7 years agoFix folly lint errors
Nicholas Ormrod [Fri, 31 Oct 2014 17:48:19 +0000 (10:48 -0700)]
Fix folly lint errors

Fix EOF whitespace
for file in `find folly -type f` ; do if [ -z "`tail -n1 $file`" ] ; then sed -i '$d' $file ; fi ; done

Test Plan: unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644130

Tasks: 5486739

Signature: t1:1644130:1414715392:b6c783851aa030ad1148f84a98139a5dca207da0

7 years agoisDetachable for HHWheelTimer
Misha Shneerson [Fri, 7 Nov 2014 20:33:14 +0000 (12:33 -0800)]
isDetachable for HHWheelTimer

Summary: title

Test Plan: unit tests

Reviewed By:

Subscribers: trunkagent, mcduff, hitesh, alandau, bmatheny, njormrod, mshneer, folly-diffs@, andreib, davejwatson

FB internal diff: D1666499

Tasks: 5563183

Signature: t1:1666499:1415347256:d58c8bbe952385c1c96f7f8cc6ae7f02216c56bb

7 years agoMove SSL socket to folly
Dave Watson [Tue, 14 Oct 2014 18:10:48 +0000 (11:10 -0700)]
Move SSL socket to folly

Summary: One of the last thrift -> folly moves.  The only change was the exception types - there are small wrapper classes in thrift/lib/cpp/async left to convert from AsyncSocketException to TTransportException.

Test Plan: run unit tests

Reviewed By:

Subscribers: jdperlow, trunkagent, doug, bmatheny, ssl-diffs@, njormrod, mshneer, folly-diffs@, fugalh, jsedgwick, andrewcox, alandau

FB internal diff: D1632425

Signature: t1:1632425:1414526483:339ae107bacb073bdd8cf0942fd0f6b70990feb4

7 years agoRemove volatile from MemoryMappingTest
Tom Jackson [Fri, 7 Nov 2014 01:20:41 +0000 (17:20 -0800)]
Remove volatile from MemoryMappingTest

Summary: Not needed, I don't know why I put them there before.

Test Plan: Run the unit test

Reviewed By:

Subscribers: njormrod, folly-diffs@

FB internal diff: D1665689

Tasks: 5487902

Signature: t1:1665689:1415323144:0998e7f700a3b40652615a36c3b9c9f661fbdadf

7 years agoAdd some SocketAddress tests
Daniel Sommermann [Thu, 6 Nov 2014 21:44:13 +0000 (13:44 -0800)]
Add some SocketAddress tests

Summary: There was a bit of gap in test coverage for bracketed ipv6

Test Plan: unit tests

Reviewed By:

Subscribers: doug, ps, bmatheny, njormrod, folly-diffs@

FB internal diff: D1664783

Signature: t1:1664783:1415316694:b17e0dc1fcfece06c6e04e5f65e2095c66d28cc4

7 years agoFix integer overflow in folly/Benchmark.cpp
Dmitry Panin [Thu, 6 Nov 2014 20:47:43 +0000 (12:47 -0800)]
Fix integer overflow in folly/Benchmark.cpp

`bm_max_secs` has type int32
So when this value is more than 2, expression
FLAGS_bm_max_secs * 1000000000;

Test Plan:
I did put a
LOG(INFO) << timeBudgetInNs;
and observed correct values (before fix they were overflowed)

Reviewed By:

Subscribers: njormrod, folly-diffs@

FB internal diff: D1663247

Signature: t1:1663247:1415261814:c9154ffde183b2a4f5403e534e47e52e8276c61b

7 years agoBMI1 support in EliasFanoCoding
Philip Pronin [Tue, 4 Nov 2014 18:21:34 +0000 (10:21 -0800)]
BMI1 support in EliasFanoCoding

This diff updates `folly::CpuId` with support of extended features (EAX =
7, ECX = 0) to provide detection logic for BMI1 introduced in Haswell, and
provides support for `BLSR` instruction in `EliasFanoReader`.

