| Age | Commit message (Collapse) | Author |
|---|
|
Fix parse fail between native and java binder.
The same way as strict mode header.
Bug: 359692915
Test: atest binderUnitTest
Change-Id: I6a488e4a98159baf79018dc90adf1c64969b5b85
|
|
Due to OBO (actually off by 4), a Status at the end of the Parcel
with exception information would be reduced to EX_TRANSACTION_FAILED
in native code.
Bug: 235006086
Bug: 132650049
Test: connectivity_native_test, binder_parcel_fuzzer (for a few minutes), aidl_integration_test
Change-Id: I5660958006d7670ad6fc3aabde725cd1d8fe188a
|
|
For errors written to the parcel late.
Bug: 211823183
Test: aidl_integration_test
Change-Id: I036a016e143408bf1c3cb4f67f622c7e0dac91bf
|
|
Java Throwable getMessage can (and will!) return null, and when this
happens, we don't get any error information in the native world. So, for
instance, Java ServiceException won't be reported to native code.
This is technically a backwards incompatible change, and we'll have to
watch for breakages, but losing service specific exceptions in such a
clearly egregious way is not acceptable. :)
Fixes: 178861468
Test: boot & aidl_integration_test
Change-Id: If70d8fb445078c2989c4288c01055bb7191a012f
|
|
stringstream extends ostream, so this is just generally more useful.
I'm tired of having to lookup that I need to use 'Status::toString8' in
order to print it out.
Test: m camera_client_test (uses API, motived last change here)
Test: grep most internal prebuilts for
_ZN7android6binderlsERNSt3__118basic_stringstreamIcNS1_11char_traitsIcEENS1_9allocatorIcEEEERKNS0_6StatusE
Change-Id: I7ed88a3ecf8b560302697e4c5c2357acc05c20fa
|
|
While it may be safe in practice, we should not rely on it.
Bug: n/A
Test: n/A
Change-Id: Id33d577f64449b966bdf6327c9a37886b435ed7c
|
|
"-2147483640" is not a clear error message. "UNEXPECTED_NULL" is.
Bug: 144534032
Test: TH
Change-Id: I235bb1abb88481ffba4c140c27223077880e4dd0
|
|
Bug: 139369544
Change-Id: I24c88d4bc056d8440d6de94488fb3d7fd6ad5200
|
|
Bug: 132650049
Test: fuzzer
Change-Id: Id230eae4316a444bc82b416b2049d5a5f589f89a
Merged-In: Id230eae4316a444bc82b416b2049d5a5f589f89a
(cherry picked from commit 509e0e02730a609c91282625c50a21f731d0f23d)
|
|
String exceptionToString(int32_t exceptionCode)
Add this method for logging purpose
Test: m -j
Bug: 127187108
Change-Id: I911c3a667ddbe39245e5c3bbbd09154b2902aad8
|
|
When EX_TRANSACTION_FAILED is set, mError will never be okay.
Bug: 116618418
Test: atest android.binder.cts
Change-Id: If44ba1fa82cce836f4ee4faad117a1e6b24bf8de
|
|
Right now, it's possible to have Status such that:
- mException = EX_TRANSACTION_FAILED
- mErrorCode = OK
When this happens, writeToParcel will return OK and also not write
anything to the parcel.
From the comment here:
// Something really bad has happened, and we're not going to even
// try returning rich error data.
The intention here is to use the status_t return from transact for
error messages and also not have clients read the parcel. For instance,
here is the AIDL generated code (primary user of this API):
::android::status_t _aidl_ret_status = ::android::OK;
switch (_aidl_code) {
case ...:
{
...
::android::binder::Status _aidl_status(TestVoidReturn());
_aidl_ret_status = _aidl_status.writeToParcel(_aidl_reply);
if (((_aidl_ret_status) != (::android::OK))) {
break;
}
if (!_aidl_status.isOk()) {
break;
}
}
...
}
return _aidl_ret_status;
As you can see, if writeToParcel returns OK for a non-okay transaction,
the transaction will be considered okay. In general, this doesn't cause
a problem since readToParcel will fail, but the error would be
confusing (out of space).
Bug: 111445392
Test: libbinder NDK's runtests.sh
Test: AIDL's runtests.sh
Change-Id: I4699999cb2482b357ae0eab278d8e4aaa66b83aa
|
|
The stack is truncated up to 5 lines at parcel time. When unparceling,
a separate RemoteException will be created and set as a cause of the
exception being thrown.
Performance results(in nanoseconds):
timeWriteExceptionWithStackTraceParceling 82175
timeWriteException 6616
timeReadException 22529
timeReadExceptionWithStackTraceParceling 34151
Test: manual + ParcelPerfTest
Bug: 36561158
Change-Id: I9c0a54a86cd82cff92d8e1a15ec19dfed58ef856
|
|
Follow the new framework changes that support sending EX_PARCELABLE
exception types.
Test: builds, boots
Bug: 33749182
Change-Id: I7a856fa89f23aab4f782c4e5ae2beb8166fa8da4
|
|
Bug: None
Change-Id: Idcb8345cda2bf18960c1c437b7a6de4ec17b59fe
|
|
The templated binder::Status stream operator doesn't work with gmock
because of ambiguous operator overloading. Limiting the stream operator
to just std::stringstream allows it to be used with gmock.
BUG=28171901
TEST=Compiles with gmock and CameraBinderTests
Change-Id: Ia674b68cbff4911b3f5cc3d8ee57d04a1d6cf6bf
|
|
Add another factory method that takes a message and service
specific error code.
Bug: 25800533
Test: system/tools/aidl integration tests pass
Change-Id: I592cb7def0538576965d14c200ab58548b3bef32
|
|
Java code expects status_t != OK to be caught at the JNI level in
android_util_Binder.cpp (see signalExceptionForError). We were
incorrectly mapping this kind of failure to a special exception type
and writing that exception type to parcels.
Instead, refuse to write EX_TRANSACTION_FAILED to a parcel and return
the status value instead.
While here, remove non-trivial constructors to push authors toward the
more explicit factory methods. Remove getException() and push authors
toward using the simpler getter methods. Fix minor camelCase issues.
Bug: 25615695
Test: system/tools/aidl integration tests still pass
Change-Id: I7cad3ac8ae8300b5ac0b466606f4934d01e503c5
|
|
This object implements equivalent functionality to the Java logic which
serializes and re-throws exceptions from services.
Bug: 25615695
Test: Integration test for generated AIDL code reveals this to work
correctly.
Change-Id: Ic80c9def0dd232582ea9c49717ec50894af6bfc2
|