Andrii Grynenko [Thu, 5 May 2016 23:41:33 +0000 (16:41 -0700)]
Allow optional atfork hook for singleton destruction
Summary: Grouping all folly::ThreadLocal atfork hooks in one. This allows proper ordering between folly::ThreadLocal hooks and hooks added by other libraries (we always want folly::ThreadLocal hooks to be registered first).
Reviewed By: ericniebler
Differential Revision:
D3262666
fb-gh-sync-id:
1aa928b9ddd484580f4a4a7b320e9e64164bb04a
fbshipit-source-id:
1aa928b9ddd484580f4a4a7b320e9e64164bb04a
Andrii Grynenko [Thu, 5 May 2016 20:22:42 +0000 (13:22 -0700)]
Implement LoopKeepAlive for EventBase
Summary: LoopKeepAlive can be useful to tell EventBase, that loop() shouldn't return even if there are no events registered, because some tasks will later be added via runInEventBaseThread. While at least one LoopKeepAlive is alive - EventBase::loop() behaves like EventBase::loopForever().
Reviewed By: yfeldblum
Differential Revision:
D3261706
fb-gh-sync-id:
d91424d3d12cae11abd72cffdbd57f136f628dae
fbshipit-source-id:
d91424d3d12cae11abd72cffdbd57f136f628dae
Andrii Grynenko [Thu, 5 May 2016 01:15:53 +0000 (18:15 -0700)]
Allow adding tasks to TaskIterator dynamically
Reviewed By: yfeldblum
Differential Revision:
D3244669
fb-gh-sync-id:
73fa4ecb0432a802e67ef922255a896d96f32374
fbshipit-source-id:
73fa4ecb0432a802e67ef922255a896d96f32374
Michael Lee [Thu, 5 May 2016 00:08:47 +0000 (17:08 -0700)]
Switch gflags to portability/GFlags.h
Summary:
There are two places that still includes gflags/gflags.h:
portability/GFlags.h and futures/exercises/Koan.h. Also removing
gflags entirely from a couple of places which never reference it at
all.
Reviewed By: yfeldblum
Differential Revision:
D3203187
fb-gh-sync-id:
e7725fbf2c4b3c475c0035e1b0e8c6e1227f277f
fbshipit-source-id:
e7725fbf2c4b3c475c0035e1b0e8c6e1227f277f
Christopher Dykes [Wed, 4 May 2016 23:45:56 +0000 (16:45 -0700)]
Create the stdlib.h portability header
Summary: Windows doesn't have these functions under these names, so be nice to the code using it.
Reviewed By: yfeldblum
Differential Revision:
D3003641
fb-gh-sync-id:
1657545af6975c96aea4a48c359250828b5921e0
fbshipit-source-id:
1657545af6975c96aea4a48c359250828b5921e0
Christopher Dykes [Wed, 4 May 2016 23:45:53 +0000 (16:45 -0700)]
Create the fcntl.h portability header
Summary: Windows has it, but all it does is define a few constants.
Reviewed By: yfeldblum
Differential Revision:
D3002161
fb-gh-sync-id:
a719faf391125522b2cb21471a56e8a9fe894225
fbshipit-source-id:
a719faf391125522b2cb21471a56e8a9fe894225
Eric Niebler [Wed, 4 May 2016 21:45:20 +0000 (14:45 -0700)]
rearrange folly::Function so that its template arguments are deducable.
Summary:
`folly::Function` was an alias to a more complex type with template arguments that could not be deduced. For example, the call to `foo` below was failing to compile.
```
template <class R, class... As>
void foo(folly::Function<R(As...)> f) {
}
int main() {
foo( folly::Function<void(int)>{ [](int i){} } );
}
```
Rearrange the code so that folly::Function is no longer an alias, thus making its template arguments deducable.
Reviewed By: luciang, spacedentist
Differential Revision:
D3256130
fb-gh-sync-id:
fb403e48d161635b3b7f36e53b1679eb46cbfe7f
fbshipit-source-id:
fb403e48d161635b3b7f36e53b1679eb46cbfe7f
Andrii Grynenko [Wed, 4 May 2016 21:32:54 +0000 (14:32 -0700)]
folly::Future integration with Python
Summary:
This diff implements two functions which can be used to bridge C++ asynchronous APIs with Python.
1. getExecutor(event_loop) allows extracting a folly::Executor-compatible object from Python event_loop.
2. wrapFuture wraps any folly::Future object in a Python Future object.
Reviewed By: ambv
Differential Revision:
D2878731
fb-gh-sync-id:
83628b6b24d6039999f88c5de441e41064b2417e
fbshipit-source-id:
83628b6b24d6039999f88c5de441e41064b2417e
Christopher Dykes [Wed, 4 May 2016 17:29:15 +0000 (10:29 -0700)]
Fix the weak references to JEMalloc under MSVC
Summary: While the previous version did compile, it would generate warnings about symbols being defined in multiple places.
Reviewed By: yfeldblum
Differential Revision:
D3255815
fb-gh-sync-id:
72f5079a3973a04bdc0b469b97ba46181e40353b
fbshipit-source-id:
72f5079a3973a04bdc0b469b97ba46181e40353b
Christopher Dykes [Wed, 4 May 2016 17:28:49 +0000 (10:28 -0700)]
Implement __builtin_ctzll for MSVC
Summary: MSVC doesn't have it, but Folly uses it, so implement it in the Builtins portability header.
Reviewed By: yfeldblum
Differential Revision:
D3256123
fb-gh-sync-id:
fd9ea1b6098d97cf1fde4732905bae9bde8cd8ad
fbshipit-source-id:
fd9ea1b6098d97cf1fde4732905bae9bde8cd8ad
Christopher Dykes [Wed, 4 May 2016 17:28:12 +0000 (10:28 -0700)]
Don't initialize kDestroyWaitTime in-class
Summary: MSVC doesn't like it when you try to initialize a static in-class non-primitive constexpr member. It was only used in one place, so stick the definition in that function instead.
Reviewed By: yfeldblum
Differential Revision:
D3255178
fb-gh-sync-id:
e4a862a2662c4f5376c19fc52827cb2953f01e52
fbshipit-source-id:
e4a862a2662c4f5376c19fc52827cb2953f01e52
Christopher Dykes [Wed, 4 May 2016 17:27:25 +0000 (10:27 -0700)]
Use std::thread::id in Fiber rather than a syscall
Summary: syscall() doesn't exist on Windows, but std::thread::id is standardized, and can do what we need it for, so use it instead.
Reviewed By: yfeldblum
Differential Revision:
D3256064
fb-gh-sync-id:
efddac82c117ccd8a53c8c93248529000b4326a5
fbshipit-source-id:
efddac82c117ccd8a53c8c93248529000b4326a5
Christopher Dykes [Wed, 4 May 2016 17:26:43 +0000 (10:26 -0700)]
Don't attempt to implicitly convert an atomic<T> to bool
Summary: MSVC doesn't like it, so the other uses were changed to .load() explicitly a while back, but this use has been added since then.
Reviewed By: yfeldblum
Differential Revision:
D3255261
fb-gh-sync-id:
11d49902eb3ef7113a7c57cc760353554a487496
fbshipit-source-id:
11d49902eb3ef7113a7c57cc760353554a487496
Marcus Holland-Moritz [Wed, 4 May 2016 10:28:44 +0000 (03:28 -0700)]
Unpoison stack memory before deallocation
Summary:
This is a workaround (maybe even the correct fix if it turns out that ASan
can't detect this FP case) for the problem described in
https://llvm.org/bugs/show_bug.cgi?id=27627
where a memory region previously allocated by a fiber stack can overlap with
the region of an mmap'd file. Accessing parts of the mmap'd file close to
the stack region will trigger a false positive ASan error.
This change makes sure each fiber explicitly unpoisons its stack memory by
calling __asan_unpoison_memory_region in an ASan-enabled build.
Reviewed By: yhfung
Differential Revision:
D3257924
fb-gh-sync-id:
484062e80af67dfd39d2eaf3cbb52fa3483924eb
fbshipit-source-id:
484062e80af67dfd39d2eaf3cbb52fa3483924eb
Pavlo Kushnir [Wed, 4 May 2016 07:27:53 +0000 (00:27 -0700)]
Optimize copy/move
Summary: this diff provides methods that allow to avoid temporaries while cloning IOBufs. Also optimizes move constructor a bit - there is no need to call `decrementRefCount`, check for self assignment, etc.
