folly.git
8 years agoSupport GroupVarint.h under MSVC
Orvid King [Mon, 27 Jul 2015 23:06:57 +0000 (16:06 -0700)]
Support GroupVarint.h under MSVC

Summary: MSVC supports the intrinsics used by it without any issue other changes, so just make sure the error doesn't trigger on MSVC.
Closes #252

Reviewed By: @yfeldblum

Differential Revision: D2282942

Pulled By: @sgolemon

8 years agoAdd support for MSVC in asm_pause in Portability.h
Orvid King [Mon, 27 Jul 2015 23:16:56 +0000 (16:16 -0700)]
Add support for MSVC in asm_pause in Portability.h

Summary: This implements support for `asm_volatile_pause` and `asm_pause`, as defined in `Portability.h`, for MSVC.
This is needed because MSVC x64 doesn't support inline assembly.
Closes #254

Reviewed By: @yfeldblum

Differential Revision: D2283006

Pulled By: @sgolemon

8 years agoAdjust a literal in AtomicHashMap.h to be correctly sized
Orvid King [Mon, 27 Jul 2015 22:09:31 +0000 (15:09 -0700)]
Adjust a literal in AtomicHashMap.h to be correctly sized

Summary: Noticed because of a warning emitted by MSVC, this fixees the literal for `kLockedPtr_` to correctly be 64-bits under MSVC, rather than 32-bit.
Closes #245

Reviewed By: @yfeldblum

Differential Revision: D2282806

Pulled By: @sgolemon

8 years agoFix the size of the enum in FBString.h under MSVC
Orvid King [Mon, 27 Jul 2015 22:17:32 +0000 (15:17 -0700)]
Fix the size of the enum in FBString.h under MSVC

Summary: MSVC defaults enum sizes to int, so explicitly use `size_t` as the base to have it correctly sized.
Closes #246

Reviewed By: @yfeldblum

Differential Revision: D2282816

Pulled By: @sgolemon

8 years agofix clang ASAN error in DelayedDestructionBaseTest
Sophia Wang [Mon, 27 Jul 2015 18:38:54 +0000 (11:38 -0700)]
fix clang ASAN error in DelayedDestructionBaseTest

Summary: fix a use after free bug. this line is stupid!

Reviewed By: @siyengar

Differential Revision: D2283041

8 years agodelete folly/wangle
James Sedgwick [Mon, 27 Jul 2015 18:11:00 +0000 (11:11 -0700)]
delete folly/wangle

Summary: she's gooooooone, oh why, oh why, i better learn how to face it...

https://www.youtube.com/watch?v=bnVXIUyshng

Reviewed By: @djwatson

Differential Revision: D2219135

8 years agoBump version to 52:0 v0.52.0
Sara Golemon [Mon, 27 Jul 2015 16:46:56 +0000 (09:46 -0700)]
Bump version to 52:0

8 years agoMake ProcessReturnCode default-constructible
Alexey Spiridonov [Mon, 27 Jul 2015 02:36:57 +0000 (19:36 -0700)]
Make ProcessReturnCode default-constructible

Summary: We have this previously-unused "NOT STARTED" status, which I recently appropriated to denote moved-out `ProcessReturnCode`s.

It's natural to also use this for default-constructed `ProcessReturnCodes`. Lacking a default constructor leads to a bunch of unnecessarily annoying use of `folly::Optional` in my upcoming diff, so I wanted to get rid of that, see e.g.

differential/diff/7657906/

Reviewed By: @tudor

Differential Revision: D2097368

8 years agoImprove waitpid error handling
Alexey Spiridonov [Mon, 27 Jul 2015 02:32:03 +0000 (19:32 -0700)]
Improve waitpid error handling

Summary: Using `checkUnixError` after `waitpid()` is confusing and useless, because the only two possible errors are `ECHILD` (some other part of the program waited for this process!?) and `EINVAL` (invalid options, which are hardcoded in both `Subprocess::wait()` and `poll()`). Neither of these are recoverable. Moreover, even if the caller catches the exception, `~Subprocess` is still booby-trapped to `abort()` since its status remains `RUNNING`.

In short, just abort.

Reviewed By: @yfeldblum

Differential Revision: D2079415

8 years agoFix some warnings in folly
mwilliams [Fri, 24 Jul 2015 18:44:20 +0000 (11:44 -0700)]
Fix some warnings in folly

Summary: Remove a couple of unused variables, and move one
that was only used inside an ifdef

Reviewed By: @yfeldblum

Differential Revision: D2279989

8 years agoadded a missing hook to IndexedMemPool's testing harness
Nathan Bronson [Fri, 24 Jul 2015 21:57:33 +0000 (14:57 -0700)]
added a missing hook to IndexedMemPool's testing harness

Summary: IndexedMemPool had one bare usage of std::atomic, rather than
the templated type Atom.  This doesn't affect any non-testing template
instantiations, because those two are usually synonyms, but it could cause
spurious failures of DeterministicSchedule tests.  Found via inspection,
not via failed tests.

Reviewed By: @yfeldblum

Differential Revision: D2277424

8 years agoPrint type of singleton that is requested after destruction
Misha Shneerson [Fri, 24 Jul 2015 18:57:21 +0000 (11:57 -0700)]
Print type of singleton that is requested after destruction

Summary: As titled
Trying to troubleshoot shutdown problems in HHVM and I see:
  EventBase::runAfterDelay() callback threw St13runtime_error exception: Raw pointer to a singleton requested after its destruction.
Would be much easier to troubleshoot if I know what is it being requested.

