Summary: Pull Request resolved: https://github.com/facebook/folly/pull/1489

Differential Revision: D25156288

Pulled By: yfeldblum

fbshipit-source-id: 01c9e6898158e9651367144c22f215ef75ff7ae3

Summary: Added yaml-cpp library to the fboss OSS build

Reviewed By: shri-khare

Differential Revision: D25126509

fbshipit-source-id: 1e88150dbf8c96a5ac696e003265dc1f14bec105

Summary: The new library firmware_storage needs the yaml-cpp. For OSS build this diff adds the yaml-cpp library to manifest

Reviewed By: shri-khare

Differential Revision: D25115387

fbshipit-source-id: 9ed1c374df7e87f284c48d366095a20183719534

Reviewed By: yfeldblum

Differential Revision: D24739525

fbshipit-source-id: 1ad55b470dd305dea6f3c7eeb56890ea6b59d252

Differential Revision: D24757714

fbshipit-source-id: 1a494c5eb758480431d5aa70861359e67e9f831c

Summary:
The fix was part of D22249713 / D24571536. However, the error is not specific to
uint8/16 support. Let's fix it right away.

Reviewed By: ot

Differential Revision: D24629540

fbshipit-source-id: 8cbc3226259d49ec3e64c03ca55f3ef1ac5b981d

Summary:
Updates the co_awaitTry() adapters to support the co_withAsyncStack()
CPO.

This required updating the WithAsyncStackAwaitable to also support
co_awaitTry() but having its wrapper awaiter type also forward through
the .await_resume_try() method if one exists.

Modified the is_awaiter trait check to no longer check for a valid
await_suspend() call as some awaiters returned from co_withAsyncStack()
now require coroutine_handle<Promise> and no longer accept the
coroutine_handle<void> parameter previously checked by the trait.

Reviewed By: andriigrynenko

Differential Revision: D24499833

fbshipit-source-id: fdcb99a2b85c4235bcd4b5dcc12cf79cb414f034

Summary:
The stack-traces for child worker tasks launched to process each
incoming AsyncGenerator now have the same common consumer-task
as the parent frame, which is then a detached-task, rather than
each of the child workers being their own detached tasks.

Reviewed By: andriigrynenko

Differential Revision: D24480161

fbshipit-source-id: c02887abf676bde9b297ee6d8f0fcabe959616e3

Summary:
This should allow stack-traces for a chain of async work that was launched
using blockingWait() to continue tracing the normal stack-frames of the caller
of blockingWait().

Reviewed By: andriigrynenko

Differential Revision: D24479522

fbshipit-source-id: e0604ed43edbd9df51daf679e0adf06d934a3ca6

Summary:
Modify the AsyncScope::add() method so that it records the call-site of .add()
as the first frame's return address instead of recording .add() itself as the
first frame.

This should hopefully give a bit more insight into where the detached task was
launched.

Reviewed By: andriigrynenko

Differential Revision: D24478808

fbshipit-source-id: 0723891e77af5e52c7f8c13e218e4fb9120867e3

Summary:
The collectAll() family of algorithms now hooks up async-stack frames
of child operations to ensure that they trace back up through to the
caller of collectAll() rather than stopping at the call to .start()
inside collectAll().

This also adds the helper `folly::coro::detail::co_current_async_stack_frame` which
allows the coroutine to get access to its own AsyncStackFrame
object - which is needed for algorithms like collectAll() which
then need to pass this on to the BarrierTask when it is started.

Reviewed By: andriigrynenko

Differential Revision: D24466130

fbshipit-source-id: f7058388ebc8fa212684eb0edc0a83deef6fc2a2

Summary:
It now saves and restores the async stack frame natively which
avoids creation of a wrapper coroutine to do this.

Reviewed By: andriigrynenko

Differential Revision: D24464577

fbshipit-source-id: 965893a7f2e30c0759fb1970784500c96ed8a8ac

Summary:
This should avoid creating two temporary wrapper coroutines to apply
both co_viaIfAsync() and co_withAsyncStack() to awaitables when awaited
within a Task/AsyncGenerator coroutine.