Reviewed By: yfeldblum
Differential Revision:
D3250456
fb-gh-sync-id:
32b0180c28f36151c6786dba6f511b491b224200
fbshipit-source-id:
32b0180c28f36151c6786dba6f511b491b224200
Yedidya Feldblum [Wed, 4 May 2016 04:32:23 +0000 (21:32 -0700)]
Let via take a functor by universal reference
Summary:
[Folly] Let `via` take a functor by universal reference.
And perfect-forward the functor to the call to `then`.
Reviewed By: fugalh
Differential Revision:
D3255684
fb-gh-sync-id:
1b46a70fb9de88d21d785c0ce4d9dd3078829f13
fbshipit-source-id:
1b46a70fb9de88d21d785c0ce4d9dd3078829f13
Christopher Dykes [Wed, 4 May 2016 00:38:24 +0000 (17:38 -0700)]
Create the sockets portability header
Summary: The one header that doesn't split well. This is the last piece of the portability headers for Windows and is also the piece that has had the most thought put into it's design so-as to be as unintrusive as possible.
Reviewed By: yfeldblum
Differential Revision:
D3046525
fb-gh-sync-id:
761a5e359bc130b95585608ee18fda94a30dae2c
fbshipit-source-id:
761a5e359bc130b95585608ee18fda94a30dae2c
Marcelo Juchem [Tue, 3 May 2016 22:35:30 +0000 (15:35 -0700)]
More useful logging upon failed AsyncServerSocket::bind()
Summary: `AsyncServerSocket::bind()` would not give information like port or family name when failing to bind a socket. This diff addresses that by including this information in the exception. Two additional helper methods were added to `SocketAddress` to retrieve both the port and the family name from a `sockaddr` structure.
Reviewed By: ckwalsh, yfeldblum
Differential Revision:
D3249778
fb-gh-sync-id:
4edb28af5c211b7bf8d525b40844a5b0b6261e07
fbshipit-source-id:
4edb28af5c211b7bf8d525b40844a5b0b6261e07
Sven Over [Tue, 3 May 2016 21:27:37 +0000 (14:27 -0700)]
fix passing move-only types to via(Executor*, Func)
Summary: This diff fixes a problem with passing move-only types to folly::via.
Reviewed By: ericniebler
Differential Revision:
D3254906
fb-gh-sync-id:
8a9c703a8db0ccf20b9fb4fe9b80ad6cdcb3e388
fbshipit-source-id:
8a9c703a8db0ccf20b9fb4fe9b80ad6cdcb3e388
Yedidya Feldblum [Tue, 3 May 2016 09:15:22 +0000 (02:15 -0700)]
Deprecate boost::mpl
Summary:
[Folly] Deprecate `boost::mpl`.
The `std` has equivalents; prefer them.
Reviewed By: Gownta
Differential Revision:
D3247294
fb-gh-sync-id:
873697b7ea3ad069ff14662bc458a2e1cc840339
fbshipit-source-id:
873697b7ea3ad069ff14662bc458a2e1cc840339
Michael Lee [Mon, 2 May 2016 23:14:23 +0000 (16:14 -0700)]
Use portability for SYS_gettid
Summary: It's there for a reason, so use it.
Reviewed By: yfeldblum
Differential Revision:
D3249849
fb-gh-sync-id:
e9fe330693a173acd987e0c8fd1f03a982c8e082
fbshipit-source-id:
e9fe330693a173acd987e0c8fd1f03a982c8e082
Christopher Dykes [Mon, 2 May 2016 19:22:28 +0000 (12:22 -0700)]
Don't try to use the malloc_usable_size portability implementation if we have JEMalloc
Summary:
This is only an issue for the OSS HHVM build on OSX, where we have JEMalloc headers included directly.
In the OSS HHVM build, we define `USE_JEMALLOC` globally if we have JEMalloc, so gate the portability implementation behind that.
Reviewed By: mzlee
Differential Revision:
D3237449
fb-gh-sync-id:
07cf9b31114723ffd9f2b29cf449bee0ed8f5703
fbshipit-source-id:
07cf9b31114723ffd9f2b29cf449bee0ed8f5703
Neel Goyal [Mon, 2 May 2016 16:30:18 +0000 (09:30 -0700)]
Set the appropriate AsyncSocketExceptionType from SSLException
Summary:
We should use the appropriate AsyncSocketExceptionType when firing SSLExceptions. This derives the type from the constructor args, and introduces a `NETWORK_ERROR` exception type. Open to suggestions on reusing something there.
Additionally, rename getType() to getSSLError() to prevent hiding the AsyncSocketException::getType.
Reviewed By: siyengar
Differential Revision:
D3241204
fb-gh-sync-id:
de631dfb3554177d5bd561f4b91e767c882767d3
fbshipit-source-id:
de631dfb3554177d5bd561f4b91e767c882767d3
Anirudh Ramachandran [Mon, 2 May 2016 16:12:01 +0000 (09:12 -0700)]
Simplify CryptoSSLRequestManager using unified futures API
Summary:
This diff unifies the nearly-replicated call paths for ECDSA and RSA
sign request/response/fallback/etc. into one generic futures-based sign API.
Once this is tested, we should be able to remove ~500 lines of code from
CryptoSSLRequestManager
Depends on
D3094660
Reviewed By: siyengar
Differential Revision:
D3116535
fb-gh-sync-id:
11c50d81dab39e5426679b5a172c1ece2039fc10
fbshipit-source-id:
11c50d81dab39e5426679b5a172c1ece2039fc10
Yedidya Feldblum [Mon, 2 May 2016 09:26:07 +0000 (02:26 -0700)]
Avoid glog when warning about Singleton double-registration
Summary:
[Folly] Avoid glog when warning about `Singleton` double-registration.
Because registration happens at static initialization time, long before main, and possibly long before glog is initialized.
This makes a difference because, in some cases of double-registration, we can get a SIGSEGV with no message, because we are attempting to `LOG(FATAL)` when glog is uninitialized. What we would much rather have is a SIGABRT with a message.
Reviewed By: meyering
Differential Revision:
D3245047
fb-gh-sync-id:
4c5dd9d25025f197d7c490ffbb429af5ccb82182
fbshipit-source-id:
4c5dd9d25025f197d7c490ffbb429af5ccb82182
Andrii Grynenko [Sat, 30 Apr 2016 00:50:50 +0000 (17:50 -0700)]
Don't require folly::Unit in addTaskFinally
Summary:
In
D3241498 addTaskFuture was fixed to work with function returning void. This however changed addTaskFinally API to not allow finally functor which accepts Try<void>.
folly::fibers generally supports Try<void> so there's no reason to force users to use folly::Unit instead of void for addTaskFinally too.
Reviewed By: yfeldblum
Differential Revision:
D3243893
fb-gh-sync-id:
d1df54738157d2019120103956f59b3971ba25ff
fbshipit-source-id:
d1df54738157d2019120103956f59b3971ba25ff
Martin Martin [Fri, 29 Apr 2016 18:54:34 +0000 (11:54 -0700)]
Make addTask[Remote]Future() work for functions returning void.
Summary:
There's no Future<void>, you're supposed to use Future<Unit>
instead. Unit has a "Lift" template to do the conversion.
Reviewed By: andriigrynenko
Differential Revision:
D3241498
fb-gh-sync-id:
db12d4f343685bc613b701e023c860c2c903ed4c
fbshipit-source-id:
db12d4f343685bc613b701e023c860c2c903ed4c
Martin Martin [Fri, 29 Apr 2016 17:50:23 +0000 (10:50 -0700)]
Clang-format in preparation for other change
Summary: Clang-format in preparation for other change
Reviewed By: andriigrynenko
Differential Revision:
D3241297
fb-gh-sync-id:
b7b26812e9e61c291d5c7bdb523df8f28f2d9b4f
fbshipit-source-id:
b7b26812e9e61c291d5c7bdb523df8f28f2d9b4f
Philip Pronin [Fri, 29 Apr 2016 00:21:06 +0000 (17:21 -0700)]
don't destory defaultElfCache in exit handlers
Summary:
As it might be used during global destruction (for example, to
symbolize unhandled exception happened at that time).
