folly.git
7 years agoExpose folly::symbolizer::dumpStackTrace().
Martin Martin [Fri, 18 Nov 2016 17:59:37 +0000 (09:59 -0800)]
Expose folly::symbolizer::dumpStackTrace().

Summary:
Expose folly::symbolizer::dumpStackTrace() for use with
custom signal handlers.

Reviewed By: luciang

Differential Revision: D4174004

fbshipit-source-id: 510b77edef652f3e9d10f0acfb4998b64a15fad5

7 years agoExplain crash when exception is thrown from Scope Guard
Phil Willoughby [Fri, 18 Nov 2016 16:08:19 +0000 (08:08 -0800)]
Explain crash when exception is thrown from Scope Guard

Summary:
Print a message to `std::cerr` when the current program is about to call `std::terminate` because a `folly::ScopeGuard` callback threw an exception. This goes to `std::terminate` in the (common) cases when the `ScopeGuard` destructor is `noexcept`

This gives the user a small clue as to what just happened, since the default diagnostics for this on some platforms do not help at all.

Reviewed By: nbronson

Differential Revision: D4061096

fbshipit-source-id: c3b534d4a36b095e08e46f375251b6fd416ccd68

7 years agoforce read for doNotOptimizeAway(*ptr_to_small_trivial)
Nathan Bronson [Fri, 18 Nov 2016 16:07:09 +0000 (08:07 -0800)]
force read for doNotOptimizeAway(*ptr_to_small_trivial)

Summary:
doNotOptimizeAway's "X" input operand constraint is interpreted
more loosely by gcc than by clang, resulting in surprising behavior
for doNotOptimizeAway(*ptr) and a difference in behavior between gcc
and clang benchmarks.  clang also is more aggressive about placing an
input operand into a register even when the constraint would allow it to
be in memory, so an "r,m" constraint has a similar problem.  This diff
changes the input constraint so that register-sized values must actually
be copied into a register, which makes the behavior more intuitive and
more consistent across platforms.

Reviewed By: davidtgoldblatt

Differential Revision: D4199767

fbshipit-source-id: aa56a7b11cb3229b95da87295f0dfc38476959d2

7 years agoFix Build: folly with -Wmissing-braces
Yedidya Feldblum [Fri, 18 Nov 2016 07:01:42 +0000 (23:01 -0800)]
Fix Build: folly with -Wmissing-braces

Summary:
[Folly] Fix Build: `folly` with `-Wmissing-braces`.

Construction of `std::array` with list-initialization for the contained C array requires double braces, per the standard.

Compilers accept it with single braces, but will reject it when `-Wmissing-braces` is passed.

Reviewed By: igorsugak, meyering

Differential Revision: D4202629

fbshipit-source-id: e5b87a655e7f25e6cddb931dda28b172c768f227

7 years agofolly/portability: move some macro definitions to `CPortability.h`
Andrew Gallagher [Fri, 18 Nov 2016 05:16:07 +0000 (21:16 -0800)]
folly/portability: move some macro definitions to `CPortability.h`

Summary: These macros apply to both C and C++.

Reviewed By: yfeldblum

Differential Revision: D4199905

fbshipit-source-id: 8dc22959d0a6349c6c76415d1a38eda52cab83a5

7 years agoFaster unhexlify
Tom Jackson [Fri, 18 Nov 2016 02:14:14 +0000 (18:14 -0800)]
Faster unhexlify

Summary:
Using already-available lookup table instead of a bunch of branching, this improves read throughput from ~750MB/s to 2.1GB/s in benchmarks.

```
before:        (time)   (rate)
BM_unhexlify   1.39ns  719.26M
after:
BM_unhexlify 470.59ps    2.13G
```

Reviewed By: philippv

Differential Revision: D4201352

fbshipit-source-id: 8393e066c45c402ccb8b537fdb25417e8e6a9511

7 years agoDon't rely on implicit lookup of member names in dependent base classes
Christopher Dykes [Fri, 18 Nov 2016 00:30:31 +0000 (16:30 -0800)]
Don't rely on implicit lookup of member names in dependent base classes

Summary:
The standard says this shouldn't be done, but GCC, Clang, and MSVC all do, at least, MSVC does if you don't pass the `/permissive-` switch to disable this non-compliant lookup.
This just qualifies it with `this->` which solves the problem.

MSVC plans to eventually (some time after the 2017 release) enable `/permissive-` by default, so let's start compiling with it now.

Reviewed By: yfeldblum

Differential Revision: D4200725

fbshipit-source-id: 8545c6bcda9ba6006733c15f668805cb1260517a

7 years agoInclude atomic in SingletonTestStructs.cpp
Christopher Dykes [Fri, 18 Nov 2016 00:30:12 +0000 (16:30 -0800)]
Include atomic in SingletonTestStructs.cpp

Summary: This was relying on it being transitively included by system headers, which isn't the case under MSVC.

Reviewed By: yfeldblum

Differential Revision: D4199882

fbshipit-source-id: c25dc1e86687e19851923a1a660a91bcf63f0e6d

7 years agoMark a couple more local constexpr values as static
Christopher Dykes [Fri, 18 Nov 2016 00:29:09 +0000 (16:29 -0800)]
Mark a couple more local constexpr values as static

Summary: Because MSVC doesn't let you access it without capturing it from within a lambda otherwise.

Reviewed By: yfeldblum

Differential Revision: D4199867

fbshipit-source-id: 7b829b2cc373edbebe3498901fa323c89a211818

7 years agoUpdate AtomicHashMap.md
Srinivas \"Teja\" Rao [Thu, 17 Nov 2016 22:55:03 +0000 (14:55 -0800)]
Update AtomicHashMap.md

Summary:
it seems second is the boolean that we need to compare.
Closes https://github.com/facebook/folly/pull/518

Differential Revision: D4197096

Pulled By: yfeldblum

fbshipit-source-id: 71987dafac0ddbdb7817b55732619edbe9aed81b

7 years agoReplace deprecated LZ4 functions
Nick Terrell [Thu, 17 Nov 2016 21:18:48 +0000 (13:18 -0800)]
Replace deprecated LZ4 functions

Summary:
Replace deprecated functions with their functionally equivalent counterparts.

See https://github.com/lz4/lz4/blob/dev/lib/lz4.c#L1405 and https://github.com/lz4/lz4/blob/dev/lib/lz4hc.c#L634

Reviewed By: Cyan4973

Differential Revision: D4194834

fbshipit-source-id: aa4f934c46fe764fcec8ea29221e3882da2b5cdf

7 years agoVirtualEventBase
Andrii Grynenko [Thu, 17 Nov 2016 20:44:52 +0000 (12:44 -0800)]
VirtualEventBase

Summary:
This implements a view onto an EventBase, which can keep track of all tasks created through it and join them on destruction. Multiple VirtualEventBases can be backed by the same EventBase. This can be useful to have the same IO thread/thread-pool be shared between multiple libraries, allowing them to each have it's own VirtualEventBase and control its lifetime.