The default co_viaIfAsync() implementation now customises the
co_withAsyncStack() CPO and saves and restores the async frame
itself if the wrapped awaitable does not support async stacks,
or otherwise passes the async frame through to the child awaitable
if it does support async stacks.

Reviewed By: andriigrynenko

Differential Revision: D24464513

fbshipit-source-id: f9751030f3c1d2725a00cb803ee46ee12dddb2ed

Summary:
The DetachedInlineTask coroutine type now has its own AsyncStackFrame.

While this is largely an internal implementation detail used only
for the TaskWithExecutor's `.start()` family of methods, this now
means that detached tasks launched using TaskWithExecutor's
`.start()` methods will now record the return-address of the call to
`.start()` in the stack-trace and will also have the async-stack begin
with the `detached_task` frame at the top-level.

Reviewed By: andriigrynenko

Differential Revision: D24440817

fbshipit-source-id: 861f1ca46ec4a28b5b7e388bb8b243e4d597ee0d

Summary:
AsyncGenerator coroutines now have their own AsyncStackFrame and so can
participate in async call-stack tracing.

AsyncGenerator now applies co_withAsyncStack() to all awaitables
awaited from within the generator coroutine.

The NextAwaiter now implements the co_withAsyncStack() CPO to advertise
that it will save/restore the awaiting coroutine's AsyncStackFrame if
it provides one.

This allows the AsyncGenerator to symmetric-transfer to awaited Task
coroutines and vica-versa. This fixes some stack-overflow issues with
some synchronously-completing use-cases of AsyncGenerator.

Reviewed By: andriigrynenko

Differential Revision: D24437278

fbshipit-source-id: f9612d54558b961bb023460d4a118a773022737b

Summary:
`BarrierTask` and `DetachedBarrierTask` coroutines now have their own `AsyncStackFrame`.

This means that when a `Task` executes `co_await barrier.arriveAndWait()` that we
can use symmetric transfer to resume the `Task` since the `BarrierTask` will
already have an active `AsyncStackRoot` that we can reuse.

This fixes the test for stack-overflow of synchronously completing coroutines that
use `collectAll()` algorithms (which internally use `BarrierTask`).

Reviewed By: andriigrynenko

Differential Revision: D24435340

fbshipit-source-id: 87b622e8083f6c8f2b12f6b2f4b0bccd8bd5a595

Summary:
Initial diff that adds support for tracing through a chain of folly::coro::Task coroutines.

This adds some scaffolding that handles saving/restoring the active AsyncStackFrame
when a Task awaits some awaitable type that does not know about AsyncStackFrame
objects but also allows awaitables to opt-in to AsyncStackFrame awareness by
customising the new `folly::coro::co_withAsyncStack()` CPO.

Currently only `Task` and `TaskWithExecutor` awaiters have customised this CPO.

Also updated the awaiters for `Task` and `TaskWithExecutor` to handle being awaited
from coroutines that are not async-stack aware - in which case they just record a
null parent-frame for the Task.

The Task's `final_suspend()` then either deactivates or pops the frame depending
on whether there was a parent frame recorded.

BUG: This change currently breaks the symmetric-transfer stack-overflow avoidance
when awaiting synchronously-completing coroutines from an AsyncGenerator or from
a BarrierTask (eg. inside collectAll implementations).

Reviewed By: andriigrynenko

Differential Revision: D24428736

fbshipit-source-id: 5722e511ad10d95198ae70a5afe567d83cb06285

Summary:
Paths are not necessarily legal glob syntax. In particular, backslash
is used for escaping. This caused problems on Windows, where we tried
to pass a backslash-delimited path into `eden prefetch --no-prefetch`.