Reviewed By: luciang, ot
Differential Revision:
D3238303
fb-gh-sync-id:
39d8a6f99d95e700586349436e961f3624c95718
fbshipit-source-id:
39d8a6f99d95e700586349436e961f3624c95718
Eric Niebler [Thu, 28 Apr 2016 23:39:30 +0000 (16:39 -0700)]
use std::forward instead of std::move on objects whose types have been deduced; don't take the sizeof incomplete types
Summary: Pretty sure std::forward is needed here instead of std::move. If you trace the call chain to see where the types of the objects come from, you'll see they can be deduced to be lvalues, so std::forward is the right choice. Also, moved some dicey looking code that appeared to be taking the size of some incomkplete types.
Reviewed By: spacedentist
Differential Revision:
D3214199
fb-gh-sync-id:
778190ffb25a648b839760a3dddfad8dc6d41c88
fbshipit-source-id:
778190ffb25a648b839760a3dddfad8dc6d41c88
Andrew Birchall [Thu, 28 Apr 2016 23:39:10 +0000 (16:39 -0700)]
Create ReadHolder::unlock
Summary:
Currently you need to depend on the destructor of `ReadHolder` (using closures as in code block #1 below or empty assignment as in code block #2 below) to ensure that a `ReadHolder` goes out of scope (and unlocks) in order to subsequently acquire a write lock via `WriteHolder` without deadlocking.
This diff introduces a way of unlocking a `ReadHolder` while it's still in scope such that a `WriteHolder` can be acquired. This makes the code more straight forward (reducing the risk of deadlock due to a programmer's misunderstanding of how `SharedMutex` / the holder framework works) => see code block # 3 below
Also add some documentation about why `WriteHolder::WriteHolder(ReadHolder&&)` doesn't exist
Code Block #1 : Use of closures
```
class foo {
public:
std::string getMemoizedData() {
{
folly::SharedMutex::ReadHolder readHolder(lock_);
if (!data_.empty()) {
// important to return by value, otherwise caller might access
// data_ after we release the read lock
return data_;
}
}
{
// try again with a write lock
folly::SharedMutex::WriteHolder writeHolder(lock_);
if (data_.empty()) {
data_ = "my awesome string";
}
return data_;
}
}
private:
folly::SharedMutex lock_;
std::string data_;
};
```
Code Block #2 : Use of empty assignment
```
class foo {
public:
std::string getMemoizedData() {
folly::SharedMutex::ReadHolder readHolder(lock_);
if (!data_.empty()) {
// important to return by value, otherwise caller might access
// data_ after we release the read lock
return data_;
}
readHolder = {};
// try again with a write lock
folly::SharedMutex::WriteHolder writeHolder(lock_);
if (data_.empty()) {
data_ = "my awesome string";
}
return data_;
}
private:
folly::SharedMutex lock_;
std::string data_;
};
```
Code Block #3 : Use of unlock()
```
class foo {
public:
std::string getMemoizedData() {
folly::SharedMutex::ReadHolder readHolder(lock_);
if (!data_.empty()) {
// important to return by value, otherwise caller might access
// data_ after we release the read lock
return data_;
}
readHolder->unlock();
// try again with a write lock
folly::SharedMutex::WriteHolder writeHolder(lock_);
if (data_.empty()) {
data_ = "my awesome string";
}
return data_;
}
private:
folly::SharedMutex lock_;
std::string data_;
};
```
Reviewed By: yfeldblum
Differential Revision:
D3176025
fb-gh-sync-id:
c7d47ca71df08673c7c1f1fd5ed9e01a663c1797
fbshipit-source-id:
c7d47ca71df08673c7c1f1fd5ed9e01a663c1797
Lucian Grijincu [Thu, 28 Apr 2016 21:34:27 +0000 (14:34 -0700)]
folly: Function: in-class init, drop (void*) casts
Summary: A few code cleanups, no functionality changes.
Reviewed By: ericniebler
Differential Revision:
D3231369
fb-gh-sync-id:
1a1c8508aab6dce7279e9dfc3f3da6add5496c67
fbshipit-source-id:
1a1c8508aab6dce7279e9dfc3f3da6add5496c67
Subodh Iyengar [Thu, 28 Apr 2016 19:10:12 +0000 (12:10 -0700)]
Better exception types to SSLexception
Summary:
Better enum exception types for SSLException.
These enum types are fairly high level, and more
info can be gleaned from the error.what()
Reviewed By: knekritz
Differential Revision:
D3234501
fb-gh-sync-id:
7cd4fbccd7f4367354dc3bd1fe4cd480d58d6173
fbshipit-source-id:
7cd4fbccd7f4367354dc3bd1fe4cd480d58d6173
Subodh Iyengar [Thu, 28 Apr 2016 19:10:01 +0000 (12:10 -0700)]
Stop abusing errno
Summary:
We abuse errno to propagate exceptions from AsyncSSLSocket.
Stop doing this and propagate exceptions correctly.
This also formats the exception messages better.
Reviewed By: anirudhvr
Differential Revision:
D3226808
fb-gh-sync-id:
15a5e67b0332136857e5fb85b1765757e548e040
fbshipit-source-id:
15a5e67b0332136857e5fb85b1765757e548e040
Subodh Iyengar [Thu, 28 Apr 2016 18:12:54 +0000 (11:12 -0700)]
Strict validation for certs
Summary: Add X509 param types
Reviewed By: knekritz
Differential Revision:
D3230008
fb-gh-sync-id:
660a2f5bdc9355f5b42c2ccfb1d7cab83df34fc2
fbshipit-source-id:
660a2f5bdc9355f5b42c2ccfb1d7cab83df34fc2
Nicholas Ormrod [Thu, 28 Apr 2016 17:41:19 +0000 (10:41 -0700)]
Fix asan build
Summary:
The array-bounds checker in gcc complains that r could be -1 after leaving CHECK_ERR in ASAN mode. Switch to DCHECK to avoid this possibility. The code was not initially compiling; now it does.
Error from gcc:
array subscript is below array bounds [-Werror=array-bounds]
Reviewed By: yfeldblum
Differential Revision:
D3232694
fb-gh-sync-id:
17f579eb152b661a8e5e4ed29ef91eec405fb90a
fbshipit-source-id:
17f579eb152b661a8e5e4ed29ef91eec405fb90a
Christopher Dykes [Thu, 28 Apr 2016 17:17:19 +0000 (10:17 -0700)]
Use the portability header for malloc_usable_size in small_vector.h
Summary: `small_vector.h` had it's own thing being defined inline, and it was breaking the OSX build of HHVM, so have it use the portability header instead.
Reviewed By: yfeldblum
Differential Revision:
D3232885
fb-gh-sync-id:
2078a54d1ec9700e81e0a9f4baed8e2d623fa5f3
fbshipit-source-id:
2078a54d1ec9700e81e0a9f4baed8e2d623fa5f3
Andrii Grynenko [Thu, 28 Apr 2016 04:00:57 +0000 (21:00 -0700)]
Fix races in TLRefCount
Summary:
This fixes 2 races in TLRefCount:
1. Thread-local constructor race, exposed by the stress test. It was possible for LocalRefCount to be created (grabbing collectGuard), but not be added to the thread-local list, so that accessAllThreads wasn't collecting it. collectAll() was then blocking waiting on baton to be posted, causing a dead-lock.
2. LocalRefCount::count_ has to be made atomic, because otherwise += operation may be not flushed (nbronson explained the race in
D3133443).
Reviewed By: djwatson
Differential Revision:
D3166956
fb-gh-sync-id:
17d58a215ebfc572f8316ed46bafaa5e6a9e2368
fbshipit-source-id:
17d58a215ebfc572f8316ed46bafaa5e6a9e2368
Giuseppe Ottaviano [Thu, 28 Apr 2016 00:47:43 +0000 (17:47 -0700)]
Further Function simplification
Summary:
Use tag dispatching instead of `enable_if`: it is clearer, it
sidesteps the GCC mangling bug, and more importantly the conditional
doesn't leak into the symbol, making stack traces and profiles more
readable.
Testing on a compilation unit with 1000 `Function`s from simple lambdas.