Since VirtualEventBase also supports LoopKeepAlive and onDestruction functionality, it can be easily integrated with FiberManagerMap.

Reviewed By: yfeldblum

Differential Revision: D4158719

fbshipit-source-id: 7df964f36e5276e2b5665fd8394ea2e187aa298c

7 years agofolly: ExceptionWrapper: remove <iostream> from header
Lucian Grijincu [Thu, 17 Nov 2016 19:06:36 +0000 (11:06 -0800)]
folly: ExceptionWrapper: remove <iostream> from header

Summary: #accept2ship

Reviewed By: yfeldblum

Differential Revision: D4192095

fbshipit-source-id: eb0cad875bcc24d1c87a99890c51aea31f7024c9

7 years agoAdd folly::getCurrentThreadID()
Christopher Dykes [Thu, 17 Nov 2016 03:20:16 +0000 (19:20 -0800)]
Add folly::getCurrentThreadID()

Summary: And also use it in a couple of tests, so that they can be compiled on Windows, where `pthread_t` is a struct rather than a pointer or integer.

Reviewed By: yfeldblum

Differential Revision: D4191560

fbshipit-source-id: 5bcf0a2952109b2a9bc5220c4640d42e2cdf8977

7 years agoRemove unneeded iostream include
Delyan Kratunov [Thu, 17 Nov 2016 00:48:47 +0000 (16:48 -0800)]
Remove unneeded iostream include

Summary:
Including `iostream` in a header causes anything that uses this header to gain
an extra static initializer for an `std::ios_base::Init` object.
If there are N translation units in a library that use the header, that's N
more static initializers that need to run.

Reviewed By: yfeldblum

Differential Revision: D4191798

fbshipit-source-id: 65b47401b9177177999cff65a47ac8dffa0e0546

7 years agoAdjust the Function::NonCopyableLambda test to use a large struct rather than a C...
Christopher Dykes [Thu, 17 Nov 2016 00:41:51 +0000 (16:41 -0800)]
Adjust the Function::NonCopyableLambda test to use a large struct rather than a C Array

Summary:
Because MSVC doesn't like it when you attempt to explicitly capture a c-style array in a lambda capture list.
See: https://developercommunity.visualstudio.com/content/problem/2444/cannot-explicitly-capture-c-style-array-in-lambda.html

Reviewed By: yfeldblum

Differential Revision: D4191400

fbshipit-source-id: 305f8086c29f079ccf2c322f20da6393235bc76d

7 years agoBroken link to gtest 1.7.0 zip replaced with correct one
Geert Van Laethem [Wed, 16 Nov 2016 22:50:58 +0000 (14:50 -0800)]
Broken link to gtest 1.7.0 zip replaced with correct one

Summary: Closes https://github.com/facebook/folly/pull/517

Reviewed By: igorsugak

Differential Revision: D4191764

Pulled By: Orvid

fbshipit-source-id: 5ab921bd0df0b9dbaca46cdf9ecdeea9ecda6f5a

7 years agoDisable the CArray RangeFunc tests under MSVC
Christopher Dykes [Wed, 16 Nov 2016 22:50:29 +0000 (14:50 -0800)]
Disable the CArray RangeFunc tests under MSVC

Summary: Because it won't compile due to an MSVC bug. See the link in the comment for more info.

Reviewed By: yfeldblum

Differential Revision: D4191233

fbshipit-source-id: 523deebe12379ff56934a4259b601d4ab677b9b6

7 years agofolly: fbvector: ubsan: avoid memset(nullptr, 0, 0)
Nathan Bronson [Wed, 16 Nov 2016 22:01:49 +0000 (14:01 -0800)]
folly: fbvector: ubsan: avoid memset(nullptr, 0, 0)

Summary:
Constructing an empty FBVector by length results in a call to
memset with a null destination, which is undefined behavior.  This diff
fixes it.

Reviewed By: luciang, meyering, Gownta

Differential Revision: D4191612

fbshipit-source-id: 3dcc091396fc33ac2230bd2d90906325131b0a3b

7 years agoMark a few constexpr locals in ThreadLocalTest as static
Christopher Dykes [Wed, 16 Nov 2016 21:20:32 +0000 (13:20 -0800)]
Mark a few constexpr locals in ThreadLocalTest as static

Summary:
Because, as discussed in past diffs, MSVC currently requires them to be static in order to resolve them from within the scope of a lambda when they aren't captured.
The only one that actually needs to be marked as `static` is `numReps`, but I like consistency, so I marked them all as static.

Reviewed By: yfeldblum

Differential Revision: D4191610

fbshipit-source-id: 7246170f30551d2079b4373eefbf3d02aab6dfa9

7 years agoHint that the port is numeric when binding an AsyncServerSocket
Christopher Dykes [Wed, 16 Nov 2016 21:19:59 +0000 (13:19 -0800)]
Hint that the port is numeric when binding an AsyncServerSocket

Summary: Because it had better be numeric or something is very wrong.

Reviewed By: yfeldblum

Differential Revision: D4190568

fbshipit-source-id: 52a5ebfcedb545201191fc48f66a4419eee8f748

7 years agofolly: MemoryMapping: less verbose warning on TRY_LOCK+ENOMEM failures
Lucian Grijincu [Wed, 16 Nov 2016 04:24:38 +0000 (20:24 -0800)]
folly: MemoryMapping: less verbose warning on TRY_LOCK+ENOMEM failures

Reviewed By: philippv

Differential Revision: D4184097

fbshipit-source-id: 2f1f82b845528fea4667af612c92c348f8d9ec25

7 years agoQualify a field name in NotificationQueueTest
Christopher Dykes [Wed, 16 Nov 2016 02:46:23 +0000 (18:46 -0800)]
Qualify a field name in NotificationQueueTest

Summary: MSVC currently has a bug that causes it to incorrectly resolve it to `std::queue`: https://developercommunity.visualstudio.com/content/problem/2342/c-failure-to-resolve-instance-member-in-lambda-con.html

Reviewed By: lbrandy

Differential Revision: D4187256

fbshipit-source-id: 207eafa339f7255ae881d8ab80f4847222c3b8c9

7 years agoDon't assume standard library types are relocatable
Christopher Dykes [Wed, 16 Nov 2016 01:30:09 +0000 (17:30 -0800)]
Don't assume standard library types are relocatable

Summary: Although currently the only one of these that isn't relocatable is `std::string`, making assumptions about the standard library is a terrible idea, so disable all of them.

Reviewed By: yfeldblum

Differential Revision: D4186024

fbshipit-source-id: 4d2a54d232df0cb05959a8d9cb1a2d2cf34d1302

7 years agoThread-safe version of loopKeepAlive()
Andrii Grynenko [Tue, 15 Nov 2016 22:52:48 +0000 (14:52 -0800)]
Thread-safe version of loopKeepAlive()

Reviewed By: yfeldblum

Differential Revision: D4152380

fbshipit-source-id: 8b3c6dc4b14b9138bb5012e05f50496e51c0fa4b

7 years agoMark FingerprintPolynomial.size() constexpr
Christopher Dykes [Tue, 15 Nov 2016 19:38:18 +0000 (11:38 -0800)]
Mark FingerprintPolynomial.size() constexpr

Summary: Because it is, and so that MSVC doesn't complain about things in the fingerprint table generator.

