Age | Commit message (Collapse) | Author |
|
For some reason, there is a conflict in the forward declaration of
NSUserActivity. This allows newer versions of OS X than what the
buildbots have to be able to build libchrome.
Bug: 28985443
Change-Id: I3ba3af04fcaba086862f3114250324348c0c9cf2
Test: mmma -j external/libchrome # in El Capitan
|
|
These are causing problems in Android.
Change-Id: I21a78391a770646a418d57e062dc47bf5f2ab7f2
|
|
Even more files are needed.
Change-Id: I9eb75819e952b804efdd7486951525cba8e5c948
|
|
The shim allocator is not supported in OSX. Don't select it when
building for it.
Change-Id: Ib8fb23c44a79f3b500509fc68f8c5b869bd4fa6e
|
|
A file was missing.
Change-Id: I54622c5697d8d3d9f5c719aab2f98fc0575ae712
|
|
Pulled the latest and greatest version of libchrome from Chromium.
The merge was done against r395517 which corresponds to git commit
ebdcb576bb346af95b8ad219f6250daf63122f98 of May 23, 2016
Notable changes are:
- scoped_ptr was removed in favor of std::unique_ptr
- base/thread_task_runner_handle.h was moved to base/threading.
BUG: 28985443
TEST: All tests in libchrome_test pass on dragonboard-eng build
Change-Id: Ic9f9ed1cafe754c96cd2f007984514e091aaba39
|
|
On Android, prefer the libbase definitions of macros. This prevents
compiler errors about macro redefinition.
Allow consumers of libchrome to see libbase headers implicitly by
exporting headers on libbase's include path.
Fix a small bug in how libchrome-dbus was defined. It needs to export
libchrome's headers, since to work with libchrome-dbus, you need all the
headers necessary to consume libchrome as well.
Bug: 28117776
Test: Compiles
Change-Id: I35e969d7037d3374af9d2e931814740e29691029
|
|
This fixes a conflict with Android's libbase over who gets to
define this very uninteresting macro.
Bug: 28117776
Change-Id: I9ef69957d0ba1469225c9701a8355d43f64d6eee
Test: Compiles.
|
|
This is to fix an AOSP build break on Mac host.
Change-Id: I6c8e28ded5639b647a5674ea3da7d7401f71380b
|
|
Change-Id: If20d7b230d1fa21c753bf50b8dcdb29e0932c7ad
|
|
This is to fix an AOSP build break on Mac host.
Change-Id: I3d2a361a09dc4a78fef6bbcc4183318b970e639f
|
|
TEST=Build all
Change-Id: I962a83a044a102515c316d04bb65e7c9302e0f0c
|
|
When we run this test on builders, we sometimes see that
the child is holding an FD open after signalling that all
open FDs have been closed. This is potentially the FD of
file being used to signal that all FDs are closed.
Bug: 27434105
Test: This test still passes.
Change-Id: I229677f8b0a7eb6bc3c6d9c9d56dd8b05e8fde70
|
|
It is difficult to reproduce the listed bug, so just get some debug
output when we encounter it on the builder. This change can/should
be reverted later.
Bug: 27434105
Test: commented out logic to close fds, ran this against an emulator
locally, saw that it printed open fds in a useful way before
failing.
Change-Id: I859e8ccd344dd4afc9a6c582472c479da5960903
|
|
This fixes the -Wexpansion-to-defined warning
BUG=chromium:587491
TEST=it builds with latest clang.
Change-Id: I2ee6a9c4ea47450f8ec352746d48fd5b3eaf6af1
|
|
When calling MessageLoopForIO::current() without a message loop set
and with debug checks enabled, a DCHECK() will dereference a null
pointer causing the program to segfault.
This patch makes the failure evident adding another DCHECK() with a
clear message.
Bug: None
TEST=FEATURES=test emerge-link libchrome
Change-Id: If69d2e07d0bd93a78001be3e683159380bdd5ff1
|
|
These files are needed for Chrome IPC. These will be compiled on
Chrome OS, so no impact on AOSP.