Before:
```
folly::impl::Function<int (), false>::Function<main::{lambda()#1}, {lambda()#1}>(main::{lambda()#1}&&, std::enable_if<std::integral_constant<bool, ((sizeof (std::decay<main::{lambda()#1}>::type))<=(sizeof folly::detail::function::Data::small))&&std::is_nothrow_move_constructible<std::decay<main::{lambda()#1}> >::value>::value, folly::detail::Tag>::type)::Ops::call(folly::detail::function&)
```
After:
```
folly::impl::Function<int (), false>::OpsSmall<main::{lambda()#1}>::call(folly::detail::function::Data&)
```
Note that the function type is repeated 5 times before, and only once after. This makes a large difference with long namespaces.
Binary size is almost unaffected, compile times slightly better:
Before:
GCC opt: 22.3 seconds,
4435232 bytes
Clang dev: 7.7 seconds,
5257344 bytes
After:
GCC opt: 18.6 seconds,
4493920 bytes
Clang dev: 7.2 seconds,
5469136 bytes
Reviewed By: ericniebler
Differential Revision:
D3231530
fb-gh-sync-id:
6aa76e7f780a8afdbfed8a378f257ceb86dce704
fbshipit-source-id:
6aa76e7f780a8afdbfed8a378f257ceb86dce704
Pavlo Kushnir [Thu, 28 Apr 2016 00:45:07 +0000 (17:45 -0700)]
Faster onDestroy
Summary: there is no need to call `std::function::invoke` for every DestructorGuard.
Reviewed By: yfeldblum
Differential Revision:
D3229345
fb-gh-sync-id:
c42f8cd05576d56b6a9b2f9d06878d9b01a36e94
fbshipit-source-id:
c42f8cd05576d56b6a9b2f9d06878d9b01a36e94
Christopher Dykes [Wed, 27 Apr 2016 21:34:33 +0000 (14:34 -0700)]
Have the GFlags portability header stick things in namespaces
Summary:
Just like the real thing does.
This was causing some issues when trying to link against glog, which expects the flags to be namespaced like this.
Reviewed By: mzlee
Differential Revision:
D3230630
fb-gh-sync-id:
a73ab3044560d561a39eb91ceee1588c147a46c5
fbshipit-source-id:
a73ab3044560d561a39eb91ceee1588c147a46c5
Eric Niebler [Wed, 27 Apr 2016 17:02:00 +0000 (10:02 -0700)]
Reimplement folly::Function to improve compile times.
Summary:
folly::Function is causing significant compile time regressions. Reimplement it in a simpler way.
These are the times for a file containing 1000 instantiations of folly::Fuction (old), folly::Function (new), and std::function with **g++ 4.8 -O3** on my CentOS7 server.
| | Old `folly::Function` | `std::function` | New `folly::Function` |
|--------|-----------------------|-----------------|-----------------------|
| Time | 10m37s | 0m16.81s | 0m14.75s |
And for the executable size:
| | Old `folly::Function` | `std::function` | New `folly::Function` |
|--------|-----------------------|-----------------|-----------------------|
| Size | 10,409,504 | 732,150 | 562,781 |
That's a **43X** improvement in compile times and an **18X** reduction in executable bloat over the old implementation.
The times for **clang (trunk)** are very different:
| | Old `folly::Function` | `std::function` | New `folly::Function` |
|-------|-----------------------|-----------------|-----------------------|
| Time | 4m6s | 0m45.27s | 0m11.78s |
That's a **20X** improvement over the old implementation and almost a **4X** improvement over `std::function`.
For **gcc-5.3.0**, compile times are again different:
| | Old `folly::Function` | `std::function` | New `folly::Function` |
|-------|-----------------------|-----------------|-----------------------|
| Time | 2m49s | 0m18.99s | 0m20.70s |
With gcc-5.3, the new implementation "only" compiles 8x faster than the old one, and is roughly the same as `std::function`.
Reviewed By: spacedentist, ot, luciang
Differential Revision:
D3199985
fb-gh-sync-id:
b97982a9dc3a63140510babea34988932e89f2d9
fbshipit-source-id:
b97982a9dc3a63140510babea34988932e89f2d9
Kyle Nekritz [Wed, 27 Apr 2016 16:42:56 +0000 (09:42 -0700)]
Add recursion limit to folly::parseJson.
Summary: Without this, malicious inputs can crash anything using folly::parseJson.
Reviewed By: yfeldblum
Differential Revision:
D3219036
fb-gh-sync-id:
3604a060170c0201473c420035b21b018383789c
fbshipit-source-id:
3604a060170c0201473c420035b21b018383789c
Subodh Iyengar [Wed, 27 Apr 2016 16:13:48 +0000 (09:13 -0700)]
Remove dead code in AsyncSSLSocket
Summary: Remove dead code in AsyncSSLSocket.
Reviewed By: knekritz
Differential Revision:
D3226948
fb-gh-sync-id:
e85823f311de2539c6aa2d6bcc3ff3aee07045bf
fbshipit-source-id:
e85823f311de2539c6aa2d6bcc3ff3aee07045bf
Sven Over [Wed, 27 Apr 2016 13:46:08 +0000 (06:46 -0700)]
fix access-after-free bug in ProducerConsumerQueueTest
Summary: This bug causes failure when the test is run under ASAN.
Reviewed By: meyering
Differential Revision:
D3229494
fb-gh-sync-id:
a43c8332cc45f7892ac86cd0abb799616bca7779
fbshipit-source-id:
a43c8332cc45f7892ac86cd0abb799616bca7779
Pavlo Kushnir [Wed, 27 Apr 2016 07:20:04 +0000 (00:20 -0700)]
Cheaper bumpHandlingTime
Summary: no need in virtual call from EventHandler.
Reviewed By: yfeldblum
Differential Revision:
D3226960
fb-gh-sync-id:
eb9c191630e1a1ac022666201100e3778eb7b611
fbshipit-source-id:
eb9c191630e1a1ac022666201100e3778eb7b611
Jon Maltiel Swenson [Wed, 27 Apr 2016 01:12:20 +0000 (18:12 -0700)]
Optimize away IOBuf clones on hot path
Summary:
'get' is the most frequent memcached operation. On a get reply, we should go to heap
as little as possible. This diff optimizes for this scenario, where replies
have only one IOBuf field.
Reviewed By: pavlo-fb
Differential Revision:
D3226592
fb-gh-sync-id:
92e1a1fac5735bd268691cf11990a96ae6fa8309
fbshipit-source-id:
92e1a1fac5735bd268691cf11990a96ae6fa8309
Aditya Muttur [Tue, 26 Apr 2016 22:53:40 +0000 (15:53 -0700)]
Adding a constructor to UTF8Range that uses std::string
Summary:
Currently UTF8Range has a constructor that allows you construct an object of type UTF8Range using only an object of type folly::Range. Adding a constructor so that we can construct an UTF8Range object using a std::string.
Currently,
void sampleMethod(UTF8StringPiece sp) {...}
/*
...
*/
std::string str = "example";
folly::UTF8Range utf8Range(str);
folly::StringPiece sp(str);
sampleMethod(utf8Range); // works
sampleMethod(sp); // works
sampleMethod(str); // doesn't work
This diff hopes to fix this issue.
Reviewed By: ddrcoder
Differential Revision:
D3221144
fb-gh-sync-id:
dd6ec4d7790d4602dccb3b63a4861d358aed3608
fbshipit-source-id:
dd6ec4d7790d4602dccb3b63a4861d358aed3608
Mirek Klimos [Tue, 26 Apr 2016 16:58:33 +0000 (09:58 -0700)]
API to set folly::RequestContext for current scope, try 2
Summary: same as
D3156698, without changes in Cpp2Connection (which was the only real change in the diff)
Reviewed By: haijunz
Differential Revision:
D3222792
fb-gh-sync-id:
245c7add837c0fc6d0bc84aa7d80b929ba2ce386
fbshipit-source-id:
245c7add837c0fc6d0bc84aa7d80b929ba2ce386
Christopher Dykes [Tue, 26 Apr 2016 16:38:18 +0000 (09:38 -0700)]
Define DECLARE_VARIABLE in portability/GFlags.h
Summary: If we don't, then anywhere we include `glog/logging.h` after the portability header will get the `DECLARE_*` and `DEFINE_*` macros undefined -_-....