Reviewed By: xavierd

Differential Revision: D25072784

fbshipit-source-id: 9ce8e5ccc8f28581512c39d04922889da0bc1bf6

Differential Revision: D25018383

fbshipit-source-id: eefc622ad60d663efd174ad341df7697c548b3bd

Summary: Fix a double delete issue

Reviewed By: danobi

Differential Revision: D25005179

fbshipit-source-id: 03ae8f0722316294979ec7511ef9ac794196a7ce

Summary: Fix for: https://github.com/facebook/wdt/issues/203

Reviewed By: vitaut

Differential Revision: D24954835

fbshipit-source-id: 92f29f71cd6eef9bdcbe92ed1ecb8f1f5d24326e

Reviewed By: andriigrynenko

Differential Revision: D24880761

fbshipit-source-id: 643da2b0fd461955ea33b4ba48f377c30720d442

Summary: The worker thread just needed to be killed before the `AcceptCallback`. `AsyncServerSocket` calls a method on this callback class in its dtor, so it must either be alive or removed using `removeAcceptCallback`.

Differential Revision: D24972293

fbshipit-source-id: 2c7b6d64c2b9b8c00d03107002fa31e565df2a01

Summary:
Added a timestamp to `AsyncServerSocket::QueueMessage` so that `NotificationQueue` can ignore new connection messages which are deemed *expired*. Expired messages represent sockets which have *probably* timed out already.

The TTL is configured per `AsyncServerSocket` instance and is applied to all future messages that are queued by it. By default, messages do not expire. This can be configured with `AsyncServerSocket::setQueueTimeout`.

Reviewed By: andriigrynenko

Differential Revision: D24667870

fbshipit-source-id: 0f9d6c235627393d964e280a0d5956676010c7aa

Summary:
If `Consumer::operator()` returns `AtomicNotificationQueueTaskStatus` then this value indicates if the consumer actually consumed the task, or if it should be discarded.

Enumerating all cases:

`operator()` return type is `void`:
*All tasks are considered consumed and counted towards `maxReadAtOnce`.*

`operator()` returns `AtomicNotificationQueueTaskStatus::CONSUMED`:
*Same as above.*

`operator()` returns `AtomicNotificationQueueTaskStatus::DISCARD`:
*Task does not count towards `maxReadAtOnce` and is silently discarded. The consumer is responsible for any cleanup other than calling the destructor.*

Reviewed By: andriigrynenko

Differential Revision: D24698221

fbshipit-source-id: 0da2364d18b67468addf737b67cae573c89b7e9c

Summary:
Add an API to enable setting signature algorithms in string form.
An example string form: `"ECDSA+SHA256:RSA+SHA256"`.
OpenSSL supports TLS1.3 Signature Scheme format,
e.g.; `"ecdsa_secp256r1_sha256:rsa_pss_rsae_sha256"`.

Reviewed By: mingtaoy

Differential Revision: D24745481

fbshipit-source-id: 018be0a242d922f36a81d051791679a20fe08893

Summary:
Fixes https://github.com/facebook/folly/issues/1486

Pull Request resolved: https://github.com/facebook/folly/pull/1487

Reviewed By: Orvid

Differential Revision: D24851586

Pulled By: yfeldblum

fbshipit-source-id: ae683ba73336c562a8912ed35ef2b3318ab53519

Summary: To support cases where the child process may outlive the parent process.  Detaching the process is an option as well, but in that case the pid for the child process is not captured.

Reviewed By: yfeldblum

Differential Revision: D24409176

fbshipit-source-id: f74f64e3ae21baa40468c6f39ac6b4d0152db03a

Summary:
We often run into cases where system CPU util is low but SR eventbase threads are backing up. During S213134 debugging, it took us a while to figure out that requests with http transport were expensive and causing issues on intern-rpc clients and later realized it was a known issue for http transport to be CPU intensive. We got lucky that http transport requests used different stack we could see in strobelight, but this will likely not be the case for other similar issues.

For ease of debugging these issues, one idea is to log SR Eventbase cost of each request to SR scuba. **The closest proxy for this is the wall clock time spent by SR's request-dispatch task in fiber, excluding all the preemptions.** This metric does have couple of drawbacks:

- It wont capture all the work a request does via runInMainContext. This should be Ok since SR tries not to do lot of work in mainContext of fiber manager.
- If system context switching the threads a lot and if eventbase thread is preempted by OS, we will count that time as time spent in task. I am hoping that new metric will still be useful even with this caveat.