Reviewed By: @chipturner

Differential Revision: D2274497

8 years agoimprovements to DeterministicSchedule
Nathan Bronson [Fri, 24 Jul 2015 15:20:45 +0000 (08:20 -0700)]
improvements to DeterministicSchedule

Summary: This diff isolates the CacheLocality back end for DeterministicSchedule
from one invocation to the next (the last one was deterministic
across the entire program but not per-DeterministicSchedule instance),
and makes it easy to do extensive tracing of memory accesses during
a deterministic test.  These changes were made while tracking down a
bug, but don't fix any bugs on their own (and in fact don't affect any
production code at all).

Reviewed By: @yfeldblum

Differential Revision: D1842390

8 years agoFix oss build
Andre Pinto [Fri, 24 Jul 2015 00:15:26 +0000 (17:15 -0700)]
Fix oss build

Summary: Add program_options to fix oss build

Reviewed By: @tudor

Differential Revision: D2274863

8 years agoDelete functions that return a pointer when the dynamic object is a rvalue.
Nick Terrell [Fri, 24 Jul 2015 00:46:30 +0000 (17:46 -0700)]
Delete functions that return a pointer when the dynamic object is a rvalue.

Summary: This diff is not yet complete, I want to see the contbuild before I change the
functions that return references to member functions.

It is unsafe to return a pointer when the dynamic object is a rvalue, because if
the pointer escapes the expression after the object is destroyed, we go into
segfault / undefined behavior land.
I have deleted these overloads.  The amount of valid code that is now disallowed
is minimal.  The only valid case I can think of is returing a pointer and
passing it to a function in the same expression that does not save the pointer.
However, this case is also dangerous, because if the function you pass it to
decides to save the pointer for later, we are in trouble, e.g.

  save_ptr(dynamic("str").c_str())

Since there are simple workarounds (naming the object), I think that is a small
price to pay for the greatly increased safety.

The next step is to overload all members that return a reference to a member
to move the member out if the dynamic is a rvalue:

  const dynamic& at(dynamic const&) const&;
        dynamic& at(dynamic const&)      &;
        dynamic  at(dynamic const&)      &&;  // Move out

I also need to go over the code more carefully to make sure that nothing went
wrong.

Reviewed By: @marcinpe

Differential Revision: D2257914

8 years agoHelper for writing nested command line apps
Tudor Bosman [Thu, 23 Jul 2015 14:30:52 +0000 (07:30 -0700)]
Helper for writing nested command line apps

Summary: Many command line apps are of the form
"program [--global_options] command [--command_options] args..."

Make writing such things less painful.

+jdelong because smcc

Reviewed By: @meyering

Differential Revision: D2217248

8 years agoMake gflags and boost::program_options play nice with each other
Tudor Bosman [Thu, 23 Jul 2015 14:30:41 +0000 (07:30 -0700)]
Make gflags and boost::program_options play nice with each other

Summary: GFlags is useful for global program options. boost::program_options makes it
easier to write command-line utilities. They don't, unfortunately, play
very nicely with each other.

Add a addGFlags() function to convert all GFlags to boost::program_options
options; you can then use boost::program_options::parse_command_line() to
parse all arguments, GFlags or not.

Also add a helper function to make parsing nested command lines easier.

Reviewed By: @fugalh

Differential Revision: D2215285

8 years agofix racy assert in SharedMutex
Nathan Bronson [Wed, 22 Jul 2015 22:14:58 +0000 (15:14 -0700)]
fix racy assert in SharedMutex

Summary: SharedMutex purposely allows the inline reader count to underflow in some
situations while preserving proper locking behavior (the inline reader
count is only trusted if all deferred locks have been applied), but there
was an additional way that this could occur that wasn't documented or
allowed by the asserts.  The effect was a false positive assert in rare
conditions or the possibility of losing track of a deferred lock slot.
This diff fixes both the over-aggressive assert and the potential loss
of the shared slot.  If the assert didn't fire for you then this diff
won't change the correctness of your program.

Reviewed By: @yfeldblum

Differential Revision: D2269018

8 years agoAdd executable_path() and convert SubprocessTest to use it
Tudor Bosman [Wed, 22 Jul 2015 19:43:36 +0000 (12:43 -0700)]
Add executable_path() and convert SubprocessTest to use it

Summary: Because I didn't want to write it again.

Reviewed By: @yfeldblum

Differential Revision: D2217246

8 years agomaking thrift and folly header files compile clean with -Wunused-parameter
Ondrej Lehecka [Wed, 22 Jul 2015 18:57:18 +0000 (11:57 -0700)]
making thrift and folly header files compile clean with -Wunused-parameter

Summary: fixing compiler errors when compiling with Wunused-parameter

Reviewed By: @yfeldblum

Differential Revision: D2267014

8 years ago(Wangle) window, fix test
Hannes Roth [Wed, 22 Jul 2015 16:35:19 +0000 (09:35 -0700)]
(Wangle) window, fix test

Summary: Thanks @​peijinz for finding this.

Reviewed By: @​peijinz

Differential Revision: D2268532

8 years agoAdd bounds check in get_ptr.
Nick Terrell [Tue, 21 Jul 2015 18:34:55 +0000 (11:34 -0700)]
Add bounds check in get_ptr.

Summary: `get_ptr()` checks if the index is past the end of the array, but it
doesn't check if it is less than 0, while `at()` does.

Reviewed By: @yfeldblum

Differential Revision: D2258548

8 years ago(Wangle) within should raise TimedOut()
Hannes Roth [Tue, 21 Jul 2015 02:42:45 +0000 (19:42 -0700)]
(Wangle) within should raise TimedOut()

Summary: A timeout now raises `TimedOut()` on the current Future.

