Summary:
- Manually swap bucket_ entries in HHWheelTimer to work around a bug in
libc++

Test Plan:
- fbconfig -r folly/io/async && fbmake runtests --extended-tests
- Sync to OS X and build there

Reviewed By: davejwatson@fb.com

Subscribers: trunkagent, doug, seanc, folly-diffs@

FB internal diff: D1766601

Signature: t1:1766601:1420566916:24a2911e997ee04f0934884bdf7b82345cb2b752

Summary:
Fixes T5922800
(I think)

Test Plan:
TBH I don't know why detaching a temporary thread variable might cause the program to abort but this seems to be the most likely cause so I'm making this change and we'll see in a day or two if it fixes it.

The test still passes and doesn't hang

Reviewed By: davejwatson@fb.com

Subscribers: fugalh, exa, folly-diffs@

FB internal diff: D1764374

Tasks: 5922800

Signature: t1:1764374:1420485233:2e4c81776ef6b6bdae18fbf2e99f0deea37b7879

Summary:
* folly/Format-inl.h (BaseFormatter): Move decls of P and END down
past the shadowed decls.  Otherwise, gcc-4.9 would emit this:
./folly/Format-inl.h:209:10: error: declaration of 'p' shadows a previous local [-Werror=shadow-local]
./folly/Format-inl.h:202:8: error: shadowed declaration is here [-Werror=shadow-local]
./folly/Format-inl.h:210:10: error: declaration of 'end' shadows a previous local [-Werror=shadow-local]
./folly/Format-inl.h:203:8: error: shadowed declaration is here [-Werror=shadow-local]

Test Plan:
Run this and note there are fewer errors than before:
fbconfig --platform-all=gcc-4.9-glibc-2.20 tao/server && fbmake dbgo

Reviewed By: andrewjcg@fb.com

Subscribers: trunkagent, net-systems@, folly-diffs@

FB internal diff: D1766673

Tasks: 5941250

Signature: t1:1766673:1420569376:22c5c3a012715e479b27ba6f0bde1a3b079a3be3

Summary: Observer methods, so users of IOThreadPoolExecutor can do stuff when threads are added/removed.  As a use case, previously the thrift server only used the threads already started when it started up, and assumed iothreadpool was never resized.

Test Plan: Added several unittests

Reviewed By: jsedgwick@fb.com

Subscribers: trunkagent, doug, fugalh, alandau, bmatheny, mshneer, folly-diffs@

FB internal diff: D1753861

Signature: t1:1753861:1420236825:54cbdfee0efb3b97dea35faba29c134f2b10a480

Summary:
* folly/Range.cpp (scanHaystackBlock): This method mistakenly
declared its "idx" (aka blockStartIdx) parameter to have signed type.
It is logically an unsigned type and is compared only with other
variables of unsigned type, so make it unsigned.
Here's the diagnostic:
folly/Range.cpp:202:44: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
folly/Range.cpp: In instantiation of 'size_t folly::detail::scanHaystackBlock(folly::StringPiece, folly::StringPiece, int64_t) [with bool HAYSTACK_ALIGNED = true; size_t = long unsigned int; folly::StringPiece = folly::Range<const char*>; int64_t = long int]':

Test Plan:
Run this and note there are fewer errors than before:
fbconfig --platform-all=gcc-4.9-glibc-2.20 tao/server && fbmake dbgo

Reviewed By: robbert@fb.com, philipp@fb.com

Subscribers: folly-diffs@

FB internal diff: D1765590

Signature: t1:1765590:1420506036:7cbe2c454ad1f018a1c0aa5112a38bed1b2ac673

Summary:
see https://www.facebook.com/groups/738749156173702/permalink/798168656898418/

I suppressed "comparison between signed and unsigned integer expressions" warnings for this first pass because there a whole lot (dozens) of those. Do we suppress those with 4.8 or something? Never seen the warnings before.

