Summary: [Folly] Generate group varint tables at compile time in C++, rather than as a separate custom build step in Python.

Reviewed By: simpkins

Differential Revision: D6832563

fbshipit-source-id: fa218c512cbac383f153e7b6fd4df7f181bcb0de

Summary: [Folly] Generate format tables at compile time in C++, rather than as a separate custom build step in Python.

Reviewed By: ot

Differential Revision: D6830372

fbshipit-source-id: 25770676e59c8070eaef7cbb691a5ba1c4b0a8f8

Summary: [Folly] Generate format tables at compile time in C++, rather than as a separate custom build step in Python.

Reviewed By: ot

Differential Revision: D6829681

fbshipit-source-id: 160b5a7616b015cc37ff389adabfec095cf136ec

Summary: Fleshes out the SemiFuture API with waiting versions of xVia calls that take a TimedDrivableExecutor that can be driven until some time point.

Reviewed By: yfeldblum

Differential Revision: D6658862

fbshipit-source-id: 74906d8101d2a762a27da1a1ec8348f106037f5d

Summary:
[Folly] Use `invoke` in `applyTuple` implementation.

V.s. reimplementing `invoke` as `makeCallable`

Reviewed By: spacedentist

Differential Revision: D6843993

fbshipit-source-id: 82f3e121486972245b75f808654d0a1299b7eafe

Summary:
[Folly] Fix missing "last-minute-reprieve" load-acquire in `Baton` and `SaturatingSemaphore`.

If the CAS fails, it is because `post` with its store-release was invoked just after spinning but just before parking the thread. So the failure load needs a matching load-acquire.

We cannot actually move the load-acquire into the CAS until C++17, since the failure memory order must be not stronger than the success memory order until C++17. Therefore, simulate the effect with an acquire fence in the CAS failure case.

The CAS success case does not need a load-acquire; there is already a load-acquire after the thread is unparked.

Reviewed By: davidtgoldblatt

Differential Revision: D6851304

fbshipit-source-id: a43243ad6a35bd9a53a9d2a5958b50a6c8806419

Summary:
It can be dangerous to set a CMake variable and not wrap the values as strings at times. The arguments can be forwarded to the compiler in use which will not know how to handle multiple arguments getting treated as one. For an example error using CMake 3.9.6 and Clang 5.0.1, if you do not wrap `CMAKE_CXX_FLAGS_COMMON` using string literals, you will get an error:

```
Problems were encountered while collecting compiler information:
	clang-5.0: error: unknown argument: '-g;-Wall;-Wextra'
```

Closes https://github.com/facebook/folly/pull/746

Reviewed By: simpkins

Differential Revision: D6854510

Pulled By: yfeldblum

fbshipit-source-id: c3ae626033e6733d0ce6f19fc6c06190315bf0db

Summary: [Folly] `make_array_with`, for making `std::array` given size and item ctor

Reviewed By: ot

Differential Revision: D6831280

fbshipit-source-id: 448bbf4843ffa70324f1a1d2d9e2790dc296548d

Summary: Flakyness was caused by a small chance that the function would still be in the heap when adding a function. Increased delay to 3 seconds.

Reviewed By: yfeldblum

Differential Revision: D6855408

fbshipit-source-id: 47d4035698a5d14a7ed71902656210f0e17b0b5d

Summary: Add steps to hazptr_obj construction and retirement to catch misuse bugs such as double retire.

Reviewed By: yfeldblum

Differential Revision: D6851965

fbshipit-source-id: 026e578b595792e3118b6b4b9a22c8c1dce6acba

Summary: In future chains destruction of the future then the promise will cause an exception to propagate. If all the work builds up in the executor and the executor does not drain then the exception does not propagate.

Reviewed By: yfeldblum

Differential Revision: D6837229

fbshipit-source-id: 095650b8aae0de4ddbe86c044d78a6fac45f9c46

Summary:
The googletest framework does not have a built-in mechanism for skipping tests
at run-time.  Folly uses a `SKIP()` macro for some tests to report themselves
as skipped.  For our internal builds these are reported as failures with a
special message that gets handled by our test runner framework.  However for
open source builds it is better to report these tests as passing rather than as
failing.

Reviewed By: yfeldblum

Differential Revision: D6843331

fbshipit-source-id: f74f29354305703448e5757ddc0ec3e72380a8f7

Summary:
This reverts commit 5ad448f0aad8f51f4328bfe9af37d63023089e88

bypass-lint

An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
cause_a_sev_many_files

Differential Revision: D6851455

fbshipit-source-id: d96ac9c2f907f09a4086ccf65ec4a3ef47cad205

Summary:
[Folly] Fix when the Timekeeper is destroyed with in-flight delays.

Previously, the Timekeeper dtor would immediately trigger the delayed callbacks early with a success result. But there are various situations where that is wrong behavior, such as when the callback enqueues more delayed callbacks in the success path. Higher-level, a future returned by `Timekeeper::after` or `sleep` would be completed early with a value.

Now, let it immediately trigger the delayed callbacks early with an error result. Higher-level, a future returned by `Timekeeper::after` or `sleep` will be completed early with an error.

