Andrii Grynenko [Wed, 10 Feb 2016 19:14:26 +0000 (11:14 -0800)]
Fix EventBase destruction race in FiberManagerMap
Summary:
Previously we could be reading from thread-local FiberManagerMap while it was modified.
This is now fixed by keeping a per-thread list of EventBases which need to be removed from local maps. On the fast-path no action is taken, since list will be empty.
This is second try, since
D2853921 got reverted.
The new implementation is simpler and does not rely on AtomicLinkedList.
Reviewed By: yfeldblum
Differential Revision:
D2908018
fb-gh-sync-id:
4d7aed974c19761f7e2732ddbf8694af57c69bd6
shipit-source-id:
4d7aed974c19761f7e2732ddbf8694af57c69bd6
Yedidya Feldblum [Wed, 10 Feb 2016 00:28:21 +0000 (16:28 -0800)]
stripLeftMargin
Summary:
[Folly] `stripLeftMargin`.
So you can do:
TEST(MyClass, doSomethingWithString) {
// Multiline string literal is indented properly inside the test case
// instead of being aligned all the way to the left, which would make
// the test case read poorly.
auto input = folly::stripLeftMargin(R"TEXT(
first line
second line
third line
)TEXT");
auto expected = //...
auto actual = MyClass::doSomethingWithString(input);
EXPECT_EQ(expected, actual);
}
Reviewed By: markisaa
Differential Revision:
D2909736
fb-gh-sync-id:
ebb07da05e1a788535064cfcd9e07f617a007800
shipit-source-id:
ebb07da05e1a788535064cfcd9e07f617a007800
Alan Frindell [Tue, 9 Feb 2016 17:44:21 +0000 (09:44 -0800)]
Relax HHWheelTimer::destroy assertion to accommodate SharedPtr
Summary:
HHWheelTimer's auto-pointers are kind of funny. You can do something like this:
```
HHWheelTimer::UniquePtr p = ...;
// create a SharedPtr from UniquePtr
HHWheelTimer::SharedPtr s(p);
// create another SharedPtr from raw ptr
HHWheelTimer::SharedPtr s(p.get());
// No problem.
If you do this:
HHWheelTimer::SharedPtr p = ....;
// this leaks
```
Why? Because SharedPtr is only have of std::shared_ptr. It's the refcounting half. But when the last SharedPtr goes out of scope, it **does not** invoke HHWheelTimer::destroy().
So code like this is possible/expected:
```
MySweetObj::MySweetObj(HHWheelTimer::SharedPtr s) {
s_ = s;
s_.scheduleTimeout(a, b);
}
{
HHWheelTimer::UniquePtr p = ...;
obj = MySweetObj(p)
// But what if I decide to kill p:
p.reset();
}
```
Since MySweetObj takes a SharedPtr and holds it, it can reasonbly expect that it can schedule timeouts on it, and the HHWheelTimer will not be deleted until it releases the SharedPtr. This is true, but the above code would hit the assert that count_ == 0.
Instead, relase the check that count_ == 0 only if there are no extra outstanding SharedPtrs.
Reviewed By: viswanathgs, chadparry
Differential Revision:
D2908729
fb-gh-sync-id:
9abd4a7d692fe952c5514dbb8d85dfbad95a3cac
shipit-source-id:
9abd4a7d692fe952c5514dbb8d85dfbad95a3cac
Michael Lee [Mon, 8 Feb 2016 20:53:55 +0000 (12:53 -0800)]
Fix bad merge.
Summary:
FBStringTest.cpp was refactored, but the TestUtil and
TemporaryFile was not properly removed.
Reviewed By: yfeldblum
Differential Revision:
D2913045
fb-gh-sync-id:
710c3f5b808acb634dfcd65219484ddc257ed52c
Yedidya Feldblum [Mon, 8 Feb 2016 02:14:17 +0000 (18:14 -0800)]
Fix Build: conditionally_existent_test in folly/test/Makefile.am
Summary:
[Folly] Fix Build: `conditionally_existent_test` in `folly/test/Makefile.am`.
HT: https://github.com/shindo.
Reviewed By: meyering
Differential Revision:
D2910782
fb-gh-sync-id:
9593f026e7c58b8644abc5996da65005bcc2d095
Sven Over [Sun, 7 Feb 2016 13:24:01 +0000 (05:24 -0800)]
folly/futures: fix early release of non-embedded callbacks
Summary:
folly::Future (more precisely folly::detail::Core) can store
callback functors (typically lambdas) up to a certain size
(8*sizeof(void*)) inside the main object. Only bigger functors
are stored in the std::function object in folly::detail::Core,
which will put them on the heap.
The behaviour of folly::Future is slightly different depending
on whether the callback can be embedded in the main object
or not. Small functors will be destructed after the callback
is executed. Functors too big to fit in the lambda space in
the Core object will be deleted when Core is deleted.
Some code assumes that functor objects are deleted as soon
as the callback has been executed. The implementations of
folly::Future::collect and variants do so. If you switch
off this optimisation temporarily (which can be done easily
by setting lambdaBufSize in folly/futures/detail/Core.h to
0), a number of unit tests fail.
Given that the lambda buffer is reasonably large, most
functors can be stored in the Core object. The different
behaviour for very large lambdas may not have been observed.
This diff fixes the inconsitent behaviour.
Firstly, it changes the unit test Future:finish to explicitly
check that the callback functor is deleted after the callback
has been executed. This should be tested, because this
behaviour is assumed in other parts of the futures
implementation (e.g. Future::collect et al.).
Secondly, it adds another unit test Future:finishBigLambda,
similar to Future:finish. The lambda captures much more data
to make sure the lambda won't be stored in the Core object,
but in the std::function object inside Core. The test verifies
that the behaviour is the same, i.e. the callback functor
is destructed after the callback has been executed.
Thirdly, it fixes Core and makes sure that functors of any
size are destructued after the callback has been called. The
new unit test fails without this.
Reviewed By: fugalh
Differential Revision:
D2883772
fb-gh-sync-id:
21a410f6592b3e090772a7b55bef6729d8739922
Den Raskovalov [Sat, 6 Feb 2016 05:31:32 +0000 (21:31 -0800)]
Call destructor for non-trivial destructors if arena allocator is used in ConcurrentSkipList.
Reviewed By: philippv
Differential Revision:
D2900534
fb-gh-sync-id:
c711a160d541d937ada24f2b524016dbceb40ee2
Francis Ma [Fri, 5 Feb 2016 23:00:39 +0000 (15:00 -0800)]
Loosen restriction to get folly::future::CoreTest pass on android
Summary: Due to different way of padding, CoreTest failed on android. Loosen the check to make it pass.