I can't find anything in boost 1.57 docs/changelog indicating why implicit conversion of boost::optional<bool> -> bool doesn't compile anymore, but it doesn't.

Test Plan: fbconfig -r --platform-all gcc-4.9-glibc-2.20 --extra-cxxflags="-Wno-sign-compare" folly && fbmake opt && fbmake runtests_opt

Reviewed By: rhysparry@fb.com

Subscribers: trunkagent, fugalh, folly-diffs@

FB internal diff: D1765093

Signature: t1:1765093:1420500411:062acf841641c7f4142c9b48e8daa45327de6822

Summary:
This is a bootstrap script for building folly on osx. I attempted to make it Just Work™ for anyone who uses Homebrew (http://brew.sh/) but even if it doesn't work out of the box it should be instructive for the motivated developer on OSX.

Facebook:
In #5134151 Wangle used to not build, because something was using a futex. But that seems to have been rectified (or the build problems are hiding in headers that aren't used in the libfolly build).

I'll also update https://our.intern.facebook.com/intern/wiki/Folly/BuildingOnMac to point people to this script instead.

Test Plan:
Tested on my laptop which is running Yosemite (10.10.1)
Build, then inspect `otool -L .libs/*dylib`

Reviewed By: davejwatson@fb.com

Subscribers: folly-diffs@, exa

FB internal diff: D1765248

Tasks: 5134151

Signature: t1:1765248:1420500905:bc5da97a0446f502cb2badc85c9004ff0a62027a

Summary: another unittest that doesn't work without reuseport support, add a check

Test Plan: ran it on sandcastle187.prn2, works

Reviewed By: njormrod@fb.com

Subscribers: doug, fugalh, folly-diffs@

FB internal diff: D1764395

Tasks: 5931982

Signature: t1:1764395:1420490666:658aef148ebca9484534a0cebb228570e28f1002

Summary:
Can't connect from ipv6-only cluster to ipv4/ipv6 service which uses different ports.

Facebook:
A lot of tests start a service, get its port, and try to connect to 127.0.0.1:port, which breaks
after this diff since the port corresponds to IPv6 address, not IPv4. Fixing this by changing
127.0.0.1 to ::1 in places found by sandcastle.

```
fbconfig servicerouter/aggregator/tests servicerouter/client/cpp/test common/fb303/cpp/test thrift/lib/cpp2/test unicorn/hotswap/test common/client_mgmt thrift/test servicerouter/client/swig/tests unicorn/async/test servicerouter/selection/tests
```

Test Plan:
```
$ ./fastcopy_server --dir . &
$ netstat -atnlp | grep LISTEN | grep fastcopy
tcp        0      0 0.0.0.0:65478               0.0.0.0:*                   LISTEN      9348/./fastcopy_ser
tcp        0      0 :::52793                    :::*                        LISTEN      9348/./fastcopy_ser
```

Reviewed By: philipp@fb.com

Subscribers: trunkagent, vkatich, fbcode-common-diffs@, chaoyc, search-fbcode-diffs@, davejwatson, andrewcox, mcduff, hitesh, unicorn-diffs@, alandau, mshneer, folly-diffs@, bmatheny, ps, soren

FB internal diff: D1760372

Tasks: 5868818, 5886688

Signature: t1:1760372:1419992695:e7d254b2b8f730baefc169effa236b8daa9d846d

Summary:
De-couple TransportInfo fields from SSL session structs to avoid
dangling pointers.

Facebook:
We sometimes lazily copy TransportInfo in handler after
detachClientTransaction for logging. If the socket is closed, then this
creates dangling pointers to some SSL structs. This is an attempt to fix
that.

This is similar to what @ajitb did in
https://phabricator.fb.com/D1666951 which had to be abandoned because of
memory overhead. Here, instead of copying the relevant fields per
transaction, we are only doing it once per session (shared_ptr), so the
memory overhead should be negligible.

