Summary:
Angled brackets inside angled brackets is not displayed.
This makes the document misleading.
Pull Request resolved: https://github.com/facebook/folly/pull/1032

Reviewed By: shixiao

Differential Revision: D14207772

Pulled By: yfeldblum

fbshipit-source-id: 4f2e58145a5473a7b887ef1da4efbeb16d5330c6

Summary:
futures::sleep returning a Future leads to continuations easily being run on
the Timekeeper's callback. The goal is to change the return type so that
futures::sleep returns a folly::SemiFuture.

Reviewed By: yfeldblum

Differential Revision: D14069549

fbshipit-source-id: b98e2c76ccbf0a80aed28f90a1f3c63f15eb2e5c

Summary:
- Existing code disables `-Warray-bounds` warning for GCC due to a bug in
  GCC 4.9.
- As seen in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59124, the bug
  has been resolved for GCC 5.1.
- Since the bug has been resolved, we no longer need to disable the
  warning.

yfeldblum I suspect that we have to hold off on this PR until the internal use cases relying on GCC 4.9 upgrade, but I figured it is better to have some of these cleanup opportunities ready for when the time arises.
Pull Request resolved: https://github.com/facebook/folly/pull/999

Reviewed By: ot, Orvid

Differential Revision: D13729723

Pulled By: yfeldblum

fbshipit-source-id: da33704c2e2022c2dad0681e4d6152105a1cf54d

Summary:
[Folly] Address MSVC C4686 in `folly/lang/Pretty.h`.

Resolves:

```
folly\lang\pretty.h(119): warning C4686: 'folly::detail::pretty_name_zarray<T>::zarray_': possible change in behavior, change in UDT return calling convention
```

Reviewed By: Orvid

Differential Revision: D14214290

fbshipit-source-id: c5e0ebd7c100413737bdcba157ec8b5ce42895de

Summary:
Infer AL has reported valid warnings on some folly and thrift code:

```
fbcode/folly/io/async/AsyncUDPServerSocket.h:228:19:
WARNING Extra Copy: Potentially unnecessary to copy var `client` at line 228, column 19. Use 'const auto&' or 'auto&' if possible.
fbcode/folly/io/async/AsyncUDPServerSocket.h:230:19:
WARNING Extra Copy: Potentially unnecessary to copy var `socket` at line 230, column 19. Use 'const auto&' or 'auto&' if possible.
fbcode/thrift/lib/cpp/TProcessorEventHandler.h:148:15:
WARNING Extra Copy: Potentially unnecessary to copy var `ew` at line 148, column 15. Use 'const auto&' or 'auto&' if possible.
fbcode/thrift/lib/cpp2/server/ThriftServer.h:705:20:
WARNING Extra Copy: Potentially unnecessary to copy var `sockets` at line 705, column 20. Use 'const auto&' or 'auto&' if possible.
```

As these are transitively included in many services, they are fairly prevalent/annoying warnings. Let's fix them.

Reviewed By: yfeldblum

Differential Revision: D14168260

fbshipit-source-id: 12e537fe5aa73e653c15c4fef01ac618e51cd4d0

Summary:
- `F14Table.h`'s fallback when SIMD isn't used has template specializations
  to track which `std::hash` implementations are considered worth caching by
  `std::unordered_map`. This included `std::string` but was missing
  support for `std::string_view`. This commit adds support for marking
 `std::string_view` as not fast.
Pull Request resolved: https://github.com/facebook/folly/pull/1031

Reviewed By: nbronson

Differential Revision: D14197661

Pulled By: yfeldblum

fbshipit-source-id: bd634b5755e2794891200c7f71e9d6874a60cc38

Summary:
- `FOLLY_USE_CPP14_CONSTEXPR` macro is not needed as Folly requires a
  recent enough version of MSVC where its constexpr support is "good
  enough". For Clang and GCC, the min compiler versions supported would
  both evaluate the prior implementation of this macro to true in both
  cases. This is potentially slightly behavior changing since
  `FOLLY_USE_CPP14_CONSTEXPR` would be `inline` for ICC and I am not
  sure if its constexpr support is "good enough" for a min version of
  ICC we claim support for.
- Replace `FOLLY_CPP14_CONSTEXPR` with `constexpr` in all call sites and
  remove the `FOLLY_CPP14_CONSTEXPR` macro.
- Simplify how we define `FOLLY_STORAGE_CONSTEXPR` and
  `FOLLY_STORAGE_CPP14_CONSTEXPR` after cutting
  `FOLLY_USE_CPP14_CONSTEXPR`.
Pull Request resolved: https://github.com/facebook/folly/pull/1029