Reviewed By: @fugalh

Differential Revision: D2261559

8 years agoFix "reference to 'thread' is ambiguous"
Hans Fugal [Tue, 21 Jul 2015 00:46:42 +0000 (17:46 -0700)]
Fix "reference to 'thread' is ambiguous"

Summary: This fixes github issue https://github.com/facebook/folly/issues/243

I don't understand why this is broken on osx, but the solution is to just `#include <thread>` in the header instead of trying to forward declare `std::thread`.

Reviewed By: @yfeldblum

Differential Revision: D2255026

8 years agofixing compiler errors when compiling with Wunused-parameter
Ondrej Lehecka [Mon, 20 Jul 2015 23:37:52 +0000 (16:37 -0700)]
fixing compiler errors when compiling with Wunused-parameter

Summary: fixing C++ compiler warnings for the -Wunused-parameter warning switch on.
This is fixing only the header files, not the library cpp files.

Reviewed By: @yfeldblum

Differential Revision: D2259957

8 years agoFix logspew in DelayedDestruction
Viswanath Sivakumar [Mon, 20 Jul 2015 23:07:03 +0000 (16:07 -0700)]
Fix logspew in DelayedDestruction

Summary: This is causing a log of spew in proxygen logs

Reviewed By: @afrind

Differential Revision: D2262006

8 years agoRemove EmptySequence exception in favor of optional returns
Tom Jackson [Mon, 20 Jul 2015 21:05:59 +0000 (14:05 -0700)]
Remove EmptySequence exception in favor of optional returns

Summary: Forces consumers to handle the empty sequence case instead of cumbersome exceptions.

Reviewed By: @​jrichey, @yfeldblum

Differential Revision: D2219505

8 years agoFix a namespace nit in folly/futures/Future.cpp.
Yedidya Feldblum [Mon, 20 Jul 2015 19:26:32 +0000 (12:26 -0700)]
Fix a namespace nit in folly/futures/Future.cpp.

Summary: [Folly] Fix a namespace nit in folly/futures/Future.cpp.

As soon as someone introduces a namespace `folly::futures::detail` this will blow up. Fix it so it won't blow up when that happens.

Reviewed By: @​hannesr

Differential Revision: D2258799

8 years agoBump version to 51:0 v0.51.0
Sara Golemon [Mon, 20 Jul 2015 19:26:38 +0000 (12:26 -0700)]
Bump version to 51:0

8 years agoAllow access to a FiberManager's currently running fiber
Brian Watling [Mon, 20 Jul 2015 18:46:02 +0000 (11:46 -0700)]
Allow access to a FiberManager's currently running fiber

Summary: Allow access to a FiberManager's currently running fiber

Reviewed By: @alikhtarov, @andriigrynenko

Differential Revision: D2257201

8 years agoOverload Optional::value() on object reference type.
Nick Terrell [Mon, 20 Jul 2015 16:03:11 +0000 (09:03 -0700)]
Overload Optional::value() on object reference type.

Summary: `folly::Optional` returns the stored value by reference when the object
is an rvalue.  This causes three issues:

  * If the user calls `value()` on an rvalue `Optional`, and assigns the result
    to a new variable, the copy constructor gets called, instead of the move
    constructor.  This causes the added test `value_move` to not compile.
  * If the user calls `value()` on an rvalue `Optional`, and assigns the result
    to a const lvalue reference, they might expect the lifetime to be extended
    when it isn't. See the added test `value_life_extention`.
  * Assigning the results of `value()` on an rvalue `Optional` to a mutable
    lvalue reference compiled in the old code, when it shouldn't, because that
    is always a dangling reference as far as I can tell.

I'm not sure how strict `folly` is with compatibility, but I believe the
breakage would be minimal, and any code that gets broken probably deserves it.

I'm not exactly sure who I should add as a reviewer, so hopefully herald has got
my back.

Reviewed By: @yfeldblum

Differential Revision: D2249548

8 years agofolly Singleton: clear some state if creator function fails
Steve O'Brien [Sun, 19 Jul 2015 16:22:34 +0000 (09:22 -0700)]
folly Singleton: clear some state if creator function fails

Summary: The creator thread ID is saved to indicate the singleton is already being built (to help detect dependency cycles).  However if the creation function throws an error, that thread ID persists, and then if the same thread tries again to build the singleton it will be falsely detected as a dependency cycle.  This clears that state in the case of failure.

Reviewed By: @chipturner

Differential Revision: D2256441

8 years agoRevert: (Wangle) within should raise TimedOut()
Hannes Roth [Sat, 18 Jul 2015 18:58:52 +0000 (11:58 -0700)]
Revert: (Wangle) within should raise TimedOut()

Summary: This reverts commit 956351018a495af89575526536af7e7c0bb285aa.

Reviewed By: @​labrams

Differential Revision: D2258219

8 years agofolly HHWheelTimer: fix loop variable
Steve O'Brien [Fri, 17 Jul 2015 19:41:02 +0000 (12:41 -0700)]
folly HHWheelTimer: fix loop variable

Summary: In nested loop, loop condition is incorrect.  Fixed var (should be `ii` not `i`)

Reviewed By: @pgriess

Differential Revision: D2255702

8 years agoMaking Optional throw exceptions instead of assert
Tom Jackson [Thu, 16 Jul 2015 20:59:07 +0000 (13:59 -0700)]
Making Optional throw exceptions instead of assert

Summary: I'm upgrading assertions to throws, since these are fatal in all circumstances. If something is explicitly `Optional`, it makes sense to fail loudly if it is misused in this manner.

