Summary: [Folly] Benchmark for `UnboundedBlockingQueue`, starting with measuring contention of `take()`.

Reviewed By: magedm

Differential Revision: D14537953

fbshipit-source-id: ddbba705787b39377f94ad0d63e85038305116cc

Summary:
The AsyncGenerator<T> class allows you to write a coroutine that asynchronously produces a sequence of values of type T.

This has the same executor-affinity behaviour for co_await as for Task<T> except that the executor can change at co_yield points to inherit the executor from the consumer each time a new value is requested.

Reviewed By: ericniebler

Differential Revision: D14446089

fbshipit-source-id: 710ad922918a135e1fe6eef853410c9a77cc2c28

Summary:
This diff expands fbcode_builder so that it can optionally spawn a macos lego job to run a build.

I've pulled in the macOS third party dependencies used by the watchman and eden builds to support this.
Longer term I want to move to a more granular solution for the third party deps.  Medium term we can move the logic for managing these deps to somewhere more central than just the watchman code.

In the linux flavor of fbcode builder, the setup step is responsible for installing system packages, so I've inserted the LFS download of the macOS deps into that stage.

In order for those macOS deps to be picked up without modifying the folly cmake we need to adjust the environment so this diff also inserts that logic at the top of each generated sandcastle step.

A similar technique could potentially be used for windows builds, but note that fbcode_builder makes some assumptions about linux and requires that bash be available.  Cygwin vs. native paths passed through the environment may be a real concern.  That is something for a follow up diff.

Reviewed By: simpkins

Differential Revision: D14514601

fbshipit-source-id: c2882f45aa4b910b5e65ce6b246c750d20ca2948

Summary:
[Folly] Fix up fiber-related guards in futures code. Because of cyclic `#include`s the preprocessor symbol definition must be moved.

Fixes #1061.

Reviewed By: wez, andriigrynenko

Differential Revision: D14555070

fbshipit-source-id: 5e4966fc0f98351924c00a459cc2d8893df4b368

Reviewed By: lewissbaker

Differential Revision: D14554480

fbshipit-source-id: 1aea2bf01d69550d0cdb68fc61d22b2c0e42664e

Summary: The type-erasing executor wrappers and the generic executor wrappers accept extra parameters to their `schedule` member functions, but nowhere is that functionality used, and such executor types would fail to satisfy the Executor concepts anyway. Remove that functionality.

Reviewed By: kirkshoop

Differential Revision: D14541290

fbshipit-source-id: 8cf0fd7bdc6dea72e43d5decc9e70ba843ab2db8

Summary:
- The `bootstrap-osx-homebrew.sh` script is out of date. It previously
  worked when Folly used Autotools, but was not updated when CMake
  replaced Autotools for the build system.
- Update dependencies to install.
- Do not export environment variables when it is not needed; simply pass the
  defines needed directly to the `cmake` invocation.
- Update `README.md` to specify correct dependencies with CMake build
  for both Homebrew and Macports.

Closes #1038
Pull Request resolved: https://github.com/facebook/folly/pull/1041

Reviewed By: yfeldblum

Differential Revision: D14293767

Pulled By: calebmarchent

fbshipit-source-id: 4c53fa105e009090c80690eca8beb5a8c1491a48

Summary:
LogDevice uses DeterministicSchedule and JsonTestUtil from Folly make
these available in the installed library targets so that LogDevice can
compile against installed folly rather than referencing sources
directly.
Pull Request resolved: https://github.com/facebook/folly/pull/1071

Reviewed By: yfeldblum

Differential Revision: D14531954

Pulled By: calebmarchent

fbshipit-source-id: 9bd343210451ad6b2e171c6ddddb7d8677fa6290

Summary:
The current logic for "which gflags library should we use" checks for
the first library (e.g., libgflags vs. libgflags-shared) listed as a TARGET.
Unfortunately, on some number of systems, the gflags-config.cmake (read:
the file that is supposed to tell you how to use gflags with cmake) actually
defines (but doesn't use!) libraries that don't exist (!) so this logic breaks.

Instead, use the system defined gflags_LIBRARIES variable which explicitly
tells us which library to use.

Reviewed By: yfeldblum, simpkins

Differential Revision: D14512966

fbshipit-source-id: add4ecf6bade502b2d12aad2bdfcf0476eeba465

Summary: [Folly] Fix odd `UserMetric` design in `folly/Benchmark.h`.

