Summary: This makes sense on systems without `backtrace` and with LLVM libunwind implementations.

Reviewed By: yfeldblum

Differential Revision: D23625739

fbshipit-source-id: 6ec50c94c73bda227118ffc36d6856bcd8dd0d8b

Summary:
`std::shared_ptr` has an aliasing constructor, but `std::weak_ptr`
has none.  This diff implements a helper function that implements the
functionality.  For libstdc++ it uses ABI-specific hacks to avoid the
need to construct a temporary `std::shared_ptr`.

Reviewed By: yfeldblum

Differential Revision: D23580416

fbshipit-source-id: 71712270519b0e7522d8019a5c356679c7dc5397

Summary:
[Folly] Decay char-array and trivial args to `throw_exception` and `terminate_with`.

The motivation is to ensure smaller call-site code size in the common cases, where the solitary argument is a string literal. Extend to cases where arguments are numbers and the like as well, since that is simple enough and is something that `Function` already does.

Reviewed By: aary

Differential Revision: D23550096

fbshipit-source-id: d3a2066dc28dc4cfac6b65e2b8972613ce55d06c

Summary: [Folly] `byte`, backported from `std::byte` in C++17.

Reviewed By: Alfus, vitaut

Differential Revision: D23535105

fbshipit-source-id: 056c634362b2ca226306a97ae9994a4631d07ee5

Summary: In newer versions of the windows STL, this gets included in <thread>, so explicitly include it in Unistd.h and remove the definition of getpid.

Reviewed By: yfeldblum

Differential Revision: D23498073

fbshipit-source-id: 5fe07dbbaa9fb1bf5c904286bd15beccbf1d1615

Summary:
[Folly] Revise invoke to be an invoker rather than a function, and therefore insusceptible to ADL, and revise invocability traits for compile-time perf.

Do the same for apply and applicability traits.

Reviewed By: Mizuchi

Differential Revision: D23334895

fbshipit-source-id: 4d3241a3f34a7bede44f8ab54a30a3074d9fd3ed

Summary: [Folly] Tweaks to hash primitives: twang, jenkins_rev, fnv

Reviewed By: ot, luciang

Differential Revision: D23290185

fbshipit-source-id: 080dfa8a5f357065473f3973654c91d01c683c3c

Summary:
When grabbing an iterator when the map is empty, size is zero but we
still create a copy of the elem_ pointer which is invalid to do when the size
is zero (since the writer could overwrite the pointer upon first insertion).
This diff checks whether the iterator would be empty and if so avoid grabbing
that pointer.

Reviewed By: yfeldblum

Differential Revision: D23520616

fbshipit-source-id: 5b3f2bc3a9b935920e7fd616e64fadafd7775a75

Summary: Honestly I haven't spent time to understand why this order needs to be acq_rel, but (1) it's the same as the one used in the other keepAliveRelease in this file (2) this ordering is free for read-modify-write operations (3) TSAN said this works :)

Reviewed By: yfeldblum

Differential Revision: D23495784

fbshipit-source-id: 8bdcafb14b27e6c0f8838ed478364821cd39f9a9

Summary:
[Folly] Fix the `reentrant_allocator` internal loop, which when finding a new head node would incorrectly preserve the free `offset` corresponding to the previous head.

Hat tip to hpmv for the report.

Differential Revision: D23593517

fbshipit-source-id: b4fc701f08056184225e7feac32d60aabf7ea6cc

Summary:
Remove the old `fbcode_builder_config.py` file used by the previous
incarnation of fbcode_builder.  This file hasn't been used in a while now.
Folly instead builds with `build/fbcode_builder/getdeps.py`, which uses the
configuration file at `build/fbcode_builder/manifests/folly`

Reviewed By: wez

Differential Revision: D23591355

fbshipit-source-id: b2e193c7dc54dcf481557984b5833fa92ff4049c

Summary:
Add a small top-level build.sh script that invokes
build/fbcode_builder/getdeps.py to perform the build.  This primarily helps
make the getdeps build process more discoverable to users.