Test Plan:
I used clang to compile the logic and run unittests

Reviewed By:

Subscribers: fbcode-common-diffs@, trunkagent, chaoyc, search-fbcode-diffs@, unicorn-diffs@, njormrod, folly-diffs@

FB internal diff: D1658100

Signature: t1:1658100:1415126635:d1820b8eb41c9e9786b5c8062b801cf1e2049a97

7 years ago(wangle) fix a race condition in Core::maybeCallback
Hans Fugal [Wed, 5 Nov 2014 23:44:06 +0000 (15:44 -0800)]
(wangle) fix a race condition in Core::maybeCallback

`calledBack_` could be seen as true by both threads in this conditional. Classic rookie mistake. :-/

Test Plan: run unit tests

Reviewed By:

Subscribers: trunkagent, hannesr, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1661199

Tasks: 55429385506504

Signature: t1:1661199:1415215840:fb69f56c8cf6f59beeca809724ce015b5260d9ad

Blame Revision: D1636487

7 years agoRemove memory leak from readRandomDevice by using a raw fd
Brian Watling [Wed, 5 Nov 2014 23:17:10 +0000 (15:17 -0800)]
Remove memory leak from readRandomDevice by using a raw fd

Summary: Switch from dynamically allocating a File to using a raw fd for readRandomDevice. This prevents races at shutdown while also eliminating the memory leak

Test Plan: run unit tests

Reviewed By:

Subscribers: njormrod, folly-diffs@, tao-eng@

FB internal diff: D1662151

Signature: t1:1662151:1415229242:525b6294b27bb68b5dda70aadb8d3ba1cc61b815

7 years agoFix crashes in readRandomDevice()
Brian Watling [Wed, 5 Nov 2014 22:50:37 +0000 (14:50 -0800)]
Fix crashes in readRandomDevice()

Summary: The comment in readRandomDevice() lies - the PCHECK fails with "Bad file descriptor" in some tests. This is most likely due to a race where a background thread is calling readRandomDevice() while the main thread shuts down. Another possibility is bad destruction order, where rnadomDevice is destroyed before some other static object whose destructor calls readRandomDevice(). Either way, this fixes it without having to chace down every instance.

Test Plan: run unit tests

Reviewed By:

Subscribers: antonl, trunkagent, meyering, simpkins, njormrod, folly-diffs@, tao-eng@

FB internal diff: D1660903

Signature: t1:1660903:1415215895:7f1f96ab7416b45c4dc18e78173fb59eb27bd03b

7 years ago(wangle) fix a race in whenAll
Hans Fugal [Wed, 5 Nov 2014 19:21:49 +0000 (11:21 -0800)]
(wangle) fix a race in whenAll

Race in `++ctx->count == ctx->total`. This ordering, while not very obvious or likely, is possible:

T1            T2
--            --
delete ctx
delete ctx

Test Plan:
That's the idea, anyway. I need some sleep, and it takes 20 minutes to build and test. I had a more convoluted fix (using `shared_ptr`) and it did seem to fix the error we were seeing, but I was seeing another error.

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1660663

Tasks: 5506504

Signature: t1:1660663:1415207632:49dc224363cec27736fc71fb211fa846be9e170f

Blame Revision: D1636487

7 years agofolly: wangle: NamedThreadFactory: store of the thread prefix locally, don't assume...
Lucian Grijincu [Wed, 5 Nov 2014 00:51:03 +0000 (16:51 -0800)]
folly: wangle: NamedThreadFactory: store of the thread prefix locally, don't assume they'll live forever

Summary: There's no real reason to store this as a StringPiece, is there?

Test Plan: n/a

Reviewed By:,

Subscribers: fugalh, njormrod, folly-diffs@

FB internal diff: D1659558

Tasks: 5538418

Signature: t1:1659558:1415147560:2c5b0c996893854c3ea9d0ad02b006bcc5960ffa

7 years agoFix licenses
Nicholas Ormrod [Tue, 4 Nov 2014 00:18:41 +0000 (16:18 -0800)]
Fix licenses

There are two types of licenses used in folly, only one of
which is accepted by the linter.

