Summary: It was broken by D5139886 which added a call to `std::__throw_bad_alloc` without first including the correct portability header.

Reviewed By: yfeldblum

Differential Revision: D5273016

fbshipit-source-id: cac1772f41ae5942a29cefa69fb07d9cf94b5aab

Summary: I broke it when I moved the stats tests around.

Reviewed By: yfeldblum

Differential Revision: D5258526

fbshipit-source-id: 70982ffaaed89a548e943e4b011b013c89808666

Summary: It's not safe because of how it manipulates the linked list.  The intended usage is to declare it locally on the stack

Reviewed By: yfeldblum

Differential Revision: D5267507

fbshipit-source-id: 07a55eaee2c6a266bb233432f6a3383236594195

Summary: They were incorrectly comparing against string literals rather than the actual demangled names. MSVC includes `class/struct` as part of the mangled name, so they also appear in the demangled name, in contrast to GCC/Clang, which don't.

Reviewed By: yfeldblum

Differential Revision: D5271087

fbshipit-source-id: 41fa0cebe098b1b935e1f8b1af5882c412cf6254

Summary:
Improve documentation and use the correct pragma for warning
on the unsupported path.

Reviewed By: ot

Differential Revision: D5264582

fbshipit-source-id: 710ee46fef6d8f37f665f4bb6f7e4c5dc0b27436

Summary: This reverts commit cd94c711f0390d249b6736ab292721a477022b02

Differential Revision: D4567233

fbshipit-source-id: a10aff6b4c231666aec9f40ab65e881c3cd870dc

Summary:
get_fiber_manager_map_vevb() and get_fiber_manager_map_evb() don't
provide very useful feedback when called with an empty map (just the former) or
with a program that doesn't have debug symbols (both)

Reviewed By: andriigrynenko

Differential Revision: D5260114

fbshipit-source-id: c44b4e279e5c88dc08507b969339a4befc23d79b

Summary: Previous version failed to compile with "-Werror=shadow" flag, this commit fixes the problem.

Reviewed By: WillerZ

Differential Revision: D5255814

fbshipit-source-id: c1252474ed5415b47759022bcbabc78c1639e10a

Summary: MSVC, GCC, and Clang define the relative comparison operators by default on `enum class` values, however under MSVC defining them explicitly causes the compiler to fail to resolve which operator to use, so just remove them entirely, as they aren't needed.

Reviewed By: simpkins

Differential Revision: D5259835

fbshipit-source-id: 1b91db5de0fc47320daabfdd7132ae910887ff0f

Summary: There are a couple of places where the caught exception was being given a name that was never used, generating a warning under MSVC.

Reviewed By: simpkins

Differential Revision: D5260131

fbshipit-source-id: f82c6bd1266f6a4c916594ec3ac94edc9a2e48fe

Summary: Glog doesn't try to, so there's no real reason to try and be smart. Just do the same thing as every other platform for now until we have reason to do otherwise.

Reviewed By: simpkins

Differential Revision: D5259435

fbshipit-source-id: dcdf55f51f2e13945afd351cb45e9c1a22f56218

Summary: As MSVC correctly warns via C4266, the overload sets of functions in base classes are not part of the overload set in derived classes where you've declared an overload (or override) unless you've explicitly imported it.

Reviewed By: simpkins

Differential Revision: D5260056

fbshipit-source-id: bbbeeea3c13201a3a6eba6e62cfa5a49a9470d43

Summary: It is needed on Windows for the `localtime_r` function.

Reviewed By: simpkins

Differential Revision: D5259944

fbshipit-source-id: e380ecf42252ecabec168cf87668ace3ff4c0a9f

Summary:
The standard does not define that `std::ostream` should have a default constructor, so this was invalid and was an error on Windows.
This explicitly initializes the base class by passing a `nullptr`.