Reviewed By: Orvid

Differential Revision: D14199538

Pulled By: yfeldblum

fbshipit-source-id: 99daecf7d7ad0c4bf6735e74247112a78923602a

Summary: [Folly] Use `pretty_name` in `F14TableStats`, which may not be exact in edge cases (e.g. function-local policy types) but which works without RTTI.

Reviewed By: nbronson

Differential Revision: D14129242

fbshipit-source-id: 6f7a8e0b3af4d6b16673770228e98cd989ab0c3f

Summary: Update to use thenError rather than onError.

Reviewed By: yfeldblum

Differential Revision: D14209634

fbshipit-source-id: 5cbb66146b148654641a7a0a58717a31608a4ac7

Summary:
[Folly] Omit excess landing pads with `StaticSingletonManager` when the type of the global is nothrow-default-constructible.

This changes behavior for nothrow-default-constructible types to terminate immediately if the process is out of memory while performing the bookkeeping around attempting to instantiate a global.

Reviewed By: andriigrynenko

Differential Revision: D14203826

fbshipit-source-id: 8c9becde3b245d6b027c97bd8541c48745727dff

Summary:
- It is verbose at the call sites to cast an enum or enum class to its
  underlying type.
- Introduce `to_underlying_type`, which, given an enum or enum class,
  returns the value from static casting to the underlying type.
Pull Request resolved: https://github.com/facebook/folly/pull/1028

Reviewed By: stevegury

Differential Revision: D14199619

Pulled By: yfeldblum

fbshipit-source-id: fde85b9be50945a390b18ba1448a0137f3d7b63e

Summary: Remove all uses and remove the overload.

Reviewed By: yfeldblum

Differential Revision: D14194039

fbshipit-source-id: c3b271747200b9219b734e5a0f19c6d5d4d135a2

Summary: Fix namespace in folly::copy comments

Reviewed By: yfeldblum

Differential Revision: D14202488

fbshipit-source-id: 14ebc5a58bfd3b62e4d6ecd03e1ceffa796f4b1f

Summary: The codemod tooling has demonstrated issues distinguishing between integer and bool overloads, so add a safety net just in case.

Reviewed By: yfeldblum

Differential Revision: D13995489

fbshipit-source-id: 2a853f8c961953d4cdd24256420ecc29cb13c677

Summary: The file descriptor overload will be going away.

Reviewed By: yfeldblum

Differential Revision: D14070240

fbshipit-source-id: 20a0248fd9840629e31a3274032b889b6bb5f8de

Summary: They are dead.

Reviewed By: yfeldblum

Differential Revision: D14191874

fbshipit-source-id: 80e6b9c51aa7d29ccc01148ccb1fd1014688c764

Summary: [Folly] Shrink `StaticSingletonManager` inline slow path by passing only one argument to the outline slow path, rather than passing three arguments. In optimized builds, the argument is an immediate loaded into a register.

Reviewed By: andriigrynenko

Differential Revision: D14121341

fbshipit-source-id: f715225d0dc94df9bf2bef440ffeb08bb7a88fba

Summary:
[Folly] Cut outdated comment in `FOR_EACH_RANGE` implementation details.

It refers to a shortcoming in a specific version of boost which is no longer supported in folly.

(Note: this ignores all push blocking failures!)

Reviewed By: Orvid

Differential Revision: D14197473

fbshipit-source-id: 3379af1cc69ab96fcb77be0a56ea7a5762ff7db5

Summary:
When adjusting nextBucketStart by rounding down, we should also check if
that causes it to not overlap with the user-specified [start, end) time
interval. If it does not overlap, then return an empty return type since
this bucket does not contribute anything to the specified time-range.