Reviewed By: @yfeldblum

Differential Revision: D2247612

8 years agoadd SocketPair.h/.cpp to Makefile
James Sedgwick [Thu, 16 Jul 2015 17:16:45 +0000 (10:16 -0700)]
add SocketPair.h/.cpp to Makefile

Summary: to unbreak OSS wangle tests build

Reviewed By: @bugok

Differential Revision: D2240264

8 years agofolly::Future-istic barrier
Tudor Bosman [Thu, 16 Jul 2015 16:50:03 +0000 (09:50 -0700)]
folly::Future-istic barrier

Summary: What it says on the tin.

Reviewed By: @fugalh

Differential Revision: D2230390

8 years agoAdd an option to disable guard pages for fiber stacks
Pavlo Kushnir [Thu, 16 Jul 2015 04:06:29 +0000 (21:06 -0700)]
Add an option to disable guard pages for fiber stacks

Summary: guard pages may make debugging more painful. For example, in some cases they increased "perf" initialization time.

Reviewed By: @alikhtarov

Differential Revision: D2247188

8 years ago(Wangle) within should raise TimedOut()
Hannes Roth [Wed, 15 Jul 2015 23:39:48 +0000 (16:39 -0700)]
(Wangle) within should raise TimedOut()

Summary: I figured it out. This works. The two extra futures are a small overhead
(just two pointers). The `Core`s are allocated anyway, so this is pretty
much optimal.

A timeout now raises on the current Future, and a fulfilled promise
cancels the timeout Future.

Reviewed By: @yfeldblum

Differential Revision: D2232463

8 years agoOnly try to use F_SETPIPE_SZ if it's available
Sara Golemon [Wed, 15 Jul 2015 16:44:05 +0000 (09:44 -0700)]
Only try to use F_SETPIPE_SZ if it's available

Summary: As the comment says, we can ignore errors in setting the size.
So it stands to reason we can ignore setting the size as well.

Reviewed By: @yfeldblum

Differential Revision: D2242882

8 years agoMake DestructorGuard inherit from a base
Sophia Wang [Wed, 15 Jul 2015 00:18:18 +0000 (17:18 -0700)]
Make DestructorGuard inherit from a base

Summary: There are more use cases that the Destruction/Guard pattern can be
used than current DelayedDestruction provides. This diff makes the pattern more
general (remove self destruct) and lets DelayedDestruction derive from that.
The functionalities of DelayedDestruction is kept unchanged.

I leave destroy(), Destructor class, and destroyPending_ in DelayedDestruction
since they are not required by our CallbackGuard in proxygen.

I add a shouldDestruct() function to allow customized conditions on when to
call destructor.

I haven't made destroyNow() a std::function since I only need it to be set at
instatiation time. If there is any other use case that needs destroyNow() to be
a std::function, I can do that as well.

Reviewed By: @afrind

Differential Revision: D2202575

8 years agoActually denote when we have infinite generators
Joe Richey [Tue, 14 Jul 2015 20:50:31 +0000 (13:50 -0700)]
Actually denote when we have infinite generators

Summary: Before we didn't do anything at all with the ::infinite value for our
generators, now all the sources operators and sinks are properly notated. The
one signifcant change regarding what is allowed concerns 'cycle' which now (when
called with no arguments) always produces an infinite generator. This is fine
for all but the weirdest of generators fed into cycle.

Reviewed By: @ddrcoder

Differential Revision: D2240328

8 years agoChanging behavior of 'any' and 'all' sinks, adding in 'isEmpty' and 'notEmpty' sinks
Joe Richey [Tue, 14 Jul 2015 20:50:24 +0000 (13:50 -0700)]
Changing behavior of 'any' and 'all' sinks, adding in 'isEmpty' and 'notEmpty' sinks

Summary: When adding in the 'filter()' default behavior, I considered adding in similar
behavior for 'any' and 'all'. However, we had 'any' with no funciton call
basically check if anything was present, not testing a predicate. This can
create a confusing senario, so I removed this behavior from 'any' and added in
the 'isEmpty' and 'notEmpty' sinks. Now the calls 'any()' and 'all()' (called
with parens, so old uses won't compile) check for truthy values simlar to
'filter()'.

I also added some unit tests and changed 'static const' objects to 'constexpr'.

Reviewed By: @ddrcoder

Differential Revision: D2234637

8 years agoRequestContext support
Andrii Grynenko [Tue, 14 Jul 2015 17:29:36 +0000 (10:29 -0700)]
RequestContext support

Summary: Integrating RequestContext into fibers library. RequestContext is saved for every task, when that task is created. If RequestContext is overriden when a task is being run (within fiber, in runInMainContext, within function passed to await call) the new context will continue to be used for the task.

Reviewed By: @alikhtarov

Differential Revision: D2240152

8 years agofolly: AsyncIOOp: improve logging message: add strerror
Lucian Grijincu [Tue, 14 Jul 2015 17:31:26 +0000 (10:31 -0700)]
folly: AsyncIOOp: improve logging message: add strerror

Reviewed By: @philippv

Differential Revision: D2240628

8 years agoFix Build: folly/io/async/test/ScopedEventBasedThreadTest.cpp
Yedidya Feldblum [Tue, 14 Jul 2015 00:18:23 +0000 (17:18 -0700)]
Fix Build: folly/io/async/test/ScopedEventBasedThreadTest.cpp

Summary: [Folly] Fix Build: folly/io/async/test/ScopedEventBasedThreadTest.cpp

Failure with latest clang:

    folly/io/async/test/ScopedEventBaseThreadTest.cpp:72:8: error: explicitly moving variable of type 'folly::ScopedEventBaseThread' to itself [-Werror,-Wself-move]
      sebt = std::move(sebt);

Reviewed By: @markisaa

Differential Revision: D2238762

8 years agoFile ctor should take StringPiece.
Yedidya Feldblum [Mon, 13 Jul 2015 20:25:25 +0000 (13:25 -0700)]
File ctor should take StringPiece.