This diff changes the license notices in pre-existing folly files!
Many folly/io/async/* files have the second type of license, but
without a Copyright notice. I have added copyright notices to these

I have also added a compliant notice to
folly/test/function_benchmark/benchmark_impl.h, which was the sole file
in folly/test/function_benchmark/ that didn't have a standard license.

Test Plan:
The changes to folly are comment only.

Run all of folly against the linter, see no more license errors.

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@, sjenkins

FB internal diff: D1648489

Tasks: 5486739

Signature: t1:1648489:1415035522:3d8bd9611eb7c7117b70d5e7f68de5768639a727

7 years agoFix global statics
Nicholas Ormrod [Mon, 3 Nov 2014 20:39:11 +0000 (12:39 -0800)]
Fix global statics

Statics in headers are bad.

Test Plan: run unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1646906

Tasks: 5486739

Signature: t1:1646906:1415042805:dc4d1cec54e9320f1e609808a73622c731a4cdc9

7 years agorm protected inheritance
Nicholas Ormrod [Mon, 3 Nov 2014 18:06:41 +0000 (10:06 -0800)]
rm protected inheritance

Remove the protected inheritance from folly.

The code looks like the protected might be appropriate: there are some
derived classes which might want access to the AsyncTimeout.

Test Plan: run unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1649686

Tasks: 5486739

Signature: t1:1649686:1415035288:18efd2cf9aae8caab66d8303c22cdc26c6b54ae5

7 years agonetty-like channels
James Sedgwick [Mon, 3 Nov 2014 17:37:35 +0000 (09:37 -0800)]
netty-like channels

I wanted some foundational abstractions to start building out codecs on top of.
I also know that Blake gets very amused when I shamelessly copy Java/Netty abstractions, and I live to amuse Blake so I did it again.
So here's an implementation of something very similar to Netty's ChannelAdapters/ChannelPipelines
Only read() and write() for now, everything/anything else can easily be bolted on once the design is settled (if this is at all the direction we want to go)
I have a lot of thoughts about this design but I'm going to save my fingers and leave that to ad hoc discussions once folks take a look at this
A couple of things, though:
- It should be possible to dynamically add handlers to the chain. How I envision this working is that each original adapters keeps two lists of shared/unique ptrs to adapters of the correct type to sit next to them on either side, and dispatch to them appropriately when they're there.
- I was trying to do without separate ChannelHandlerContext objects altogether (keep the interface, but have ChannelPipeline act as the context itself) but ran into issues with virtual multiple inheritance. I might have another go at this.
- Only movable types are permitted. I hope this won't be too restrictive, because it would be a PITA to support both move-only and copy-only types.
- Why no Rx? Seems to me that any handlers that actually needs Rx (e.g. stats fanout or something) can deal with it themselves.
- If it turned out to be useful to nest these (for more flexible composition) that would also be doable. i.e. ChannelPipeline<ChannelPipeline<Handler1, Handler2>, ChannelPipeline<Handler3, Handler4>>

Test Plan: super basic test compiles and runs as expected

Reviewed By:

Subscribers: ajitb, folly-diffs@, ldbrandy, trunkagent, fugalh, njormrod

FB internal diff: D1604575

Tasks: 5002299

Signature: t1:1604575:1415034767:bc3b12fae726890aa6a55ed391286917ae23e56e

7 years agoFix -Wsign-compare
Daniel Sommermann [Mon, 20 Oct 2014 19:52:01 +0000 (12:52 -0700)]
Fix -Wsign-compare

Summary: Folly should be able to compile with strict flags.