Reviewed By: mzlee
Differential Revision:
D3220797
fb-gh-sync-id:
52907ddcd6b222fb1c6423034ea999eac5ce09ab
fbshipit-source-id:
52907ddcd6b222fb1c6423034ea999eac5ce09ab
Michael Lee [Tue, 26 Apr 2016 15:43:27 +0000 (08:43 -0700)]
Remove extra dependency on Portability.h
Summary: This dependency is in the wrong order.
Reviewed By: knekritz
Differential Revision:
D3224312
fb-gh-sync-id:
98d2c58b75cf5f16892d462ae1b877d06bce3faa
fbshipit-source-id:
98d2c58b75cf5f16892d462ae1b877d06bce3faa
Pavlo Kushnir [Tue, 26 Apr 2016 09:27:37 +0000 (02:27 -0700)]
Use std::string in folly::dynamic
Summary: Why not kill fbstring entirely? There are some legitimate use cases. For example, `folly::IOBuf` supports `moveToFbstring()` method which is impossible to implement with `std::string`.
Reviewed By: joeg, snarkmaster
Differential Revision:
D3189410
fb-gh-sync-id:
9bb9090ca6012ac32ba9fb79041b26ec4888781f
fbshipit-source-id:
9bb9090ca6012ac32ba9fb79041b26ec4888781f
Yedidya Feldblum [Tue, 26 Apr 2016 07:56:27 +0000 (00:56 -0700)]
Retire BOOST_STATIC_ASSERT in favor of static_assert
Summary:
Retire `BOOST_STATIC_ASSERT` in favor of `static_assert`.
`static_assert` is part of C++ now, so we don't need workarounds like `BOOST_STATIC_ASSERT` anymore.
Partially automated:
hg grep -lw BOOST_STATIC_ASSERT | xargs perl -pi -e 's~\bBOOST_STATIC_ASSERT\(([^;]*)\);~static_assert(\1, "");~g'
hg grep -lw 'boost/static_assert.hpp' | xargs perl -pi -e 's,^#include <boost/static_assert\.hpp>\n,,gm'
Caught most instances. Formatting and remaining instances addressed manually.
Reviewed By: meyering
Differential Revision:
D3215944
fb-gh-sync-id:
f4552d5d9bfc416ce283923abe880437a4d0cba5
fbshipit-source-id:
f4552d5d9bfc416ce283923abe880437a4d0cba5
Stepan Palamarchuk [Mon, 25 Apr 2016 21:12:21 +0000 (14:12 -0700)]
Remove unnecessary construction of IOBuf in Cursor
Summary: This assignment is unnecessary, because in `for` loop we **always** call `cloneOneInto` that would assign another IOBuf to that IOBuf.
Reviewed By: alikhtarov
Differential Revision:
D3215861
fb-gh-sync-id:
ad87b99848aaae79f0870d49e0474f1abf0f28e5
fbshipit-source-id:
ad87b99848aaae79f0870d49e0474f1abf0f28e5
Andrew Cox [Mon, 25 Apr 2016 17:51:47 +0000 (10:51 -0700)]
UndelayedDestruction virtual=>override
Summary:
Had this in another diff, but it caused too much test noise. So I've separated
it out on it's own.
Reviewed By: yfeldblum
Differential Revision:
D3213605
fb-gh-sync-id:
9ebfcf8430da7c66a31868032a0cef1e616ffc58
fbshipit-source-id:
9ebfcf8430da7c66a31868032a0cef1e616ffc58
Sven Over [Sat, 23 Apr 2016 15:07:25 +0000 (08:07 -0700)]
use forward instead of move for universal references
Summary:When dealing with universal references, std::move will move from
objects that are passed as lvalue references. Instead std::forward
should be used, which only moves from an object if it's passed
as rvalue reference.
Reviewed By: yfeldblum
Differential Revision:
D3200402
fb-gh-sync-id:
14be071e8498dd64cb8b2583c0cc2dd383bfebb8
fbshipit-source-id:
14be071e8498dd64cb8b2583c0cc2dd383bfebb8
Hans Fugal [Fri, 22 Apr 2016 21:11:29 +0000 (14:11 -0700)]
update futures README
Summary:It's been awhile.
Still has intern URLs (#
10943549) but I'll fix that separately.
Reviewed By: jsedgwick
Differential Revision:
D3213245
fb-gh-sync-id:
ff60193ff368deaac8ca233d4289f30d8f6bb223
fbshipit-source-id:
ff60193ff368deaac8ca233d4289f30d8f6bb223
Sven Over [Fri, 22 Apr 2016 18:00:39 +0000 (11:00 -0700)]
Subprocess: allow non-copyable callbacks
Summary:Instead of std::function<bool(int, int)>, use folly::Function to
pass callbacks to Subprocess::communicate. This makes it possible
to pass non-copyable callables, which is especially interesting
because Subprocess::readLinesCallback returns a non-copyable
object.
This diff also fixes the forwarding of the callback passed to
readLinesCallback in case you pass an lvalue reference.
Reviewed By: snarkmaster
Differential Revision:
D3169956
fb-gh-sync-id:
7a906f9a3ab50502fc04e0d83a23ca5e0201bb3e
fbshipit-source-id:
7a906f9a3ab50502fc04e0d83a23ca5e0201bb3e
Yedidya Feldblum [Fri, 22 Apr 2016 06:14:27 +0000 (23:14 -0700)]
get_default and get_ref_default variants taking functions
Summary:[Folly] `get_default` and `get_ref_default` variants taking functions.
Useful if the default value is computationally expensive to construct or requires IO.
Reviewed By: andriigrynenko, spacedentist
Differential Revision:
D3189247
fb-gh-sync-id:
51c64293f8712d7590348d53cbfd892a5efd9e82
fbshipit-source-id:
51c64293f8712d7590348d53cbfd892a5efd9e82
Tobias Ritzau [Fri, 22 Apr 2016 02:04:13 +0000 (19:04 -0700)]
Updates the internal representation of signals in NotificationQueue
Summary: The signals were represented using bytes in a pipe or using a count on an event fd (when available). This count was ever growing and caused the pipe to overflow, and in both cases you would get signals on empty queues. This diff only writes to the fd if it there are no bytes to read. Due to races there can still be multiple bytes in the pipe, but overflowing should not be possible. Instead of blindly signaling when there could be messages in the queue, the signals are now synchronized with the state of the queue so that the signals are drained when the queue is empty. This also made it possible to skip the semaphore behavior of the event fd which should improve perf.
Reviewed By: dcolascione
Differential Revision:
D3198252
fb-gh-sync-id:
39e620b10c254ffcacabc4c5ac36950a215d4803
fbshipit-source-id:
39e620b10c254ffcacabc4c5ac36950a215d4803
Taiyuan Zhang [Thu, 21 Apr 2016 22:42:26 +0000 (15:42 -0700)]
fix cancelAllFunctions
Summary: as titled. When I tried re-adding all functions after cancelAllFunctions, it throws error. This is because it doesn't reset currentFunction to null.
Reviewed By: jjs0
Differential Revision:
D3209237
fb-gh-sync-id:
5b6e2d967fc3c0986320d23b8d61eb1bfbff8940
fbshipit-source-id:
5b6e2d967fc3c0986320d23b8d61eb1bfbff8940
chenkan [Thu, 21 Apr 2016 22:02:44 +0000 (15:02 -0700)]
fix: readvNoInt use wrong read function
Summary: Closes https://github.com/facebook/folly/pull/401
Reviewed By: yfeldblum
Differential Revision:
D3207682
Pulled By: snarkmaster
fb-gh-sync-id:
617aeec0e6fe1f46d2464df4b72a3e91c953dc12
fbshipit-source-id:
617aeec0e6fe1f46d2464df4b72a3e91c953dc12
Michael Lee [Wed, 20 Apr 2016 22:54:54 +0000 (15:54 -0700)]
Define more pieces of portability/GFlags
Summary: Add more compatibility for missing gflags
Reviewed By: yfeldblum
Differential Revision:
D3202560
fb-gh-sync-id:
d25e4abfe8ceee9fb329c9ba12d259d1d03d4974
fbshipit-source-id:
d25e4abfe8ceee9fb329c9ba12d259d1d03d4974
Marcelo Juchem [Wed, 20 Apr 2016 21:23:31 +0000 (14:23 -0700)]
Added FOLLY_IGNORE and FOLLY_SEMICOLON helper macros
Summary: these macros are generally useful for other higher-order macros
Reviewed By: yfeldblum
Differential Revision:
D3194777
fb-gh-sync-id:
667fc51c681786ad422309ee463881dc22c972f7
fbshipit-source-id:
667fc51c681786ad422309ee463881dc22c972f7
Michael Lee [Wed, 20 Apr 2016 20:24:25 +0000 (13:24 -0700)]
Clean up deprecation warnings
Summary:Clean up a deprecation warning for JSONSchema's use of
Singleton. This is in experimental, so the interface change should be
reasonable..