Reviewed By: yfeldblum

Differential Revision: D4178562

fbshipit-source-id: 8816f413bc100b42dd72f0e991aec8559b328e9b

7 years agofolly: MemoryMapping: isolate _MSC_VER-related changes
Lucian Grijincu [Tue, 15 Nov 2016 04:39:13 +0000 (20:39 -0800)]
folly: MemoryMapping: isolate _MSC_VER-related changes

Summary: Cleaner to isolate the #ifdefs to a constant than different code based on platform.

Reviewed By: yfeldblum, Orvid

Differential Revision: D4179949

fbshipit-source-id: 5b9975a78cf234f29a379c1dbe489373b7f1c253

7 years agoMove runAfterDelay/tryRunAfterDelay into TimeoutManager
Andrii Grynenko [Tue, 15 Nov 2016 02:13:02 +0000 (18:13 -0800)]
Move runAfterDelay/tryRunAfterDelay into TimeoutManager

Summary: This is useful to add more TimeoutManager implementations.

Reviewed By: yfeldblum

Differential Revision: D4178511

fbshipit-source-id: 9167cfe40e7487da1c35c537f9db8a2bdf085626

7 years agoMake Observer stress test pass a stress run
Andrii Grynenko [Tue, 15 Nov 2016 02:07:46 +0000 (18:07 -0800)]
Make Observer stress test pass a stress run

Reviewed By: yfeldblum

Differential Revision: D4179908

fbshipit-source-id: 493b302cd7e9f7ebc851599b2ca5988af817387d

7 years agocorrect the description about the thread to run
yizhiren [Tue, 15 Nov 2016 00:19:37 +0000 (16:19 -0800)]
correct the description about the thread to run

Summary:
exchange A and B in this sentence :
"Or, maybe x will execute in Thread B, but y and/or z will execute in Thread A".

I think x in B and y,z in A is impossible. but x in A and y,z in B is possible, in that example.
Closes https://github.com/facebook/folly/pull/514

Reviewed By: yfeldblum

Differential Revision: D4176766

Pulled By: Orvid

fbshipit-source-id: cf3a47a30d7e43e5291d6b6401198025beba33f8

7 years agoRevert D4164236: [EventBase] Move runAfterDelay/tryRunAfterDelay into TimeoutManager v2016.11.14.00
Alejandro Peláez [Mon, 14 Nov 2016 00:19:32 +0000 (16:19 -0800)]
Revert D4164236: [EventBase] Move runAfterDelay/tryRunAfterDelay into TimeoutManager

Summary: This reverts commit 5f2057f6ebfbdc971bc0b5594e3bc4b5a337aaef

Differential Revision: D4164236

fbshipit-source-id: a397e6ba8c9d7a691cc6381b8f60bfcea3eb91b0

7 years agoAdd ecdsa smart ptr
Subodh Iyengar [Sun, 13 Nov 2016 01:13:44 +0000 (17:13 -0800)]
Add ecdsa smart ptr

Summary: Add ecdsa smart ptr.

Reviewed By: knekritz

Differential Revision: D4173310

fbshipit-source-id: abd5f799b3414b70128635c4390af6f2de329e11

7 years agoMove runAfterDelay/tryRunAfterDelay into TimeoutManager
Andrii Grynenko [Sat, 12 Nov 2016 03:49:34 +0000 (19:49 -0800)]
Move runAfterDelay/tryRunAfterDelay into TimeoutManager

Summary: This is useful to add more TimeoutManager implementations.

Reviewed By: yfeldblum

Differential Revision: D4164236

fbshipit-source-id: 5f2057f6ebfbdc971bc0b5594e3bc4b5a337aaef

7 years agotry to improve folly/io/async/test:async_test - AsyncSocketTest.Write
Marcin Pawlowski [Sat, 12 Nov 2016 02:37:38 +0000 (18:37 -0800)]
try to improve folly/io/async/test:async_test - AsyncSocketTest.Write

Summary:
the test fails a lot as seen on the task.
To kill the connection in said test, we depeneded
on destructor, however this uses a delayed destruction
patter so I guess we did not destroy on time and
we were failing the tests.
I close socket explicitly. No idea if this misses
the point of whole test.

Reviewed By: yfeldblum

Differential Revision: D4171111

fbshipit-source-id: 01b8c97a04a2bcdfaeeea522ca8bcae343d6bb5d

7 years agoRevert DLOG_INFO warnings about RequestContext being overriden
Mirek Klimos [Sat, 12 Nov 2016 02:17:17 +0000 (18:17 -0800)]
Revert DLOG_INFO warnings about RequestContext being overriden

Summary: This is causing a log spew on some services, let's remove it until we figure out where the problem is and fix

Differential Revision: D4171379

fbshipit-source-id: 2afdab6917a2498c376559f9171f6e93ab9c8ad7

7 years agoMake Observer.Stress test not fail under load
Andrii Grynenko [Sat, 12 Nov 2016 02:12:11 +0000 (18:12 -0800)]
Make Observer.Stress test not fail under load

Reviewed By: marcinpe

Differential Revision: D4171655

fbshipit-source-id: e6ef33c50d5d91e05250520171c959a893470460

7 years agoWhen you read from a default-constructed MPMCQueue, assert instead of SIGFPE.
Martin Martin [Sat, 12 Nov 2016 00:08:46 +0000 (16:08 -0800)]
When you read from a default-constructed MPMCQueue, assert instead of SIGFPE.

Summary:
I accidentally forgot to specify the capacity for my
MPMCQueue.  When I then did a blockingRead(), I got a SIGFPE.  Thanks
to a custom signal handler that doesn't print stack traces, and a few
more comedy of errors, I lost a day to this.  With this patch, I would
have gotten an assertion failure instead.

Reviewed By: yfeldblum

Differential Revision: D4169033

fbshipit-source-id: fab97ea0d5afc3c06885758b31a5e8c91ae75a45

7 years agoAdd warnings when overriding RequestContext
Mirek Klimos [Fri, 11 Nov 2016 19:15:55 +0000 (11:15 -0800)]
Add warnings when overriding RequestContext

Summary: We currently store the previous value on creation of RequestContextScopeGuard and reset it to the previous one when it goes out of scope - this means that if it's not unset properly somewhere, we'll prolong its life by saving it and restoring later. We need to do this because of some edge cases but a RequestContext should generally never be overridden by a different RequestContext - let's add a warning if this happens

Reviewed By: djwatson, palmtenor

Differential Revision: D3971904

fbshipit-source-id: e72b5f01102d18977d8aa5f8dca5a7802545098a

7 years agofolly/test/SingletonTestStructs: move inlined function to impl file
Andrew Gallagher [Fri, 11 Nov 2016 06:44:43 +0000 (22:44 -0800)]
folly/test/SingletonTestStructs: move inlined function to impl file

Reviewed By: yfeldblum

Differential Revision: D4162763

fbshipit-source-id: 25c9bfae2e7679f4b98a3d64ee9f2319eb6e7f4a

7 years agoLet folly::range(const std::array<...>&) be constexpr
Yedidya Feldblum [Fri, 11 Nov 2016 06:02:17 +0000 (22:02 -0800)]
Let folly::range(const std::array<...>&) be constexpr

Summary:
[Folly] Let `folly::range(const std::array<...>&)` be `constexpr`.