Test Plan: unit tests, review

Reviewed By:

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

FB internal diff: D1627280

Signature: t1:1627280:1414792755:004f5a737ece1e93bcf4331718a98afc57e4f80c

7 years agoName prefix setter for NamedThreadFactory
James Sedgwick [Mon, 3 Nov 2014 00:28:49 +0000 (16:28 -0800)]
Name prefix setter for NamedThreadFactory

Summary: it's useful to update the prefix after the construction.

Test Plan:
unit, that's it :/

Reviewed By:

Subscribers: mshneer, folly-diffs@, wch, atlas2-eng@, everstore-dev@, wormhole-diffs@, ads-dsp-eng@, bwester, trunkagent, fugalh, alandau, njormrod, bmatheny

FB internal diff: D1585087

7 years agoFix bad rebase for SSLContext
Dave Watson [Fri, 31 Oct 2014 19:49:57 +0000 (12:49 -0700)]
Fix bad rebase for SSLContext

Summary: Put back previous liger diffs

Test Plan: It builds

Reviewed By:

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

FB internal diff: D1652754

Signature: t1:1652754:1414785984:df0fc7bf59dc2e89defd2c1a4ffe3b288238ba58

7 years agoFix line too long lint errors
Nicholas Ormrod [Fri, 31 Oct 2014 21:54:30 +0000 (14:54 -0700)]
Fix line too long lint errors

Many lines that are too long contain URLs. This diff adds
url-like lines to the list of exceptions to 80-char limits. It also
fixed the locations in folly with line-too-long errors.

Test Plan: arc lint on folly, see no line-too-long errors.

Reviewed By:

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644151

Tasks: 5486739

Signature: t1:1644151:1414779408:76bcec1d14daaa8ed071c715bf26b108c8fe4b87

7 years agoIncrease size of demangled symbol buffer
Mark Drayton [Fri, 31 Oct 2014 20:42:40 +0000 (13:42 -0700)]
Increase size of demangled symbol buffer

Symbols which demangle to a string longer than 1024 bytes are
quite common in our code. This diff increases the size of the output
buffer to accommodate them.

Test Plan: run it

Reviewed By:

Subscribers: trunkagent, njormrod, folly-diffs@

FB internal diff: D1639801

Tasks: 5464222

Signature: t1:1639801:1414784601:2f59d5a58e434f4cf9df5b25b917c5094c8b133f

7 years agoUndo a local change
Sachin Kadloor [Fri, 31 Oct 2014 18:13:13 +0000 (11:13 -0700)]
Undo a local change

Summary: I had committed a change by mistake. Reverting it.

Test Plan: build it.

Reviewed By:

Subscribers: njormrod, folly-diffs@

FB internal diff: D1652341

Signature: t1:1652341:1414779013:9f363acba95fbac1988200081659edf71ac63eec

7 years agovolatile override
Nicholas Ormrod [Fri, 31 Oct 2014 18:07:44 +0000 (11:07 -0700)]
volatile override

Some uses of volatile are legit. Add a hidden override.

Test Plan: run unit tests

Run flint against folly/Malloc.cpp, see no more volatile warning.

Reviewed By:

Subscribers: sdwilsh, louisk, njormrod, folly-diffs@

FB internal diff: D1644493

Tasks: 5486739

Signature: t1:1644493:1414715317:491d0f631d8152a5b7ec66237e00c404530ab590

7 years agofix build
Dave Watson [Fri, 31 Oct 2014 17:37:55 +0000 (10:37 -0700)]
fix build

Summary: add missing ssl dep

Test Plan: build it

Reviewed By:

Subscribers: doug, njormrod, folly-diffs@

FB internal diff: D1652234

Signature: t1:1652234:1414778507:4504ea06c3076e0cdfa7772166cf48bc564c9076

7 years agoFix large size class category handling in folly::goodMallocSize()
Tomislav Novak [Fri, 24 Oct 2014 18:46:51 +0000 (11:46 -0700)]
Fix large size class category handling in folly::goodMallocSize()

jemalloc seems to now use `4064 * 1024` as `arena_maxclass`. Therefore,
allocations that fall in the range `(4064*1024, 4072*1024]` will trigger an
assert in `folly::goodMallocSize()` (`nallocx()` will round up sizes within
that range up to the next multiple of 4 MB (chunk size)).