Summary: [Folly] File ctor should take StringPiece.

This way we can use it with `std::string` and `folly::fbstring` a touch more easily.

Reviewed By: @luciang

Differential Revision: D2235870

8 years agoBump version to 50:0 v0.50.0
Sara Golemon [Mon, 13 Jul 2015 18:17:36 +0000 (11:17 -0700)]
Bump version to 50:0

8 years agoFix examples in folly::Singleton code
Andrii Grynenko [Mon, 13 Jul 2015 17:39:32 +0000 (10:39 -0700)]
Fix examples in folly::Singleton code

Reviewed By: @chipturner

Differential Revision: D2235232

8 years agosplit SmallLocks.h to get a larger portion of folly compiled on 32bit platforms
Kjell Schubert [Sat, 11 Jul 2015 16:12:08 +0000 (09:12 -0700)]
split SmallLocks.h to get a larger portion of folly compiled on 32bit platforms

Summary: This is only a simple #include file split, no functional change was made. SmallLocks.h provides 2 classes: MicroSpinLock and PicoSpinLock, with the latter not being available on 32bit platforms. There are several classes in folly that included SmallLocks.h but only needed MicroSpinLock, these also refused to compile on 32bit, e.g. folly/futures/Future.h. Now these compile fine.

Reviewed By: @yfeldblum

Differential Revision: D2234263

8 years agoAn has_member_type macro.
Yedidya Feldblum [Sat, 11 Jul 2015 04:24:17 +0000 (21:24 -0700)]
An has_member_type macro.

Summary: [Folly] An has_member_type macro.

Reviewed By: @juchem

Differential Revision: D2229775

8 years agoAdded default value for filter's predicate argument
Joe Richey [Thu, 9 Jul 2015 21:48:17 +0000 (14:48 -0700)]
Added default value for filter's predicate argument

Summary: After discussion with Tom about frozen files, we noticed that the use
case of filtering out somthing that looks like false (0, nullptr, folly::none)
is unnecessarily tedious. So folly::gen::filter now filters by the value of
the item by default.

Reviewed By: @ddrcoder

Differential Revision: D2220582

8 years agoAdd mocks for isBufferMovable
Subodh Iyengar [Thu, 9 Jul 2015 05:44:59 +0000 (22:44 -0700)]
Add mocks for isBufferMovable

Summary: Add mocks for the new isBufferMovable methods.

Reviewed By: @yfeldblum

Differential Revision: D2222642

8 years agoThreadedExecutor.
Yedidya Feldblum [Thu, 9 Jul 2015 00:54:16 +0000 (17:54 -0700)]
ThreadedExecutor.

Summary: [Folly] ThreadedExecutor.

It's an Executor that runs functions each in its own thread.

Kind of simple. Suitable for a few types of strange cases.

Reviewed By: @​hannesr

Differential Revision: D2226211

8 years agoAdd isManaged / makeManaged
Tudor Bosman [Wed, 8 Jul 2015 15:39:55 +0000 (08:39 -0700)]
Add isManaged / makeManaged

Summary: Add a way to tell whether an IOBuf (either indvidually or the whole chain)
refers to memory that's outside of IOBuf's control (via WRAP_BUFFER). That is,
can you assume that clone() will extend the lifetime of the memory? Is it safe
to use clone() instead of copying the data (and avoid a copy)?

Reviewed By: @simpkins

Differential Revision: D2191883

8 years agofolly/singleton: fatal in unrecoverable error cases
Steve O'Brien [Tue, 7 Jul 2015 23:10:07 +0000 (16:10 -0700)]
folly/singleton: fatal in unrecoverable error cases

Summary: Early in the startup process there may not be a default signal handler installed, and stack traces are not available; also during the startup process is when init-order fiascos occur.  Dump a stacktrace and fatal when an unregistered singleton is used.

Also fatals -- with glog `LOG(FATAL)`, which triggers an ABRT signal -- on other illegal and unrecoverable usage like double-registration or circular dependency.

Reviewed By: @andriigrynenko

Differential Revision: D2200408

8 years agohigh init priority for FiberManagerMap data
Mark McDuff [Tue, 7 Jul 2015 22:20:59 +0000 (15:20 -0700)]
high init priority for FiberManagerMap data

Summary: Ran into a bug in D2158570 where we dtor'ing these data structures before the event bases destructed, so the assert on line 47 (of the old code) failed.

Reviewed By: @andriigrynenko

Differential Revision: D2198318

8 years agoBump version to 49:1 v0.49.1
Sara Golemon [Tue, 7 Jul 2015 20:45:17 +0000 (13:45 -0700)]
Bump version to 49:1

8 years agoInstall Ubuntu 12.04 LTS dependencies for Travis CI
Alexey Spiridonov [Tue, 7 Jul 2015 20:36:16 +0000 (13:36 -0700)]
Install Ubuntu 12.04 LTS dependencies for Travis CI

Summary: This is the first of a series of diffs to get folly, fbthrift, and bistro to run on Ubuntu 12.04 LTS, which is what powers Travis CI.

