Summary: Skip the EventBase tests if the backend is not available

Reviewed By: yfeldblum

Differential Revision: D19002064

fbshipit-source-id: a995df02998c80de2f9076ea16a0281b6288d6b5

Summary: Fix io_uring backend tests

Reviewed By: yfeldblum, danobi

Differential Revision: D19023775

fbshipit-source-id: d696064d26b37f52529c84601f6797ef82b0ddb5

Summary: EventBase test lib rework

Reviewed By: yfeldblum

Differential Revision: D19001877

fbshipit-source-id: d99ad41007012c7bd21ee259a0a0b5ee2b709512

Summary: Add IoUringBackend::NotAvailable exception

Reviewed By: kevin-vigor

Differential Revision: D19143143

fbshipit-source-id: 7b55106d50882d34fc6100a81ff2a4e5748e9b85

Summary: This allows makeObserver to work even during shutdown.

Reviewed By: yfeldblum

Differential Revision: D19136597

fbshipit-source-id: 17e946a66743a56d24904697c02ce7165bcaf397

Summary:
Linux thread identifier is something that we can get with `gettid`.
http://man7.org/linux/man-pages/man2/gettid.2.html

This TID is super usefully for providing thread information while debugging on Linux (this is
the thing that `ps` command uses, or procfs), but is not really mappable from
`std::thread::id` (e.g. the thread handle we get with `pthread_self()`).

Reviewed By: yfeldblum

Differential Revision: D19134596

fbshipit-source-id: 32ad025c5e378e3c7b55da1aad8dd9f3e26d0892

Summary:
While running a system lots of different values will be added to lots of different `MultiLevelTimeSeries`. These values can be timestamps, lengths of time, file sizes, just about anything. There's no way for the underlying library to know before hand how large they will be. So it's always going to be possible to cause undefined behavior.

That means we could fix this one of two ways. We could make this code be ignored for ubsan, or we could make it never roll over.

I have made it so that the code won't roll over here. This seems better because it will continue to give more representative numbers. Rather than going from a very large number to a very small number, we'll remain at a very large number. That should still convey to every one that the numbers being accumulated are very large.

Reviewed By: mzlee

Differential Revision: D16970762

fbshipit-source-id: c56f6db1dbdecefd190e6ac357133886a3642571

Summary: New 20191216 liburing

Reviewed By: danobi

Differential Revision: D19125513

fbshipit-source-id: 38d2800ee75865cb76cb06edc0328feaa44bab3e

Summary: Integrate calls into logging framework to allow tracking of source and use of the global executor.

Reviewed By: andriigrynenko

Differential Revision: D18232728

fbshipit-source-id: e714b870cc041af9c99a95baf8f32c3819852b37

Summary: Adds sampled logging to track executors passed to Future::via.

Reviewed By: andriigrynenko

Differential Revision: D18233242

fbshipit-source-id: b5c004f643261a36fd0b63fb6e033263ece1662f

Reviewed By: Orvid

Differential Revision: D19053964

fbshipit-source-id: dd31937cbeb3c7ef202e4b91c1b153f849376bdf

Summary: [Folly] Prefer `TEST_F` and tests named `...Test` in `folly/experimental/coro/`.

Reviewed By: lewissbaker

Differential Revision: D18746584

fbshipit-source-id: 49156a1dca5ef115bf12add5ceb38cb7905fac3d

Summary: These files fail to build when a different version of glog is used. Notice that both cpp files use gflags but don't include its header.

Reviewed By: pixelb

Differential Revision: D18974446

fbshipit-source-id: 4cfaa6dc62c795f4fe7f6624f078cffe550ecfec

Summary:
D18890515 broke folly builds when zstd isn't present. (My bad!!)

This should fix it.

Reviewed By: lukaspiatkowski

Differential Revision: D19029293

fbshipit-source-id: 2ede5a8402e60af61a294c9db35ca15dffdf4c49

Summary:
`folly::AtomicBitSet<N>` used `N` to denote the number of blocks
(32-bit or 64-bit each) instead of number of bits as used by
`std::bitset`. This confusion of unit led to unintentional allocation
of extra storage in many cases.