Also: use `NoTimeout` consistently, and, on interrupt, pass the interrupt exception through.

Differential Revision: D6837669

fbshipit-source-id: 7d18d7f57999d56bbb5269cacdb443ac22448cf0

Summary: [Folly] Be more explicit about exceptions in `ThreadWheelTimekeeper`.

Differential Revision: D6842211

fbshipit-source-id: 9ef9c7149b610bbad76a08ef771aa13870d4467f

Summary: [Folly] When invoking `stealPromise`, capture the result.

Differential Revision: D6842173

fbshipit-source-id: 517a88744ecaff660f0e6615af541211018fa690

Summary:
[Folly] Use `stealPromise` v.s. `releasePromise` in `ThreadedWheelTimekeeper`.

This makes it easier to to combine working with the promise with releasing the promise.

Differential Revision: D6842090

fbshipit-source-id: ff3859ed25c1d2a779068603109d1fc801771377

Summary:
The code in AsyncSocketTest.ConnectTimeout attempts to pick an IP address to
connect to that will time out rather than failing immediately.  It checks to
see if IPv4 or IPv6 is available, but it cannot easily check if routes are
available.  If IPv6 is available but no routes are configured it will use IPv6,
and then the connect attempt will fail immediately rather than timing out.

This changes the test to simply report itself as skipped in this case rather
than failing.

Reviewed By: yfeldblum

Differential Revision: D6843332

fbshipit-source-id: 02ef3dd42f1d4b6e35edfe93b2e5e42c6666fa52

Summary:
The xlog test currently fails due to the way that CMake invokes the compiler
with absolute path names.  In the future we should probably update the xlog
code with the ability to strip off some directory prefixes.

Reviewed By: yfeldblum

Differential Revision: D6817547

fbshipit-source-id: 40f08d56d0199c21808a4046ebc1a3a9d70a33e9

Summary:
The logic in `AsyncSocketTest.WriteErrorCallbackBytesWritten` relied on overly
strict assumptions about how socket buffering would be implemented by the
kernel.  The tests would fail spuriously if the kernel allowed more data to be
buffered than the test assumed.

This updates the tests to send more data and allow the kernel to buffer more
data than it initially expects.

Reviewed By: yfeldblum

Differential Revision: D6843330

fbshipit-source-id: a53e19028d50c5d646654e67518a80d6b68106a2

Summary:
The UNIX builds treat all warnings as errors, and this missing `override` keyword causes a warning which is then a hard error in the build. In Boost 1.59 and later, `Boost.ProgramOptions` provides a virtual method `adjacent_tokens_only()` which the class `GFlagValueSemanticBase` wants to override. As such, mark it using `override`.
Closes https://github.com/facebook/folly/pull/743

Differential Revision: D6851455

Pulled By: yfeldblum

fbshipit-source-id: 5ad448f0aad8f51f4328bfe9af37d63023089e88

Summary:
Update OpenSSLPortabilityTest to work with older OpenSSL versions that do not
have the ASN1_TIME_diff() function.

Reviewed By: anirudhvr

Differential Revision: D6843333

fbshipit-source-id: 1d9cb104676296170df28fda9285c76680f8ef92

Summary:
Update NextProtocolTest.NpnTestNoOverlap to work with OpenSSL 1.0.2g.
The test previously had a comment explicitly indicating that it did not work
with this version of OpenSSL.

Reviewed By: anirudhvr

Differential Revision: D6843334

fbshipit-source-id: 2c179597e5b48b5284601ca84299da8b9a9c0e62

Summary: Added noexcept to drive functionality of executors as exceptions should not be leaking from these operations.

Reviewed By: yfeldblum

Differential Revision: D6846415

fbshipit-source-id: 569819ef94d28c9c8c060cc8015247c416b8ab45

Summary: Keeping moved out functions on the heap leads to segfaults. Let's fix that by calling make_heap each time we need to update. We will regress resetFunctionTimer complexity to O(n), which is exactly what it was before D5668557.

Reviewed By: simpkins

Differential Revision: D6822488

fbshipit-source-id: 94f75789b01ff85e2df271822b767160285462c3

Summary:
Update EventBaseTest.IdleTime() to drive the loop once before starting the bulk
of the test logic.  This triggers the initial NotificationQueue callbacks which
can otherwise slow down the test slightly and make it more prone to failing due
to not accounting for the time required for these callbacks.

Also replace `ASSERT_*()` checks inside a lambda with `EXPECT_*()`.
The googletest `ASSERT_*()` macros only work directly inside the main test
function, as they `return` from the current function.  When used inside the
lambda these skip the remainder of the lambda but continue running the
remainder of the test checks, which causes incorrect failures.

Reviewed By: yfeldblum

Differential Revision: D6843335

fbshipit-source-id: 879901edb6e5a5a84736705fb2d1cb8d2e0cc252

Summary:
Include the fbcode_builder sources in the folly repository under
build/fbcode_builder/.  These build scripts will make it easier. to set
up continuous integration for the opensource github repository.