Reviewed By: agoder
Differential Revision:
D3203348
fb-gh-sync-id:
3283c7d9c507a545f3217eb5afc3734eb047b833
fbshipit-source-id:
3283c7d9c507a545f3217eb5afc3734eb047b833
Michael Lee [Wed, 20 Apr 2016 20:24:18 +0000 (13:24 -0700)]
Fix use of dynamic {} initialization in JSONSchemaTest
Summary: find replace 's/{([0-9, a-z"-]+)}/dynamic::array(\1)/g'
Reviewed By: agoder
Differential Revision:
D3203454
fb-gh-sync-id:
064bb2bdd0e19a80da32e8f4eb730fc4268d7cad
fbshipit-source-id:
064bb2bdd0e19a80da32e8f4eb730fc4268d7cad
Mirek Klimos [Wed, 20 Apr 2016 20:08:48 +0000 (13:08 -0700)]
Reverted commit
D3156698
Summary:There're currently two ways to set RequestContext
- RequestContext::create() - creates new context and sets it
- RequestContext::setContext(context) - sets context previously captured by saveContext
In most cases, the RequestContext is set back after the request is processed but sometimes it's not (especially with RequestContext::create). We want to measure cpu time for a request by measuring the total cpu time when a RequestContext is set, so we need to make sure that it's properly reset after the thread is done with the request. Scope guards can help us with that.
Reviewed By: haijunz
Differential Revision:
D3156698
fb-gh-sync-id:
7c3eb06c1cc27849071625011bf64c5ad36c0612
fbshipit-source-id:
7c3eb06c1cc27849071625011bf64c5ad36c0612
Alexey Spiridonov [Wed, 20 Apr 2016 19:08:29 +0000 (12:08 -0700)]
Throw on errors in read callback
Summary: Before this fix, the callback would silently ignore `ret == -1` if the error isn't `EAGAIN`.
Reviewed By: spacedentist
Differential Revision:
D3193845
fb-gh-sync-id:
e4a7aa37de7dab8ebe0633dd9888d8adc11dd1c2
fbshipit-source-id:
e4a7aa37de7dab8ebe0633dd9888d8adc11dd1c2
Michael Lee [Wed, 20 Apr 2016 02:47:54 +0000 (19:47 -0700)]
Backward compatibility for older versions of clang.
Summary: `__attribute__((no_sanitize("address")))` is not valid in clang 3.6 and earlier
Reviewed By: skotchvail
Differential Revision:
D3197412
fb-gh-sync-id:
bd555b030fe48268988299dd88b9b7aea54c1a73
fbshipit-source-id:
bd555b030fe48268988299dd88b9b7aea54c1a73
Michael Lee [Tue, 19 Apr 2016 14:52:23 +0000 (07:52 -0700)]
`detail::Futex` wants 4 bytes but MicroLock only gives you one
Summary: ^^^ this is a stack overflow in the test, and a possible stack or heap overflow.
Reviewed By: dcolascione
Differential Revision:
D3151717
fb-gh-sync-id:
b4f0660ebbb89139dff003870e132c312068d9a8
fbshipit-source-id:
b4f0660ebbb89139dff003870e132c312068d9a8
Mark Isaacson [Tue, 19 Apr 2016 07:46:56 +0000 (00:46 -0700)]
Add make_array to folly
Summary: This function is being proposed in WG21, the C++ standards body for inclusion in the STL via the Library Fundamentals v2 TS. Using the normal constructor for a std::array with an initializer list introduces a source of coupling between the # of elements you put in the initializer list and the size of the array you specify as a template argument. Worse still, if you put less things in the initializer list than the template argument specifies, it doesn't warn you that you've probably made a pretty devious and subtle error. With this function your array size will always be the same as the # of things you actually put in it. What's more, in some cases this can deduce the type of the elements as well.
Reviewed By: yfeldblum
Differential Revision:
D3164432
fb-gh-sync-id:
beceaae2ee01cd5f93dec86cf36efdb78a28b4a3
fbshipit-source-id:
beceaae2ee01cd5f93dec86cf36efdb78a28b4a3
Mirek Klimos [Mon, 18 Apr 2016 21:28:06 +0000 (14:28 -0700)]
API to set folly::RequestContext for current scope
Summary:There're currently two ways to set RequestContext
- RequestContext::create() - creates new context and sets it
- RequestContext::setContext(context) - sets context previously captured by saveContext
In most cases, the RequestContext is set back after the request is processed but sometimes it's not (especially with RequestContext::create). We want to measure cpu time for a request by measuring the total cpu time when a RequestContext is set, so we need to make sure that it's properly reset after the thread is done with the request. Scope guards can help us with that.
Reviewed By: haijunz
Differential Revision:
D3156698
fb-gh-sync-id:
cfb678531810e8be5faaf02cb7803bd247557e42
fbshipit-source-id:
cfb678531810e8be5faaf02cb7803bd247557e42
Brian Walker [Mon, 18 Apr 2016 17:59:14 +0000 (10:59 -0700)]
NotificationQueue Logging for Android bug
Summary: Add some logging to figure out the state of the Qeueu when the write error happens.
Differential Revision:
D3190882
fb-gh-sync-id:
6515d77df8aa3086922cb4053f2e437f3e527a36
fbshipit-source-id:
6515d77df8aa3086922cb4053f2e437f3e527a36
Sven Over [Mon, 18 Apr 2016 01:37:01 +0000 (18:37 -0700)]
remove superfluous include directive
Summary:folly::fibers does not use folly::MoveWrapper anymore, so no need to
include that header file.
Reviewed By: yfeldblum
Differential Revision:
D3169374
fb-gh-sync-id:
2409b46156843fb0a1454b560c42186f1c412988
fbshipit-source-id:
2409b46156843fb0a1454b560c42186f1c412988
Sven Over [Sat, 16 Apr 2016 03:26:30 +0000 (20:26 -0700)]
remove MoveWrapper from io/async/AsyncUDPServerSocket.h
Summary:This is the last place in folly that uses MoveWrapper (other than
the MoveWrapper implementation itself and its tests, of course).
Instead of using a MoveWrapper, the object in question is moved
into the lambda using C++14 syntax, and the lambda is moved in a
call to EventBase::runInEventBaseThread which is possible now
that the EventBase methods accept non-copyable callbacks.
Reviewed By: yfeldblum
Differential Revision:
D3169316
fb-gh-sync-id:
2dcb1a523ac417f4619c607898e58b572648e3da
fbshipit-source-id:
2dcb1a523ac417f4619c607898e58b572648e3da
Andrii Grynenko [Sat, 16 Apr 2016 02:50:46 +0000 (19:50 -0700)]
Non-intrusive AtomicLinkedList
Summary:Renamed AtomicLinkedList to AtomicIntrusiveLinkedList.
AtomicLinkedList is a simple AtomicIntrusiveLinkedList wrapper, which handles intrusive list hook.
Reviewed By: yfeldblum
Differential Revision:
D3188171
fb-gh-sync-id:
0823b04a48336d65e0a6a8cd412c75f52afe02b9
fbshipit-source-id:
0823b04a48336d65e0a6a8cd412c75f52afe02b9
Michael Lee [Fri, 15 Apr 2016 22:16:46 +0000 (15:16 -0700)]
Forgot to fix the guard for String.cpp
Summary:
9ab69bc removed `__APPLE__` around the strndup definition
because it is defined on modern OSX. But there was another reference
in String.cpp. Fixing that.
Reviewed By: yfeldblum
Differential Revision:
D3186360
fb-gh-sync-id:
709c3b93cd22945e2237412637929df1979526c1
fbshipit-source-id:
709c3b93cd22945e2237412637929df1979526c1
Christopher Dykes [Fri, 15 Apr 2016 17:57:01 +0000 (10:57 -0700)]
Remove the strings.h portability header
Summary:Although, according to the manpage, these functions are defined in `strings.h`, but they are also defined in `string.h`. We never actually use these functions via `strings.h`, and instead only ever reference them via `string.h`.