Test Plan: Unit tests pass. Will canary

Reviewed By: afrind@fb.com

Subscribers: fugalh, bmatheny, ssl-diffs@, folly-diffs@, ajitb

FB internal diff: D1757318

Tasks: 5865651, 5879508

Signature: t1:1757318:1420482488:9f5144b499eb2086cf2a80243328db5715b48f88

Summary: This change simplifies Singleton API (methods don't need to accept name) and the actual implementation. It also makes it similar to folly::ThreadLocalPtr. Additionally misspelled singleton name becomes compilation error, not runtime error. Some users were actually naming singletons, when that was neccessary, this should also be fixed.

Test Plan: unit tests for all touched projects

Reviewed By: chip@fb.com

Subscribers: trunkagent, fugalh, jsedgwick, fbcode-common-diffs@, mcduff, hitesh, mshneer, folly-diffs@

FB internal diff: D1744978

Signature: t1:1744978:1419282587:bd29dd8a70d7572530ac371a96a21764229bc397

Summary: `__cxa_throw` signature didn't match C++ ABI (see 2.4.3 in http://refspecs.linuxfoundation.org/abi-eh-1.22.html), gcc 4.9 was complaining.

Test Plan: fbconfig -r folly && fbmake opt -j32

Reviewed By: soren@fb.com

Subscribers: folly-diffs@

FB internal diff: D1760492

Tasks: 5908365

Signature: t1:1760492:1420251038:6d04683f96ac889ff348ca0485420c975819b90b

Summary:
Making sure Wangle can handle circular dependencies.
There was an actual code issue in September that cause a crash in Atlas
Adserver due to this. The issue has been since fixed and Adserver's code
has been changed to avoid such circular dependency. We're adding a unit
test for this case.

Unit test is added as per our conversation on the Wangle group: https://www.facebook.com/groups/715931878455430/permalink/770180369697247/

Test Plan:
Add the same test on a code base from Sep 20th, observe that it fails:
https://phabricator.fb.com/P19352007

Run the unit test on latest code base - it succeeds

Reviewed By: hans@fb.com

Subscribers: atlas2-eng@, fugalh, folly-diffs@, leizhao, mchughj, kit, mpechuk

FB internal diff: D1761006

Tasks: 5384229

Signature: t1:1761006:1420238204:74ffb3fe8b88a25a23ade8e0990e69d405ea7f1e

Summary: For when you have a future that you want to complete within a duration, else raise a `TimedOut` exception.

Test Plan: new unit tests

Reviewed By: jsedgwick@fb.com

Subscribers: trunkagent, fugalh, exa, folly-diffs@

FB internal diff: D1756580

Tasks: 4548494

Signature: t1:1756580:1420215704:862f68816fc3a9d05a77077c439bec002aa29cf3

Summary:
I can't repro the very frequent breakage we get from the unit
test framework, and it would be helpful to get the proper error messages
from it, so this adjusts the logging and changes a CHECK to an EXPECT to
get more data when the failure occurs.

Test Plan: runtests

Reviewed By: hans@fb.com

Subscribers: lins, anca, folly-diffs@

FB internal diff: D1761800

Tasks: 5870141

Signature: t1:1761800:1420226753:27a66fb0ce47cd57529e5fd0518550fb649c9eb5

Summary:
Clang 3.5 performs more rigorous static analysis than compilers we've previously used, and accordingly comes up with some additional warnings. This diff fixes some of the issues clang-3.5 warns about.

Facebook:

Test Plan:
# after pulling in changes that allow clang-3.5 to be used
fbconfig --clang --with-project-version clang:3.5 -r common/memory && fbmake runtests_opt
fbconfig --clang --with-project-version clang:3.5 -r common/strings && fbmake runtests_opt
fbconfig --clang --with-project-version clang:3.5 folly/test && fbmake runtests_opt

Tests for common/strings and folly/test pass. common/memory has one test failure, which also occurs with clang-3.4, while eliminating a couple of test failures that clang-3.4 had.