Test Plan: Call to `folly::goodMallocSize(4161568)` no longer triggeres an assert.

Reviewed By:

Subscribers: aalexandre, trunkagent, njormrod, folly-diffs@

FB internal diff: D1638370

Signature: t1:1638370:1414703429:f3ed2cc8dcb183ff4e8367e2c1f884e122605cba

7 years agoSocketAddress logging
Sachin Kadloor [Fri, 31 Oct 2014 17:58:47 +0000 (10:58 -0700)]
SocketAddress logging

Better logging in SocketAddress

Test Plan: run it

Reviewed By:

Subscribers: ps, bmatheny, xning

FB internal diff: D1644573

Tasks: 5406091

Signature: t1:1644573:1414626398:1e23ee85ec8ee0be89cb6bdc07da23978b4d988a

7 years agoFix/override bad symbol names
Nicholas Ormrod [Fri, 31 Oct 2014 17:55:44 +0000 (10:55 -0700)]
Fix/override bad symbol names

Symbols containing two underscores, or starting with underscore
followed by a capital letter, are reserved by the implementation. Fix
the inappropriate cases, and nolint the ones which we are using

Test Plan: run unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1650435

Tasks: 5486739

Signature: t1:1650435:1414713329:7844e7802ebabcac7ef21aef0becf71ba513afb0

7 years agoComparing the same tokens
Nicholas Ormrod [Fri, 31 Oct 2014 17:52:41 +0000 (10:52 -0700)]
Comparing the same tokens

Avoid comparing-the-same-tokens error by parenthesizing one

This diff only touches folly tests

Test Plan: run unit tests

Reviewed By:

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1650386

Tasks: 5486739

Signature: t1:1650386:1414713837:cbb578d23f759aa049f4246113d9da6d40504da4

7 years agothrow() specs deprecated
Nicholas Ormrod [Fri, 31 Oct 2014 17:50:05 +0000 (10:50 -0700)]
throw() specs deprecated

Summary: Switch to noexcept

Test Plan: run unit tests

Reviewed By:,

Subscribers: aalexandre, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644135

Tasks: 5486739

Signature: t1:1644135:1414715537:22e1baf91ab7e3250b0b2a460a12d56783f2baed

7 years agoFix folly lint errors
Nicholas Ormrod [Fri, 31 Oct 2014 17:48:19 +0000 (10:48 -0700)]
Fix folly lint errors

Fix EOF whitespace
for file in `find folly -type f` ; do if [ -z "`tail -n1 $file`" ] ; then sed -i '$d' $file ; fi ; done

Test Plan: run unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644130

Tasks: 5486739

Signature: t1:1644130:1414715392:b6c783851aa030ad1148f84a98139a5dca207da0

7 years agoSSLContext
Dave Watson [Wed, 1 Oct 2014 17:51:40 +0000 (10:51 -0700)]

Move SSLContext to folly.

Test Plan: It builds.

Reviewed By:

Subscribers: jdperlow, atlas2-eng@, wormhole-diffs@, bwester, trunkagent, doug, ps, bmatheny, ssl-diffs@, alikhtarov, njormrod, mshneer, folly-diffs@, andrewcox, alandau, jsedgwick, fugalh

FB internal diff: D1631924

Signature: t1:1631924:1414616562:9a67dbf20f00eb8fbcb35880efcb94c0fae07dcc

7 years agosave/restore request context in future
Dave Watson [Thu, 30 Oct 2014 23:16:56 +0000 (16:16 -0700)]
save/restore request context in future

Generic 'threadlocal' like object that follows async calls around.

Note: Finagle's futures do this also.  It is super useful.