BUG: 26772882
TEST: make -j32 for dragonboard
./build_packages on CrOS for various boards.
Change-Id: Ibca0a04692a59f245cab2d5c9707cff109ca144f
|
|
Applying libchrome r369476 changes to Chrome OS-only files from the
upstream of libchrome (from Chromium).
Also modified SConstruct to add new files and remove deleted files.
Change-Id: Id22160e32d19b0bb228903e3df372fb82550626c
|
|
Some of the "unused" files were apparently needed for building libchrome
on Chrome OS. Reverting deletion of these files.
Change-Id: I02e32f112d16480206f43ca0087342a9de7f1e1b
|
|
The recent uprev attempt missed a couple Mac-specific files.
Adding them now.
Also, there were some other build breaks in Mac files.
Particularly, had to revert the uprev change to the following file
base/message_loop/message_pump_mac.mm because the change relied
on base::mac::CallWithEHFrame() which was implemented in assembly
and didn't compile on 32 bit platform. So, I have removed the
change introduced with https://codereview.chromium.org/1212093002
on Chromium side.
Change-Id: Ic78acbb22732e64965f99ce31bc8382f228e22bb
|
|
These are apparently used on Chrome OS and were deleted from AOSP
by mistake... Recovering.
Change-Id: I257eee62a27d1dbb6029c776ac56af78ca110213
|
|
Pulled the latest and greatest version of libchrome from Chromium.
The merge was done against r369476 which corresponds to git commit
0471d0e2e2ef4a544a63481a389e1df33ea7c00a of Jan 14, 2016
Notable changes are:
- base::scoped_ptr<T> is now almost identical to std::unique_ptr<T>
No Pass() method, now std::move() is used on scoped pointers
- basictypes.h is removed and custom int types such as int32 are now
replaced with the standard int32_t and similar from <stdint.h>
- String utility functions are cleaned up/refactored. Now all are in
base:: namespace, many now return values rather than take pointers
for results, ambiguous Booleans are replaced with enums, such as:
base::StartsWithASCII(current_url, "https://", false);
now is:
base::StartsWith(current_url, "https://",
base::CompareCase::INSENSITIVE_ASCII);
- COMPILE_ASSERT() is now replaced with standard static_assert()
- Numeric range constants such as kuint64max are removed in favor of
standard <limits> constructs such as std::numeric_limits<uint64_t>::max()
- base::Value and derived classes use scoped_ptr<> more and support
for raw pointers to base::Value is deprecated and/or removed in
many places.
- base::MessageLoopProxy is completely removed (was marked deprecated before)
- base::MessageLoop::Quit() and QuitClosure are renamed to QuitWhenIdle
and QuitWhenIdleClosure for more semantic clarity.
Change-Id: I1f5436d253a0a32b2299160a76993752d818736f
|
|
In order to help with uprev maintenance, remove all the source files
that are not being built on AOSP/ChromeOS
Change-Id: Iec3221ae8aa02a17326949085b33925382968673
|
|
These turned out be be entirely related to logging enum values, where
the compiler would get confused about the overload to pick among the
signed/unsigned/long/int stream operators. Fix this by providing
a special template specialization that is enabled only for enum types.
Special thanks to avakulenko@ who suggested this fix.
Bug: 26228533
Test: Compiles under gcc/clang
Change-Id: If0eb6ea8f2998fc826387816d54882462e526650
|
|
Fix a number of unused parameter warnings and two uninitlialized member
warnings.
Bug: 26228533
Change-Id: I46e40115c1087882821a51a978591d013f911b73
Test: Compiles on mac for x86_64-eng
|
|
Bug: 26228533
Test: Compiles under clang/gcc, unittests pass for same
Change-Id: I1cd4f5b7307f2316b2b55e8823382d52cf641826
|
|
Bug: 26228533
Change-Id: I93cc73f395d3be10fe873bc14e39644371f0fbca
Test: Compiles under clang/gcc
|
|
Bug: 26228533
Test: libchrome builds under clang and gcc with this change
Change-Id: Ie534056c7557652fbbd7c2a134131d1cd25824e5
|
|
This reverts commit b636ff6a8ac3b54b3067289f01848252ab71eceb.