Reviewed By: meyering@fb.com

Subscribers: fbcode-common-diffs@, sdwilsh, mathieubaudet, folly-diffs@

FB internal diff: D1761101

Signature: t1:1761101:1419976252:ff60166dc8986e1239f88012b5e552a7d204aede

Summary:
emutls as far as StaticMeta needs to use it is broken on
Android as well due to unspecified pthread_key cleanup order between the
emulated __thread and our manual uses of it. Use the alternative that we
use on __APPLE__ there as well.

Test Plan: compiled and ran consuming code on Android

Reviewed By: dancol@fb.com

Subscribers: fma, shikong, kmdent, benyluo, ranjeeth, subodh, folly-diffs@

FB internal diff: D1760569

Tasks: 5265754, 5907613

Signature: t1:1760569:1419955865:afe8f35dadda85393492ac9331e9f62a74f4fdad

Summary: same as title

Test Plan: - fbconfig -r folly && fbmake runtests

Reviewed By: andrei.alexandrescu@fb.com

Subscribers: folly-diffs@

FB internal diff: D1754577

Signature: t1:1754577:1419618119:4c1a59cd19b23de9a9f6574341944e5e80530d62

Summary: Add a test to check that the accept IOPool can be the same as the IO worker thread pool

Test Plan: unittest

Reviewed By: jsedgwick@fb.com

Subscribers: doug, fugalh, njormrod, folly-diffs@

FB internal diff: D1710628

Signature: t1:1710628:1417469919:6d81a9426b61e9f1b804114f895ed541e547110f

Summary: kma asked

Test Plan: tests added

Reviewed By: andrei.bajenov@fb.com

Subscribers: alandau, bmatheny, mshneer, folly-diffs@

FB internal diff: D1756731

Signature: t1:1756731:1419442987:afdccce166ef1f1e609d8894ea587915f6fea8e7

Summary:
(almost) all the necessary code has been moved to folly.  A couple last functions in SSLUtils, moved them to SSLContext

I think this is enough so that open source mcrouter doesn't need an fbthrift dep anymore - although stuff in mcrouter/facebook still uses thrift.

Test Plan:
fbconfig -r mcrouter; fbmake runtests_dev

contbuild will probably show some other projects to fix up for SSLUtils

Reviewed By: alikhtarov@fb.com

Subscribers: doug, ps, alandau, bmatheny, alikhtarov, mshneer, folly-diffs@

FB internal diff: D1747133

Signature: t1:1747133:1418946064:1c30a60e43f017213e5514d462f267f91761abbe

Test Plan: jenkins

Reviewed By: ajitb@fb.com

Subscribers: exa, folly-diffs@

FB internal diff: D1758358

Tasks: 5880251

Signature: t1:1758358:1419640865:c9e9ea5922a602c44042c358777067e025800e6e

Summary: Add some forgotten header and cpp files to `Makefile.am` for OSS build of folly.

Test Plan: jenkins

Reviewed By: ajitb@fb.com

Subscribers: exa, folly-diffs@

FB internal diff: D1758318

Tasks: 5880251

Signature: t1:1758318:1419639812:47f527fdb2b3d2c49e7557b4176c6ab2478091ae

Summary:
Add basic timeout functionality. This adds `futures::sleep` which makes an async Future that finishes after the given duration, and `Future::get` which blocks on the result and takes an optional timeout.

Introducing the folly::wangle::futures namespace (soon to be just folly::futures) which will hold our wangle utility functions, the things that live in the Future object in Twitter's scala code. We'll probably move when* and wait-ish methods in here too, and perhaps alias makeFuture-ish methods too, though James has me mostly convinced not to deprecate them at the folly::wangle level (because they're basically Future constructors and Future lives at folly::wangle)

`Future::delayed` after Twitter's helper of the same name

Test Plan: new and old unit tests

Reviewed By: davejwatson@fb.com

Subscribers: trunkagent, fugalh, exa, folly-diffs@

FB internal diff: D1748894

Tasks: 4548494

Signature: t1:1748894:1419363496:1f4a62ec8455989c1fcce845695ace1d01c101c8

Summary: Provide a way to massacre outstanding timers.