Test Plan: I should write a unittest?

Reviewed By:,

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

FB internal diff: D1650843

Tasks: 4698780

Signature: t1:1650843:1414711295:c7439733680ab4903eb9c05dcf8bf52100bf3f07

7 years ago(wangle-test) fix double ifndef in thens.rb
Hans Fugal [Thu, 30 Oct 2014 23:20:17 +0000 (16:20 -0700)]
(wangle-test) fix double ifndef in thens.rb

Summary: Looks like a bad git merge

Test Plan: builds

Reviewed By:

Subscribers: hannesr, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1649773

Tasks: 5501131

Signature: t1:1649773:1414700920:cc691283884e9654803bba3ffe78f4553194752d

7 years agooperator::* needs implicit
Nicholas Ormrod [Thu, 30 Oct 2014 21:15:02 +0000 (14:15 -0700)]
operator::* needs implicit

operator::* needs to be explicit, or explicitly labeled
implicit. The build breaks with explicit, so label.

Test Plan: run unit tests

Reviewed By:

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644176

Tasks: 5486739

Signature: t1:1644176:1414693471:e3a08e27ba27a5ae0e4b9a5e009acf0a16728c45

7 years agoMake dynamic noexcept
Nicholas Ormrod [Thu, 30 Oct 2014 20:55:13 +0000 (13:55 -0700)]
Make dynamic noexcept

Many dynamic operations, including its move constructor, are
noexcept but not labeled so. Labeled some important functions as

Test Plan: run unit tests

Reviewed By:

Subscribers: tudorb, philipp, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644380

Tasks: 5486739

Signature: t1:1644380:1414618757:fb910d8a3fe3e634da4215c432577edf7371be61

7 years agoFix -Wsign-compare
Daniel Sommermann [Fri, 17 Oct 2014 20:38:00 +0000 (13:38 -0700)]
Fix -Wsign-compare

Fixed unsigned/signed compares.

I had to suppress the warning in one place, but don't worry, clang
supports `#pragma GCC diagnostic ...` (see

Test Plan: unit tests

Reviewed By:

Subscribers: alandau, mshneer, trunkagent, doug, jdperlow, bmatheny, njormrod, cgheorghe, folly-diffs@

FB internal diff: D1624249

Tasks: 5140804

Signature: t1:1624249:1414620209:5399e8d90d8ca32b30794a7b2a4a7c2d7d437dda

7 years agoOverride #define of keyword
Nicholas Ormrod [Thu, 30 Oct 2014 18:38:36 +0000 (11:38 -0700)]
Override #define of keyword

Add a hidden option to suppress ###define [keyword]##, which is
occasionally appropriate.

Test Plan:
Build flint, run it on FBString.h, see no error.
Ran flint tests locally, see no errors.

Reviewed By:

Subscribers: sdwilsh, louisk, njormrod, folly-diffs@

FB internal diff: D1644182

Tasks: 5486739

Signature: t1:1644182:1414616294:4b6004d358cba865455366f5a943644b3542ae51

7 years agoOverride for include-guard
Nicholas Ormrod [Thu, 30 Oct 2014 18:26:45 +0000 (11:26 -0700)]
Override for include-guard

In certain circumstances, it is appropriate to omit and include
guard. Add an option to allow this to be overriden.

Test Plan: Build flint, run it on folly, see no include-guard errors.

Reviewed By:

Subscribers: sdwilsh, louisk, njormrod, folly-diffs@

FB internal diff: D1644170

Tasks: 5486739

Signature: t1:1644170:1414616595:7ac52f474c1312a0c28e89255b1151d56c680acf

7 years agoAllow catch-int to be nolinted
Nicholas Ormrod [Thu, 30 Oct 2014 18:16:39 +0000 (11:16 -0700)]
Allow catch-int to be nolinted

Add a hidden nolint option to suppress this lint warning. It is
appropriate in this case.