This may be useful for future work.

Reviewed By: simpkins

Differential Revision: D4165608

fbshipit-source-id: 8f1064c3cdc07301e198acd596ff26724d603cdd

7 years agofolly/Foreach.h: fix FOR_EACH_R and FOR_EACH_KV to avoid shadowing warnings
Jim Meyering [Fri, 11 Nov 2016 00:50:45 +0000 (16:50 -0800)]
folly/Foreach.h: fix FOR_EACH_R and FOR_EACH_KV to avoid shadowing warnings

Summary:
Fix these macros not to self-shadow for nested uses.
How? Make each local variable name include __LINE__.
This works in most cases, but not all: if you put the
nested uses all on one line, you'll still get shadowing warnings.

Reviewed By: yfeldblum

Differential Revision: D4161553

fbshipit-source-id: 9e11ae6778272c733a4c820cfb30c5db59998a5b

7 years agoAlways use inline-asm version on GCC/Clang
Christopher Dykes [Thu, 10 Nov 2016 20:46:19 +0000 (12:46 -0800)]
Always use inline-asm version on GCC/Clang

Summary:
Because the intrinsic version requires explicitly adding `__target__` attributes, which results in things not being inlined. Although the code generated with the `__target__` attribute is strictly better, ensuring it's applied on all the relevant functions is error-prone, so just use the inline assembly version for GCC/Clang so that it can be inlined elsewhere. MSVC will inline the intrinsic version without any issue.
This also marks the functions as `ALWAYS_INLINE`, as the diff that is getting reverted made that change as well.

Reviewed By: yfeldblum, philippv, ot

Differential Revision: D3963935

fbshipit-source-id: 47175d64e7be351eb455a4d053b91ce9392bf152

7 years agoAdd include files necessary to use symbolizer
Tudor Bosman [Thu, 10 Nov 2016 20:40:38 +0000 (12:40 -0800)]
Add include files necessary to use symbolizer

Summary:
Some include files are not installed, causing including symbolizer/Symbolizer.h to fail.
Closes https://github.com/facebook/folly/pull/512

Reviewed By: philippv

Differential Revision: D4161113

Pulled By: luciang

fbshipit-source-id: 5a5f8577f8d84ba33f89073781d374b99ba37df6

7 years agoDon't try to call _free_osfhnd when not compiling agains the static CRT
Christopher Dykes [Thu, 10 Nov 2016 20:32:51 +0000 (12:32 -0800)]
Don't try to call _free_osfhnd when not compiling agains the static CRT

Summary:
Because, unfortunately, it isn't exported from the dynamic CRT dlls :(
There's not really a nice way to handle this when using the dynamic CRT without doing very fragile hocus-pocus that relies on the exact layout and implementation details of the file descriptor table in the CRT, so the best we can really do is close the socket and just let the file descriptor itself leak.

Reviewed By: yfeldblum

Differential Revision: D4156558

fbshipit-source-id: 32cb4bf357f6746cf6597b66649ff9f018fb1bed

7 years agoUse std::cerr for folly/Singleton.cpp error printing
Phil Willoughby [Thu, 10 Nov 2016 12:09:21 +0000 (04:09 -0800)]
Use std::cerr for folly/Singleton.cpp error printing

Summary:
Create a local std::ios_base::Init instance to ensure that it is valid, then
use std::cerr.

Reviewed By: yfeldblum, nbronson

Differential Revision: D4139681

fbshipit-source-id: 03377dd417710e320a6b53298d507fd0168592f6

7 years agoFix wrong use of upgrade lock
Andrii Grynenko [Thu, 10 Nov 2016 04:19:57 +0000 (20:19 -0800)]
Fix wrong use of upgrade lock

Reviewed By: yfeldblum, nbronson

Differential Revision: D4149681

fbshipit-source-id: 37bd1b0b7d1ad6e6fa813228307abebfe772012f

7 years agoDon't try to install the failure signal handler in folly::init on Windows
Christopher Dykes [Wed, 9 Nov 2016 23:13:35 +0000 (15:13 -0800)]
Don't try to install the failure signal handler in folly::init on Windows

Summary: Because GTest doesn't support it on platforms that don't implement `sigaction`.

Reviewed By: yfeldblum

Differential Revision: D4155882

fbshipit-source-id: fe8df316242b293afaf94c2a5d0f8a5e60b51285

7 years agoDisable a couple of warnings generated by the GTest and GMock headers
Christopher Dykes [Wed, 9 Nov 2016 23:13:25 +0000 (15:13 -0800)]
Disable a couple of warnings generated by the GTest and GMock headers

Summary: As the comment says, this disables a couple of warnings related to exporting classes derived from stdlib classes that aren't explicitly exported.

Reviewed By: yfeldblum

Differential Revision: D4155978

fbshipit-source-id: 6e732b53c280a3bc3a39b2128cd577ba7cb70408

7 years agoDon't generate empty statements in DeterministicScheduleTest
Christopher Dykes [Wed, 9 Nov 2016 22:35:17 +0000 (14:35 -0800)]
Don't generate empty statements in DeterministicScheduleTest

Summary: Because MSVC generates warnings if you have a control-flow statement followed by a single trailing semicolon.

Reviewed By: yfeldblum

Differential Revision: D4155588

fbshipit-source-id: dfd8774fc6a62675764f7e069c26b799b622a9af

7 years agoRequest lower privilege crypt context when generating random numbers
Christopher Dykes [Wed, 9 Nov 2016 22:31:44 +0000 (14:31 -0800)]
Request lower privilege crypt context when generating random numbers

Summary: Without this flag, we were unintentionally requesting a context with access to private keys.

Reviewed By: yfeldblum

Differential Revision: D4155525

fbshipit-source-id: f33d18451b06e32b81d03bf6180efc1671cfa204

7 years agoUse ASSERT_EQ rather than CHECK_EQ in the socket tests
Christopher Dykes [Wed, 9 Nov 2016 19:20:55 +0000 (11:20 -0800)]
Use ASSERT_EQ rather than CHECK_EQ in the socket tests

Summary: Because `CHECK_EQ` is a fatal error, and makes it impossible to keep running the other tests if it fails, even if it was just a timeout.

Reviewed By: yfeldblum

Differential Revision: D4151084

fbshipit-source-id: db88dc2ef875e5d1fe31792f26a3dc2d223ded5e

7 years agoSplit the benchmarks out of UriTest
Christopher Dykes [Wed, 9 Nov 2016 19:07:51 +0000 (11:07 -0800)]
Split the benchmarks out of UriTest

Summary: So that they can be run independently of the tests.