It's important for folly to be easily usable with Travis CI, since everything C++ depends on it, and we really want all FB open-source projects to be monitored via continuous integration to avoid breaking their OSS builds.

Thanks to @int's work on `mcrouter`, this was considerably easier than it would have been otherwise.

This puts us a hop and a skip away from actually running folly's `make && make check && make install` on Travis CI. Do y'all think this is useful?

Reviewed By: @yfeldblum

Differential Revision: D2217441

8 years agoBump version to 49:0 v0.49.0
Sara Golemon [Tue, 7 Jul 2015 17:26:57 +0000 (10:26 -0700)]
Bump version to 49:0

8 years agofolly/wangle -> wangle cutover
James Sedgwick [Mon, 6 Jul 2015 18:35:30 +0000 (11:35 -0700)]
folly/wangle -> wangle cutover

Summary: Another stab at D2111063 now that we have the go ahead to actually open up the wangle github repo

Reviewed By: @djwatson, @mdordal

Differential Revision: D2178291

8 years agocleanup CompressedList / MutableCompressedList interfaces
Philip Pronin [Mon, 6 Jul 2015 00:29:39 +0000 (17:29 -0700)]
cleanup CompressedList / MutableCompressedList interfaces

Summary: Drop those `const_cast<>` hacks.

Reviewed By: @ot

Differential Revision: D2217629

8 years agofolly/futures Update README.md
Hans Fugal [Thu, 2 Jul 2015 23:03:25 +0000 (16:03 -0700)]
folly/futures Update README.md

Summary: This is a documentation dump, with updates from the past couple weeks. Most notably, the new information about `Future<Unit>` replacing `Future<void>`.

Reviewed By: @​hannesr

Differential Revision: D2211135

8 years agoLift thrift/lib/cpp/test/TNotificationQueueTest.
Yedidya Feldblum [Thu, 2 Jul 2015 20:58:58 +0000 (13:58 -0700)]
Lift thrift/lib/cpp/test/TNotificationQueueTest.

Summary: [Folly] Lift thrift/lib/cpp/test/TNotificationQueueTest.

`NotificationQueue` is already moved into folly; move its accompanying test suite as well.

Reviewed By: @simpkins

Differential Revision: D2207104

8 years agoFix typo - ac defined are prefixed with FOLLY_
Sara Golemon [Thu, 2 Jul 2015 21:02:18 +0000 (14:02 -0700)]
Fix typo - ac defined are prefixed with FOLLY_

Summary: Got this ifdef wrong in my last change here.

Reviewed By: @paulbiss

Differential Revision: D2213708

8 years agonice error message if you try to Future<void>
Hans Fugal [Thu, 2 Jul 2015 18:43:42 +0000 (11:43 -0700)]
nice error message if you try to Future<void>

Summary: If you try to make a future or promise of `void`, you will get this nice error message telling you to use `Unit` instead.

Reviewed By: @djwatson

Differential Revision: D2211239

8 years agoAdd an exception for building on OSX.
Michael Lee [Thu, 2 Jul 2015 14:49:15 +0000 (07:49 -0700)]
Add an exception for building on OSX.

Summary: The Homebrew setup fails to compile on OSX.

Reviewed By: @paulbiss

Differential Revision: D2212084

8 years agofolly::Unit::Drop.
Yedidya Feldblum [Thu, 2 Jul 2015 01:39:41 +0000 (18:39 -0700)]
folly::Unit::Drop.

Summary: [Folly] folly::Unit::Drop.

Antisymmetric to folly::Unit::Lift.

Reviewed By: @fugalh

Differential Revision: D2211725

8 years agominor Timekeeper bug
Hans Fugal [Thu, 2 Jul 2015 01:21:28 +0000 (18:21 -0700)]
minor Timekeeper bug

Summary: Fix so `Timekeeper::at(now() - something)` works. Also introduce a test which explicitly tests when <= now codepath, which wasn't broken per se here, but which test also tickled this bug.

Reviewed By: @jsedgwick

Differential Revision: D2209166

8 years agoDiscourage Duration in code comments and tests
Hans Fugal [Thu, 2 Jul 2015 01:18:53 +0000 (18:18 -0700)]
Discourage Duration in code comments and tests

Reviewed By: @yfeldblum

Differential Revision: D2209095

8 years agoFuture<Unit> global fixup
Hans Fugal [Wed, 1 Jul 2015 22:42:50 +0000 (15:42 -0700)]
Future<Unit> global fixup

Summary: This is three codemods:

* `(folly::(Future|Promise|Try))<void>` -> `\1<folly::Unit>`
* `(Future|Promise|Try)<void>` -> `\1<Unit>`
* add `using folly::Unit` statements where needed

Then
* undo false positives in javascript files and fibers::Promise

cf D2201259, which this will land with

Reviewed By: @djwatson

Differential Revision: D2201801

8 years agoFuture<Unit> wangle fixup
Hans Fugal [Wed, 1 Jul 2015 22:42:30 +0000 (15:42 -0700)]
Future<Unit> wangle fixup

Summary: Mostly this is `s/(Promise|Future|Try)<void>/\1<Unit>/`

Reviewed By: @djwatson

Differential Revision: D2201272

8 years agoNuke Future<void> (folly/futures)
Hans Fugal [Wed, 1 Jul 2015 22:42:23 +0000 (15:42 -0700)]
Nuke Future<void> (folly/futures)

Summary: Nuke Future<void> in favor of Future<Unit>, for the `folly/futures` subdirectory.