Reviewed By: phoad

Differential Revision: D14494202

fbshipit-source-id: c1be1d9b5bdfe07677713ef00401c85270bbe627

Summary:
Adds two new functions to folly::coro for concurrently awaiting a fixed number of sub-tasks.

* `folly::coro::collectAll(tasks...)`
* `folly::coro::collectAllTry(tasks...)`

Both can be used to concurrently await multiple input tasks. The difference is in how they report the results.

`collectAll()` produces a tuple of the result values for each of the input operations. If any of the input operations fails with an exception then the whole operation fails with an exception (which one is unspecified) any successful results are discarded.

`collectAllTry()` produces a tuple of `Try<T>` objects regardless of whether any of the input operations failed with an exception. The individual result objects can then be queried for the success/failure, allowing the caller to handle partial failure and/or determine which operation(s) failed.

Reviewed By: andriigrynenko

Differential Revision: D14334714

fbshipit-source-id: 22eb51e2198be42e77677a066bfbc15e1c7eb7dd

Summary:
Compile error on Linux 2.6.32-754.6.3.el6.x86_64 /gcc 7.4/NetOps.cpp (issue #1011)
Pull Request resolved: https://github.com/facebook/folly/pull/1053

Reviewed By: Orvid

Differential Revision: D14454590

Pulled By: yfeldblum

fbshipit-source-id: 9c4dccfa30291dd0fb830cc2417c17568ae4fde3

Summary: This change enables to acquire the executor from the current FiberManager when using ExecutorLoopController (which is already possible to do with the EventBase's loop controller).

Reviewed By: andriigrynenko

Differential Revision: D14515799

fbshipit-source-id: 3b9e48450bf58ea94bebab50dbbcfd7720f90d61

Summary:
- For CMake 3.12 and above, CMP0075 is available which allows the macro
  `check_include_file`, `check_include_files`, and `check_include_files_cxx`
  to prefer to link the check executable to the libraries listed in the
  `CMAKE_REQUIRED_LIBRARIES` variable.
- For CMake 3.13.4, not having this policy set results in a warning such
  as the one below.

```
CMake Warning (dev) at /usr/local/Cellar/cmake/3.14.0/share/cmake/Modules/CheckIncludeFileCXX.cmake:79 (message):
  Policy CMP0075 is not set: Include file check macros honor
  CMAKE_REQUIRED_LIBRARIES.  Run "cmake --help-policy CMP0075" for policy
  details.  Use the cmake_policy command to set the policy and suppress this
  warning.

  CMAKE_REQUIRED_LIBRARIES is set to:

    Threads::Threads;gflags_shared;/usr/lib/libssl.dylib;/usr/local/Cellar/openssl/1.0.2q/lib/libcrypto.dylib

  For compatibility with CMake 3.11 and below this check is ignoring it.
Call Stack (most recent call first):
  CMake/folly-deps.cmake:144 (CHECK_INCLUDE_FILE_CXX)
  CMakeLists.txt:90 (include)
This warning is for project developers.  Use -Wno-dev to suppress it.
```

- This patch sets the policy to the new behavior which means we will
  honor `CMAKE_REQUIRED_LIBRARIES` in the include file check macros now.
- The warning is now gone for those using CMake 3.13.4 or later.
Pull Request resolved: https://github.com/facebook/folly/pull/1066

Reviewed By: Orvid

Differential Revision: D14500811

Pulled By: yfeldblum

fbshipit-source-id: 767bafbc6e15099a85cd0a5556e7f50f7c4fd0eb

Summary:
The macOS 10.14 build is failing in homebrew with some compilation errors in the hash/detail files:

* https://github.com/Homebrew/homebrew-core/pull/37954
* https://jenkins.brew.sh/job/Homebrew%20Core%20Pull%20Requests/39386/version=mojave/testReport/junit/brew-test-bot/mojave/install___build_bottle_folly/
* https://github.com/Homebrew/homebrew-core/pull/35982

~~~
/tmp/folly-20190114-14797-1m0rah5/folly-2019.01.07.00/folly/hash/detail/ChecksumDetail.cpp:181:28:
 error: '__builtin_ia32_pclmulqdq128' needs target feature pclmul
    y0 = _mm_xor_si128(y0, _mm_clmulepi64_si128(x0, multipliers_4, 0x00));
                           ^
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/clang/10.0.0/include/__wmmintrin_pclmul.h:54:13: note: expanded from macro '_mm_clmulepi64_si128'
  ((__m128i)__builtin_ia32_pclmulqdq128((__v2di)(__m128i)(__X), \
            ^
~~~

