Summary:
Use folly::cacheline_align_v to prevent failures when running folly on platforms
where extended alignment is not supported.  The thread pools only used extended
alignment as an optimization; so it is safe to disable it when not needed

Reviewed By: yfeldblum

Differential Revision: D17077439

fbshipit-source-id: 63ba59aaf7ce45231bb25427af84292c3a3464a9

Summary:
The non-atomic field `inUse_` is protected by a lock in `getStackCache`, but not in `giveBack`. This
is a data race and it is caught by TSAN.

Reviewed By: yfeldblum

Differential Revision: D17109710

fbshipit-source-id: 7c0d5d948740516422efca0895a702566c804bb4

Summary:
Windows and macOS compilers are sensitive to this,
and this was overlooked in D16636130

Reviewed By: selat

Differential Revision: D17113825

fbshipit-source-id: 16fe9210b07a096a58f3370df64874514e28ea46

Summary:
Compiler passes rvalue reference for named local variable to return_value:
https://reviews.llvm.org/D51741

Reviewed By: lewissbaker

Differential Revision: D16963353

fbshipit-source-id: 98d846153adb2235b61bb8fb54d10f032d4c0055

Summary:
- Prior to commit 8cb615a2
  (Differential Revision: D3394612), `Range:size()` had an assert checking
  that the beginning iterator of the range is less than or equal to the end
  iterator. Unfortunately, due to GCC bug 71448, the `assert` was not allowed
  if we wanted `size()` to remain a `constexpr` function.
- This bug is no longer an issue with our current supported versions of GCC.
  As such, add the assert back in.
Pull Request resolved: https://github.com/facebook/folly/pull/1155

Reviewed By: ericniebler

Differential Revision: D15641627

Pulled By: yfeldblum

fbshipit-source-id: 0677053c18fa55aaa4727947c241f0675362861b

Summary:
Update install_fb_python_library() to work with CMake versions older than 3.7.

Previously the code used a generator expression in the directory argument
supplied to `install(DIRECTORY)`.  Support for generators in this parameter
was only added in CMake 3.5, but it was buggy and did not correctly handle
absolute paths until 3.7.

This updates the code to avoid using a generator expression.  This does
require that the corresponding `add_fb_python_library()` statement appear
before the `install_fb_python_library()` statement, but this seems like a
pretty reasonable restriction.

Reviewed By: chadaustin

Differential Revision: D17103597

fbshipit-source-id: 54d6af93a3f530373cfe3137db99436d8484f308

Summary:
[Folly] RTM nits.

* No nested namespace.
* All status values supported.
* Tests included in cmake builds.
* API functions avoid `SIGILL` at runtime.

Reviewed By: nbronson

Differential Revision: D16636130

fbshipit-source-id: 9f9dc99b0826abbda295b069eb4823b3416af5d1

Summary:
This is a useful testing utility that people can use if they use timeout APIs
like `folly::futures::sleep()`

Reviewed By: LeeHowes

Differential Revision: D17065352

fbshipit-source-id: f9ea8366af7fc9429bfb1065328ed746c824f810

Summary: The sorted vector types have constructors which "steal" capacity from a pre-filled container.  Add tests to cover the cases where this is a folly:small_vector, and a std::pmr::vector.

Reviewed By: nbronson

Differential Revision: D17074150

fbshipit-source-id: e94b7e136553653081abe86559c4c013b9aac031

Summary:
Adds static factory `CancellationSource::invalid()`.

This can be used to construct a `CancellationSource` that can't be cancelled but that doesn't perform any heap-allocations.

Reviewed By: kirkshoop

Differential Revision: D16816764

fbshipit-source-id: bb87273903a6a8d245d207f4e6bda94e5c3d3dad

Summary: Tweak futures::within to be completely deferred, including the timeout path, allowing it to support a SemiFuture-returning TimeKeeper.

Reviewed By: yfeldblum

Differential Revision: D17006080

fbshipit-source-id: cbed58579d33f6d6ecf60dce5f9c8b886fc4d4c4

Summary:
wait() is guaranteed to succeed and not throw for either Baton type, so
using a reference here is safe. When the future is completed on another thread,
the shared_ptr causes cache contention.

```
Thread 1            Thread 2
make_shared/inc
inc
...
                    dec ref
dec ref/free
```

Reviewed By: yfeldblum

Differential Revision: D17077672

fbshipit-source-id: a6010ed159353decefbd5c51f4347370b16f7f60

Summary:
Update the generated `run_cmake.py` script to use `os.execve()` rather than
`subprocess.call()`, so that it now propagates CMake's return code back to its
caller.

Reviewed By: chadaustin

Differential Revision: D17089206

fbshipit-source-id: e01f05f492ccb842d4967e59fd0bc9a3e59b8a42

Summary:
Fix printing the manifest context in the error message if we cannot find a
project fetcher.  Previously the context in the message would be printed as
something like `<getdeps.manifest.ManifestContext object at 0x7fcce987e610>`,
now it shows instead as something like
`{distro=ubuntu, distro_vers=18.04, fb=off, os=linux, test=off}`

Reviewed By: chadaustin

Differential Revision: D17089208

fbshipit-source-id: c16549b61030d813b7b5ff9f65966436dc1e1898

Summary:
This just strips off a trailing '/' from the `INSTALL_DIR` property that we
set on python targets, to avoid having double-slashes in the paths that we use
during installation.  This shouldn't really have any material difference other
than cleaning up the paths that get printed during the installation phase.

