Previously, Looper internally kept track of the requests to add fds
using the fd value itself. It used an internal sequence number
associated with each request to add a callback to avoid a situation
where a callback is unexpectedly called. However, since it used the fd
value rather than the sequence number to register events into epoll,
there was still a way where unintended hangups could occur.
This exact sequence of events caused unintended behavior in Looper:
- An fd (FD) is added to the looper.
- Looper registers FD into epoll using its FD number.
- FD is closed.
- A hangup event arrives from epoll_wait while the Looper is polling.
Looper is waiting for the lock to process the callback for FD, because
it is blocked by:
- A new fd is created and added to the looper. Since the lowest number
fd is reused, this new fd has the same value as FD.
- The poll request for Looper is now unblocked, so it looks up the
callback associated with FD to process the hangup.
- Since FD is already associated with the new callback, the new callback
is called unintentionally.
This CL uses the sequence number to register fds into epoll. That way,
when we get a hangup from epoll that is associated with a sequence
number, there is no way an unexpected callback will called.
This CL also adds a test to verify this behavior. Due to the
nondeterministic nature of this multi-thread scenario, the test verifies
this scenario repeatedly. Without the fix in Looper, the test is flaky,
but should never fail after the fix.
Bug: 195020232
Bug: 189135695
Test: atest libutils_test --rerun-until-failure
Ignore-AOSP-First: Topic CL aosp/1799831 has a merge conflict with
internal master, resolved in ag/15613419.
Change-Id: Ib4edab7f2407adaef6a1708b29bc52634f25dbb6
A lot of things had moved out of system/core/ without their TEST_MAPPING
entries having gone with them, reducing the amount of presubmit coverage
for those things.
In order to reduce the likelihood of that happening again, I've pushed
all that remained in the system/core/ TEST_MAPPING down into the
individual subdirectories.
Test: treehugger
Change-Id: Ib75d65f9200fa64ae1552471da6fbe5b7023cf94
This reverts commit 70d9fb63e6.
Reason for revert: Outstanding usage of this method removed internally
Change-Id: Idcc00ec261aa1d97f11e47abdb08b10a37b5d20f
Test: Local build; treehugger (which I'll manually confirm runs on the appropriate targets)
String16's ctors already handle static strings, so we don't need a
specialized constructor which accepts StaticString16.
Bug: n/a
Test: libutils_test
Change-Id: I93a1ba70d743ff9c73f113d53ffba73cef6adade
String16 is still in use by AIDL compiler. Because String16 is not
noexcept-move-constructible, the C++ compiler will complain when it is
used with non-copyable types (such as ParcelFileDescriptor).
For example, when vector<Foo> is resized, copy-ctor of Foo is called,
which is not available.
parcelable Foo {
String s;
ParcelFileDescriptor[] pfds;
}
By providing noexcept move-ctor for String16, vector<Foo> can be resized
with no problem.
Btw, copy from StaticString16 is specialized for efficiency and move
from StaticString16 don't need to be different from copy.
Bug: 192136980
Test: libutils_test
Change-Id: I13744a2ceebf5781c3ef7f3a04237a6750b0db0a
Fixes a compiler warning for implicit conversion changes from signed
to unsigned which surfaced when refactoring native input libraries.
Add an explicit cast to avoid adding -Wno-sign-conversion compile
flags.
Test: b libsurfaceflinger_unittest
Bug: b/188792659
Change-Id: I1b9309c3530d6063d5cefcaa31af84d9cebc3f5d
Merged-In: I8866aef7f09ca5173604abe18c586b68bbf12ed6
Fixes a compiler warning for implicit conversion changes from signed
to unsigned which surfaced when refactoring native input libraries.
Add an explicit cast to avoid adding -Wno-sign-conversion compile
flags.
Test: b libsurfaceflinger_unittest
Change-Id: I8866aef7f09ca5173604abe18c586b68bbf12ed6
This function, ironically, is being removed.
Even more amusing, it was never "remove" anyway --- it literally did
the opposite, and removed everything *except* the range you passed to
it, and should probably have been called "keep"!
I'm looking at reimplementing much of libutils, but first I'm improving
test coverage, and literally every test I wrote for this failed. And
then when I fixed the "obvious bugs" in the implementation, I found
there actually were a couple of existing unit tests --- that mostly
served to demonstrate just how counter-intuitive this function was.
Bug: http://b/156999009
Test: treehugger
Change-Id: I41fd85f7c0988070f4039f607d2e57523d862ed9
These tests are only in internal right now, preventing me from removing
this function entirely. So let's get them out of the way...
Test: treehugger
Ignore-AOSP-First: workaround for presubmit merge conflict
Bug: http://b/35267372
Change-Id: Ib5de178ffad833c87e8e93bfd33e85a0ff504ac0
It's not tested, and it's not used. Also remove the fuzzer which is just
wasting CPU cycles.
This gets us to 100% function coverage, 100% line coverage, and N/A
branch coverage for StopWatch.
Test: treehugger
Change-Id: Ib5e08510ef1046a6f2af3f0b8a1c317a8bb39fd4
This dates from when libutils was still part of frameworks, and talks
about stuff that *didn't* move out of frameworks.
Test: treehugger
Change-Id: Icf22eebc417bcb37abdbfdc89e9a3f14894e18c2
Allow LightRefBase to be used with
ANDROID_UTILS_REF_BASE_DISABLE_IMPLICIT_CONSTRUCTION, mainly for
libhwui.
Bug: N/A
Test: libutils_test
Change-Id: I251c874a80f0a069572bc51da45f8f8e74ba6f5b
This test previously expected failure - a machine which does not have
2GiB of memory. However, while today is becoming the past,
2GiB allocations working is no longer a dream of the future!
Fixes: 186569165
Test: libutils_test
Change-Id: I6a9ed608c0989d415b4e7735b8a451b8928b4083