Design: Allow fiberManager.AddTask() to optionally take in a bool `logRunningTime` to decide whether it can log the running time of the task. If `logRunningTime` is set, fiber will keep track of elapsed time since most recent run in  `currStartTime_` and track the running time of all previous runs together in  `prevDuration_`.

I looked into using ExecutionObserver for this purpose. I did not go that route because
1. ExecutionObserver's callbacks return fiber IDs, we want a callback to return function/task ID that we can tie back to Sr request.  See here https://fburl.com/diffusion/qh9ioeq3
2. Caller (SR) will need to maintain the separate cache of all outstanding tasks with their running_times and manage their lifetime, which feels more complex than current approach.

I have a way to turn off this change using new killswitch I added in SR.

Reviewed By: yfeldblum

Differential Revision: D24823396

fbshipit-source-id: 62f79b8c340cc48a22684c9cb3cdddadc260a0ab

Reviewed By: andriigrynenko

Differential Revision: D24698220

fbshipit-source-id: aab8c96440d60bb653f0c537af56081819129306

Summary:
Allowing a bounded `AtomicNotificationQueue` is helpful to migrate `AsyncSocketServer` over from `NotificationQueue`.

`AtomicNotificationQueue::tryPutMessage` is the same as `putMessage` but also accepts a `maxSize`. If the queue size is >= `maxSize`, the message will not be queued. The user of the queue is responsible maintaining state regarding the max size of the queue.

Reviewed By: andriigrynenko

Differential Revision: D24698219

fbshipit-source-id: 1e3d1dc0073c65cd1803629f6a4a9b015e111b08

Summary: Add retryingUnsafe as a duplicate of retrying that forces the return type to be a Future rather than a SemiFuture.

Reviewed By: yfeldblum

Differential Revision: D24707194

fbshipit-source-id: cb2a107f51d4d4dd47a41ce6f8a009c4d82e3138

Summary:
`add_fbthrift_cpp_library` should honor `BUILD_SHARED_LIBS`, and call
`add_library` with the right setting (`SHARED` if enabled, `STATIC` otherwise)

Reviewed By: yns88

Differential Revision: D24911124

fbshipit-source-id: 79df7640a758a592a3df3e9e79bb129dd57f2d47

Summary:
needed for QUIC/H3 interop

also make it possible to run getdeps build with extra arguments (--no-tests for
example)

Reviewed By: mjoras

Differential Revision: D24925777

fbshipit-source-id: fbdc1aa56e398d295ef8dac0ad0bab03bd7bd803

Reviewed By: zertosh

Differential Revision: D24880203

fbshipit-source-id: 2034cdfc2712209e86d3d05c119c58f979b05c52

Summary:
Right now proxygen hardcodes a static build when using cmake  and
ignores BUILD_SHARED_LIBS. Fix that, and enable PIE on the shared libs so they
can be linked properly

Closes: https://github.com/facebook/proxygen/issues/335

Reviewed By: mjoras, lnicco

Differential Revision: D24787944

fbshipit-source-id: 7a654af7cb43227ca913a1bed67f2432703a343d

Summary: This header is needed by python/futures.h

Reviewed By: yfeldblum

Differential Revision: D24861734

fbshipit-source-id: 913ad13e6ca155250238020538af762c2360f137

Summary: Allow recycling of std::unique_ptr<IOBuf>

Reviewed By: yfeldblum

Differential Revision: D24650191

fbshipit-source-id: 85d219052cc62c35098085abb3eed6cfe00beefc

Summary:
Honor `DESTDIR` on Unix-like platforms (important when building packages),
and use `CMAKE_INSTALL_PREFIX` instead of the default prefix.

Reviewed By: yfeldblum

Differential Revision: D24821935

fbshipit-source-id: 42a311f920ed74760597f8908cd339b914ecb92e

Summary: `exception_wrapper::handle` uses the signature of `Fn::operator()` to pick apart the function type and deduce the exception type for the handler. We use specializations of an `arg_type_` class template to do this. In C++17, `noexcept` is a part of the type system, but there were no specializations of `arg_type_` to handle `noexcept`-qualified callables.

Reviewed By: ispeters

Differential Revision: D24733871

fbshipit-source-id: 344e735e653296c9344b6c09662fccd94953e8a7