Note that for folly this is potentially slightly confusing, since this
will now add a top-level build directory in addition to the existing
folly/build subdirectory.  However, it seems easiest for now to follow
the default fbcode_builder conventions.

This does not yet include configuration files for the folly build--I
will add those in a subsequent diff.

Test Plan:
Confirmed the CI build work with the upcoming configuration files.

Reviewed By: yfeldblum

Differential Revision: D6700052

fbshipit-source-id: ba026835e866bda5eed0776ec8bbf1ac51014034

Summary:
Some compilers such as Apple Clang do not support the compiler flag `-fopenmp`. So, do not add it by default to the list of `CXX_FLAGS` that are expected for *every* compiler to support. Instead, check if the `CXX_COMPILER` specified by `CMake` supports `-fopenmp` by using `CHECK_CXX_COMPILER_FLAG`.
Closes https://github.com/facebook/folly/pull/741

Reviewed By: simpkins

Differential Revision: D6843171

Pulled By: yfeldblum

fbshipit-source-id: 671958d867400b5802d7d563036639d61d11f615

Summary: Update CMakeLists.txt to support CMake versions as old as 3.0.2.

Reviewed By: yfeldblum

Differential Revision: D6843104

fbshipit-source-id: 43c2fc685c63df373ed33183f8144de779c97edf

Summary:
Within() is not properly propegating cancellations and other
interrupts via raise() since it technically breaks the normal propegation
mechnaism of copying the interrupt handler pointer.

Reviewed By: yfeldblum

Differential Revision: D6838285

fbshipit-source-id: 8f3efbd3eebe21c2f348b6aba991b0997be4f14f

Summary: [Folly] Move `folly/PicoSpinLock.h` to `folly/synchronization/`.

Differential Revision: D6830101

fbshipit-source-id: 529df14d9cda46bae5989f901ca73effbe0022e5

Summary: [Folly] Fixed initialization of `Tearable` with libstdc++-v4.9.2, which is missing definitions of `atomic_init`.

Reviewed By: davidtgoldblatt

Differential Revision: D6833004

fbshipit-source-id: 88d4cc91f90618a98df110bec5e434c35e750b16

Summary:
The SingletonTestGlobal code asserted that there was exactly one singleton
object registered (the one defined in this test file).  However if folly is
built as a shared library, all of it will be loaded, and several other
singletons are also defined by other parts of folly.  (Currently 7 singletons
get registered.)

Reviewed By: yfeldblum

Differential Revision: D6817544

fbshipit-source-id: de3947770690ffb68d097040bbbfbe42bce3bf94

Summary: Add more types and a macro to make defining them easier.

Reviewed By: mingtaoy

Differential Revision: D6823450

fbshipit-source-id: eb13a8b48efb1625fc3f8035356812f918703c6d

Summary: Foolishly #defined __has_feature when it wasn't and let it leak, but something later must have been relying on something similar. This replaces the uses with FOLLY_HAS_FEATURE.

Reviewed By: yfeldblum, magedm

Differential Revision: D6829724

fbshipit-source-id: 40635dd67135316f27be2aa2a6b2d579dbf843e1

Summary: thread_local does not work on certain mobile builds. This will disable it if FOLLY_MOBILE is set.

Reviewed By: yfeldblum, magedm

Differential Revision: D6785773

fbshipit-source-id: 6a74edec739f42562045aadb4e1ee39d121ce90d

Summary:
This Diff adds new `FOLLY_SDT_WITH_SEMAPHORE` Macro that allows the (potentially expensive) Static Tracepoints be gated with Semaphore and be enabled on-demand.
The Semaphores are compatible with SystemTap's Tracepoint Semaphores, with the following difference:
- Instead of having build Process generate a header file containing the Semaphore variable and check function, and to be included, this Diff uses the approach that users need to explicitly define the semaphore using `FOLLY_SDT_SEMAPHORE_DEFINE` before using `FOLLY_SDT_WITH_SEMAPHORE`. Then the check function could be used in the same file, or in any other linked module with doing `FOLLY_SDT_SEMAPHORE_DECLARE` first. This is inspired from GFlag's defining and declaring.
- We keep allowing original `FOLLY_SDT` with no Semaphore for flexibility.

Differential Revision: D6814268

fbshipit-source-id: e1e4463c71539b4d2071b21880d4ee7689fedce8

Summary:
Since C++14, constexpr non-static member function will not be implicitly marked
const. See the blog post below for more details.
https://akrzemi1.wordpress.com/2013/06/20/constexpr-function-is-not-const/

Reviewed By: mzlee

Differential Revision: D6819658

fbshipit-source-id: 4a756191fca3e744bda3c63c2669e24d6b16861e

Summary: [Folly] Move `folly/SmallLocks.h` to `folly/synchronization/`.

Differential Revision: D6806563

fbshipit-source-id: 26c7be6b6a01673a4f58a0952b5c531528f8ca7f

Summary: Create a "fast" stack trace printer that caches symbol lookups.

Reviewed By: yfeldblum

Differential Revision: D6000178

fbshipit-source-id: e81be1db21cfcd2cd26524c9ef6262337153ae36