A simple fix would be to make `N` denote number of bits. However,
changing meaning of `N`could lead to silent breakage of code during
run time for some users who used `N` to denote number of blocks.

Therefore, to make the new meaning of `N` more explicit,
`AtomicBitSet` is renamed to `ConcurrentBitSet` & `N` now denotes the
number of bits.

Reviewed By: al13n321

Differential Revision: D18866751

fbshipit-source-id: 17741646862776e1a080eaac7877f495bb56b656

Differential Revision: D19001755

fbshipit-source-id: bda1a278017059e83c9a529aac7b429f074e42c1

Summary: We should be able to associate rootId with the TID the request is running on.

Reviewed By: yfeldblum

Differential Revision: D18918492

fbshipit-source-id: 539cd78729b438928a9a4b0c93e8e17295b642d5

Summary:
What:
exposed an API `folly::RequestContext::getRootsFromAllThreads` that reports all "root" RequestContexts for all threads in the process.
(where "root" is defined as the original RequestContext created for the request but which might have been overridden with `ShallowCopyRequestContextScopeGuard)`

Reviewed By: andriigrynenko

Differential Revision: D18849187

fbshipit-source-id: f3c67becf1e38fb8d2198c81ed0ce77bac860d63

Summary:
Fixing couple more failures in folly when compiling wasm:
1) making posix_memalign used instead of memalign (wasi-libc has memlaign as legacy P123789170)
2) disabling clearenv() that is using environ, which wasi-libc has off by ```#ifdef _GNU_SOURCE```, not sure why, but it's already off for linux and we won't use it our wasm code either.

Reviewed By: shodoco

Differential Revision: D18944271

fbshipit-source-id: ffa838591ada1248fdb8ae24f7eac45896a311e2

Summary:
This is intentionally not hidden in `folly/compression/Zstd.cpp`, so that it
can be used by other Zstd users as well.

Reviewed By: bimbashrestha

Differential Revision: D18890515

fbshipit-source-id: e87eb6acd03f4b4d5f68201ef924984047d0160e

Summary:
This reduces contention on the synchronized stack in the
`CompressionContextPool`, as well as potentially helping cache locality under
heavy compression workloads.

I have plans to make a singleton of this, and to use it in `folly::io::Codec`
and some other places, so they all use the same contexts.

Reviewed By: bimbashrestha

Differential Revision: D18863824

fbshipit-source-id: 63c2f3dd0abe0aa4c88350767e9e77acb763c9a5

Summary: `F14Vector{Map,Set}` have specified iteration order: inverse insertion order. When serialized, however, the order does not round-trip (in fact, it flips). This diff introduces an extension point to specify the iteration order for serialization, and it specializes it for F14 to make the iteration order round-trip.

Reviewed By: vitaut

Differential Revision: D17108774

fbshipit-source-id: 466b746baacff861242729bfa62447c138094cee

Reviewed By: yfeldblum

Differential Revision: D18926899

fbshipit-source-id: d16c1b5b54a75fe70514f25b648de70ae66b6415

Summary: [Folly] Define support macro in all branches v.s. as a fallback for style.

Differential Revision: D18900489

fbshipit-source-id: 9f9cb3c70889c8d6800070221bd8fe9ce091745c

Summary: Stub interface for logging functions to integrate with the GlobalExecutor and Futures.

Reviewed By: andriigrynenko

Differential Revision: D18229089

fbshipit-source-id: ad57de08a8bd6a80e2d0af9a9792de86a8267176

Summary:
This should fix the Travis CI builds. It adds rust toolchain support inside docker and sets the required THRIFT env variable.
Pull Request resolved: https://github.com/facebookexperimental/rust-shed/pull/3

Reviewed By: krallin

Differential Revision: D18905608

Pulled By: lukaspiatkowski

fbshipit-source-id: 5db1eff6f215a6617d8acaa0c99a62d45225956b

Summary: [Folly] Helper variable templates for tuple applicability traits: `is_applicable_v`, `is_applicable_r_v`, `is_nothrow_applicable_v`, `is_nothrow_applicable_r_v`.