This broke trunk with messages like:
In file included from external/libchrome/base/time/time_posix.cc:29:0:
external/libchrome/base/lazy_instance.h:51:36: error: missing initializer for member 'base::LazyInstance<base::Lock, base::internal::LeakyLazyInstanceTraits<base::Lock> >::private_instance_' [-Werror=missing-field-initializers]
#define LAZY_INSTANCE_INITIALIZER {}
^
external/libchrome/base/time/time_posix.cc:39:38: note: in expansion of macro 'LAZY_INSTANCE_INITIALIZER'
g_sys_time_to_time_struct_lock = LAZY_INSTANCE_INITIALIZER;
^
external/libchrome/base/lazy_instance.h:51:36: error: missing initializer for member 'base::LazyInstance<base::Lock, base::internal::LeakyLazyInstanceTraits<base::Lock> >::private_buf_' [-Werror=missing-field-initializers]
#define LAZY_INSTANCE_INITIALIZER {}
^
external/libchrome/base/time/time_posix.cc:39:38: note: in expansion of macro 'LAZY_INSTANCE_INITIALIZER'
g_sys_time_to_time_struct_lock = LAZY_INSTANCE_INITIALIZER;
^
cc1plus: all warnings being treated as errors
Change-Id: I0c0308e716bd1ed7914e2a032e439a9261d38e56
|
|
Also fix compiler warnings in unit tests, except for unused parameter.
There are a ton of unused parameters in unit tests, and those errors
have no risk of spreading through the platform.
Bug: 26228533
Test: libchrome builds, unittests pass
Change-Id: I50431c8e143902c4b71b4381e4dbbc67cdc4507c
|
|
This is an outdated library that added support for thread/memory safety
annotations before clang was extended to support the same. We don't
run those tools on Chrome OS or Android, and it is easiest just to
remove the library and includes.
Bug: 26253162
Test: Compiles, libchrome unittests pass, brilloemulator_arm64 builds
Change-Id: I72f3bdad82fa830c9c76922d98b16dfeebfb5389
|
|
Replace usages of these headers with includes to the Android tree
provided valgrind headers. On Chrome OS, since we don't have that
luxury, we will need to drop the previous headers back in the
ebuild.
Bug: 26253162
Test: Compiles, unittests pass on Android host
Change-Id: I2848d6cb0e7bc8d0430766f9ad5298953e6d59be
|
|
We don't actually compile this in on either Chrome OS or Android.
Bug: 26253162
Test: Compiles, unittests pass on Android (host tests)
Change-Id: Id665b22286e16dd38cee4f98a2000f0e231daae6
|
|
We can use std::hash instead on brillo
Bug: 26253162
Test: brilloemulator_arm64 compiles
Change-Id: I3d363a77c55e9ed0e53ef40aa0b4eba81813d211
|
|
Bug: 26253162
Change-Id: I22a10863d856c60240ee8e5139bbb680bed2807e
Test: Compiles
|
|
We can depend on standard primitives for these operations in chrome os.
Note the formatting of DoubleToString() changes a little since the
standard primitives are a little less pretty.
The behavior of StringToDouble changed a little with respect to numbers
not representable as a double (too large or small). StringToDouble
still reports a broken conversion, but the value returned is infinity
rather than the maximum finite value.
Bug: 26253162
Change-Id: I5ba87d0787bb832d188f6460d952fd852e27fed6
Test: compiles, unittests pass on host
|
|
When a service sends a histogram with invalid construction, print an
error.
Bug: 25670685
Change-Id: I405befe42d81746c0e31abcd0b1626f516d2d000
Test: Sends incorrect histogram. A message is logged.
|
|
Other libraries define this macro already (libnativehelper for example).