To keep things sane, lets just move the functions into `string.h` and kill `strings.h`.
Reviewed By: yfeldblum
Differential Revision:
D3181596
fb-gh-sync-id:
8a474df510ddafc4c595b08b809a7c33e3256177
fbshipit-source-id:
8a474df510ddafc4c595b08b809a7c33e3256177
David Goldblatt [Fri, 15 Apr 2016 17:14:13 +0000 (10:14 -0700)]
Correctly deduce RNG type in folly::Random
Summary:Fix a bug in folly where it can't infer any RNG type except the
default-supplied parameter.
Differential Revision:
D3163421
fb-gh-sync-id:
23d3963ba19dac93fa3407f3a2dfd1d9aa39ea44
fbshipit-source-id:
23d3963ba19dac93fa3407f3a2dfd1d9aa39ea44
Michael Lee [Fri, 15 Apr 2016 00:20:26 +0000 (17:20 -0700)]
Fix build for no FOLLY_TLD_USE_FOLLY_TLS
Summary: unique_ptr doesn't cast to void*
Reviewed By: andriigrynenko
Differential Revision:
D3181346
fb-gh-sync-id:
5b912d2b5d46f3c9953b44d5ced7a8e1b4320678
fbshipit-source-id:
5b912d2b5d46f3c9953b44d5ced7a8e1b4320678
Andrii Grynenko [Thu, 14 Apr 2016 18:19:59 +0000 (11:19 -0700)]
Improve folly::ThreadLocal perf
Summary: This fixes folly::ThreadLocal perf after several refactorings which touched the code.
Reviewed By: ericniebler
Differential Revision:
D3170761
fb-gh-sync-id:
d44198f19aebc9f9f2588a65950efba0694e2a11
fbshipit-source-id:
d44198f19aebc9f9f2588a65950efba0694e2a11
Andrii Grynenko [Thu, 14 Apr 2016 04:30:45 +0000 (21:30 -0700)]
Fix GuardPageAllocator to do mprotect lazily
Summary:Currently GuargPageAllocator mmap's a memory block enough for 100 fiber stacks and then protects a bottom page for each stack. If FiberManager is used to process a single task, protecting 100 stacks doesn't make much sense, but is costly.
This change makes sure we protect a bottom page of every stack the first time given stack is requested.
Reviewed By: alikhtarov
Differential Revision:
D3139944
fb-gh-sync-id:
d9d724eaa0e65a227eac1d09a33018e6cb098aae
fbshipit-source-id:
d9d724eaa0e65a227eac1d09a33018e6cb098aae
Den Raskovalov [Thu, 14 Apr 2016 02:46:08 +0000 (19:46 -0700)]
remove duplicate licence prolog
Reviewed By: yfeldblum
Differential Revision:
D3174112
fb-gh-sync-id:
00a01f64fb43d2333818097bb2212e26e4dd5775
fbshipit-source-id:
00a01f64fb43d2333818097bb2212e26e4dd5775
Christopher Dykes [Wed, 13 Apr 2016 17:19:24 +0000 (10:19 -0700)]
Fix the includes in portability/Malloc.h
Summary:We use `size_t` but didn't include `<stdlib.h>`.
Fix #385
Reviewed By: markisaa
Differential Revision:
D3171643
fb-gh-sync-id:
1f915f9602c4be59a21cdf77b31fe18816f35f81
fbshipit-source-id:
1f915f9602c4be59a21cdf77b31fe18816f35f81
Neel Goyal [Wed, 13 Apr 2016 15:04:45 +0000 (08:04 -0700)]
Add method to get the connect timeout used for an AsyncSocket
Summary: Have the AsyncSocket keep track of the timeout used for connecting and add a getter to retrieve it.
Reviewed By: hiteshk
Differential Revision:
D3170625
fb-gh-sync-id:
61d0ecd8d975c49978a1cf222671aa16a2160499
fbshipit-source-id:
61d0ecd8d975c49978a1cf222671aa16a2160499
Michael Lee [Tue, 12 Apr 2016 17:59:21 +0000 (10:59 -0700)]
Sorry, forgot to add portability/Memory.cpp
Summary: Add it to the Makefile.am
Reviewed By: Orvid
Differential Revision:
D3168833
fb-gh-sync-id:
7184acc565ffebde0f93bc8bdce4d38ba1a79923
fbshipit-source-id:
7184acc565ffebde0f93bc8bdce4d38ba1a79923
Alexander Shaposhnikov [Mon, 11 Apr 2016 20:15:45 +0000 (13:15 -0700)]
Fix the order of EXPECT_EQ parameters
Summary:Change the order of EXPECT_EQ parameters to
EXPECT_EQ(expected, actual). It will make gtest print
the correct error message if a test fails.
Reviewed By: yfeldblum
Differential Revision:
D3161878
fb-gh-sync-id:
69e4536c6f396858d189a9d25904ef4df639ad49
fbshipit-source-id:
69e4536c6f396858d189a9d25904ef4df639ad49
Yedidya Feldblum [Sun, 10 Apr 2016 21:51:09 +0000 (14:51 -0700)]
Some cleanups to folly::EventBase after converting to folly::Function
Summary:[Folly] Some cleanups to `folly::EventBase` after converting to `folly::Function`.
* Fix up some comments referring to `std::function`.
* Remove the `SmallFunctor` bits - `folly::Function` takes over for that.
* Remove `runFunctionPtr` - it's unused.
Reviewed By: spacedentist
Differential Revision:
D3155511
fb-gh-sync-id:
38c75d1254993f59c8eaa7826dc8d9facb50a3a1
fbshipit-source-id:
38c75d1254993f59c8eaa7826dc8d9facb50a3a1
Adam Norton [Sat, 9 Apr 2016 02:30:43 +0000 (19:30 -0700)]
Add <new> header for placement new
Summary: This header is necessary for the placement new operator used in the construct method (line 265)
Reviewed By: yfeldblum
Differential Revision:
D3156741
fb-gh-sync-id:
84a64821553b42d46fb70b5ee267a401028c94f2
fbshipit-source-id:
84a64821553b42d46fb70b5ee267a401028c94f2
Tom Jackson [Thu, 7 Apr 2016 21:01:28 +0000 (14:01 -0700)]
Extensibility for folly::to<> through ADL
Summary: Primarily to support slightly more flexible implementations of `split()`;
Reviewed By: ot
Differential Revision:
D3116763
fb-gh-sync-id:
69023c0f26058516f25b9c1f9824055efc7021f9
fbshipit-source-id:
69023c0f26058516f25b9c1f9824055efc7021f9
Alexey Spiridonov [Thu, 7 Apr 2016 16:53:34 +0000 (09:53 -0700)]
DynamicParser to reliably and reversibly convert JSON to structs
Summary:We have a bunch of code that manually parses `folly::dynamic`s into program structures. I can be quite hard to get this parsing to be good, user-friendly, and concise. This diff was primarily motivated by the mass of JSON-parsing done by Bistro, but this pattern recurs in other pieces of internal code that parse dynamics.
This diff **not** meant to replace using Thrift structs with Thrift's JSON serialization / deserialization. When all you have to deal with is correct, structured plain-old-data objects produced by another program -- **not** manually entered user input -- Thrift + JSON is perfect. Go use that.
However, sometimes you need to parse human-edited configuration. The input JSON might have complex semantics, and require validation beyond type-checking. The UI for editing your configs can easily enforce correct JSON syntax. Perhaps, you can use `folly/experimental/JSONSchema.h` to have your edit UI provide type correctness. Despite all this, people can still make semantic errors, and those can be impossible to detect until you interpret the config at runtime. Also, as your system evolves, sometimes you need to break semantic backwards-compatibility for the sake of moving forward ? thus making previously valid configurations invalid, and requiring them to be fixed up manually.
So, people end up needing to write manual parsers for `dynamic`s. These all have very similar recurring issues:
- Verbose: to get an int field out of an object, typical code: (i) tests if the field is present, (ii) checks if the field is an integer, (iii) extracts the integer. Sometimes, you also want to handle exceptions, and compose helpful error messages. This makes the code far longer than its intent, and encourages people to write bad parsers.
