id: I787d291a4843a3192a84853d105d57377b774b23
|