Reviewed By: mzlee
Differential Revision:
D2907987
fb-gh-sync-id:
81d715725d6908ec7b259b0a67789a91ed63df71
Francis Ma [Fri, 5 Feb 2016 22:42:28 +0000 (14:42 -0800)]
Make folly/future portable on android
Summary: Some small changes to port folly/futures onto android
Reviewed By: mzlee
Differential Revision:
D2787564
fb-gh-sync-id:
8c161942c24c2b7b0484339eaa51c5a356f17de5
Yedidya Feldblum [Fri, 5 Feb 2016 20:27:56 +0000 (12:27 -0800)]
ConditionallyExistent<typename>
Summary:
[Folly] `ConditionallyExistent<typename>`.
For when we need extra member fields in dbg builds, but want to save the space in opt builds. Or other situations along similar lines, but with perhaps another the statically-known condition.
Conditional compilation can have some nasty side-effects. Best to avoid it. Let the compiler see everything, rather than having the preprocessor ruin our day. Let the optimizer remove code that will never be called, after the compiler has seen it. Let us have a single AST in our source file, not one AST for dbg and one for opt, or other statically-known conditions.
Reviewed By: andriigrynenko
Differential Revision:
D2879397
fb-gh-sync-id:
d631141a984eebd46674f27a40a97f670eb33f54
Andrii Grynenko [Fri, 5 Feb 2016 20:01:26 +0000 (12:01 -0800)]
Revert
D2853921
Summary: This was causing memory leaks. I assume that happens if FiberManager was once requested from a thread (so that thread local was initialized), but the thread was never used since then, accumulating EventBase* in the queue.
Reviewed By: simonmar
Differential Revision:
D2906752
fb-gh-sync-id:
71ab14cb051a9cee3684a30eaf6729ef65888a52
Michael Lee [Fri, 5 Feb 2016 15:27:02 +0000 (07:27 -0800)]
Turn off a FBStringTest for Android.
Summary:
The wstring support on KitKat is not exactly working. Works
on Lollipop though.
Reviewed By: markisaa
Differential Revision:
D2901732
fb-gh-sync-id:
97b57fb4d8c645192be62dca25d8fb1b2397ad7d
Louis Brandy [Fri, 5 Feb 2016 01:23:56 +0000 (17:23 -0800)]
Greatly expand the components overview in Overview.md
Summary: See title. This is create a big flat mess, which I may clean up in a follow on to highlight the biggest and most generally useful libraries.
Reviewed By: fugalh
Differential Revision:
D2902152
fb-gh-sync-id:
028633448ab97f47a3419f31fa58c79b77a89e52
Louis Brandy [Fri, 5 Feb 2016 00:55:29 +0000 (16:55 -0800)]
promote a bunch of high level txt to the main Readme.md
Summary: See title
Reviewed By: snarkmaster
Differential Revision:
D2901817
fb-gh-sync-id:
5b9ede7b941cf18e7be352a93ec5686e6f1f50a7
Lee Howes [Fri, 5 Feb 2016 00:19:09 +0000 (16:19 -0800)]
Adding addTaskFuture and addTaskRemoteFuture to FiberManager.
Summary: Adds functionality to fibermanager to add tasks locally and remotely that will return futures. As a side effect, also wraps the function in addTaskRemote in a move wrapper so that it behaves correctly with move-only types such as promises.
Reviewed By: andriigrynenko, yfeldblum
Differential Revision:
D2892898
fb-gh-sync-id:
1666c103db35d9c149c36f8b8c3058cd6e0465fc
Andrii Grynenko [Thu, 4 Feb 2016 21:58:25 +0000 (13:58 -0800)]
Improve RequestContext::getStaticContext() perf
Summary: Avoid using folly::ThreadLocal on the fast-path. This uses the fact that it is a static object.
Reviewed By: yfeldblum
Differential Revision:
D2873374
fb-gh-sync-id:
978108da0c9e8576fda6849058b0262478c2841e
Neel Goyal [Thu, 4 Feb 2016 18:43:06 +0000 (10:43 -0800)]
Allow SSLSessionCallbacks to be used on SSL* that isn't attached to AsyncSSLSocket
Summary: The session callbacks assumed that the SSL* was associated with a folly::AsyncSSLSocket when it didn't need to. This enables apps that manage their own SSL* to use these callbacks.
Reviewed By: yfeldblum
Differential Revision:
D2896426
fb-gh-sync-id:
c51df6b4cb3f4cc188a6411c1f3e7e89e96e8a67
Giuseppe Ottaviano [Thu, 4 Feb 2016 01:32:42 +0000 (17:32 -0800)]
Build workaround for nvcc in Malloc.h
Summary: See comment.
Reviewed By: juancarabina, luciang
Differential Revision:
D2896262
fb-gh-sync-id:
cb884651d47b86c0a5b3c2c22f421b2b1f39dc35
Alan Frindell [Tue, 2 Feb 2016 23:31:32 +0000 (15:31 -0800)]
Make HHWheelTimer take a TimeoutManager rather than EventBase
Summary: TimeoutManager is a Mockable parent class
Reviewed By: maxgeorg
Differential Revision:
D2892671
fb-gh-sync-id:
d20635b18a7c748009ce4880c80a13ac23b9222a
Peter Griess [Tue, 2 Feb 2016 23:27:21 +0000 (15:27 -0800)]
Add MockTimeoutManager
Summary: - Add new MockTimeoutManager mock. For testing.
Reviewed By: yfeldblum
Differential Revision:
D2891814
fb-gh-sync-id:
6d15feb533e5a8984484d1efdf58d2c77b7b3f49
Wez Furlong [Tue, 2 Feb 2016 22:54:56 +0000 (14:54 -0800)]
folly: add bser encode/decode for dynamic
Summary:
To support consuming Watchman from within fbcode and hhvm
in particular, these functions add a BSER serialization for the
folly::dynamic data type.
Reviewed By: bhamiltoncx
Differential Revision:
D2876539
fb-gh-sync-id:
bc49d6bc453cc66cebda7185a5907a6f70970b24
Denis Samoylov [Tue, 2 Feb 2016 22:38:25 +0000 (14:38 -0800)]
fixed Ubuntu 12 build script dependencies
Summary: added missed boost dependency and changed CMake build order
Reviewed By: snarkmaster, yfeldblum
Differential Revision:
D2878880
fb-gh-sync-id:
2a5cded97afde4fd0d741168c3296f2c71426374
Martin Martin [Tue, 2 Feb 2016 22:12:48 +0000 (14:12 -0800)]
Add some quick comments about folly::fiber implementation classes.