- Unsystematic: sometimes, we use `if (const auto* p = dyn_obj.get_ptr("key")) { ... }`, other times we use `dyn_obj.getDefault()` or `if (dyn_obj.count())`, and so on. The patterns differ subtly in meaning. Exceptions sometimes get thrown, leading to error messages that cannot be understood by the user.
- Imperative parses: a typical parse proceeds step by step, and throws at the earliest error. This is bad because (i) errors have to be fixed one-by-one, instead of getting a full list upfront, (ii) even if 99% of the config is parseable, the imperative code has no way of recording the information it would have parsed after the first error.
`DynamicParser` fixes all of the above, and makes your parsing so clean that you might not even bother with `JSONSchema` as your first line of defense -- type-coercing, type-enforcing, friendly-error-generating C++ ends up being more concise. Besides all the sweet syntax sugar, `DynamicParser` lets you parse **all** the valid data in your config, while recording *all* the errors in a way that does not lose the original, buggy config. This means your code can parse a config that has errors, and still be able to meaningfully export it back to JSON. As a result, stateless clients (think REST APIs) can provide a far better user experience than just discarding the user?s input, and returning a cryptic error message.
For the details, read the docs (and see the example) in `DynamicParser.h`. Here are the principles of `DynamicParser::RECORD` mode in a nutshell:
- Pre-populate your program struct with meaningful defaults **before** you parse.
- Any config part that fails to parse will keep the default.
- Any config part that parses successfully will get to update the program struct.
- Any errors will be recorded with a helpful error message, the portion of the dynamic that caused the error, and the path through the dynamic to that portion.
I ported Bistro to use this in
D3136954. I looked at using this for JSONSchema's parsing of schemas, but it seemed like too much trouble for the gain, since it would require major surgery on the code.
Reviewed By: yfeldblum
Differential Revision:
D2906819
fb-gh-sync-id:
aa997b0399b17725f38712111715191ffe7f27aa
fbshipit-source-id:
aa997b0399b17725f38712111715191ffe7f27aa
Yedidya Feldblum [Thu, 7 Apr 2016 10:30:56 +0000 (03:30 -0700)]
Use folly::Function in folly::EventBase
Summary:[Folly] Use `folly::Function` in `folly::EventBase`.
`folly::Function` is moveable but noncopyable and therefore supports wrapping moveable but noncopyable lambdas - like the kind that arises when move-capturing a `std::unique_ptr`.
`std::function` is copyable - therefore it does not support wrapping such noncopyable lambdas.
Switching `folly::EventBase` to use it will allow callers to pass such noncopyable lambdas, allowing, e.g.:
```
auto numptr = folly::make_unique<int>(7); // unique_ptr is noncopyable
folly::EventBase eb;
eb.runInLoop([numptr = std::move(numptr)] { // therefore lambda is noncopyable
int num = *numptr;
});
eb.loop();
```
This allows us to move away from the `folly::MoveWrapper` hack, which worked like:
```
auto numptr = folly::make_unique<int>(7); // unique_ptr is noncopyable
auto numptrw = folly::makeMoveWrapper(std::move(numptr)); // MoveWrapper is "copyable" - hacky
folly::EventBase eb;
eb.runInLoop([numptrw] { // therefore lambda is "copyable" - hacky
int num = **numptrw;
});
```
We needed to do that hack while:
But neither condition is true anymore.
Reviewed By: spacedentist
Differential Revision:
D3143931
fb-gh-sync-id:
4fbdf5fb77eb5848ed1c6de942b022382141577f
fbshipit-source-id:
4fbdf5fb77eb5848ed1c6de942b022382141577f
Giuseppe Ottaviano [Wed, 6 Apr 2016 18:57:19 +0000 (11:57 -0700)]
Remove leftover comment in Portability.h
Reviewed By: luciang
Differential Revision:
D3145033
fb-gh-sync-id:
f37d0a2c6e22c1866134159b78722ff43d70918b
fbshipit-source-id:
f37d0a2c6e22c1866134159b78722ff43d70918b
Bert Maher [Wed, 6 Apr 2016 18:31:41 +0000 (11:31 -0700)]
malloc.h doesn't exist on OSX
Summary: Use malloc/malloc.h there instead
Reviewed By: Orvid
Differential Revision:
D3143768
fb-gh-sync-id:
3258b68df49cf7fbff28a3853663d39c147aee8b
fbshipit-source-id:
3258b68df49cf7fbff28a3853663d39c147aee8b
Michael Lee [Wed, 6 Apr 2016 15:00:14 +0000 (08:00 -0700)]
Start compiling a bit of `-Wshadow`
Summary: Sometimes variable shadowing is fine, sometimes it can cause subtle mistakes.
Reviewed By: yfeldblum
Differential Revision:
D3140450
fb-gh-sync-id:
acf7195ad65614d4f012bef8ce7dccb3a3038456
fbshipit-source-id:
acf7195ad65614d4f012bef8ce7dccb3a3038456
Sven Over [Wed, 6 Apr 2016 10:55:14 +0000 (03:55 -0700)]
folly::fibers::Fiber: use folly::Function instead of std::function
Summary:We are jumping through some ugly hoops in the implementation of
fibers to allow for non-copyable functors/lambdas. We can get rid of
those by using folly::Function instead of std::function to store
functions in folly::fibers::Fiber.
This involves one observable interface change: the virtual function
folly::fibers::InlineFunctionRunner::run must take a
folly::Function<void()> argument instead of a std::function<void()>.
Reviewed By: andriigrynenko
Differential Revision:
D3102711
fb-gh-sync-id:
56705b972dc24cc0da551109ed44732b97cb6e13
fbshipit-source-id:
56705b972dc24cc0da551109ed44732b97cb6e13
Dmitry Pleshkov [Wed, 6 Apr 2016 04:39:18 +0000 (21:39 -0700)]
Allow usage of Symbolizer options for ExceptionStats printing
Summary: For exceptions aggregation conveniece
Reviewed By: philippv
Differential Revision:
D3128132
fb-gh-sync-id:
48c72df364228c1d2c8f29161c2b85c131b4fea8
fbshipit-source-id:
48c72df364228c1d2c8f29161c2b85c131b4fea8
James Sedgwick [Tue, 5 Apr 2016 22:05:33 +0000 (15:05 -0700)]
Function::asStdFunction()
Summary:This is an ugly but occassionally useful hack to accomodate cases
where we want to propagate folly::Function but run into the brick wall that is
std::function
@override-unit-failures
Reviewed By: spacedentist
Differential Revision:
D3118432
fb-gh-sync-id:
80ed56494dfcf60e0f266013d880107acabc7dcf
fbshipit-source-id:
80ed56494dfcf60e0f266013d880107acabc7dcf
Giuseppe Ottaviano [Tue, 5 Apr 2016 19:30:17 +0000 (12:30 -0700)]
Update libgcc and boost symlinks (Disable SSO on ASan builds)
Summary: Allow ASan to detect access to invalidated strings even when they are small (in-situ). See
D3114022 for details.
Reviewed By: luciang
Differential Revision:
D3118161
fb-gh-sync-id:
3b83e3309d510ee9721dd505d176b09bdb7fd42d
fbshipit-source-id:
3b83e3309d510ee9721dd505d176b09bdb7fd42d
Sven Over [Tue, 5 Apr 2016 18:34:21 +0000 (11:34 -0700)]
folly/futures: replace MoveWrappers with generalised lambda capture
Summary:Now that folly::Future uses folly::Function, we can use non-copyable
callbacks. That allows us to get rid of folly::MoveWrapper in the
implementaion.
This diff also enforces perfect forwarding in the implementation of
folly::Future, thereby reducing the number of times that a callable
that is passed to Future::then et al. gets passed by value.
Before folly::Function, Future::then(callback) has invoked the move
constructor of the callback type 5 times for small callback objects
(fitting into the in-place storage inside folly::detail::Core) and
6 times for large callback objects. This has been reduced to 5 times
in all cases with the switch to UniqueFunction. This diff reduces it
to 2 times.
Reviewed By: yfeldblum
Differential Revision:
D2976647
fb-gh-sync-id:
9da470d7e9130bd7ad8af762fd238ef9a3ac5892
fbshipit-source-id:
9da470d7e9130bd7ad8af762fd238ef9a3ac5892