Test Plan:
Build flint, run it on ExceptionWrapperTest.cpp, see no catch
lint failure

Reviewed By:

Subscribers: sdwilsh, louisk, njormrod, folly-diffs@

FB internal diff: D1644165

Tasks: 5486739

Signature: t1:1644165:1414616664:a6b9dc34660df84b33ed8faaf48ec048a02bad01

7 years agoMove constructors should be noexcept
Nicholas Ormrod [Thu, 30 Oct 2014 17:18:27 +0000 (10:18 -0700)]
Move constructors should be noexcept

The compiler is able to make some nice optimizations when it
knows that move constructors are noexcept. (In particular, it helps a
lot inside of std::vector if you don't need to worry about the
possibility of exception during reallocation.) Some move constructors in
folly are obviously moveable: change them.

Test Plan: unit tests

Reviewed By:

Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644296

Tasks: 5486739

Signature: t1:1644296:1414618605:d9a0db5193c82650b96e9c62b019d5da218b15c5

7 years agoFuture<T>::then([](T&&)), aka next()
James Sedgwick [Wed, 29 Oct 2014 23:33:14 +0000 (16:33 -0700)]
Future<T>::then([](T&&)), aka next()

variants of then() that bypass Try and forward any exceptions up to the next future. like next() from

this could go a long way towards reducing spurious rethrows wherever people don't ever actually catch exceptions in their continuations, which is probably often enough.

anyone know if there's something in folly or standard library that does what my silly VoidWrapper struct does?

there's a lot of copypasta here but i'm not sure consolidating into helpers would actually be useful considering the amount of template crap, i don't feel strongly about it though.

Test Plan: added unit

Reviewed By:

Subscribers: trunkagent, sammerat, fugalh, njormrod, folly-diffs@, bmatheny

FB internal diff: D1641776

Signature: t1:1641776:1414610434:c742563b8061a748fca9292fc2765081edcf9d52

7 years agoBump version to 14:0
dcsommer [Wed, 29 Oct 2014 23:08:59 +0000 (16:08 -0700)]
Bump version to 14:0

7 years agoReally fix the clang warning in Format-inl.h
Sean Cannella [Wed, 29 Oct 2014 21:02:55 +0000 (14:02 -0700)]
Really fix the clang warning in Format-inl.h

The previous diff added the safety check but didn't actually
remove the compiler warning. This is what I get for doing an incremental

Test Plan: fbmake clean ; fbmake dbg with clang:dev

Reviewed By:

Subscribers: trunkagent, mathieubaudet, njormrod, folly-diffs@, bmatheny, ranjeeth, subodh, kmdent, fma, benyluo

FB internal diff: D1646519

Signature: t1:1646519:1414602292:cfb908ae094caaef02e64eb2c42544fd34fa1389

7 years agoFix spelling mistakes
Nicholas Ormrod [Wed, 29 Oct 2014 20:16:38 +0000 (13:16 -0700)]
Fix spelling mistakes

Summary: @override-unit-failures

Test Plan: n/a

Reviewed By:

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644139

Tasks: 5486739

Signature: t1:1644139:1414553949:626fdca07a583c3d5017abe19408a5e6f6d7674a

7 years agoFix multiple includes
Nicholas Ormrod [Wed, 29 Oct 2014 20:10:39 +0000 (13:10 -0700)]
Fix multiple includes

Add nolint comments to appropriate multiply-included files.
Removed raw duplicates from StlVectorTest.

Test Plan:
fbconfig -r folly && fbmake runtests

Ran arc lint, see no "included multiple times" messages

Reviewed By:

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644144

Tasks: 5486739

Signature: t1:1644144:1414599280:1e403ea3675d2173b249c4ce6db0dad621a139bc

7 years agooperator bool()
Nicholas Ormrod [Wed, 29 Oct 2014 20:03:02 +0000 (13:03 -0700)]
operator bool()

Among all of the implicit conversions, conversion to bool is
particularly tricky. For this reason, it has a special call-out in flint
among the casts, and canno be overriden.