I noticed a comment by passy in https://github.com/facebook/flipper/pull/125 referencing this same error message "needs target feature pclmul", and the comment also said that it "should be easily fixable by ensuring that -mpclmul is added to the CFLAGS". I had trouble identifying the system architecture as he suggested, but figured that we could just check if the compiler supports that flag and add it for those specific source files if so.

This fixes the 10.14 build for me. It would be nice to set up CI to keep this build working, but it's difficult to currently set up CI for 10.14 without your own hardware, since travis and azure pipelines only support up to 10.13.
Pull Request resolved: https://github.com/facebook/folly/pull/1067

Reviewed By: Orvid

Differential Revision: D14500840

Pulled By: yfeldblum

fbshipit-source-id: 939fb6d495b05491af6e95adc16a0d0bbe86829f

Summary:
- Fix a few shellcheck warnings on `docker_build_with_ccache.sh`.
- This will help when new changes are made to this script as
  shellcheck is run internally.
Pull Request resolved: https://github.com/facebook/folly/pull/1054

Reviewed By: yns88

Differential Revision: D14454554

Pulled By: yfeldblum

fbshipit-source-id: f4b692f21d2478f9db06b478ce56c1ba57fa59e0

Summary: This allows us to always use a dummy keep-alive for all awaitables from coro::Task. We also don't need a custom await_transform for coro::Task anymore.

Reviewed By: lewissbaker

Differential Revision: D14491068

fbshipit-source-id: 0e931c2ae9f52da770b08207e551401e4a3778cb

Summary: Title, remove debugging code that does not seem to be needed anymore

Reviewed By: yfeldblum, davidtgoldblatt

Differential Revision: D14267545

fbshipit-source-id: df57e4b85d850e344b12ce1768a1e1a56ae72bce

Summary:
[Folly] Fix fibers build - `Semaphore::future_wait` declared but undefined.

The `#if` guarding related code is overly aggressive.

Reviewed By: Orvid

Differential Revision: D14498252

fbshipit-source-id: 4183fc67fece0e2bdad1fab5c83e7b9a50ae847b

Summary: Add a basic transport cert with only the needed fields.

Reviewed By: knekritz

Differential Revision: D14429070

fbshipit-source-id: 6728310f75a17e3640025f689b69ffcc3f829522

Summary: as title. Disabling -Wconversion since it can be triggered pretty easily. We've run into this a lot in cachelib code. Until now, we've manually disabled -Wconversion around the header for folly/Format.h in our own code. But I think it's much easier just to disable it inside Format-inl.h itself.

Reviewed By: yfeldblum

Differential Revision: D14471124

fbshipit-source-id: df6bcbd57cbbdb0e6a762eaa5d11548c9e614e1b

Summary:
- Implement `contains(F14HashToken const& token, key_type const& key)` similar
  to the C++20 `contains(key_type const& key, size_t hash)` interface. Parameter order follows existing F14 interface, even though it differs from the C++20 interface.
- Implement heterogeneous lookup overloads of `contains` of the above.
Pull Request resolved: https://github.com/facebook/folly/pull/1063

Reviewed By: nbronson

Differential Revision: D14500835

Pulled By: yfeldblum

fbshipit-source-id: 85b0fbb66c2db948958e67105cf8ef98851ee0a6

Summary: [Folly] `conditional_t`, like `std::conditional_t` but where the result can be used in deducible contexts.

Reviewed By: nbronson

Differential Revision: D14477367

fbshipit-source-id: d0a9da7f0e1584a45a0709e80e2b8afb2851214c

Summary: as title

Reviewed By: yfeldblum

Differential Revision: D14476923

fbshipit-source-id: 5ce16adb2402720b16acd12ab0003c1e93395f6c

Reviewed By: amajeed1

Differential Revision: D14496345

fbshipit-source-id: 985e5299930c3812e7510f6dfe7bdcd3ce3519cf

Summary: [Folly] Remove include of `glog` from various places.