Reviewed By: @djwatson

Differential Revision: D2201259

8 years agoDo an explicit test for XSI strerror_r
Sara Golemon [Wed, 1 Jul 2015 21:56:12 +0000 (14:56 -0700)]
Do an explicit test for XSI strerror_r

Summary: The current #if directive is a laundry list of
"strerror_r defaults to XSI style on these platforms",
and we keep missing some, so let's just do a proper test
and simplify the code directives.

Closes #232

Reviewed By: @paulbiss

Differential Revision: D2208843

8 years ago(easy) wrong test suite name
Hans Fugal [Wed, 1 Jul 2015 22:07:31 +0000 (15:07 -0700)]
(easy) wrong test suite name

Reviewed By: @yfeldblum

Differential Revision: D2209197

8 years agoImplement a BitVector-based list
Vignesh Gowda [Wed, 1 Jul 2015 21:07:37 +0000 (14:07 -0700)]
Implement a BitVector-based list

Summary: Implemented a BitVector Encoder and BitVector Reader with the same interface as EliasFanoCoding.h

Reviewed By: @ot, @philippv

Differential Revision: D2198935

8 years agofix? AutoTimerTest
Andrey Goder [Wed, 1 Jul 2015 17:37:34 +0000 (10:37 -0700)]
fix? AutoTimerTest

Summary: Apparently this test sometimes fails. @Gownta suggests it's because we're
storing temproaries in StringPiece. Change it to store in a std::string, which
should hopefully fix it.

Reviewed By: @Gownta

Differential Revision: D2208643

8 years agomake it really clear that wangle/rx is deprecated
Hans Fugal [Wed, 1 Jul 2015 17:24:18 +0000 (10:24 -0700)]
make it really clear that wangle/rx is deprecated

Summary: More people are starting to use Rx. Yay! But they're trying to use `wangle/rx`. So make sure there's no doubt about its deprecated status.

Reviewed By: @yfeldblum

Differential Revision: D2196273

8 years agoAdd missing include for flock()
Sara Golemon [Wed, 1 Jul 2015 16:00:48 +0000 (09:00 -0700)]
Add missing include for flock()

Summary: flock() expects sys/file.h which on many platforms
is getting included as a side-effect, but on at least one
this isn't the case.

Add the explicit include to fix builds on those platforms.

Closes #232

Facebook

Unrelated runtime failures.

Reviewed By: @JoelMarcey

Differential Revision: D2196306

8 years agomake folly/detail/FunctionalExcept.* work if FOLLY_HAVE_BITS_FUNCTEXCEPT_H is set...
Marc Horowitz [Tue, 30 Jun 2015 19:39:50 +0000 (12:39 -0700)]
make folly/detail/FunctionalExcept.* work if FOLLY_HAVE_BITS_FUNCTEXCEPT_H is set or not

Summary: When doing android development, we need to compile with
FOLLY_HAVE_BITS_FUNCTEXCEPT_H set for some flavors, and
FOLLY_HAVE_BITS_FUNCTEXCEPT_H unset for others (primarily for tests
run in the local (xcode) development platform).  This diff adds some
more conditionals so it all works.

In this repro, these files are never used, so this is more or less a
noop, but this is the upstream so I'm making these changes here.

Reviewed By: @Gownta

Differential Revision: D2202107

8 years agocheck ReadCallback before using it
Woo Xie [Tue, 30 Jun 2015 16:21:08 +0000 (09:21 -0700)]
check ReadCallback before using it

Summary: before the first setReadCallback(), the readCallback_ is nullptr.  Thanks @siyengar for pointing out this!

Reviewed By: @siyengar

Differential Revision: D2203489

8 years agomove the socket setting into setReadCB()
Woo Xie [Mon, 29 Jun 2015 23:43:14 +0000 (16:43 -0700)]
move the socket setting into setReadCB()

Summary: refactor D2201181

Reviewed By: @afrind

Differential Revision: D2202245

8 years agoBump version to 48:0 v0.48.0
Sara Golemon [Mon, 29 Jun 2015 18:01:35 +0000 (11:01 -0700)]
Bump version to 48:0

8 years agoa subset of folly now compiles with Windows mingw64 toolchain
Kjell Schubert [Mon, 29 Jun 2015 15:33:54 +0000 (08:33 -0700)]
a subset of folly now compiles with Windows mingw64 toolchain

Summary: a subset of folly now compiles with Windows mingw64 toolchain

Reviewed By: @djwatson

Differential Revision: D2171872

8 years agocleanup includes in folly/Format.h
Philip Pronin [Sun, 28 Jun 2015 01:09:24 +0000 (18:09 -0700)]
cleanup includes in folly/Format.h

Summary: `folly/Format.h` pulls in `folly/small_vector.h` and
`folly/FBVector.h`, which (according to @​oleksandr's analysis)
are the most expensive includes for multiple fbcode projects.

Reviewed By: @ot

Differential Revision: D2198904

8 years agoFix resplit | batch
Tom Jackson [Fri, 26 Jun 2015 20:30:07 +0000 (13:30 -0700)]
Fix resplit | batch

Summary: Resplit violated the contract of `apply()`; it returned `false` even though the input sequence was fully consumed.

Reviewed By: @philippv

Differential Revision: D2195889

8 years agoRemove double-conversion as a source dep in OS X script
Mike Arpaia [Fri, 26 Jun 2015 18:37:16 +0000 (11:37 -0700)]
Remove double-conversion as a source dep in OS X script

Summary: double-conversion was recently added to homebrew, making it the last
folly dependency to do so. this diff modifies the existing OS X
bootstrap tool to account for that. I also added an rm for the gtest
wget, so there wasn't a leftover zip file in your repo. alternatively,
we could remove that and add it to the gitignore.