Reviewed By: simpkins

Differential Revision: D5259690

fbshipit-source-id: b8914d73ff2682e1a4447b8338860259778b2247

Differential Revision: D5251218

fbshipit-source-id: 1ceb37dd727e8ac2fd842e2c437cdaa9017221c8

Summary:
Fix build problems in the open source Makefiles that cause issues when running
`make check`

- The test subdirectory needs to be processed before experimental, since some
  tests in experimental/ depend on libfollytestmain from test/
- The stats/test Makefile had an incorrectly copy-and-pasted line for
  libgtest_la_SOURCES but didn't actually build a libgtest.la library.
- The test/ Makefile defined thread_id_test but forgot to add it to TESTS so
  that it would actually be run as part of "make check"

Reviewed By: Orvid

Differential Revision: D5249132

fbshipit-source-id: 5a71e1f72a39d5407b843a5876891c67238ec006

Summary:
Add a logging/Init.h header file with a couple convenience functions for
initializing log levels and log handlers.

This is pretty basic for now, but simplifies usage for small programs that just
want to easily initialize the logging library.

Reviewed By: wez

Differential Revision: D5083106

fbshipit-source-id: 73c1fd00df2eaf506b9c1485d6afd12570412a0f

Summary:
Add a LogFormatter implementation that logs messages using a similar format to
the glog library.

Reviewed By: wez

Differential Revision: D5083108

fbshipit-source-id: 75f0a6b78ce5406b4557d6c4394f033d5e019f71

Summary:
Add an AsyncFileWriter class that implements the LogWriter interface using a
separate I/O thread to write the log messages to a file descriptor.

This LogWriter implementation ensures that normal process threads will never
block due to logging I/O.  By default it will buffer up to 1MB of data.  If
log messages are generated faster than they can be written to the file, log
messages will be discarded once the buffer limit is exceeded.  The LogWriter
will emit a message into the file recording how many messages were discarded
where the dropped messages should have been.

The downside of this class is that unwritten log messages still in the buffer
will be lost when the program crashes.

Reviewed By: wez

Differential Revision: D5083107

fbshipit-source-id: c67226f4d0726675d480b03eae83a29c5c3431b2

Summary:
Add an ImmediateFileWriter class that implements the LogWriter interface by
immediately writing data it receives to a file descriptor.

This LogWriter implementation can be used by users that want to ensure log
messages are immediately flushed to a file as soon as they are written, at the
expense of possibly having logging I/O block their normal process threads.

Reviewed By: wez

Differential Revision: D5083105

fbshipit-source-id: 15cd071834e1784fe50e1d6d6ce403b9bc907047

Summary:
A few places in the logging code need to report issues when something goes
wrong in the logging code itself.  This adds a new LoggerDB::internalWarning()
method for use in these places.

This provides a default implementation that logs these warnings with
_CrtDbgReport() on Windows, and writes to stderr on non-Windows platforms.
The default implementation can also be overridden for programs that want some
alternative behavior.

Reviewed By: wez

Differential Revision: D5157731

fbshipit-source-id: e27f5587d58b3941fca8d5b38c8c3edbcb75d7f3

Summary:
Add a new RateLimiter API to the logging library, to support rate limiting
messages in the future.  I have included a single IntervalRateLimiter
implementation in this diff.  In the future we can add more implementations if
necessary, to mimic the functionality available with our older logging code, to
make it easier for users to convert to the new framework.

Note that RateLimiter is inside a `folly::logging` namespace, unlike most of
the other code in the logging library that lives directly in the `folly`
namespace.  I intentionally chose this since RateLimiter is a fairly generic
class name, and I wanted to distinguish it from other possible generic class
names in folly.  On the other hand, most of the other class names already start
with `Log`, so there seems to be no need to put them in a `logging` namespace.

Nothing is using this new API yet, but I will use it for some internal logging
APIs in an upcoming diff.  Later on I also plan to use it to implement
per-LogCategory rate limiting, and possibly per-LogHandler rate limiting.