Reviewed By: yfeldblum

Differential Revision: D4150331

fbshipit-source-id: 3ab39024795add84885436b757f2505527297a3d

7 years agoReturn the correct resolution for clock_getres
Christopher Dykes [Wed, 9 Nov 2016 19:07:23 +0000 (11:07 -0800)]
Return the correct resolution for clock_getres

Summary: When the implementation of `clock_gettime` for `REALTIME` and `MONOTONIC` was switched to use `std::chrono`, this didn't get switched over to the resolution of the new implementation.

Reviewed By: yfeldblum

Differential Revision: D4150297

fbshipit-source-id: a4578af85cae538a3ebb6da7e759058927c31d3e

7 years agoExpose isFamilyInet() to check if SocketAddress is a valid ipv4 or ipv6
Ali Zaveri [Wed, 9 Nov 2016 08:29:32 +0000 (00:29 -0800)]
Expose isFamilyInet() to check if SocketAddress is a valid ipv4 or ipv6

Summary: Expose isFamilyInet() in SocketAddress. This cleans up code and provides a way to validate if SocketAddress is a ipv4 or ipv6

Differential Revision: D4150650

fbshipit-source-id: dc8883e904b6577fc27bdd54064fcc50de93b0d2

7 years agoMake NeedleFinderTest::Needles256 actually do things
Christopher Dykes [Tue, 8 Nov 2016 18:40:14 +0000 (10:40 -0800)]
Make NeedleFinderTest::Needles256 actually do things

Summary:
The min value of a `char` is `-128`, but we're initializing our `size_t` i's with it, so, as MSVC quite rightly warns (as C6294) the condition is never true and the loop never loops.
Fix it by signing our i's.

Reviewed By: yfeldblum

Differential Revision: D4145419

fbshipit-source-id: 87591d0f84c54472b1b5847e025d091de3f98bca

7 years agoadd tryReadUntil and make fixes along the way
James Sedgwick [Tue, 8 Nov 2016 07:13:30 +0000 (23:13 -0800)]
add tryReadUntil and make fixes along the way

Summary:
this diff adds tryReadUntil, which is a mirror of tryWriteUntil in both function and implementation.
Two bugs were exposed in the process of implementing and testing tryWriteUntil; they are fixed as well and are as follows:
  1. tryObtainPromisedPopTicket didn't assign to the passed ticket return reference in the failure case
  2. TurnSequencer::tryWaitForTurn() didn't distinguish between past turns and timeouts in the failure case; they need to be
     differentiated because SingleElementQueue::tryWaitFor{De/En}queue() should only fail in the timeout case, not if the turn has passed.

The two added unit tests are admittedly clumsy, but making the obvious simplifications to them keeps them from triggering the premature timeout race caused by bug 2 above, so I kept them as is.

Reviewed By: magedm

Differential Revision: D4050515

fbshipit-source-id: b0a3dd894d502c44be62d362ea347a1837df4c2f

7 years agoMark implicitly inlined static locals with default visibility
Andrew Gallagher [Tue, 8 Nov 2016 00:10:41 +0000 (16:10 -0800)]
Mark implicitly inlined static locals with default visibility

Summary:
These functions contain static locals which need to be exported to
the global dynamic symbol table to get deduplicated at runtime.  Since
using `-fvisibility-inlines-hidden` demotes the visibility of these
static locals, explicitly tag them.

Reviewed By: yfeldblum

Differential Revision: D4139154

fbshipit-source-id: 08b81721c6c945ac5556472481ae771d0eeb5f88

7 years agoRemove noexcept from AsyncSocket::prepareReadBuffer
Stepan Palamarchuk [Mon, 7 Nov 2016 20:37:52 +0000 (12:37 -0800)]
Remove noexcept from AsyncSocket::prepareReadBuffer

Summary:
`prepareReadBuffer` is calling a callback function that is not a noexcept
also `prepareReadBuffer` is being called only from one place, that wraps this function call with try/catch.

Reviewed By: jmswen

Differential Revision: D4138424

fbshipit-source-id: a05bdd6f741761567a7d0291017c285b0bf15522

7 years agoAnnotate some static locals as "library-local" v2016.11.07.00
Andrew Gallagher [Mon, 7 Nov 2016 04:06:46 +0000 (20:06 -0800)]
Annotate some static locals as "library-local"

Summary:
This tags certain static locals, wich may get duplicated via dynamic
linking and inlining, as safe to have multiple copies at runtime,
making this subtle behavior explicit.

Reviewed By: igorsugak

Differential Revision: D4082929

fbshipit-source-id: ba4e23cb121e957a43f5659a1ca3432cb89cb8f3

7 years agoLet folly::range be constexpr
Yedidya Feldblum [Sun, 6 Nov 2016 21:09:44 +0000 (13:09 -0800)]
Let folly::range be constexpr

Summary:
[Folly] Let `folly::range` be `constexpr`.

It will be useful to have this helper function be `constexpr` to support further compile-time code.

Reviewed By: juchem

Differential Revision: D4135943

fbshipit-source-id: a62be93f9e492ca302130d1a3b190ed4b69f7a80

7 years agooperator== for ReadMostlySharedPtr to nullptr_t
Christopher Small [Sat, 5 Nov 2016 19:36:36 +0000 (12:36 -0700)]
operator== for ReadMostlySharedPtr to nullptr_t

Summary: add operator== against nullptr_t for ReadMostlySharedPtr

Reviewed By: andriigrynenko, nbronson, yfeldblum

Differential Revision: D4128722

fbshipit-source-id: 40d08bb85bfe2f96ad4ed4995f5829e0616d0661

7 years agoAdd StringKeyedMap/Set.h to OSS build
Zach Amsden [Fri, 4 Nov 2016 19:59:59 +0000 (12:59 -0700)]
Add StringKeyedMap/Set.h to OSS build

Summary: Eden uses this and won't build with OSS folly without it.

Reviewed By: Orvid

Differential Revision: D4132103

fbshipit-source-id: 68d6c7dc642a7401eca60b0713db9f90cc11ced1

7 years agoTrying to improve the efficiency of RequestContext callbacks
Mirek Klimos [Fri, 4 Nov 2016 18:10:46 +0000 (11:10 -0700)]
Trying to improve the efficiency of RequestContext callbacks

Summary: Changed RWSpinLock to a SharedMutex in the RequestContext to see if this would lower the overhead of the onSet/onUnset callbacks.

Reviewed By: djwatson

Differential Revision: D3780396

fbshipit-source-id: 3d91f8194357533072fe053d23e0454a36faecff

7 years agoAdd splitAtMost to IOBufQueue.
Kyle Nekritz [Thu, 3 Nov 2016 22:53:52 +0000 (15:53 -0700)]
Add splitAtMost to IOBufQueue.

Summary: This allows getting up to n bytes from the queue without first having to check the length.

Reviewed By: yfeldblum

Differential Revision: D4083484

fbshipit-source-id: 2a468992c97f036c22f1a0d9f830e6d5286a4bc2

7 years agoAllow folly::io::Cursor to move backwards.
Kyle Nekritz [Thu, 3 Nov 2016 22:53:50 +0000 (15:53 -0700)]
Allow folly::io::Cursor to move backwards.