Reviewed By: lewissbaker

Differential Revision: D18746853

fbshipit-source-id: 9d01b24030f95ab5b4963660dc4ef471b5e0ad02

Summary: Add support for io_uring based EventBase backend

Reviewed By: kevin-vigor

Differential Revision: D18682173

fbshipit-source-id: d6e02879e2b493680caaa3097d75970d0b5d7d07

Summary: Helps test resumption failure/success in other tests that use AsyncSSLSocketTest classes.

Reviewed By: mingtaoy

Differential Revision: D18561383

fbshipit-source-id: 29852ede80ecd59f662f536368ce2efe7d26745d

Summary: This fixes a timing issue that has been observed in rare cases under very heavy load.

Differential Revision: D18875069

fbshipit-source-id: cd47abfbd01c10e2568693f876dd3e6c2cb3bb24

Summary: Add support for setting a zerocopy enable function

Reviewed By: kevin-vigor

Differential Revision: D18579527

fbshipit-source-id: dc3ab7bb13b26528bc964c7eb616517e444487a4

Summary: To enable building with -Werror,-Wundef.

Reviewed By: guangyfb

Differential Revision: D18850905

fbshipit-source-id: 95555781fec991eac119c95a6a6cd7751b5ba742

Summary:
When compiling each file individually, bring the includes
closer as they are referenced (e.g., rely less on transitive includes
and include what you use)

Reviewed By: bithree

Differential Revision: D18864197

fbshipit-source-id: 24adef418fd2eb944aa283a55411ec5a7c6f549b

Summary: Currently if compiler decides to inline fiber function, we'll end up allocating variables on fiber stack even though we use `runInMainContext`.

Reviewed By: andriigrynenko

Differential Revision: D18842065

fbshipit-source-id: 847c4f95c44ef49f5c7537af878ca8340f2b7b16

Summary:
Update the getdeps builders to save all command output to a log file in the
build directory.  This makes it easier to go back after the fact and look at
the config options selected by the build run.

Reviewed By: wez

Differential Revision: D18777059

fbshipit-source-id: 86d3ffa48ec992fe90c59a3835e18f08310ed88c

Summary: Add check for `__cpp_lib_atomic_is_always_lock_free` for using `is_always_lock_free`.

Reviewed By: davidtgoldblatt

Differential Revision: D18827013

fbshipit-source-id: 796d1ce3c279a807198897c092464dc3aff734e3

Summary: `folly/portability/Config.h` serves this purpose now. Clean up references

Reviewed By: Orvid

Differential Revision: D18800127

fbshipit-source-id: 8e1300af74d63f028b83f5e2d8963a674e77e573

Reviewed By: mitrandir77

Differential Revision: D18850917

fbshipit-source-id: f211dfbfadf4b5b0cb372cf8aeb9a232c4b3a739

Summary:
Pull Request resolved: https://github.com/facebook/folly/pull/1273

[Folly] Skip some `is_constexpr_default_constructible` tests under gcc7.

In gcc7 specifically, i.e. not gcc6 and not gcc8, some of the expectations fail. The compiler is wrong and this affects the outcome of `is_constexpr_default_constructible`, but this is mostly harmless as far as existing uses go. Fix the tests by skipping them when using gcc7.

Reviewed By: vitaut

Differential Revision: D18820312

fbshipit-source-id: a2d0d53ed2716fffba7840e2b167e9ad4ac012ff

Summary: [Folly] Add `kGnuc` C++ constant representing `__GNUC__` preprocessor constant.

Reviewed By: vitaut

Differential Revision: D18827910

fbshipit-source-id: 43e07e12017df1e7cf4d6a5747e7a85d8a7616f3