Reviewed By: wez

Differential Revision: D5162805

fbshipit-source-id: 9b81c2f4544006cd392152a768296bce0c5daaa1

Summary:
This simplifies the LogHandler interface to a single generic `handleMessage()`,
and adds a `StandardLogHandler` implementation that defers to separate
`LogFormatter` and `LogWriter` objects.

The `LogFormatter` class is responsible for serializing the `LogMessage` object
into a string, and `LogWriter` is responsible for then doing something with
the serialized string.

This will make it possible in the future to have separate `LogWriter`
implementations that all share the same log formatting code.  For example, this
will allow separate `LogWriter` implementations for performing file I/O
immediately versus performing I/O asynchronously in a separate thread.

Reviewed By: yfeldblum

Differential Revision: D5083103

fbshipit-source-id: e3f5ece25e260c825d49a5eb30e942973d6b68bf

Summary:
This adds new `XLOG()` and `XLOGF()` macros to the logging library.  These are
similar to `FB_LOG()` and `FB_LOGF()`, but do not require a Logger argument.
Instead, the log category is picked automatically based on the current file
name.  The basic algorithm for picking the category name is to replace
directory separators with '.', and to strip off the filename extension.

For instance, all `XLOG()` statements in `src/foo/bar.cpp` will log to the
category `src.foo.bar`.  This also works correctly in header files: `XLOG()`
statements in `src/foo/mylib.h` will log to `src.foo.mylib`

This should generally result in a good log category hierarchy without the user
having to spend additional time picking category names--we simply re-use the
decisions that they already made for their directory layout.

In general I expect the `XLOG()` macros to be convenient enough that users will
use `XLOG()` in almost all cases rather than using `FB_LOG()`.

The log category name used by `XLOG()` statements can be overridden using
`XLOG_SET_CATEGORY()`, but this only works in .cpp files, not in header files.

Reviewed By: wez

Differential Revision: D4920257

fbshipit-source-id: 7ffafd9a4c87e6fb5eb35d86e0eb86ef1ed5be95

Summary:
Added support for thread caching of hazard pointers.
Added thread caching benchmarks.
Removed function call from hazptr_domain constexpr constructor.
Optimizations of memory order and code refactoring.

Reviewed By: davidtgoldblatt

Differential Revision: D5249070

fbshipit-source-id: 487fb23abccde228c3c726de4ac8e9f07bfa9498

Summary: We would always up_ref the ctx before setting it as the initial_ctx.  This causes a leak in 1.1.0 since the initial_ctx isn't set in this version of OpenSSL.  We'll move the up_ref for the initial_ctx into the OpenSSLUtils helper.

Reviewed By: anirudhvr

Differential Revision: D5227823

fbshipit-source-id: b4490b317bd4dc8752a8d7e244fd153100a52aa6

Summary: According to internal linter

Reviewed By: Orvid

Differential Revision: D5051010

fbshipit-source-id: febdeca05ac1cf3ad82617169f90912a445cf173

Summary: Flag caused compilation error on unused param, asserts optimized out in opt-asan/ubsan builds.  Replaced with DCHECK as per meyering's advice.

Reviewed By: meyering

Differential Revision: D5246089

fbshipit-source-id: dc0abda91f900dd98af31410f4667c52404997f5

Summary: `ZSTD_decompress()` doesn't verify the uncompressed length in the frame as I thought, so throw an exception instead of `DCHECK()`.

Reviewed By: meyering

Differential Revision: D5234576

fbshipit-source-id: f72cf085a7267de32ce13553ce7ebbfe3b8a3f05

Summary:
The removed comment cites a gcc compiler bug as a reason to spell
out the return type of the three operator() methods. While that
compiler bug can lead to compiler errors in some situations,
it is not the only reason why we need the return type arrow
notation.