Summary: Add some quick comments about folly::fiber implementation classes.
Reviewed By: yfeldblum
Differential Revision:
D2891940
fb-gh-sync-id:
b9899bb63df03f32763f93a5dae8dff94c8e354e
Giuseppe Ottaviano [Tue, 2 Feb 2016 18:27:10 +0000 (10:27 -0800)]
Kill writeTerminator in fbstring
Summary:
`writeTerminator` needs to determine the string category, and in the small case, also its size. This information is often known, but the compiler cannot optimize it away especially if `capacity_` was just overwritten. Also, some occurrences are just redundant. We can remove the latter and specialize the former.
Small operations such as `push_back` are up to 40% faster.
Before/after:
```
$ ./fbstring_benchmark_using_jemalloc --bm_regex '_fbstring' --bm_min_usec 200000 --bm_max_secs 2
============================================================================ ===================
./folly/test/FBStringTestBenchmarks.cpp.h relative time/iter iters/s time/iter iters/s
============================================================================ ===================
BM_initRNG_fbstring(0) 0.00fs Infinity 0.00fs Infinity
BM_defaultCtor_fbstring(0) 6.82us 146.60K 6.26us 159.74K
BM_copyCtor_fbstring(32768) 16.86ns 59.32M 15.50ns 64.52M
BM_ctorFromArray_fbstring(32768) 2.25us 444.78K 2.12us 471.58K
BM_ctorFromTwoPointers_fbstring(0) 9.76ns 102.44M 7.85ns 127.35M
BM_ctorFromTwoPointers_fbstring(7) 9.68ns 103.32M 8.10ns 123.47M
BM_ctorFromTwoPointers_fbstring(15) 9.77ns 102.34M 8.12ns 123.17M
BM_ctorFromTwoPointers_fbstring(23) 9.46ns 105.68M 8.78ns 113.88M
BM_ctorFromTwoPointers_fbstring(24) 31.23ns 32.02M 30.71ns 32.56M
BM_ctorFromChar_fbstring(
1048576) 40.04ns 24.97M 35.68ns 28.03M
BM_assignmentOp_fbstring(256) 66.76ns 14.98M 63.93ns 15.64M
BM_assignmentFill_fbstring(0) 8.23ns 121.47M 7.02ns 142.49M
BM_resize_fbstring(524288) 4.09us 244.63K 3.94us 253.84K
BM_equality_fbstring(65536) 2.47ms 404.63 2.19ms 455.92
BM_replace_fbstring(256) 172.36ns 5.80M 159.10ns 6.29M
BM_push_back_fbstring(1) 9.32ns 107.27M 7.79ns 128.40M
BM_push_back_fbstring(23) 252.44ns 3.96M 158.31ns 6.32M
BM_push_back_fbstring(127) 721.50ns 1.39M 515.08ns 1.94M
BM_push_back_fbstring(1024) 5.21us 191.87K 3.14us 318.03K
BM_short_append_fbstring(23) 431.71ns 2.32M 335.74ns 2.98M
BM_short_append_fbstring(1024) 11.96us 83.64K 9.19us 108.78K
BM_getline_fbstring(23) 57.06ns 17.53M 39.78ns 25.14M
BM_getline_fbstring(1000) 232.26ns 4.31M 203.24ns 4.92M
============================================================================ ===================
```
(`find` benchmarks were removed due to high variance caused by randomness)
Reviewed By: luciang
Differential Revision:
D2879646
fb-gh-sync-id:
10fd8573495d285220ae98858d11de9ec6d97e54
Orvid King [Tue, 2 Feb 2016 17:49:08 +0000 (09:49 -0800)]
Allow building without JEMalloc under MSVC
Summary: This was falling back to a hard depency on JEMalloc, and MSVC doesn't have proper weak linkage, so emulate it with some linker magic to get things building.
Reviewed By: yfeldblum
Differential Revision:
D2887695
Pulled By: Orvid
fb-gh-sync-id:
50c812a07c4e4e5bbe71263b86386783bf76ba82
Lucian Grijincu [Tue, 2 Feb 2016 01:00:24 +0000 (17:00 -0800)]
folly: #define UNDEFINED_SANITIZER in ubsan mode
Summary:
Undefined Sanitizer doesn't define any macro to detect that it's active.
The build system should provide `-DUNDEFINED_SANITIZER` when building
with -fsanitize=undefined or any of the other flags from
http://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html
FWIW: Chrome defines the same preprocessor symbol:
https://chromium.googlesource.com/chromium/src/+/
ac18f489dca8c902e4dfaa1a28d716b7914121d0%5E%21/build/common.gypi
Reviewed By: andrewjcg
Differential Revision:
D2885167
fb-gh-sync-id:
e1129c0863bfde5d032c32e7d5cea7c43d82009f
Dmitry Pleshkov [Mon, 1 Feb 2016 23:35:18 +0000 (15:35 -0800)]
Implementing callback functionality for exception routines
Summary: Depends on
D2865911
Reviewed By: luciang
Differential Revision:
D2742158
fb-gh-sync-id:
3e7866a742575ee4f7501cff0abbd5c21e26a46e
Lucian Grijincu [Mon, 1 Feb 2016 23:19:55 +0000 (15:19 -0800)]
folly: ubsan: Benchmark: avoid division by zero (fsanitize=float-divide-by-zero)
Reviewed By: philippv
Differential Revision:
D2886132
fb-gh-sync-id:
a845d657fb920847df52fe5ec077e91554544f5b
Igor Sugak [Mon, 1 Feb 2016 22:35:10 +0000 (14:35 -0800)]
folly: fix one more -Wunused-parameter
Summary: The usage is guarded by a preprocessor macro, and I didn't noticed this earlier.
Reviewed By: yfeldblum
Differential Revision:
D2887016
fb-gh-sync-id:
791c4d16475aab77235792953997a281354018e9
Scott Wolchok [Mon, 1 Feb 2016 21:17:55 +0000 (13:17 -0800)]
Remove unnecessary includes from Optional.h
Summary:
I want to be able to include this in
fbobjc/xplat/folly:headers_only, but it currently pulls in a boost
header and Portability.h. Fortunately, neither of those includes are
needed.
Reviewed By: ddrcoder, yfeldblum, mhorowitz
Differential Revision:
D2880443
fb-gh-sync-id:
409561b8fb555ac3946d59cd6657c6c643b67c86
Christopher Dykes [Mon, 1 Feb 2016 20:36:43 +0000 (12:36 -0800)]
Support constexpr_strlen under MSVC.
Summary: MSVC doesn't support evaluating strlen at compile time, so implement our own version instead.