We already have a guard there to ensure we don't define the macros
twice. We should have one in libchrome too.
Bug: 25862968
Change-Id: I2e238eb06d58c3877745841ec1f7da79bf58a34e
|
|
Instead, pull the build date of the system from the ro.build.date system
property. Then this library will be identical as long as the sources and
dependencies don't change, and we won't have to update it on every OTA.
Bug: 24204119
Change-Id: Ie5368ec0bbbc635dc6b86f9259d6567fe26ca2ba
|
|
The base::CommandLine singleton is usually initialized from
main(), but some gtest-based unit tests don't have control
over that function. Guard CommandLine accesses in
base/logging.cc to handle this gracefully.
Bug: 24536569
Change-Id: If3eab0f36c28bb6b3cf3569c61c7d7cec5cf7b10
|
|
When libchrome is compiled in Brillo, using Android bionic, uid_t is
actually defined as 'unsigned int'.
This is Chromium r260569, and is the sole code fix required to allow
building the Fluoride unit tests and test tools for a Mac host.
BUG=22771504, chromium:534675
Change-Id: I16e5ab954d1fb8540c0b875ce6b31d48a787a612
|
|
To make it easier to attribute log messages, use argv[0]'s
basename (as supplied by base::CommandLine::GetProgram()) as
the log tag. This only affects libchrome builds for Android
(i.e. when __ANDROID__ is defined but OS_ANDROID isn't).
Bug: 23168740
Change-Id: I5a95602877c393d1e05b3ae8a31a25c7b47d3ce1
|
|
Temporarily revert base::WriteFile to the behavior in older
revision of libchrome until we sort out the expected file
permissions at all call sites of base::WriteFile in Chrome
OS code (crbug.com/412057).
This is benchan@chromium.org's
base-334380-revert-writefile-permissions.patch from Chrome
OS. The original commit was
https://chromium-review.googlesource.com/216584.
Bug: 22317122
Change-Id: Ia86ef66122b6cf148bad56d13ef82624e1db8a5b
|
|
CallocOverflow test has some difficult issues preventing it from passing.
See b/23428680. Disabling it until these issues are addressed.
BUG: 23428680, 22884411
Change-Id: Ie254ae63e9cb61ff4a575ee0369bff8e85dbe29c
|
|
Drop the __BRILLO__ #define, instead using the existing
__ANDROID__ define set by the toolchain and setting a new
__ANDROID_HOST__ define when building for the host.
Also use Android logging when __ANDROID__ is defined.
Bug: 23358460
Change-Id: I0d86eac8af381b002a3d46d8a95434c7e24518bd
|
|
Submitted before chromium cl 1273243002 went through, CQ found
issue that is fixed here.
BUG: 22884411, 22947438
Change-Id: Ib91967ce6cdd60d4c923822b0fa52c40661070fe
|
|
See discussion on bug. Summary:
* When building for brillo not on host tmp dir now defaults to the
android "/data/local/tmp"
* Command line is temporary fix because current test runner
doesn't Init.
* Dir reader posix was using an outdated method of getting a temp
dir. (checked with agl)
* String conversions used incorrect capitalization of locales.
* File path needed localization on host machine (checked with rvargas)
BUG: 22884411, 22947438
Change-Id: Ib7ffa58a84c8d93c819ba572e2fbd04dd1719e55
|
|
libchromeTests added to Android.mk, along with many unit test files.
Once built and synced, the tests can be run on the device/emulator via
/data/nativetest/libchrome_test/libchrome_test
Moved the gtest & gmock testing directory out of base/ since all the files
were looking for it directly under libchrome/.
Bug: 22857514
Change-Id: Ib65897b32979160768dd1e45092bb8f520833666
Signed-off-by: Ari Hausman-Cohen <arihc@google.com>
|
|
|
|
modp_b64 was moved to a subdirectory. We need to change the import path
too.
Change-Id: I3b9e75895bf93dc6c128944264e205240ddf33ca
|