Reviewed By: simpkins

Differential Revision: D14121401

fbshipit-source-id: bbfb1d6a71c9fb99244cdff887a7a9ee18741f25

Summary: [Folly] Revert `StaticMeta` deleted dtor, which does not work well with standard constructibility tests like `std::is_constructible<StaticMeta<...>, ...>`.

Reviewed By: andriigrynenko

Differential Revision: D14181902

fbshipit-source-id: de75fe2fc864c9a32a74b8058532ef49412ac838

Summary: [Folly] `FOLLY_TYPE_INFO_OF` for cases where `folly::type_info_of` is insufficient.

Reviewed By: andriigrynenko

Differential Revision: D14181350

fbshipit-source-id: 65d63424a5feaa6597eaddb670f760f0223717ae

Summary:
Add XCHECK_EQ(), XCHECK_NE(), XCHECK_LT(), XCHECK_LE(), XCHECK_GT(), and
XCHECK_GE(), plus corresponding XDCHECK_*() versions that only fail in debug
buildk.

These are similar to XCHECK()/XDCHECK(), but on failure they also log the
values of the two expressions being compared.

Reviewed By: therealgymmy

Differential Revision: D14016494

fbshipit-source-id: dba02d748f78d306227ec734d9e52ef0bc73919b

Summary:
Fix an `invalid-shift-base` UndefinedBehaviorSanitizer failure.
Previously all of the DistributedMutex-inl.h tests would fail with the
following message on my system:

  runtime error: left shift of 94500093116194837 by 8 places cannot be
  represented in type 'long'

It might be slightly nicer in the long run to change this code to use
`std::chrono::duration<std::uint64_t, std::nano>` throughout rather than
`std::chrono::nanoseconds`.  Currently the `time()` function casts the
`uint64_t` value returned by `folly::hardware_timestamp()` into a signed
value.

Reviewed By: yfeldblum, aary

Differential Revision: D14180336

fbshipit-source-id: b199ae22d951162dc6f31d7f1c41a7d67cbcc935

Summary: [Folly] Better constrain `EvictingCacheMap` iterator conversions.

Reviewed By: shixiao

Differential Revision: D14162009

fbshipit-source-id: 8a3011450620fcf1ee6a137834ebbd1e5953e0ed

Summary:
We observed some RSS regression for certain use cases - setting decay time to 0 should fix this as freed regions will be released back to the kernel immediately instead of delayed.

The risk is that we lose regions before we can reallocate them and thus over time end up with holes in the huge page region.
If we see this becoming an issue for some use cases, we can add a parameter to control the decay time.

Releasing immediately should be OK for huge pages in most cases. For cases with frequent allocation and deallocation of large areas there can be a perf impact however.

Reviewed By: interwq

Differential Revision: D13853171

fbshipit-source-id: 83fb95ed9c9bcb6ebdd1f80d78347e5eec61c084

Summary:
- When Folly supported autotools builds, a different include for
  `dwarf.h` was required in `folly/experimental/symbolizer/Dwarf.cpp`.
- Since Folly does not support autotools  builds as of `1d58fd57`, cut
  this include and just assume `<dwarf.h>` works.
Pull Request resolved: https://github.com/facebook/folly/pull/1023

Reviewed By: simpkins

Differential Revision: D14119901

Pulled By: yfeldblum

fbshipit-source-id: f676f72ec91e25390cb3bbbe544de7cbc68e1d73

Summary: [Folly] `type_info_of`, which returns `typeid` when RTTI is available, otherwise `nullptr`.

Reviewed By: swolchok

Differential Revision: D14159675

fbshipit-source-id: b64ab770f12c7385bb286f658627d085e561def6

Summary: [Folly] Simplify `is_negative`.

Reviewed By: stevegury

Differential Revision: D14141289

fbshipit-source-id: 88bfd5923d90c5ac23d431110c0f2dc6c476d868

Summary: demangle.h isn't available on macOS, so only include it on Linux.

Reviewed By: andrewjcg

Differential Revision: D14114593

fbshipit-source-id: 59b12bfb4a52ada636b3648115dec1383f58c22f

Summary:
[Folly] Avoid unary negation of unsigned in `FormatValue`, which some compilers may warn against.

```
folly\format-inl.h(452): error C4146: unary minus operator applied to unsigned type, result still unsigned
```

Reviewed By: Orvid

Differential Revision: D14141989

fbshipit-source-id: 3bff06fe09fd4ec22d93b72a74f086ad9c576692

Summary: [Folly] Remove a legacy guard around an `exceptionStr` overload protecting against platforms lacking `std::exception_ptr`.