Reviewed By: yfeldblum, lbrandy
Differential Revision:
D2856926
fb-gh-sync-id:
22222350b57d9eff6a06c9d0f37d43a3cb1f2534
Lucian Grijincu [Mon, 1 Feb 2016 19:51:51 +0000 (11:51 -0800)]
folly: Hash: use loadUnassigned to avoid UB in hsieh_hash32_buf (fsanitize=alignment)
Reviewed By: philippv
Differential Revision:
D2886152
fb-gh-sync-id:
64c3543db831c72c7f4cf307867223270293066e
Lucian Grijincu [Mon, 1 Feb 2016 19:45:53 +0000 (11:45 -0800)]
folly: ubsan: HashTest: avoid invalid shift (sanitize=shift)
Summary:
[ RUN ] Hash.TWang_Unmix64
folly/test/HashTest.cpp:125:20: runtime error: shift exponent 32 is too large for 32-bit type 'unsigned int'
Reviewed By: philippv
Differential Revision:
D2886144
fb-gh-sync-id:
8d7963c087c9db34b08c07451d35e5568c750520
Igor Sugak [Mon, 1 Feb 2016 18:44:42 +0000 (10:44 -0800)]
folly: fix clang -Wmissing-field-initializers
Summary: Fix a few `-Wmissing-field-initializers` exposed by clang.
Reviewed By: yfeldblum
Differential Revision:
D2881902
fb-gh-sync-id:
67fdffd39d3ccca64f84055adae1f3b47fdec633
Neel Goyal [Mon, 1 Feb 2016 13:52:01 +0000 (05:52 -0800)]
Comments for SSLVerifyPeerEnum
Summary:
Document what the enum settings mean since they can be
somewhat confusing.
Reviewed By: shamdor-fb
Differential Revision:
D2882929
fb-gh-sync-id:
74ec30132bf5d2dce42f51a0b7b30cf2fae12dbf
Igor Sugak [Fri, 29 Jan 2016 23:53:26 +0000 (15:53 -0800)]
folly: fix -Wunused-parameter in opt mode
Summary: Fix a few remaining unused parameters, that are exposed only in opt build.
Reviewed By: markisaa
Differential Revision:
D2878865
fb-gh-sync-id:
d0d9761362220973cda14d99ab7342fbe8b1a469
Igor Sugak [Fri, 29 Jan 2016 23:33:50 +0000 (15:33 -0800)]
folly: fix clang build with -Wunused-const-variable
Summary: Fix a few unused const variables exposed by clang's `-Wunused-const-variable`.
Reviewed By: yfeldblum
Differential Revision:
D2878944
fb-gh-sync-id:
f5500fda4782eac2964761c0398d016d57716269
Orvid King [Fri, 29 Jan 2016 22:37:19 +0000 (14:37 -0800)]
Adjust AsyncServerSocket to not use getsockname before it's connected
Summary:
Winsock doesn't like it when you try to call getsockname on a socket that hasn't yet been connected or bound, because that socket doesn't have a name yet.
This only occurred because we were trying to get the family of the socket.
To solve this, I just passed the family in from the parent methods that already knew what the family was.
Reviewed By: yfeldblum
Differential Revision:
D2871859
Pulled By: Orvid
fb-gh-sync-id:
7674f565a968aa0258355fc977c185a416e4fbe4
Michael Lee [Fri, 29 Jan 2016 18:32:31 +0000 (10:32 -0800)]
Handle tmp dir on Android
Summary:
There is no universally accessible tmpdir on Android which
means mkstemp fails. This diff calls a function that the test runner
should provide to fill a valid temporary directory.
Reviewed By: yangchi
Differential Revision:
D2842034
fb-gh-sync-id:
9b826757bd750af016a18adccd5a21174be644d6
Denis Samoylov [Fri, 29 Jan 2016 18:18:01 +0000 (10:18 -0800)]
Make advanced init function for test optional
Summary: In order to avoid mandatory dependncy on elf, dwarf, unwind made new init function optional
Reviewed By: markisaa
Differential Revision:
D2878545
fb-gh-sync-id:
f66d3884a531dcf56fc1432330325ab45b149d7f
Michael Lee [Fri, 29 Jan 2016 15:23:19 +0000 (07:23 -0800)]
Exclude based on __APPLE__
Summary: This was showing up on a Linux compile to my sadness.
Reviewed By: francis-ma
Differential Revision:
D2876010
fb-gh-sync-id:
f638d593712c0aeb50177d96aaacb26575820359
Giuseppe Ottaviano [Fri, 29 Jan 2016 15:02:58 +0000 (07:02 -0800)]
Optimize getline(istream&, fbstring&) implementation
Summary:
Current `getline` implementation in `fbstring` always allocates, even if the passed string is already large enough. Furthermore, the growing strategy relies on outdated assumptions about the allocator.
This implementation reuses the existing allocation as much as possible, then uses exponential growth.
Reviewed By: luciang, philippv
Differential Revision:
D2871976
fb-gh-sync-id:
8db9512030be3f4953efa8f008747827504c032c
Denis Samoylov [Thu, 28 Jan 2016 19:18:54 +0000 (11:18 -0800)]
A common init function for binaries and a default main function for tests
Summary: Added initialization routines to test main function that can help with debugging tests
Reviewed By: markisaa, yfeldblum
Differential Revision:
D2839759
fb-gh-sync-id:
71cad45f3747336c8c7f8706db139cd060e1442b
Lucian Grijincu [Thu, 28 Jan 2016 19:07:44 +0000 (11:07 -0800)]
folly: ubsan: replace undefined call through reinterpret-cast fn-pointer with std::function
Summary:
This code casts function pointers to void(*fn)(void*) and calls
functions through that type. Calling functions through a different
pointer type is undefined behavior. Casts were used to avoid memory
allocations.
`std::bind` needs a memory allocation - so it's avoided in EventBase.
std::function<void()> x = std::bind(fn, args);
`std::function` should use small object optimizations when possible and embed the functor in the body of `std::function`.
On GCC 5.0 and forward small lambdas don't need memory allocations
(lambdas being tiny structures - two pointers in this case - and a
function operator).
std::function<void()> y = [=] { fn(args); };
On GCC 4.7 .. 4.9 functors for which __is_location_invariant<Func> is
true also don't need memory allocations.
Remove undefined behavior by using a `SmallFunctor` which leverages `std::function`'s small object optimization.
Reviewed By: philippv
Differential Revision:
D2864895
fb-gh-sync-id:
ab40f60b3519ce38f43fecebf88ccdbf09d9bea9
Michael Lee [Thu, 28 Jan 2016 18:52:42 +0000 (10:52 -0800)]
Add (void)ssl back to SSLContext for -Wunused-parameter
Summary:
I'm not sure what accepted practice is to make the
-Wunused-parameter not balk.
