diff options
author | 2024-10-15 07:33:57 +0000 | |
---|---|---|
committer | 2024-10-15 10:03:24 +0000 | |
commit | f2d6bfd7d6b36366ece4a7dea15e66a6e5970615 (patch) | |
tree | b2760a92a0c0a6ceac50d9f00fe7be4045d40a03 /compiler/optimizing/scheduler_test.cc | |
parent | b544a1017753bfef7b91a2ea3024ffbde6cd5bb9 (diff) |
verifier: Do not resolve types in other class loaders.
Remove performance optimizations introduced in
fc0e94bed3f88ed7e50854fd8dfaf5dcb345250f StringPiece clean up.
https://android-review.googlesource.com/103969
(the first is not available in AOSP code review but it can
be seen with `git log` in a checked out ART repository).
These optimizations allowed the return type to be resolved
in the resolved method's class loader which may differ from
the class loader of the class being verified. We do not do
that for arguments, so we should not do that for return type
either. Similarly, clean up resolved field's type retrieval.
Note that `ClassLinker::ValidateSuperClassDescriptors()`
checks for some mismatches during class initialization but
it only checks virtual and interface methods within the
class hierarchy and not for every invoke (and it does not
check direct methods at all).
Test: m test-art-host-gtest
Test: testrunner.py --host --optimizing
Change-Id: I3b4313ac6cca5071f94709bb1bde50a6a253d18d
Diffstat (limited to 'compiler/optimizing/scheduler_test.cc')
0 files changed, 0 insertions, 0 deletions