Summary: This is helpful for parsing data IOBufs in reverse.

Reviewed By: siyengar

Differential Revision: D4082810

fbshipit-source-id: 6a201d25e7d22befe28f92e4a1d7aa41ad7d6817

7 years agoquiet -Wmaybe-uninitialized
Louis Brandy [Thu, 3 Nov 2016 17:35:18 +0000 (10:35 -0700)]
quiet -Wmaybe-uninitialized

Summary: gcc-4.9 doesn't appreciate this clever loop and worries about `prev` being uninitialized on first use. Assuming the rest of the program behaves properly and preserves the invariant this code is expecting, this isn't bugged. But let's initialize it anyway.

Reviewed By: meyering

Differential Revision: D4120746

fbshipit-source-id: 1a7c323a4b6296a81768a62dd7d72337c5c35bf0

7 years agofolly/test/RangeFindBenchmark.cpp: avoid shadowing warning: rename inner "n": s/...
Jim Meyering [Thu, 3 Nov 2016 02:02:20 +0000 (19:02 -0700)]
folly/test/RangeFindBenchmark.cpp: avoid shadowing warning: rename inner "n": s/\<n\>/c/

Reviewed By: yfeldblum

Differential Revision: D4120553

fbshipit-source-id: 87c202232bcfabeff7eb62a8133b9b5577b09cbc

7 years agofolly/test/ThreadLocalBenchmark.cpp: avoid shadowing warning
Jim Meyering [Thu, 3 Nov 2016 02:02:00 +0000 (19:02 -0700)]
folly/test/ThreadLocalBenchmark.cpp: avoid shadowing warning

Summary:
There was a doubly-nested loop, both with index "i".
Change the inner for-loop index from i to j.

Reviewed By: yfeldblum

Differential Revision: D4120589

fbshipit-source-id: c786d8f7bb1ee05fc990db898b15836032e93940

7 years agofolly/test/ExceptionWrapperBenchmark.cpp: avoid shadowing warnings
Jim Meyering [Thu, 3 Nov 2016 02:00:24 +0000 (19:00 -0700)]
folly/test/ExceptionWrapperBenchmark.cpp: avoid shadowing warnings

Reviewed By: jsedgwick

Differential Revision: D4120563

fbshipit-source-id: c97645a068662e41d5038bcd1e9805bdf6f0490c

7 years agoAllow getting the sequence number of a token via a read only getter function
Ameya Limaye [Thu, 3 Nov 2016 01:01:53 +0000 (18:01 -0700)]
Allow getting the sequence number of a token via a read only getter function

Summary: Allow getting the sequence number of a token via a read only getter function

Reviewed By: yfeldblum

Differential Revision: D4120900

fbshipit-source-id: b664dda29d13a1755e041ab9c33c0d7324690ba7

7 years agoUse loopKeepAlive() mechanism in FiberManager
Andrii Grynenko [Thu, 3 Nov 2016 00:39:29 +0000 (17:39 -0700)]
Use loopKeepAlive() mechanism in FiberManager

Reviewed By: yfeldblum

Differential Revision: D4086486

fbshipit-source-id: bd0cca8dd2c9b74d5c30e4cd095191c1d1ecab79

7 years agoCompile RangeSse42.cpp with -msse4.2
Tudor Bosman [Wed, 2 Nov 2016 22:00:45 +0000 (15:00 -0700)]
Compile RangeSse42.cpp with -msse4.2

Summary:
Solves #509 by creating a libtool convenience library that's compiled with -msse4.2.
Closes https://github.com/facebook/folly/pull/510

Differential Revision: D4120400

Pulled By: yfeldblum

fbshipit-source-id: bf5d937dc37ef8707e755c89f1883d65b4d21b0f

7 years agofix race between StaticMetaBase::destroy() and StaticMetaBase::onThreadExit()
Philip Pronin [Wed, 2 Nov 2016 04:24:55 +0000 (21:24 -0700)]
fix race between StaticMetaBase::destroy() and StaticMetaBase::onThreadExit()

Summary:
We would like to guarantee that after `folly::ThreadLocal<>` dtor
returns no per-thread instances are longer alive.  Currently this is not a case:

* T1 is excuting `StaticMetaBase::onThreadExit()`, it acquired all per-thread
  instances and erased them from meta under `accessAllThreadsLock_`.
* T2 destroys `folly::ThreadLocal<>`, it executes `StaticMetaBase::destroy()`,
  collects all per-thread instances (thus missing the ones being destroyed by
  T1), destroys them and returns.
* T1 executes dtor of per-thread instances, after parent `folly::ThreadLocal<>`
  dtor is finished.

Reviewed By: ot

Differential Revision: D4109820

fbshipit-source-id: d547b8cc77c9871126538c38644c2e98ddccf220

7 years agoMake sequence number in AtomicBatchDispatcher<....>::Token non-const
Ameya Limaye [Wed, 2 Nov 2016 03:28:09 +0000 (20:28 -0700)]
Make sequence number in AtomicBatchDispatcher<....>::Token non-const

Summary:
Make sequence number in AtomicBatchDispatcher<....>::Token non-const:
- The field is already private and having it as const prevents a
  move assignment operator from getting defined

Reviewed By: yfeldblum

Differential Revision: D4114851

fbshipit-source-id: 9f7a4003daff02c007c01747cca6ae9da17fcb79

7 years agoDon't declare the extern FingerprintTable specializations under MSVC
Christopher Dykes [Tue, 1 Nov 2016 23:50:28 +0000 (16:50 -0700)]
Don't declare the extern FingerprintTable specializations under MSVC

Summary:
MSVC doesn't allow uninitialized non-`extern` `const` declarations, but we can't mark them `extern` because they are part of a template specialization. We also can't initialize them, because they'd conflict with the definition specializations in the generated tables.
As the forward-declarations aren't required for it to work under MSVC, as the specializations are found appropriately, just don't declare the specializations under MSVC.

Reviewed By: yfeldblum

Differential Revision: D4098282

fbshipit-source-id: 84b73d63bbe6cf9c54b7fe7a3bfc2488699df7c2

7 years agoFix CLANG compilation: folly/TokenBucket.h
David Callahan [Tue, 1 Nov 2016 19:56:08 +0000 (12:56 -0700)]
Fix CLANG compilation: folly/TokenBucket.h

Summary:
In the next revision of clang, compilation generates multiple erroes of the form

   folly/TokenBucket.h:127:29: error: exception specification is not available until end of  class definition
         double nowInSeconds = defaultClockNow()) {

which are eliminated by reordering the method declaration for defaultClockNow()

Refer https://llvm.org/bugs/show_bug.cgi?id=30860

Reviewed By: philippu, yfeldblum

Differential Revision: D4107443

fbshipit-source-id: ce64f2ae7983e533c2fcb5cb043dbdd3da5c00f7

7 years agoEliminate more VLAs
Christopher Dykes [Tue, 1 Nov 2016 18:24:18 +0000 (11:24 -0700)]
Eliminate more VLAs

Summary:
MSVC doesn't support them, so use dynamic allocation instead.
This also fixes a place where MSVC wasn't propogating the constexpr-ness of a local into the body of a lambda, resulting in an attempt to create a VLA.