Reviewed By: igorsugak
Differential Revision:
D2874705
fb-gh-sync-id:
8765bae504fd90dfd8896857f4bf865ca8f64a8a
Marcus Holland-Moritz [Thu, 28 Jan 2016 16:05:32 +0000 (08:05 -0800)]
Fix unused parameter errors under -Werror
Summary:
D2872406 enables -Werror=unused-parameter, which in conjunction with mode/opt
(which eliminates asserts) causes two files in folly/io/async to fail to compile.
This change works around the error.
Reviewed By: vchalyshev
Differential Revision:
D2874625
fb-gh-sync-id:
97104679f964390c5df88ee7831af7df243a152a
Igor Sugak [Thu, 28 Jan 2016 03:34:02 +0000 (19:34 -0800)]
folly: build with -Wunused-parameter
Summary: Mechanical changes (using custom clang-tidy) to fix all of the `-Wunused-parameter` violations in folly.
Reviewed By: yfeldblum
Differential Revision:
D2872406
fb-gh-sync-id:
bdb1941f3dadf6ab854e7a9f271f50fda93f9480
Igor Sugak [Thu, 28 Jan 2016 03:32:21 +0000 (19:32 -0800)]
folly/test/FBVectorTestBenchmarks.cpp.h: remove unused argument
Summary: Remove unused argument `iters`.
Reviewed By: meyering
Differential Revision:
D2872551
fb-gh-sync-id:
0989e5b256d4d80dd3ad02401e3fc65d0f749ba6
Francis Ma [Thu, 28 Jan 2016 00:33:36 +0000 (16:33 -0800)]
Fix FOLLY_TLS under macosx and clang
Summary: Only include guard iphone simulator and iphone devices for folly_tls
Reviewed By: ldemailly
Differential Revision:
D2872383
fb-gh-sync-id:
00fb8c1ee03a97037e92d20aeda75f2435d71f5a
Igor Sugak [Thu, 28 Jan 2016 00:12:47 +0000 (16:12 -0800)]
folly: use -Wheader-hygiene with clang
Summary: `-Wheader-hygiene` warns on using namespace directive in global context in header. Fix all of the violations.
Reviewed By: yfeldblum
Differential Revision:
D2867655
fb-gh-sync-id:
46840f8ece99e7af262058e631635d870bd51149
Andrii Grynenko [Thu, 28 Jan 2016 00:08:08 +0000 (16:08 -0800)]
Fix EventBase destruction race in FiberManagerMap
Summary:
Previously we could be reading from thread-local FiberManagerMap while it was modified.
This is now fixed by keeping a per-thread list of EventBases which need to be removed from local maps. On the fast-path no action is taken, since list will be empty.
Reviewed By: yfeldblum
Differential Revision:
D2853921
fb-gh-sync-id:
f05e1924dd2b97bfb359537de1909bbe193e0cb9
Francis Ma [Wed, 27 Jan 2016 22:37:05 +0000 (14:37 -0800)]
Make folly::detail::CacheLocality portable on apple
Summary:
This is one of the series steps to port folly::future on ios. Apple doesn't support __thread. Adding a HashingThreadId as a fallback
on apple.
Reviewed By: nbronson
Differential Revision:
D2832068
fb-gh-sync-id:
c3389245f3c0bbd36de6260680f7ac6110b3206c
Lucian Grijincu [Wed, 27 Jan 2016 22:33:34 +0000 (14:33 -0800)]
folly: ubsan: reduce vector size to avoid UBSAN timeout
Summary:
Based on diff where this was introduced in {
D360195} it seems like
```
// This value should we multiple of word size.
static size_t const kHeapifyCapacitySize = sizeof(
typename std::aligned_storage<
sizeof(InternalSizeType),
alignof(value_type)
>::type);
// Threshold to control capacity heapifying.
static size_t const kHeapifyCapacityThreshold =
100 * kHeapifyCapacitySize;
```
So anything above 100*sizeof(SizeType) should do.
Reviewed By: philippv
Differential Revision:
D2871422
fb-gh-sync-id:
a69e47286c53887ac05e89dab565b9d609e183a0
Orvid King [Wed, 27 Jan 2016 21:57:12 +0000 (13:57 -0800)]
Implement a generalized mechanism for pushing/popping warnings
Summary: Folly synchronized requires disabling the shadow warning in a macro, but that doesn't work under MSVC, so abstract a mechansim out that allows them to be handled gracefully.
Reviewed By: yfeldblum
Differential Revision:
D2870357
Pulled By: Orvid
fb-gh-sync-id:
a4b0e425736ddd5293f020b360244554571d397f
Maxim Georgiev [Wed, 27 Jan 2016 21:08:50 +0000 (13:08 -0800)]
Add mechanizm for caching local and peer addresses in AsyncSSLSocket.
Summary: This change adds a flag to AsyncSSLSocket which forces the socket to cache local and remote addresses right after TCP connection is established. Cached address values will be available after the connection is closed. Caching addresses can be halpful in SSL handshake failure investigations.
Reviewed By: yfeldblum
Differential Revision:
D2863274
fb-gh-sync-id:
d7b415292988c2fb187a80422e8ccbf8ba8ab0e3
Giuseppe Ottaviano [Wed, 27 Jan 2016 09:54:59 +0000 (01:54 -0800)]
Fix typo in comment
Reviewed By: yfeldblum
Differential Revision:
D2869494
fb-gh-sync-id:
889957a92dd7f59c4b9564d1946e9f4058293839
Yedidya Feldblum [Wed, 27 Jan 2016 07:18:09 +0000 (23:18 -0800)]
Extract endianness checks into Portability.h
Summary: [Folly] Extract endianness checks into Portability.h.
Reviewed By: fugalh
Differential Revision:
D2857924
fb-gh-sync-id:
23ecd2a3cad661024acb62769cd85df394786c59
Christopher Dykes [Tue, 26 Jan 2016 16:59:32 +0000 (08:59 -0800)]
Use FOLLY_DEPRECATED rather than directly using GCC specific attributes.
Summary: Without this MSVC can't compile.