Reviewed By: spalamarchuk

Differential Revision: D14494311

fbshipit-source-id: 122b63a726a08802d7b4ef6b55b304e9800c258d

Summary:
- Fix a few shellcheck warnings on `bootstrap-osx-homebrew.sh`.
- This will help when new changes are made to this script as
  shellcheck is run internally.
Pull Request resolved: https://github.com/facebook/folly/pull/1058

Differential Revision: D14493361

Pulled By: yfeldblum

fbshipit-source-id: e46f9d0e329fd1ebe16820bdb5614c53363c5dfe

Reviewed By: A5he

Differential Revision: D14396454

fbshipit-source-id: a00b1eab527237c885e04413db1a0a413121d91b

Summary: This makes it possible to add future APIs to primitives built using fibers::Baton.

Reviewed By: A5he

Differential Revision: D14396094

fbshipit-source-id: bf64a31c25ed4dd92b8c484cb4256078efdc6262

Summary:
- Remove unneeded typename for a using alias in `F14Table.h`.
Pull Request resolved: https://github.com/facebook/folly/pull/1056

Reviewed By: nbronson

Differential Revision: D14473219

Pulled By: yfeldblum

fbshipit-source-id: a6ac26c42dbca219ce123aff00b1ee7079068007

Summary:
contains() is part of the c++20 spec.  This implements it, and
refactors count() to use contains().

Reviewed By: nbronson

Differential Revision: D14426805

fbshipit-source-id: 2435124e9c0a8eeb4e53109f0ddac884c227c781

Summary:
The heterogenous count() method currently returns size_type, unlike
similar methods (normal count(), F14Map.count() methods).  This
standardizes it.

Reviewed By: nbronson

Differential Revision: D14426807

fbshipit-source-id: 140be39fcda99c9f82ced3962ddcbdd1d5f53ba6

Summary:
When a user counter is added to benchmarks, the counter's value is written as a bare integer.
This change gives chance to user to either still keep it reported as bare integer, a time value or a metric value.

Differential Revision: D14451341

fbshipit-source-id: 95e64a89c201bb385d9a403f3442fa70c2ac7e2a

Summary: Make AsyncIOTest.cpp conditionally depend on $LIBAIO_FOUND in CMakeLists.txt.

Reviewed By: yfeldblum

Differential Revision: D14365682

fbshipit-source-id: 5ac7dc25174ab48ed1de8846956e69c7759fc9bd

Summary: convert all executors in pushmi and all executor concepts and usage in pushmi to use `schedule()`

Reviewed By: ericniebler

Differential Revision: D13463902

fbshipit-source-id: 01eac4fe0b5cd49afbd95118db769db59a3b968c

Summary: It's dead.

Reviewed By: yfeldblum

Differential Revision: D14425042

fbshipit-source-id: 098de614955eedd7ccc733858db85b997d01d9e0

Summary: At least make it possible to do things correctly.

Reviewed By: yfeldblum, LeeHowes

Differential Revision: D14426753

fbshipit-source-id: 63b3572175224dbbda828981a5c9f067a5e35f33

Summary: Fix typo

Reviewed By: aary

Differential Revision: D14411090

fbshipit-source-id: 4ca0273201facebc815db26d23662d0d192c831f

Summary:
[Folly] Move `FOLLY_DEFINE_CPO` to `folly/lang/CustomizationPoint.h`.

And move `StaticConst` to top-level `folly` namespace and into `folly/lang/`.

And Americanize the spellings of words in the comments.

Reviewed By: ericniebler, lewissbaker

Differential Revision: D14402315

fbshipit-source-id: e13dbc9d471ee73ba0a289d375c235734334b920

Summary:
future.then([] -> T)  currently returns Future<T>
future.then([] -> Try<T>)  currently returns Future<Try<T>>

This changes that behavior, such that future.then([] -> Try<T>) will
return Future<T>.

This is a more desirable interface for futures. It also simplifies
implementation of a separate task (propagating exceptions more
efficiently through chained callbacks) because it enables us to internally
substitute lambdas returning T with lambdas returning Try<T>.

Reviewed By: yfeldblum, ryantimwilson

Differential Revision: D14318624

fbshipit-source-id: 51363e82ceb86bfecf87ae661188be9ca3dd8002