Also document this script in README.md

Reviewed By: yfeldblum

Differential Revision: D23484822

fbshipit-source-id: 6013ace6d80d7d3d80aa87d4c6269ac106814cf0

Summary:
Update the `README.md` file in the fbcode_builder subdirectory to document the
current getdeps.py build system rather than the older fbcode_builder scripts
that required each project to keep their own python-based build files in their
project directory.

Reviewed By: chadaustin, wez

Differential Revision: D23591356

fbshipit-source-id: 75a099a10793e68a2b59696682010c4dff47ec69

Differential Revision:
D23489967 (https://github.com/facebook/folly/commit/65bb1505d3f64d9248bc1f7a04abc212db518676)

Original commit changeset: db11c9c5946e

fbshipit-source-id: 4970fc7de479a51d4acf91b2871ff0c15712130e

Summary:
The code in `FindDoubleConversion.cmake` called
`FIND_PACKAGE_HANDLE_STANDARD_ARGS()` with the wrong arguments.  The first
argument must match the package name: since this file is called
`FindDoubleConversion` the first argument must be `DoubleConversion`.

Previously this code was passing in `DOUBLE_CONVERSION`, causing it to set
`DOUBLE_CONVERSION_FOUND` instead of `DoubleConversion_FOUND`.  Since this
module did not set `DoubleConversion_FOUND` the `find_package()` call assumed
it always succeeded, even when the double-conversion library was not found.

Also remove the extra status message.  `FIND_PACKAGE_HANDLE_STANDARD_ARGS()`
already prints out a status message as long as `find_package()` was called
without the `QUIET` argument.

Reviewed By: yfeldblum

Differential Revision: D23588980

fbshipit-source-id: 8c088930b1b01c027403c3e425f1924f6149f72a

Summary:
Update FindGflags.cmake to work on recent CentOS and RedHat distributions.

On these distributions the CMake package configuration installed by the
gflags-devel RPM is slightly broken, and sets `gflags_INCLUDE_DIR` to a
directory that does not exist.  This happens because `/lib64` is symlinked to
`/usr/lib64`, and CMake ends up searching `/lib64` before `/usr/lib64` by
default when searching for packages.  Therefore it finds `gflags-config.cmake`
via the `/lib64` symlink.  However, `gflags-config.cmake` computes the include
directory relative to where its config file was found, and this relative path
computation only works when using the actual `/usr/lib64` path where it was
installed.  When found via `/lib64` it returns a bogus `//include` path that
does not exist.

This updates `FindGflags.cmake` to verify if the `gflags_INCLUDE_DIR` path
actually exists, and set it to a sane location instead.

This also updates the code to use the `gflags-shared` target that is exported
by the default gflags package configuration if it was only built as a shared
library.

Pull Request resolved: https://github.com/facebook/folly/pull/1409

Reviewed By: yfeldblum

Differential Revision: D23588288

fbshipit-source-id: b68a717953ae0521f568d7bcfd05ca33cd6dc578

Summary:
FOLLY_USE_SYMBOLIZER is a poor guard for the existence of execinfo.h
and backtrace(3), which is widely available on unix platforms, but not
available on Android NDK. Instead, introduce FOLLY_HAVE_BACKTRACE and
have CMake detect its availability.

Reviewed By: yfeldblum, luciang

Differential Revision: D23361495

fbshipit-source-id: 4409363f3eaa754977a93e1f006c5911d8e32020

Summary: Darwin libc doesn't implement quick_exit, so use std::_Exit instead.

Reviewed By: yfeldblum

Differential Revision: D23592838

fbshipit-source-id: bda318216239965e14b4c97ca184a2303353ed73

Summary: For consistency with get() make getTryVia r-value qualified and move the result out. This has the benefit of removing surprise copies in code that already operated on a temporary future but ended up with an l-value reference to copy from as a result of this operation.

Reviewed By: yfeldblum

Differential Revision: D23489967

fbshipit-source-id: db11c9c5946edd402103b3aa4140c0a1db5aae5e

Summary:
VERSION_GREATER 3.18 also matches 3.18.2, which is wrong as the target argument was only added in 3.19.
After noticing that folly requires an ancient CMake version I converted over from VERSION_GREATER_EQUAL and introduced the comparison error.