Reviewed By: lbrandy
Differential Revision:
D2856598
fb-gh-sync-id:
0e146afe844b0ce5d3782528ed9c3de53f7c8b05
Arjen Roodselaar [Tue, 26 Jan 2016 01:39:11 +0000 (17:39 -0800)]
Debian DEBs build script
Summary:
This scripts uses fpm to build a somewhat reproducible set of debs which can be deployed to Ubuntu 14.04 hosts. The main library package carries the Folly major version number, allowing multiple versions to be installed alongside. The -dev deb is simply called libfolly-dev and will be upgraded as Folly moves forward. In accordance to the Debian packaging policies the shared libraries have their (debug) symbols stripped and saved to external symbol files, contained in the -dev deb.
sgolemon, yfeldblum you guys are my best guess to review this. Feel free to suggest additional folks if needed.
Reviewed By: yfeldblum
Differential Revision:
D2806082
fb-gh-sync-id:
42605acccdec781f7a6b59a925121e6ed7c7cdf5
Christopher Dykes [Mon, 25 Jan 2016 20:21:24 +0000 (12:21 -0800)]
Initialize LifoSem's padding to allow for its constexpr constructor.
Summary: MSVC correctly gives an error about the constexpr constructor not initializing all members.
Reviewed By: yfeldblum
Differential Revision:
D2856806
fb-gh-sync-id:
cef97639906dd3c39e3d3dc2ba939021e15edcb9
Michael Bejda [Mon, 25 Jan 2016 17:45:49 +0000 (09:45 -0800)]
Thread-safe RequestContext putIfAbsent operation
Summary:
Adds a thread-safe putIfAbsent operation to the RequestContext. The current setContextData() is not sufficent to do it safely.
Just like setContextData, this method is unfair, as a high volume of reads will block the spinlock.
Reviewed By: fugalh
Differential Revision:
D2850752
fb-gh-sync-id:
2ff22ea9e9bd8f27f6ae7a57214a6dbc4fdcd4c5
David Callahan [Fri, 22 Jan 2016 21:34:32 +0000 (13:34 -0800)]
avoid aggressive optimization
Summary: GCC will now dead-code eliminate the folly-based version of this test without a mechanisms to force the result to be live.
Reviewed By: ttsugriy
Differential Revision:
D2854633
fb-gh-sync-id:
0e3841ed22c040fda7653bcfb5a3f19ca3d1f835
Tom Jackson [Thu, 21 Jan 2016 22:19:37 +0000 (14:19 -0800)]
Add PrintTo for dynamic
Summary: Making `EXPECT_EQ(dyn1, dyn2)` easier to debug
Reviewed By: luciang, ot, yfeldblum
Differential Revision:
D2848318
fb-gh-sync-id:
0c7cdd292665a493f2b792798df4e6966c1f28db
Kyle Nekritz [Thu, 21 Jan 2016 21:54:28 +0000 (13:54 -0800)]
Add getPeerCert() to AsyncTransport.
Reviewed By: elindsey
Differential Revision:
D2850760
fb-gh-sync-id:
60dbc3117e658d2fd083a87884892924bf313019
Kyle Nekritz [Thu, 21 Jan 2016 21:54:26 +0000 (13:54 -0800)]
Adding OpenSSLPtrTypes.h.
Summary:
So that these deleter and unique_ptr types don't have to be redeclared every single place they are used.
To be expanded on.
Reviewed By: mzlee
Differential Revision:
D2850376
fb-gh-sync-id:
e7f8bba320163b8b12a93b5cf3cd9a5921d38edc
Dmitry Pleshkov [Thu, 21 Jan 2016 06:23:30 +0000 (22:23 -0800)]
Haswell-specific implementation of Select64 routine.
Summary: k-th bit selection could be efficiently implemented via new BMI2 instruction set.
Reviewed By: ot, philippv
Differential Revision:
D2843311
fb-gh-sync-id:
4c0cf52176a03422aef276ce5f677080f67f5fdf
Michael Lee [Wed, 20 Jan 2016 15:10:31 +0000 (07:10 -0800)]
Move StringTest benchmarks into StringBenchmark.cpp
Summary: StringTest.cpp is a mix of benchmarks and normal unittests. Fixing this.
Reviewed By: yfeldblum
Differential Revision:
D2840804
fb-gh-sync-id:
d3efc357f5f09385e9f69b70e38b64d59045ff0d
Michael Lee [Wed, 20 Jan 2016 15:09:04 +0000 (07:09 -0800)]
Remove unecessary main functions.
Summary: Starting with this, but I will keep digging through these. The tests will be compiled into one combined test with a single main.
Reviewed By: yfeldblum
Differential Revision:
D2841391
fb-gh-sync-id:
78fd153e282f1ca2dbe7ada942dc04fc8ba5d42d
Alan Frindell [Tue, 19 Jan 2016 22:21:15 +0000 (14:21 -0800)]
Add Range::erase
Summary: Needed to use StringPiece with boost::algorithm::trim()
Reviewed By: yfeldblum
Differential Revision:
D2833657
fb-gh-sync-id:
3430b1a2540279b2f69f04c871df3bca748f2cb1
Alexey Spiridonov [Fri, 15 Jan 2016 20:53:27 +0000 (12:53 -0800)]
Print null correctly
Summary: Null by itself was printing as 0, now it prints as null, which is consistent with the 'pseudo json' output.
Reviewed By: yfeldblum
Differential Revision:
D2789284
fb-gh-sync-id:
f318b8d0f8349f4b36f868c419842fb50bee9517
Aaron Balsara [Fri, 15 Jan 2016 20:02:40 +0000 (12:02 -0800)]
Fix buck build for SSLContext
Summary:
D2800746 broke buck with an unsigned/signed compare
Reviewed By: dkgi
Differential Revision:
D2835102
fb-gh-sync-id:
a0b8311b38a199e089d3ed5a69b12f8f8abe38b1
Aaron Balsara [Fri, 15 Jan 2016 18:50:33 +0000 (10:50 -0800)]
Allow SSLContext to read certificates and keys from memory
Summary: Added the ability for SSLContext to load X509 Certificates and private keys from memory
Reviewed By: yfeldblum
Differential Revision:
D2800746
fb-gh-sync-id:
14cad74f8d761b9b0f07e2827b155cec9ba27f50
Francis Ma [Thu, 14 Jan 2016 22:07:21 +0000 (14:07 -0800)]
Decouple future and fiber for mobile
Summary:
folly::fibers have a bunch of dependencies that are shaky on mobiles.
Decoupling it by using folly::Baton instead of folly::fibers::Baton. Should have
zero effect on fbcode cases.
Reviewed By: djwatson, yfeldblum
Differential Revision:
D2821603
fb-gh-sync-id:
0ce3472c9eedf97224e8584d25e04515396cd18e
Rachel Kroll [Thu, 14 Jan 2016 01:40:01 +0000 (17:40 -0800)]
match new signal text from prior diff; squash colors (!)
