| Age | Commit message (Collapse) | Author |
|---|
|
This reverts commit a3b8c73d8ae9943c8aaabf49349302b4c05cea71.
Reason for revert: Reland
Test: ./art/test.py -b --host --optimizing --64
Test: device boots with container dex enabled for framework
Test: tests pass with compax dex re-enabled
Change-Id: I100a796f5105c351a531a45dcbea8b0245f598a8
|
|
This reverts commit 854b363b284592d0bbfdbe31eb487fa64a4b9471.
Reason for revert: PlayStore Scanner failures
Change-Id: I7807aba1bfe1f662129f90c870c82427ca6d121f
|
|
Allow multiple dex files within single "container"
(either a zip entry or a plain on-disk dex file).
This allows sharing of string (and other) dex data,
since the offsets can point to shared data payload.
Bug: 266950186
Test: test.py -b --host
Change-Id: I4f5901fd2f26a5a9dba427eb48c0fa5ddb6243d8
|
|
Move some of the helper methods to the test instance so that
they can be used from the per-test run scripts.
Test: Build artifacts are identical.
Change-Id: Ieebd40390c34a9a9bb123e54e0b2b6cd410735f1
|
|
Refactor the script to use python Path object rather than strings.
Keep all paths absolute to make the script independent of working dir.
This allows us to use threads rather than forks to compile the tests.
Test: All build artifacts are identical.
Change-Id: Id38881f6457c697d8b5f7ce509beab6e56bd3277
|
|
Use python variables rather than environment to pass state.
Cleanup the environment use left over from the bash version.
Test: build artifacts are same as before
Change-Id: I3ba7f9e31a9a7046c62d67c6c66af4988d5c469a
|
|
Follow the same convention as run-test run scripts.
The python files shall export build function,
rather than doing the work when the script is loaded.
This has the advantage that we now pass context/arguments
to the function (rather than using environment variables).
Test: the generated build artefacts are identical
Change-Id: I90e0ef0d2e31b27813042d51d07b5ae132e1e704
|
|
Tests use the build script for two distinct purposes:
* Some tests generate their source code before completion.
Keep this code in bash for now and rename it "generate-sources".
* To customize arguments passed to the default compilation script.
Since the default compilation script is in python, make this
a normal python call, and add any customization as arguments.
This speeds up the run-test compilation. There was small cost to
pay due to the switching from python to bash to python.
This added up for the thousands of run-test compilations that we do.
It also removes the need for argument parser in the default script.
We could also remove the data passing via environment (to-do later).
It moves the definition to more build-system-like look and feel.
Test: The outputs are bit-for-bit identical as before.
Change-Id: Idca4181a4676036f06aae0a8f6eea3a3c30d390e
|
|
The smali compilation in run-tests are non-deterministic.
However, we don't observe that much, as most tests are
dex-merged by D8 which normalize it back to deterministic.
The non-determinism is only observed for few tests where
well-intended short-cut bypasses the dex merger. Let's
remove that short-cut to get deterministic build output.
Also pass the proper test's API level to the dex merger.
This avoids bumping dex file version to maximum as side
effect of the no-op merge. It also avoids excessive log
spam as the previous input was invalid. And it avoids
bumping the dex file version to max for tests that need
merge, which gives us more version variety for tests.
Bug: 147814778
Test: test.py --host
Change-Id: I4af262aef3305a90d8cb8d14d87f7c52b9dd734a
|
|
Ensure 089 build script "passes" (i.e. expected error is seen).
This was the only (intentionally) failing test.
Use more of prebuilt tools rather than system tools.
Bug: 147814778
Test: test.py --host
Change-Id: I1401edb363312105ecc0da5a5ec41b8be5420798
|
|
Introduce expected standard error files for ART run-tests. Collect
tests' standard output and standard error separately and check them
against the corresponding expectation file.
Test: Run ART run-tests on host and device using `testrunner.py`.
Test: atest --test-mapping art:all
Test: atest --test-mapping cts/hostsidetests/jvmti:all
Bug: 171865375
Bug: 147812905
Change-Id: Ie95bec4a4684ff6791d464124ce8976339432d1f
|
|
This is in preparation for the addition of `expected-stderr.txt` files
in all ART run-test directories, which will record the expected
standard error for each of these tests, and which will be verified
along with `expected-stdout.txt`.
Test: Run ART run-tests on host and device using `testrunner.py`.
Test: atest --test-mapping art:all
Bug: 171865375
Bug: 147812905
Change-Id: I37e3c4b8409475790e2fc63514cdf57156d47ff4
|
|
This reverts commit 09261a8c5cd36a8c7a1ae5107da554dd35008b97.
Fixed redefine-stress failures. Worked around CTS failures
by reducing the cases where we throw the ClassFormatError.
Test: 180-native-default-method.
Test: m test-art-host-gtest
Test: testrunner.py --host --optimizing
Test: aosp_taimen-userdebug boots.
Test: testrunner.py --host --optimizing --redefine-stress \
--debug --debugggable --cdex-fast
Test: cts-tradefed run cts --m vm-tests-tf
Bug: 157170505
Bug: 157718952
Change-Id: I95264af9041836fd6bc54e85263e2a405e877d30
|
|
This reverts commit adf0f2e55c285f3fbe0a70c218a8f267029e736c.
Bug: 157170505
Bug: 157575227
Reason for revert: b/157575227
Change-Id: Ibc0aa5505a59deed717555441a9723b5b6561115
|
|
Fix DCHECK() failure for copied native default methods.
For dex file version 37+ we reject the dex file.
For older versions, throw the ClassFormatError like the RI.
Test: 180-native-default-method.
Test: m test-art-host-gtest
Test: testrunner.py --host --optimizing
Test: aosp_taimen-userdebug boots.
Bug: 157170505
Change-Id: I5d2aefdcbf0b807c1cb482d0fda2e77316783cb4
|