If the compiler tries to instantiate operator() with a set of
parameters that the embedded function does not accept, then the
substitution failure is inside decltype, which is not a compiler
error (SFINAE) but just means that the Partial class doesn't define
operator() for that parameter signature. Without decltype, the body
of operator() is ill-formed and that is a compiler error.

Reviewed By: meyering

Differential Revision: D5236537

fbshipit-source-id: 7daa0fbcf128a2c7c5c36532a5402deba394e1b8

Summary: A common use case for `IPAddress::toFullyQualified()` is to print a `<ip>:<port>` string. It is faster to reserve enough space for both beforehand than create 2 strings.

Reviewed By: yfeldblum

Differential Revision: D5224454

fbshipit-source-id: 4536f89a9d51d39dd9fd970c753ecb8ecced5d22

Summary: During decompression, when the data is truncated, `StreamCodec::doUncompress()` loops forever, since it doesn't check forward progress. `Bzip2Codec` does the same.

Reviewed By: chipturner

Differential Revision: D5233052

fbshipit-source-id: 8797a7f06d9afa494eea292a8a5dc980c7571bd0

Summary:
If the folly::sformat() call fails in an FB_LOGF() statement, make a
best-effort attempt to log the format arguments as well, in addition to the
format string.  For each argument, folly::to<std::string>() is use if it is
supported for this argument.

This will help ensure that the arguments that were being logged aren't lost
even if the format string was incorrect.

Reviewed By: WillerZ

Differential Revision: D5082978

fbshipit-source-id: 0d56030e639cd7e8f2242bb43646ab4248c6a877

Summary:
Update the logging library so that FB_LOG() and FB_LOGF() also accept
streaming style arguments:

  FB_LOG(logger) << "logs can be written like this now: " << 1234;
  FB_LOG(logger, "both styles can be used") << " together";
  FB_LOGF(logger, "test: {}, {}", 1234, 5678) << " and with FB_LOGF() too";

Streaming support does make the upcoming XLOG() macros more complicated to
implement, since the macro has to expand to a single ternary `?` expression.
However I was able to come up with a solution that works and has minimal
overhead.

Reviewed By: wez

Differential Revision: D5082979

fbshipit-source-id: 11734e39c02ad28aceb44bbfbd137d781caa30cf

Summary:
This begins adding a new logging library for efficient, hierarchical logging.

This includes the basic library framework, plus a README file with a brief
overview and the motivation for creating a new logging library.

Reviewed By: wez

Differential Revision: D4911867

fbshipit-source-id: 359623e11feeaa547f3ac2c369bf806ee6996554

Summary: Closes https://github.com/facebook/folly/pull/609

Differential Revision: D5228342

Pulled By: Orvid

fbshipit-source-id: edb81284a0401a4ea267310e1ca01c390bcdfb32

Summary: 2_ms seems too short. I am not sure whether it's feasible to check whether thread is waiting for an address. We could wait for longer to reduce false alarm.

Reviewed By: nbronson

Differential Revision: D5220819

fbshipit-source-id: 42f31206e9cb7f9addaa049d0e7cd995f6735f6c

Summary:
[Folly] Let `SubprocessError` inherit `std::runtime_error`.

As an added bonus, this gives it `std::runtime_error`'s refcounted string for cheap copies.

Reviewed By: ericniebler

Differential Revision: D5216758

fbshipit-source-id: 43298e06f02cfd88abf2d73f7aa16117a6cb052b

Summary: The source and tests for the stats directory was spread across folly/detail and folly/test, move them into folly/stats directly instead.

Reviewed By: yfeldblum

Differential Revision: D5216810

fbshipit-source-id: 00a8bb95a4f7830d0bc46b3f914f256a37833b78

Summary: Use clang-tidy to mechanically add missing `override` and remove redundant `virtual`.

Reviewed By: igorsugak

Differential Revision: D5211868

fbshipit-source-id: 6a85f7c4a543a4c9345ec5b0681a8853707343dc