Reviewed By: yfeldblum
Differential Revision:
D2828789
fb-gh-sync-id:
71e45ffe31061220e06d7e7762058f8a1501b5cc
Michael Lee [Wed, 13 Jan 2016 00:22:47 +0000 (16:22 -0800)]
Add a preprocessor guard around __STDC_LIMIT_MACROS in IOBuf.cpp
Summary:
Adding the standard preprocessor guard around the `#define
__STD_LIMIT_MACROS` in IOBuf.cpp.
public
Reviewed By: knekritz
Differential Revision:
D2825816
fb-gh-sync-id:
191824ec8d1f837e2b8380a2f8451972c421ceaa
Michael Lee [Wed, 13 Jan 2016 00:22:02 +0000 (16:22 -0800)]
Use PRIuMAX for Android format strings.
Summary: Provide a more Android compatible version of sformat.
Reviewed By: sgolemon
Differential Revision:
D2825105
fb-gh-sync-id:
ce328d17b9f8008d81bc7dd5bf7f0905e560dfe1
Rachel Kroll [Tue, 12 Jan 2016 22:57:55 +0000 (14:57 -0800)]
Log pid/uid of sending process in signal handler, too
Reviewed By: yfeldblum
Differential Revision:
D2816450
fb-gh-sync-id:
3d3b0bf20a3f20570f847e3c5ec91a5e3786a370
Kyle Nekritz [Mon, 11 Jan 2016 00:33:45 +0000 (16:33 -0800)]
Add security protocol trace event.
Reviewed By: siyengar
Differential Revision:
D2805183
fb-gh-sync-id:
9ed320dd3762658e993daa658100d0c3c617d210
Yedidya Feldblum [Sat, 9 Jan 2016 02:40:42 +0000 (18:40 -0800)]
SingeltonVault::ScopedExpunger - RAII class to clear singletons
Summary:
[Folly] `SingeltonVault::ScopedExpunger` - RAII class to clear singletons.
Clears all singletons in the given vault at ctor and dtor times. Useful for unit-tests that need to clear the world.
This need can arise when a unit-test needs to swap out an object used by a singleton for a test-double, but the singleton needing its dependency to be swapped has a type or a tag local to some other translation unit and unavailable in the current translation unit.
Other, better approaches to this need are "plz 2 refactor" ....
Reviewed By: andriigrynenko
Differential Revision:
D2802459
fb-gh-sync-id:
c24cebd3a464ed5da29ea1d9e7b86c51d61cf631
Andrii Grynenko [Sat, 9 Jan 2016 01:15:21 +0000 (17:15 -0800)]
Fix Python FiberTask to properly init/reset Python thread state
Reviewed By: alikhtarov
Differential Revision:
D2813735
fb-gh-sync-id:
ab3c3e18618ed8bf15f478bcfca008786834e65c
Pallab Bhattacharya [Fri, 8 Jan 2016 04:45:32 +0000 (20:45 -0800)]
folly::Symbolizer can be enhanced for dumpStackTrace
Summary:
added functionality in Symbolizer.cpp to handle multi-segment binary and to allow symbol lookup via /proc/self/exe when some or whole text is relocated to ANON pages such as during hugification. An example of multi-segment binary : P56071432
cc markw65 - a part of the change include logic pulled from hphp/runtime/base/stack-logger.cpp:symbolize_huge_text so that hugified consumers of dumpStackTrace need not fork off.
Reviewed By: edwardc
Differential Revision:
D2802837
fb-gh-sync-id:
577ab1b4ef8f22059894bfdd9c0526a22ee89ca8
Giuseppe Ottaviano [Fri, 8 Jan 2016 01:34:42 +0000 (17:34 -0800)]
Fix two fbstring operator+ overloads
Summary: `insert()` returns `fbstring` in most cases, but `iterator` (that is, `value_type*`) when the first argument is an iterator. Two overloads of `operator+` used `insert` as if it returned `fbstring`, which by chance works anyway unless the resulting string contains a `'\0'` (plus it does an extra string copy). This diff fixes the bug.
Reviewed By: philippv, luciang, Gownta
Differential Revision:
D2813713
fb-gh-sync-id:
015188b72813da2dabe23980f50f00832d62aa14
David Callahan [Wed, 6 Jan 2016 04:27:15 +0000 (20:27 -0800)]
parse args for -json flag
Reviewed By: yfeldblum
Differential Revision:
D2803911
fb-gh-sync-id:
65023cae7fd8e06c30fed100826b4b834ee2e9b1
Gustavo Serra Scalet [Tue, 5 Jan 2016 20:18:23 +0000 (12:18 -0800)]
Enable GroupVarint on PPC64
Summary:
This PR is necessary for both Folly and HHVM to compile on PPC64 (https://github.com/PPC64/hhvm) and other platforms that are not compatible with SSE instructions.
It also removes GroupVarint32 tables generator dependency on x86 platform.
Closes https://github.com/facebook/folly/pull/339
Reviewed By: mxw
Differential Revision:
D2760156
fb-gh-sync-id:
b4e93b54b62d6f68ccf684d34840678f677bf276
David Callahan [Tue, 5 Jan 2016 19:38:11 +0000 (11:38 -0800)]
parse args for -json flag
Reviewed By: lbrandy
Differential Revision:
D2803778
fb-gh-sync-id:
0cb21413bb1c78d256b4408322f8d585f53cb0ac
Neel Goyal [Tue, 5 Jan 2016 17:33:55 +0000 (09:33 -0800)]
Add canAdvance to cursor and tests
Summary:
Determine if the cursor can advance N bytes. This is useful if
applications want to check before reading so an exception isn't thrown.
It tries to walk the minimal amount of links needed in the chain.
I had a task that could have used this, though caching totalLength and
macro magic ended up being the implementation chosen. I think this just
adds to the cursor API.
Reviewed By: djwatson
Differential Revision:
D2728498
fb-gh-sync-id:
8657653b82a48828cccab143653dc169ef715702
Michael Lee [Mon, 4 Jan 2016 23:35:02 +0000 (15:35 -0800)]
Switch back to SYS_gettid, and fix
Summary:
SYS_gettid is different on Linux vs. OSX. `__NR_gettid` is
only sometimes present and `SYS_gettid` is only sometimes present, but
we can pick one name and just follow that one.
Reviewed By: dcolascione
Differential Revision:
D2800515
fb-gh-sync-id:
4245de4b9184ac4233ade9da297409c1031869a3
Michael Lee [Mon, 4 Jan 2016 17:42:21 +0000 (09:42 -0800)]
Clean up folly tests.
Summary: Clean up and remove unecessary gflags use.