Issue reported [here](https://github.com/facebook/folly/pull/1433#discussion_r484278801).

Pull Request resolved: https://github.com/facebook/folly/pull/1439

Reviewed By: yfeldblum

Differential Revision: D23564812

Pulled By: Orvid

fbshipit-source-id: 19881876147ec73ad70610cf97c408740f998340

Summary:
From Dwarf4 standard

http://dwarfstd.org/doc/DWARF4.pdf#page=53

```name=2.17.3 Non-Contiguous Address Ranges
The applicable base address of a range list entry is determined by the closest preceding base
address selection entry (see below) in the same range list. If there is no such selection entry, then
the applicable base address defaults to the base address of the compilation unit (see Section
3.1.1).

In the case of a compilation unit where all of the machine code is contained in a single
contiguous section, no base address selection entry is needed.
```

http://dwarfstd.org/doc/DWARF4.pdf#page=58

```name=3.1.1 Normal and Partial Compilation Unit Entries
Compilation unit entries may have the following attributes:

1. Either a DW_AT_low_pc and DW_AT_high_pc pair of attributes or a DW_AT_ranges
attribute whose values encode the contiguous or non-contiguous address ranges, respectively,
of the machine instructions generated for the compilation unit (see Section 2.17).
A DW_AT_low_pc attribute may also be specified in combination with DW_AT_ranges to
specify the default base address for use in location lists (see Section 2.6.2) and range lists
(see Section 2.17.3).
```

http://dwarfstd.org/doc/DWARF4.pdf#page=61

```name=3.1.1 Normal and Partial Compilation Unit Entries
The base address of a compilation unit is defined as the value of the DW_AT_low_pc attribute,
if present; otherwise, it is undefined. If the base address is undefined, then any DWARF entry or
structure defined in terms of the base address of that compilation unit is not valid.
```

http://dwarfstd.org/doc/DWARF4.pdf#page=52
```name=2.17 Code Addresses and Ranges  - 2.17.1 Single Address
When there is a single address associated with an entity, such as a label or alternate entry point
of a subprogram, the entry has a DW_AT_low_pc attribute whose value is the relocated address
for the entity.

While the DW_AT_entry_pc attribute might also seem appropriate for this purpose, historically
the DW_AT_low_pc attribute was used before the DW_AT_entry_pc was introduced (in DWARF
Version 3). There is insufficient reason to change this.
```

Reviewed By: luciang

Differential Revision: D23422082

fbshipit-source-id: 4de79fa81e6c323095bb7c79faf87369998138c6

Summary: Avoid an expensive `__vdso_getcpu` on every `get()`.

Reviewed By: terrelln

Differential Revision: D23469601

fbshipit-source-id: 6c7e62a136b7cd80baab6dcfb10cba13a2e472c0

Summary:
The function folly::get_cached_pid() saves and reuses the result of calling getpid().
At fork, the value is updated for the child process.

It is thread-safe. The first call is async-signal-unsafe but subsequent calls are async-signal-safe.

Reviewed By: yfeldblum, davidtgoldblatt

Differential Revision: D23433702

fbshipit-source-id: a026300206c1726fc1c2c6a588828f2e6559c4e2

Add getTry back into folly::Future, with blocking behaviour consistent with get() and with SemiFuture::getTry().

Summary:
Formerly Future::getTry was non-blocking, and exhibited confusing behaviour of throwing on timeouts, but returning a try containing exceptions related to the workload of the future. This was inconsistent with Future::get, and thus surprising.

An earlier change safely removed this broken version of the operation. Here we add it back in with consistent behaviour.

Reviewed By: yfeldblum

Differential Revision: D23477042

fbshipit-source-id: afda6b37c86531dab43ac31ec6d605cc6c5d670c

Summary: [Folly] A few nits for `constexpr_strlen`, `constexpr_strcmp`.

Reviewed By: ericniebler

Differential Revision: D23547224

fbshipit-source-id: f232861f5cc4dbd5eb77c4bbe3b9766fd233e9f5

Summary:
In the "Complex" test we can have nodes running at the same time. So
appending to the same vector concurrency isn't safe. Protecting that with a
mutex.