Test Plan: new test, and using in wangle (different diff)

Reviewed By: davejwatson@fb.com

Subscribers: jsedgwick, exa, folly-diffs@

FB internal diff: D1753866

Tasks: 4548494

Signature: t1:1753866:1419361846:58ff6ca4d01b0d546495b032b454c8bf0fdb0277

Summary:
futexWaitUntil could generate an invalid timespec when presented
with a time_point from before the epoch, which resulted in an EINVAL from
the futex syscall.  Debug builds crashed on this unexpected return value,
while release builds soldiered on.  This path happened to be exercised
by the unit test.  This diff fixes the unintentional deadline overflow
in the test, adds explicit testing of overflow behavior, and makes
futexWaitUntil handle invalid time_points in a sensible manner.

Test Plan:
1. new unit tests
2. fbmake runtests_dbg

Reviewed By: mssarang@fb.com

Subscribers: strager, njormrod, folly-diffs@

FB internal diff: D1747972

Tasks: 5853949

Signature: t1:1747972:1419382193:862c193a13428d96acb33c85f962f59203661f40

Summary: This reverts commit fc83e983e2376bccc80dc4b21890e3a41dbc211f.

Test Plan: none

Reviewed By: lucian@fb.com, robot9@fb.com

Subscribers: ps, bmatheny, folly-diffs@

FB internal diff: D1756753

Tasks: 5883894, 5886688

Signature: t1:1756753:1419374764:977160379cfa49530babb38022b9d4cc80573a7d

Summary:
Sometimes when the socket is destroyed from a destructor, we wouldn't
want further callbacks on shutdown. We can unregister the readCallback_
by calling setReadCB(nullptr), but if the state is CONNECTING, we can
still get connectErr() callback. I found an ASAN trace (https://phabricator.fb.com/P18837265)
that turned out to be because of this inability to cancel this callback.
This provides a way to unregister the connect callback as well.

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: afrind@fb.com

Subscribers: folly-diffs@

FB internal diff: D1751778

Tasks: 5852935

Signature: t1:1751778:1419363638:26967c2d4fc5819e4d65ae706d217a954dfd784f

Summary:
The 'test' command uses one equal sign for string equality, not
two.  bash apparently is okay with two, but zsh isn't, causing this test
to pass if your shell was bash and fail if it was zsh.  The test now
passes with bash, bash in sh mode, zsh, and even dash.

Test Plan:
runtests, also, these pass (zsh failed before):

SHELL=/bin/bash _build/dbg/folly/test/subprocess_test
SHELL=/bin/sh _build/dbg/folly/test/subprocess_test
SHELL=/bin/zsh _build/dbg/folly/test/subprocess_test
SHELL=/bin/dash _build/dbg/folly/test/subprocess_test

Reviewed By: njormrod@fb.com

Subscribers: lins, anca, folly-diffs@

FB internal diff: D1756090

Signature: t1:1756090:1419360674:8576e61a6e0ee102612c5eae0e1fbd79cc397bfa

Summary:
Simplify the logic slightly to better handle joining paths.
This makes DwarfTest.cpp pass.

Test Plan: runtests

Reviewed By: njormrod@fb.com

Subscribers: lins, anca, folly-diffs@

FB internal diff: D1756036

Tasks: 5871014

Signature: t1:1756036:1419359254:123cc508b8836ea0b1485abe361b1c158538aa08

Summary: advancePastEmpty() was called for all created iterators. It only makes sense for begin(). For find() it's harmful: find() shouldn't return an iterator to the next element if the key was removed. I suspect that the same race condition was possible for insert(), but I haven't tried to reproduce it.

Test Plan: Added a test for it. It fails without these changes.