Reviewed By: yfeldblum
Differential Revision:
D2795904
fb-gh-sync-id:
99cccb4dc32a051b3d552b72cbc9243e20ba8127
Giuseppe Ottaviano [Wed, 30 Dec 2015 20:41:32 +0000 (12:41 -0800)]
Close AsyncServerSocket sockets in reverse order
Summary:
When a large number of processes concurrently bind and close `AsyncServerSocket`s with `port=0` (for example in tests) binding the IPv4 socket on the same port bound with the IPv6 socket can currently fail, because sockets are closed in the same order as they are opened, which makes the following scenario possible:
```
P0: close IPv6 port
P1: open IPv6 port (succeed)
P1: open IPv4 port (FAIL)
P0: close IPv4 port
```
This diff reverses the closing order, and also fixes a couple of outdated comments.
Reviewed By: philippv
Differential Revision:
D2795920
fb-gh-sync-id:
0b5157a56dfed84aed4a590c103050a2727847b3
Samrat Bhattacharya [Mon, 28 Dec 2015 23:03:16 +0000 (15:03 -0800)]
Add Notification Queue Size in verbose logs of EventBase
Summary: As above
Reviewed By: alandau
Differential Revision:
D2791829
fb-gh-sync-id:
8ae5b875464dfa910b88dc12465879a9420d398c
Dominik Gabi [Wed, 23 Dec 2015 23:43:26 +0000 (15:43 -0800)]
wrapper for nullable attribute
Reviewed By: yfeldblum
Differential Revision:
D2784472
fb-gh-sync-id:
84c7426cc82edabb04c662fa699764ffc0864b7e
Dominik Gabi [Wed, 23 Dec 2015 23:43:24 +0000 (15:43 -0800)]
clang attribute wrappers
Summary: adding GCC compatible wrappers for clang attributes to folly.
Reviewed By: yfeldblum
Differential Revision:
D2706913
fb-gh-sync-id:
17db34089d48a6a01c20904321a47ff98141c2fe
Michael Lee [Wed, 23 Dec 2015 20:38:29 +0000 (12:38 -0800)]
Change `SYS_gettid` to __NR_gettid instead
Summary: SYS_gettid is not available on all platforms, but __NR_gettid should be.
Reviewed By: dcolascione
Differential Revision:
D2787003
fb-gh-sync-id:
ee024437ac95281a3573e2440653847f6f7d1738
Giuseppe Ottaviano [Wed, 23 Dec 2015 20:07:16 +0000 (12:07 -0800)]
Fix test Dynamic.ArrayReassignment with Clang 3.7
Summary:
`dynamic` is one of the few unfortunate recursive types that have a constructor of the form `T(initializer_list<T>)`. After the Defect Report 95 (http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#1467) a statement `T t = {v};` where `v` is of type `T` invokes the copy constructor instead of the initializer list constructor. Clang 3.7 implements the new behavior, and GCC probably will soon.
This diff only fixes a test where the ambiguous syntax is used, but a better solution would be preferable.
Reviewed By: philippv
Differential Revision:
D2786455
fb-gh-sync-id:
7fa5235e4041cbc8dd3ce685b5a0c23c343f78d2
Mike Kolupaev [Wed, 23 Dec 2015 10:05:01 +0000 (02:05 -0800)]
Added missing instantiation for HistogramBuckets::computeTotalCount()
Summary:
D2078239 added a template method to HistogramBuckets but didn't add an instantiation for it, similar to getPercentileBucketIdx() and getPercentileEstimate() (see comment in Instantiations.cpp). This diff adds it, making `computeTotalCount()` usable without including `Histogram-defs.h`.
Also removes the weird `const` in return type.
Reviewed By: simpkins
Differential Revision:
D2783534
fb-gh-sync-id:
9226489820116e0cbcb1f6a631b389439558061e
Michael Lee [Tue, 22 Dec 2015 23:48:46 +0000 (15:48 -0800)]
Remove extra FLAGS_seed definition
Summary: This is not used within the file. Cleaning up.
Reviewed By: yfeldblum
Differential Revision:
D2784500
fb-gh-sync-id:
60270cb20c40b79f988c5536fa00e0a0f1d08e46
Michael Lee [Mon, 21 Dec 2015 15:55:45 +0000 (07:55 -0800)]
Move ThreadLocal.h into the implementation
Summary: Move ThreadLocal.h out of Random.h so we don't over include.
Reviewed By: yfeldblum
Differential Revision:
D2770085
fb-gh-sync-id:
e6934d606236c2b5cdde728639f372641ee716dc
Kosov Eugene [Fri, 18 Dec 2015 18:17:19 +0000 (10:17 -0800)]
fixed build with g++ 5+ and ASAN enabled
Summary:
Fixing by deleting unnecessary macro checks - folly already requires at least g++4.8
Closes https://github.com/facebook/folly/pull/338
Reviewed By: yfeldblum
Differential Revision:
D2742024
Pulled By: fredemmott
fb-gh-sync-id:
c0ecc7edceeebd1f972b0be4f4740214211f1c24
Andrii Grynenko [Thu, 17 Dec 2015 21:03:15 +0000 (13:03 -0800)]
Fix EventBaseLoopController destruction races
Summary:
Existing scheduleThreadSafe implementation had 2 potential races on destruction:
1. (very unlikely) insertHead is complete, but fiber loop is already running on another thread, so it finishes processing all of the fibers, destroys FiberManager or EventBase or both. By the time we get to scheduleThreadSafe EventBaseLoopController is already destoyed
2. (more likely) scheduleThreadSafe is complete, but FiberManager loop which is already running, picks complete fiber, finishes the processing. After that FiberManager may be destoyed. So when EventBase actually executes the callback FiberManager is already dead.
This solution fixes both races. Holding the alive shared_ptr when completing sheduleThreadSafe assures EventBase can't be destoyed until its completed (or it won't try to schedule anything after EventBase was destroyed). Locking alive weak_ptr in the EventBase loop callback ensures FiberManager and thus EventBaseLoopController were not destroyed yet (they can be destoyed only by the same thread which is running EventBase loop).
Reviewed By: spalamarchuk
Differential Revision:
D2763206
fb-gh-sync-id:
1972d6c0c11aa931747ebdaed4029a209130f69c
Michael Lee [Thu, 17 Dec 2015 18:57:47 +0000 (10:57 -0800)]
Allow override for unaligned reads in SpookyHashV2
Summary: Add configuration for disabling unaligned reads in SpookyHashV2
Reviewed By: yfeldblum
Differential Revision:
D2750885
fb-gh-sync-id:
9f48dcdfd5af05478a38e354f6fa0332b1332c14