Reviewed By: chadaustin

Differential Revision: D17089207

fbshipit-source-id: ab36bb76c19fa60fe037f7a5290ccfd6bdbf13b0

Summary:
Update fboss to use `ThriftCppLibrary.cmake` directly from fbcode_builder
rather than maintaining its own copy of this file.

Reviewed By: chadaustin

Differential Revision: D17005421

fbshipit-source-id: a64df426118df6088b47f09410dad7b8b7e79a43

Summary:
Update `add_thrift_cpp2_library()` to pass in the correct `-I` flags when
invoking the thrift compiler so that it can find all of the other thrift files
that this library depends on.

D16062657 was a previous attempt to do this, but suffered from a few problems:
- It required all dependencies to be defined before
  `add_thrift_cpp2_library()` was called.  This requires users to carefully
  order their CMake files and subdirectory include ordering.
- It only handled one level of dependencies, and did not propagate include
  paths for deeper dependencies.
- It set the include path for dependencies to the source directory path where
  the dependency was originally built, rather than the directory where the
  thrift file for that dependency would be installed.

This change does require CMake 3.8+.  Previous versions of CMake do not
support using generator expressions to generate multiple arguments for a
custom command.

Reviewed By: strager

Differential Revision: D17005381

fbshipit-source-id: 31190beba94b4d1010445375a5e2791450230f7d

Summary: Allow futures::when to pass through a SemiFuture-returning function.

Reviewed By: yfeldblum

Differential Revision: D17001974

fbshipit-source-id: dbe8b347f0a4c3fd19774c68c4e58e3b523b971f

Reviewed By: yfeldblum

Differential Revision: D17086382

fbshipit-source-id: 318834a1d82025bc0c34bcf7a95fedfffdb02ddb

Summary: times is hard to use with SemiFutures, leading to potential use of inline executors just to make it work. This adds native support.

Reviewed By: yfeldblum

Differential Revision: D16992774

fbshipit-source-id: 483110d923ca51e037439ccc80173ffa9299cd64

Summary: Use FOLLY_CL_USE_FOLLY_TLS for CacheLocality to avoid an iOS build issue

Reviewed By: mzlee

Differential Revision: D17085507

fbshipit-source-id: acd407c9e0045eb4d8a59348a20874dbe75fd651

Summary: Earlier change to duration waiting broke propagation of the interrupt handler. This change correctly propagates it into the new future that is returned.

Reviewed By: yfeldblum

Differential Revision: D17061618

fbshipit-source-id: b228221acbfab3ae94118c1197a485e01aa2e41e

Summary: whileDo is hard to use with SemiFutures, leading to potential use of inline executors just to make it work. This adds native support.

Reviewed By: yfeldblum

Differential Revision: D16988569

fbshipit-source-id: abaf36180d6ff8a5277eebb277f7f0a254315d36

Summary:
All of the requisite headers are now open sourced and eden
has been converted to use them. This removes any logic that references
this directory and kills the now unused cmake file.

Reviewed By: chadaustin

Differential Revision: D16994732

fbshipit-source-id: 573b8b50540ee64590682be2a54aa94659f12368

Summary:
Most stat code in fbcode was moved in to open source under
fbcode/fb303. The only exception that we use internally is the
MonotonicCounter class. This diff moves all of our stats code to use
the open source headers and updates cmake files to pull in the open
source fb303.

Reviewed By: chadaustin, shri-khare

Differential Revision: D16969960

fbshipit-source-id: 12b7abb54c956c242c8e27eb69fd96925c7e61f7

Summary:
Latest and C++17 are all we currently intend to support.

Created from Diffusion's 'Open in Editor' feature.

Reviewed By: yfeldblum

Differential Revision: D17053747

fbshipit-source-id: 7ef2404b82242e363c9ccf1c5d7136eb52f356dc

Summary: In preparation for making TimeKeeper::at and TimeKeeper::after return SemiFutures to make more explicit the need to keep work off of the TimeKeeper's executor, add *Unsafe versions of both functions to migrate code to.

Differential Revision: D17052948

fbshipit-source-id: 4fb138862e417f02746ad4eaf71f091d37302e7a

Summary:
Fixing a comment in Evicting Cache Map, no functional code change.

(Note: this ignores all push blocking failures!)

Reviewed By: yfeldblum

Differential Revision: D17051293

fbshipit-source-id: 7a1197790bd9a0905934888dabeecde9f7c9c165

Summary:
Update add_thrift_cpp2_library() to add a PUBLIC_HEADER property to the
generated library target, so that the generated headers can be installed using
an `install()` call.  Also add a `HEADER_INSTALL_DIR` property which indicates
the directory they should be installed into.  A `INCLUDE_DIR` argument was
added to customize the include installation path, should anyone care to do so.