Reviewed By: yfeldblum

Differential Revision: D4098932

fbshipit-source-id: 742b2fcd96f7f6eceb5043159403d557f7fc9673

7 years agouint8_t isn't a valid template argument for uniform_int_distribution
Christopher Dykes [Tue, 1 Nov 2016 18:24:09 +0000 (11:24 -0700)]
uint8_t isn't a valid template argument for uniform_int_distribution

Summary: Clang and GCC allow it, but the spec, and MSVC, don't, so use uint16_t instead.

Reviewed By: yfeldblum

Differential Revision: D4099285

fbshipit-source-id: cdf7ba14b6c00e1bcdfc3650b76e6e4a47a61929

7 years agoSilence a couple of warnings for MSVC in Optional.h
Christopher Dykes [Tue, 1 Nov 2016 18:23:40 +0000 (11:23 -0700)]
Silence a couple of warnings for MSVC in Optional.h

Summary: Silence a couple of warnings that get very spammy for MSVC.

Reviewed By: yfeldblum

Differential Revision: D4098682

fbshipit-source-id: b410b56062cdf82367675ea9c3dd22975e7b91bd

7 years agoFix BoostContextCompatibility for boost >= 1.61
Andrii Grynenko [Tue, 1 Nov 2016 01:16:50 +0000 (18:16 -0700)]
Fix BoostContextCompatibility for boost >= 1.61

Reviewed By: yfeldblum

Differential Revision: D4100221

fbshipit-source-id: 7778b48f1bba9be85f5712f61d6e1731524901c3

7 years agoSimplify the API exposed by BoostContextCompatibility
Andrii Grynenko [Tue, 1 Nov 2016 01:05:31 +0000 (18:05 -0700)]
Simplify the API exposed by BoostContextCompatibility

Summary: Instead of exposing raw jumpContext, it now exposes a higher-level FiberImpl class, which can be extended for newer versions of jump_fcontext.

Reviewed By: yfeldblum

Differential Revision: D4099849

fbshipit-source-id: 28c7ce32284a0109cf040c264d46a31a45867934

7 years agoSimplify activateFiber()/deactivateFiber()
Andrii Grynenko [Mon, 31 Oct 2016 23:08:47 +0000 (16:08 -0700)]
Simplify activateFiber()/deactivateFiber()

Summary: A cleanup in preparation to make folly::fibers work with newer boost.

Reviewed By: jsedgwick

Differential Revision: D4093947

fbshipit-source-id: e9a0aaeb915c40af96282169d4758afe572f2dd8

7 years agoUpdate hazard pointers interface and implementation
Maged Michael [Mon, 31 Oct 2016 21:47:41 +0000 (14:47 -0700)]
Update hazard pointers interface and implementation

Summary:
The main purpose of this diff and this library at this point is to be a public reference for the C++ standard committee and whoever is interested in the proposal to the committee.

This diff aims to be consistent with the latest version of the proposal (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0233r2.pdf).

The current interface proposal focuses on the core components (domain, object base, and raii owner). Once, that part is settled we can add to the interface:
- Thread local options (thread caching of hazard pointers and private storage of retired objects)
- Programmer control of reclamation (when and by which threads)
Also, at this point the implementation does not optimize memory ordering.

I removed hazptr_domain::try_reclaim() from the public interface at this point. The latest update to interface aims to relieve the programmer from the need to take spacial precautions against shutdown fiascos involving reclamation functions of objects stored by the default domain. Please let me know if you have any concerns about this.

Having said that about the current purpose of this library, I really appreciate any comments (in this diff or separately) on the interface in general and any suggestions for the eventual optimized implementation.

Reviewed By: davidtgoldblatt

Differential Revision: D4104381

fbshipit-source-id: df98adf6fd9b7a93406cb8eeca8fe2ad12388139

7 years agofolly: AsyncSignalHandler: update comments to match code
Lucian Grijincu [Mon, 31 Oct 2016 19:56:26 +0000 (12:56 -0700)]
folly: AsyncSignalHandler: update comments to match code

Summary: #accept2ship

Reviewed By: jsedgwick

Differential Revision: D4105140

fbshipit-source-id: eca1f7066e514ec3c5d316f974edc34e03872784

7 years agoAccess the individual words of an IPv6 address correctly in IPAddressTest on Windows
Christopher Dykes [Mon, 31 Oct 2016 19:00:28 +0000 (12:00 -0700)]
Access the individual words of an IPv6 address correctly in IPAddressTest on Windows

Summary: Winsock defines the internals of `in6_addr` in its own way, so we have to account for that.

Reviewed By: jsedgwick

Differential Revision: D4099453

fbshipit-source-id: c0ebb4e2017f61bed7d5d63058161ef3f16f9a65

7 years agoIn the portability implementation of sendmsg return a partial message send if it...
Christopher Dykes [Mon, 31 Oct 2016 18:59:37 +0000 (11:59 -0700)]
In the portability implementation of sendmsg return a partial message send if it would block

Summary: If multiple iovs are passed to sendmsg, sendmsg is supposed to return the number of bytes sent if it would block, but only if the number of bytes sent is greater than 0.

Reviewed By: yfeldblum

Differential Revision: D4099691

fbshipit-source-id: e58fa71604966129b1fbd418c24b1bf012060428

7 years agoImplement AtomicBatchDispatcher in folly::fibers
Ameya Limaye [Mon, 31 Oct 2016 16:26:16 +0000 (09:26 -0700)]
Implement AtomicBatchDispatcher in folly::fibers

Summary:
Implement AtomicBatchDispatcher in folly::fibers
- Details about how to use the added functionality can be found
  in the doc comment for class AtomicBatchDispatcher.

Reviewed By: andriigrynenko

Differential Revision: D4054148

fbshipit-source-id: 090272eeab8c8abb15d5e400e52725853fcfc364

7 years agoUse MSG_MORE instead of 2 setsockopt calls on every AsyncSSLSocket write.
Kyle Nekritz [Mon, 31 Oct 2016 15:09:06 +0000 (08:09 -0700)]
Use MSG_MORE instead of 2 setsockopt calls on every AsyncSSLSocket write.

Summary: Previously we set the cork option on the socket before making multiple writes, and then unset it after, which elip found was hurting perf with 2 extra syscalls. The cork logic was also not the same as the buffer combining logic, which made us often set cork even when only doing one write.

Reviewed By: djwatson

Differential Revision: D4058357

fbshipit-source-id: 1a07447ff5e027751e455a2403e0042bf67cb1c5

7 years agoimprovements to folly::doNotOptimizeAway v2016.10.31.00
Nathan Bronson [Mon, 31 Oct 2016 01:12:26 +0000 (18:12 -0700)]
improvements to folly::doNotOptimizeAway

Summary:
A common failure mode for folly::doNotOptimizeAway is to pass it a
pointer to a complex object that the caller wants to ensure is fully
calculated, which doesn't actually ensure that.  Also, the GCC and
MSVC implementations used inline assembly with an in-out parameter,
which causes a store.  A survey of many usages of doNotOptimizeAway
found only 1 that relies on this make-unpredictable behavior.