Reviewed By: delong.j@fb.com

Subscribers: folly-diffs@, lovro

FB internal diff: D1751280

Tasks: 5841499

Signature: t1:1751280:1419107193:71311ff68d92d0a4dcf1941dacdfdc23c25255cc

Summary: Can't connect from ipv6-only cluster to ipv4/ipv6 service which uses different ports.

Test Plan:
```
./fastcopy_server --dir . &

# netstat -atnlp | grep LISTEN | grep fastc
tcp        0      0 0.0.0.0:65478               0.0.0.0:*                   LISTEN      9348/./fastcopy_ser
tcp        0      0 :::52793                    :::*                        LISTEN      9348/./fastcopy_ser
```

Reviewed By: philipp@fb.com, sdoroshenko@fb.com

Subscribers: ps, bmatheny, folly-diffs@

FB internal diff: D1752846

Tasks: 5868818

Signature: t1:1752846:1419043494:7cc0646882249f17258ade5ce7ae5619b13148a0

Summary: Some overloads, particularly with rvalue references, were missing. Also this fixes https://our.intern.facebook.com/intern/tasks/?t=5849579

Test Plan: unittests

Reviewed By: simpkins@fb.com

Subscribers: trunkagent, las, net-systems@, njormrod, folly-diffs@

FB internal diff: D1746319

Tasks: 5849579

Signature: t1:1746319:1418875228:febb965cf52710a5e76b7c1cce5aec601086ad90

Summary: Allow conversion from folly::StringPiece to folly::StringPiece

Test Plan: - fbconfig -r folly && fbmake runtests

Reviewed By: mpawlowski@fb.com

Subscribers: folly-diffs@

FB internal diff: D1752609

Signature: t1:1752609:1419037172:04846f658901306b53fb60c9eafbdf0f4d4d7005

Summary:
For servers using GLOG, all thread ID are get from gettid system call. So it
is very useful if we print this ID so during debugging I can find the log
of crashing thread.

I'm not sure who should review this diff. Just found you guys from previous
diffs touched this file :)

Test Plan:
See the crash log.
*** Aborted at 1419029404 (Unix time, try 'date -d @1419029404') ***
*** Signal 11 (SIGSEGV) (0x18) received by PID 18852 (pthread TID 0x7f52a1fff700) (linux TID 27908), stack trace: ***

Reviewed By: tudorb@fb.com

Subscribers: seanc, folly-diffs@

FB internal diff: D1752301

Signature: t1:1752301:1419032887:e43c2ace6914e54812149a8b8e84eaffaf42a09c

Summary:
- Allow both `baseDir` and `subDir` to be empty.  This can happen
if DW_AT_comp_dir is `.` and gets simplified to the empty string.
(The DWARF standard doesn't appear to disallow relative dirs here,
but I couldn't find anthing definitive).
- Fix `./` prefix  stripping to avoid making paths like `.///hello` absolute.
- Fix `/` suffix stripping to avoid making the root dir reference (`/`) empty.
- Do `baseDir` and `subDir` swapping after simplification.

Test Plan: Added unittests.

Reviewed By: njormrod@fb.com

Subscribers: trunkagent, sdwilsh, folly-diffs@

FB internal diff: D1747978

Signature: t1:1747978:1419014942:87f14cb31b8c19b524d7a95b14d63cf5661a8634

Test Plan: no, it's a comment

Reviewed By: zanfur@fb.com

Subscribers: folly-diffs@

FB internal diff: D1752850

Tasks: 5823969

Signature: t1:1752850:1419043661:673c68c0daad28b3fde9709e767b528b0dec8ad1

Summary: A potential "real life" scenario that maybe can hit this bug is if we erase almost all elements and then iterate over the whole map.

Test Plan: Added a test for it.

Reviewed By: mwilliams@fb.com

Subscribers: folly-diffs@, lovro

FB internal diff: D1751455

Tasks: 5866368

Signature: t1:1751455:1419016611:b44c41d348f54397844460cb38002ad0d9704972