Reviewed By: mzlee

Differential Revision: D14129260

fbshipit-source-id: 2947c5b00d901ccaf58aefeef51e442121777eba

Summary:
[Folly] Use `bit_cast` in `Endian`.

Requires moving `bit_cast` to the top of the header.

Reviewed By: aary

Differential Revision: D14080604

fbshipit-source-id: cae28003895f1326138459c3a951d38e27e27506

Summary: [Folly] Cut unnecessary `assume_unreachable` in `exceptionStr`.

Reviewed By: meyering

Differential Revision: D14129281

fbshipit-source-id: a7ae13188bb19d829a6c977243719285186dbfec

Summary: Removes references to custom patches to OpenSSL 1.0.2 for async from AsyncSSLSocket.

Reviewed By: anirudhvr

Differential Revision: D13908917

fbshipit-source-id: a01c1a8a7523dd7b4b95a7f701cc6b356c645332

Summary:
- Backporting `std::index_sequence`, `std::integer_sequence`, and other
  associated template aliases such as `index_sequence_for` and
  `make_index_sequence` is not needed as Folly requires C++14. These
  functions and template aliases are available in C++14 standard libraries,
  including GCC 4.9.
Pull Request resolved: https://github.com/facebook/folly/pull/1022

Reviewed By: Orvid

Differential Revision: D14119891

Pulled By: yfeldblum

fbshipit-source-id: 5ba55cc91495cf488cebfa1ebe82c0ff919e9840

Summary: getKeepAlive can get a KeepAlive object and return a copy of itself

Reviewed By: yfeldblum

Differential Revision: D14095402

fbshipit-source-id: 9d00ebecddc938e8cf02bcb7cee0642a666db349

Summary:
I'm not entirely sure the original motivation for this, but for the current
executable, we've been normalizing addresses from the virtual address referenced
by the first "LOAD" program header instead of 0.  This appears to be a bug, as
it forms invalid addresses into the ELF file.

In the case of the signal handler test, this happens to work with gold, as the
initial section lookup just happens to hit in the `.debug_str` section (as opposed
to the expected `.text` section), and the following name translation then undoes
the base address normalization to get a correct lookup.

While this works for gold, it breaks for LLD and meant we had to opt-out folly
out (D13279459).

Reviewed By: yfeldblum

Differential Revision: D14119804

fbshipit-source-id: 53dcaeb72ed75e58f9a46b15799ce1af5ff80531

Summary:
OpenSSL 1.1.1 adds TLSv1.3 support, but changes several semantics (e.g.
assumptions on SSL_get_session() returning resumable sessions) that require
some work to address. A lot of our AsyncSocket tests fail (e.g. tests for
resumption will automatically fail, since TLSv1.3 has no resumption support),
and would need to be updated to explicitly disable TLSv1.3.

The plan is to eventually remove this after these items are addressed.

Reviewed By: yfeldblum

Differential Revision: D14073093

fbshipit-source-id: 181b05395ed35aaa7deb00b8968ff4d371b683f4

Summary:
- Backporting `std:exchange` is no longer needed as Folly requires C++14
  support and C++14-compliant vendors have `std::exchange` in their
  standard library, notably including GCC 4.9.
- Cut backport of `std::exchange` and fix call sites to explicitly use
  `std::exchange` where they were previously relying on `exchange` being
  found in the `folly` namespace.
Pull Request resolved: https://github.com/facebook/folly/pull/1021

Reviewed By: Orvid

Differential Revision: D14119875

Pulled By: yfeldblum

fbshipit-source-id: 686a8e812f6728281b432eae1586de28d21da9dd

Summary:
A specialization for reading into enums from serialized form was added a while back. Unfortunately, no corresponding change to serialize from enums into a dynamic was added.

This change adds the corresponding DynamicConstructor specialization for enums to allow serializing enums and objects containing enums.

Reviewed By: ot, shixiao

Differential Revision: D14091417

fbshipit-source-id: 15ef8bc46843cd668b4b2c991e84167dabaf69fa