This diff makes doNotOptimizeAway(&x) equivalent to doNotOptimizeAway(x)
for GCC and clang, and makes it a read-only sink.  For the rare
case that the benchmark wants to disable subexpression elimination,
constant propagation, or power reduction, there is a new function
makeUnpredictable.  It also merges the clang and GCC implementations,
removing a potential bias in our microbenchmarks.

Reviewed By: davidtgoldblatt

Differential Revision: D4074670

fbshipit-source-id: 43f02e7fe149147bb172babe77787dea06e098fa

7 years agoInclude the Time.h portability header where needed
Christopher Dykes [Sat, 29 Oct 2016 00:20:16 +0000 (17:20 -0700)]
Include the Time.h portability header where needed

Summary: In this case, the SpookyHash tests, which call `clock_gettime`.

Reviewed By: yfeldblum

Differential Revision: D4098997

fbshipit-source-id: e402fd6626e59ccc555513384e93ef0ef891bbf3

7 years agoDon't abort when calling qfind_first_byte_of_sse42 without compiler support for SSE 4.2
Christopher Dykes [Fri, 28 Oct 2016 21:20:34 +0000 (14:20 -0700)]
Don't abort when calling qfind_first_byte_of_sse42 without compiler support for SSE 4.2

Summary:
If Folly was compiled without support for SSE 4.2, but this function ends up getting called, we'd abort due to the check failing.
To solve this, just don't do the check and fall back to the non-sse version.

This is reasonable because it means that the code calling `qfind_first_byte_of_sse42` only needs to care whether there is CPU support for SSE 4.2, not whether we were able to actually implement the search with SSE 4.2 based on compiler support.

Reviewed By: yfeldblum

Differential Revision: D4096645

fbshipit-source-id: d26a791c09f61595e40c84c47a7a2596b9b69c1e

7 years agofolly/experimental/hazptr: mark static local as "library-local"
Andrew Gallagher [Thu, 27 Oct 2016 18:38:46 +0000 (11:38 -0700)]
folly/experimental/hazptr: mark static local as "library-local"

Summary:
When using `-fvisibility-inlines-hidden` or using C/C++ shared libs
transitively from a Python binary, static locals in inline functions
can potentially have more than one definition at runtime.  This diff
explicitly marks this static local as safe for this scenario.

Reviewed By: djwatson

Differential Revision: D4082831

fbshipit-source-id: 75616ac26a39f44691ef59b889ea555e9daa391b

7 years agoAdd default constructor to dynamic, make other constructors simpler and stricter
Giuseppe Ottaviano [Wed, 26 Oct 2016 23:47:00 +0000 (16:47 -0700)]
Add default constructor to dynamic, make other constructors simpler and stricter

Summary:
Now that the initializer list syntax has been removed we can add a default constructor.
Also,
- The `dynamic(T)` constructor was unconstrained, so it would match any type but then fail to compile (as a side effect, `is_convertible<T, dynamic>` would be always true). This also leaked the implementation details of `Array` and `Object`, as they were accepted as arguments. The diff makes the constructor accept only integral and float arguments, and all other types are SFINAEd out.
- `dynamic(Iterator, Iterator)` is made `explicit` to avoid accepting statements like `dynamic d = {"a", "b"};`.
- `object(...)` methods are simplified.

Reviewed By: luciang

Differential Revision: D4065021

fbshipit-source-id: ac289da7bece67c674b7036b7b51d5e016b297e5

7 years agoMove inline functions with static-locals to implementation files
Andrew Gallagher [Wed, 26 Oct 2016 08:24:23 +0000 (01:24 -0700)]
Move inline functions with static-locals to implementation files

Summary:
This refactors inline functions defined in headers which had static locals
so that their definition is moved to the corresponding implementation file.

Reviewed By: yfeldblum

Differential Revision: D4049175

fbshipit-source-id: 56eeb82eb23b04c3b9940d803d05050949aa5ef9

7 years agoSmall rewording in folly Benchmark comment
Matthieu Martin [Tue, 25 Oct 2016 23:50:39 +0000 (16:50 -0700)]
Small rewording in folly Benchmark comment

Summary: A bit more clear

Reviewed By: yfeldblum

Differential Revision: D4075931

fbshipit-source-id: 1398500336d412b5e7663ef1e67b20a329ce2979

7 years agoFix maybe-uninitialized warning in Conv.h
Michael Lee [Tue, 25 Oct 2016 16:45:49 +0000 (09:45 -0700)]
Fix maybe-uninitialized warning in Conv.h

Summary: Some compilers think result may be uninitialized

Reviewed By: francis-ma

Differential Revision: D4074526

fbshipit-source-id: 5df4568cfa5eb8f3b2993a82b831fd5e4028a281

7 years agofix Expected<..., eUnion> move / copy ctor
Philip Pronin [Tue, 25 Oct 2016 06:27:16 +0000 (23:27 -0700)]
fix Expected<..., eUnion> move / copy ctor

Summary:
Consider `Expected<..., eUnion>(Expected&&)` ctor, where
`ExpectedUnion` move ctor would be called first, which is noop leaving
`which_`  uninitialized; then `MoveConstructible` ctor is executed, calling
`assignValue()`, which performs action depending on uninitialized `which_`.

Reviewed By: yfeldblum

Differential Revision: D4073199

fbshipit-source-id: 623660e7047afcebf9f72d83f91f84ff6078090f

7 years agoprevPowTwo / faster bit operations
Ben Maurer [Tue, 25 Oct 2016 03:14:08 +0000 (20:14 -0700)]
prevPowTwo / faster bit operations

Summary:
Add a prevPowTwo method to bits.h and optimize the current code for
GCCs output.

Reviewed By: ot

Differential Revision: D4072341

fbshipit-source-id: 6e949d0bfcf88ff8500022939d08a2b5aa9e00c9

7 years agoNo need for libdwarf, only the header file matters
Tudor Bosman [Tue, 25 Oct 2016 03:04:41 +0000 (20:04 -0700)]
No need for libdwarf, only the header file matters

Summary:
folly/experimental/symbolizer doesn't actually link with libdwarf, it only requires dwarf.h.
Closes https://github.com/facebook/folly/pull/501

Differential Revision: D4068438

Pulled By: djwatson

fbshipit-source-id: 6339eabf8fceac300caafb478c003833768125fd

7 years agoAdd smart pointer types for ec
Subodh Iyengar [Mon, 24 Oct 2016 20:51:50 +0000 (13:51 -0700)]
Add smart pointer types for ec

Summary:
Add smart pointer types for ec groups
and points

Reviewed By: yfeldblum

Differential Revision: D4066848

fbshipit-source-id: f8a5c59cf902584e5a4bea7265834fcc8898677b