This also removes code that was previously incorrectly installing the
generated headers into the source tree.

Reviewed By: wez

Differential Revision: D16853404

fbshipit-source-id: 6f640b2bd347c99804e074fca0209dda37c8a9cf

Summary:
`folly::detail::std_pmr` uses either `std::pmr` or `std::experimental::pmr`.

"Polymorphic Memory Resources" has been standardized in C++17, but `pmr::resource_adaptor` was [left behind](https://github.com/cplusplus/papers/issues/33), which means `resource_adaptor` is not available when a C++17 ready standard library is detected by folly.

This commit reimplements `test_resource` without using the non-standard adaptor.

```
folly/test/sorted_vector_test.cpp:862:31: error: ‘folly::detail::std_pmr::resource_adaptor’ has not been declared
  862 | using folly::detail::std_pmr::resource_adaptor;
      |                               ^~~~~~~~~~~~~~~~
```
Pull Request resolved: https://github.com/facebook/folly/pull/1207

Reviewed By: nbronson

Differential Revision: D17003049

Pulled By: yfeldblum

fbshipit-source-id: 7bcd5b4bb4368366cd57d89b9ef88deaaff808d6

Summary: Previous within changes changed how long the executor keepalive existed for. This explicitly grabs the executor keepalive before calling within to ensure it still exists when we try to use it in `via`.

Reviewed By: LeeHowes

Differential Revision: D17054622

fbshipit-source-id: 7826c6ac0ad332a8f5f1331cd2d798619cf318c3

Summary:
The external Travis CI builds for fbzmq are still using the older
fbcode_builder spec (as opposed to the newer getdeps manifest file).

D16577367 moved the CMakeLists.txt file to the top-level directory, but the
fbzmq spec file was still looking for it in the fbzmq subdirectory.

Reviewed By: jstrizich

Differential Revision: D17005361

fbshipit-source-id: 3f7664eadfb60ec7606124a14445b44ae586b8a7

Summary:
Fix the fbzmq pathmap listed in its manifest file to match the path map
actually used by ShipIt.  The fact that this was broken was why internal
getdeps builds did not detect that D16577367 broke the build.

Reviewed By: wez, jstrizich

Differential Revision: D17005360

fbshipit-source-id: 046ff58ad70c03b860c3fccebaba975808df244d

Summary:
Update ThriftCppLibrary.cmake to record that the output depends on the thrift
`cpp2` template files.  The `THRIFT_TEMPLATE_FILES` variable was defined in
D16062657 but doesn't appear to be used anywhere.  This uses it, and also
updates it to only match the C++ template files.

Reviewed By: wez

Differential Revision: D16738442

fbshipit-source-id: aa4097e9ec1d009ef1e3ee2efd03ffa19a82ed12

Summary:
Add explicit double quotes around arguments that are intended to be exactly a
single parameter.  Also line wrap to 80 characters.

Reviewed By: wez

Differential Revision: D16738443

fbshipit-source-id: b8c9b6d50da72b44e23eaf5effbc384b6cbbc1c0

Summary:
Change ThriftCppLibrary to use `fb_cmake_parse_args()`

This code doesn't really care much about handling empty arguments, so we could
use the standard `cmake_parse_arguments()` function here instead of
`fb_cmake_parse_args()`.  However, it is slightly more convenient that
`fb_cmake_parse_args()` automatically errors out on unhandled arguments, so
that we don't have to check for this ourselves.

Reviewed By: wez

Differential Revision: D16738445

fbshipit-source-id: 67b9761f2ceaf7665bd336290bb44af32fec6f4a

Summary:
The manifest file for iproute2 previously listed the commit to check out as
7ca63aef7d1b0c808da0040c6b366ef7a61f38c1.  However, this is not a commit, but
a tag object that refers to commit 92a0236a3cdf3438000834121b7ea8a09f1f52b1.

This caused getdeps to think that iproute2 was always out-of-date, even when
it was already on the correct commit, since it was comparing current commit ID
against the tag ID.

This updates the manifest to refer to the commit ID instead of the tag ID.

Reviewed By: shri-khare

Differential Revision: D17005418

fbshipit-source-id: c7145c73b66ae93a2e9097ddcf8c53237b135d32

Summary: Enable FOLLY_TLS on FOLLY_MOBILE too

Reviewed By: guangyfb

Differential Revision: D16883165

fbshipit-source-id: 1a8f0991088f6963dc2acf84b74b294be998b877

Summary: [Folly] Avoid `__forceinline` for `catch_exception` under MSVC, which warns about functions which contain `try`-and-`catch` blcoks having the `__forceinline` modifier.

Reviewed By: akrieger

Differential Revision: D16982707

fbshipit-source-id: 68c443c4ce8ca77edd49d509a88ef4b0ab9488a9

Summary:
The definition of RLockedPtr should use a non-const Subclass type, and not a
const one.

Reviewed By: ot

Differential Revision: D15356827

fbshipit-source-id: b8ad41e263f0e15ffa25b0698aa85eab8ca2ccb8