Reviewed By: @fugalh

Differential Revision: D2196000

Pulled By: @sgolemon

8 years agoFix passing RequestContext to executor thread
Alex Landau [Wed, 24 Jun 2015 23:20:08 +0000 (16:20 -0700)]
Fix passing RequestContext to executor thread

Summary: If x->add() executes its lambda on a different thread and
doesn't pass the context on its own, the callback wouldn't
have the correct context set.

Reviewed By: @djwatson

Differential Revision: D2189318

8 years agoFix stop_ thread race
Dave Watson [Fri, 26 Jun 2015 16:19:32 +0000 (09:19 -0700)]
Fix stop_ thread race

Summary: make it std::atomic

Reviewed By: @tudor

Differential Revision: D2191861

8 years agoto_shared_ptr.
Yedidya Feldblum [Fri, 26 Jun 2015 06:11:30 +0000 (23:11 -0700)]
to_shared_ptr.

Summary: [Folly] to_shared_ptr.

So you can write this:

    auto sptr = to_shared_ptr(getSomethingUnique<T>());

Instead of this:

    auto sptr = shared_ptr<T>(getSomethingUnique<T>());

Useful when `T` is long, such as `T = foobar::cpp2::FooBarServiceAsyncClient`.

Reviewed By: @meyering

Differential Revision: D2193572

8 years agoopt proxygen with newly added OpenSSL functions
Woo Xie [Fri, 26 Jun 2015 03:11:55 +0000 (20:11 -0700)]
opt proxygen with newly added OpenSSL functions

Summary: this taks advantage of SSL_write_iovec and SSL_read_buf functions to improve CPU.  This eliminates one malloc+memcpy+free operation for all HTTPS traffic, and save up to two for small packets.

It saves 10~15% CPU
https://fburl.com/99681397
https://fburl.com/99427544

Reviewed By: @djwatson

Differential Revision: D1993764

8 years agoGroupBy
Tom Jackson [Thu, 25 Jun 2015 23:39:22 +0000 (16:39 -0700)]
GroupBy

Reviewed By: @ot

Differential Revision: D2188740

8 years ago(easy) remove cruft comment
Hans Fugal [Thu, 25 Jun 2015 19:17:51 +0000 (12:17 -0700)]
(easy) remove cruft comment

Summary: We have this now.

Reviewed By: @jsedgwick

Differential Revision: D2191642

8 years ago(Wangle) Fix possible race in updating FSM state
Hannes Roth [Thu, 25 Jun 2015 15:44:29 +0000 (08:44 -0700)]
(Wangle) Fix possible race in updating FSM state

Summary: Storing the new state could be a memory race according to C++ (but wasn't in practice). I only checked GCC though.

Reviewed By: @nbronson

Differential Revision: D2189287

8 years ago(Wangle) Clean up some data races in tests
Hannes Roth [Thu, 25 Jun 2015 15:41:28 +0000 (08:41 -0700)]
(Wangle) Clean up some data races in tests

Summary: Not 100% about the `ThreadWheelTimekeeper` test. It makes TSAN happy
though.

Reviewed By: @yfeldblum

Differential Revision: D2187901

8 years ago(Wangle) Fix bug with CrappyExecutors, and bad PriorityExecutor
Hannes Roth [Thu, 25 Jun 2015 15:38:14 +0000 (08:38 -0700)]
(Wangle) Fix bug with CrappyExecutors, and bad PriorityExecutor

Summary: 1) We forgot to `--attached_` if `x` throws an exception
2) `PriorityExecutor` didn't execute `Func`, causing leaks in the test (not a bug in Futures)
3) I moved up the initialization for an empty `Core` into the constructor to make it easier to see

Reviewed By: @jsedgwick

Differential Revision: D2187343

8 years agoDon't hardcode thread limit in ServerBootstrap
Sara Golemon [Wed, 24 Jun 2015 17:48:20 +0000 (10:48 -0700)]
Don't hardcode thread limit in ServerBootstrap

Summary: This was set to 32 which probably works fine for the
majority of FB servers, but not everyone's system is so endowed.

Closes #119

Reviewed By: @fredemmott

Differential Revision: D2187008

8 years ago(Wangle) Remove unnecessary constructor argument to CollectAnyContext
Hannes Roth [Wed, 24 Jun 2015 17:20:39 +0000 (10:20 -0700)]
(Wangle) Remove unnecessary constructor argument to CollectAnyContext

Summary: See title.

Reviewed By: @fugalh

Differential Revision: D2186833

8 years agoFix tests for glog and gflags
Sara Golemon [Wed, 24 Jun 2015 16:31:49 +0000 (09:31 -0700)]
Fix tests for glog and gflags

Summary: The existing tests were trying to find undocumented
symbols in glog/glags which aren't guaranteed to be there on all
platforms.

Use AC_RUN_IFELSE() to try building and executing some code
typical of the usage of these libraries.

Closes #117

Reviewed By: @JoelMarcey

Differential Revision: D2183526

8 years agoChange visibility of folly::Arena constructors to public
Sara Golemon [Wed, 24 Jun 2015 16:32:24 +0000 (09:32 -0700)]
Change visibility of folly::Arena constructors to public

Summary: In the case of the copy constructor, this change
doesn't really matter, since it's deleted anyway.

In the case of the move constructor, it fixes what was probably
an unintentional hiding.  The class certainly looks like
it supports moving.

Closes #121

Reviewed By: @yfeldblum

Differential Revision: D2184131