Reviewed By: yfeldblum

Differential Revision: D23545746

fbshipit-source-id: b6f6e8b7204cf63c7e8fbbd8fc5134cb66c8180a

Summary: Variable length arrays are an extension to C++ standard that is not universally supported. Use a fixed size array instead.

Differential Revision: D23563131

fbshipit-source-id: 743e65838e83286ea28d8de10cd9d4416e6e4e91

Summary: Add the ability to specify the min number of threads without changing a gflag

Reviewed By: stuclar

Differential Revision: D23557854

fbshipit-source-id: d4394e979c123e7f312816b1dec2849d46b0de9a

Summary: This explains why the commonly used `int(*s1 - *s2)` was changed to a slightly less obvious alternative.

Reviewed By: yfeldblum

Differential Revision: D23537394

fbshipit-source-id: 152554bc9864cdf25c569216d21b0c04bea67e9d

Summary: This is used so needs to be included. MSVC 2019's STL exposes this.

Reviewed By: yfeldblum

Differential Revision: D23497796

fbshipit-source-id: 612e0fb91c5d5637c25d062a4b3afc0c67344012

Summary: [Folly] Fix `StaticSingletonManager::create_<true>` which is missing the `noexcept` specification.

Reviewed By: Mizuchi, xavierd

Differential Revision: D23552042

fbshipit-source-id: 629ad215602125a3e6322b36948df78c62fbef15

Summary: [Folly] Limit `StaticSingletonManager` check functions to `StaticSingletonManager` calls.

Reviewed By: Alfus

Differential Revision: D23544598

fbshipit-source-id: 081d324064025d5fc86bb20b9b685c2591f01642

Summary: [Folly] Let `noexcept` member function template `StaticSingletonManager::create` be outline to avoid pathological inline path behavior.

Reviewed By: iahs

Differential Revision: D23522488

fbshipit-source-id: 5ec9ed271b814f4496812e226e0f0929f65565a2

Summary: [Folly] Add missing `< 0` tests for `constexpr_strcmp` to verify that the implementation is symmetric.

Reviewed By: Mizuchi

Differential Revision: D23482134

fbshipit-source-id: d3cb1fe8ce4690a34a12620eb57e7a8ff57ad8a1

Summary: [Folly] Make `is_same` failures easier to diagnose in `constexpr_strlen` tests.

Reviewed By: Mizuchi

Differential Revision: D23481769

fbshipit-source-id: a990e1b96381f976feb7b4d3efda860f38fb6fe3

Summary:
These tests pass in our build system, and fail on other platforms (Ubuntu, Darwin). Specifically, they cause our folly's CI to fail, since it's run on Ubuntu.

The tests fail because TLS 1.3 is being enabled for these tests. In retrospect, that *should* cause session resumption to fail, since no additional reads are being done on the socket after the handshake. I'm not quite sure yet why they pass in our build system.

Reviewed By: yfeldblum

Differential Revision: D23513991

fbshipit-source-id: 0a401d19300d8ca7da3f9b5f8263959eac826f8c

Summary:
[Folly] Some machine code check functions for `terminate` and `terminate_with`.

Demonstrates the impact of adding the extra argument.

Reviewed By: magedm

Differential Revision: D23523207

fbshipit-source-id: c37ae61e59ca060e7e370db425cb937341bd6e09

Summary: This is pretty self-explanatory.

Reviewed By: yfeldblum

Differential Revision: D23514563

fbshipit-source-id: 39b5796c742bfacc830c0154bf72bbd190501e33

Summary: Windows needs it

Reviewed By: yfeldblum

Differential Revision: D23497837

fbshipit-source-id: e4fa8fadb3a7ac26e602ff5fbddaef0976d47b0d

Summary: Avoid a needless use of the preprocessor in FixedString.h by using `kIsDebug` instead of the `NDEBUG` preprocessor macro in `checkNullTerminated`.

Reviewed By: yfeldblum

Differential Revision: D23411248

fbshipit-source-id: 78990ade25bcfa3956f917b2b285c096b76ecb09