However, there are legitimate use cases for operator bool. In this case,
we have a iterator that logically represents a bool but in practice
represents a packed integral. This class needs to be implicitly
convertible to bool, and warrants an override.

Push blocking tests have had over a day to complete, but haven't.
Further, the folly change is comment-only.

Test Plan:
Build flint, run it on BitIteratorDetail.h, see no more lint

Reviewed By:

Subscribers: sdwilsh, louisk, njormrod, folly-diffs@

FB internal diff: D1644141

Tasks: 5486739

Signature: t1:1644141:1414607351:29740da758b75187f4f7b6b5a5ad28a523e7080d

7 years ago__attribute__s need double underscore
Nicholas Ormrod [Wed, 29 Oct 2014 19:55:12 +0000 (12:55 -0700)]
__attribute__s need double underscore

Summary: See D1508946 for the rationale behind this lint message.

Test Plan:
fbconfig -r folly && fbmake dbg

Reviewed By:

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

FB internal diff: D1644190

Tasks: 5486739

Signature: t1:1644190:1414532704:87d5851f363cf37c58f7e7c3d57c8210a792d4a1

7 years agomove clang disable of Thens.cpp to thens.rb
Hans Fugal [Wed, 29 Oct 2014 18:30:02 +0000 (11:30 -0700)]
move clang disable of Thens.cpp to thens.rb

Summary: So when we re-generate code we don't cause clang test failures (until #4412111 is done)

Test Plan: git diff

Reviewed By:

Subscribers: net-systems@, fugalh, mathieubaudet, exa, njormrod, folly-diffs@

FB internal diff: D1644550

Tasks: 4412111

Signature: t1:1644550:1414604128:8184e7ce1f3e417c170ef67346d553caecc1f013

7 years agoMove AsyncSocket to folly (try 2)
Dave Watson [Tue, 28 Oct 2014 17:52:49 +0000 (10:52 -0700)]
Move AsyncSocket to folly (try 2)


Test Plan:

Reviewed By:

Subscribers: trunkagent, mcduff, hitesh, doug, alandau, bmatheny, njormrod, mshneer, folly-diffs@

FB internal diff: D1644071

Signature: t1:1644071:1414526899:c41dd55e2957a7e1fcc54508e20cdb4a9c8c30d4

7 years agoAddress clang sign warning in Format-inl.h
Sean Cannella [Wed, 29 Oct 2014 02:09:35 +0000 (19:09 -0700)]
Address clang sign warning in Format-inl.h

This has been bothering me for a while when I've been trying to
verify builds with clang. This generates a ton of spew and the current
code can't actually handle non-default negatives so let's just throw if
we see one.

Test Plan: fbmake runtests

Reviewed By:

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

FB internal diff: D1643881

Signature: t1:1643881:1414523777:085035ce8c280fbd6481b2ce0354b53b96fdc50e

7 years agokill threadgate
Hans Fugal [Tue, 28 Oct 2014 21:49:16 +0000 (14:49 -0700)]
kill threadgate

Test Plan: This diff is actually the test plan. It was contbuild bait. I fixed things that broke (in separate diffs), and then I rebased this on top, everything is green, and we pulled the trigger.

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1632937

Tasks: 5409538

Signature: t1:1632937:1414531009:4dbaa9513fbfba45a9c63e7be3de3646a6e926a2

7 years agoAdd MemoryIdler suppot to IOThreadPoolExecutor
Dave Watson [Mon, 27 Oct 2014 17:11:03 +0000 (10:11 -0700)]
Add MemoryIdler suppot to IOThreadPoolExecutor

Idle memory in IO threads.   If loop is unused for a period of time, free associated memory, and call epoll again.

Had to add a new list of callbacks that don't make the loop nonblocking (i.e. using runInLoop() instead would use the nonblocking version of epoll).

Could bake this in to EventBase directly, but that seems like the wrong abstraction, since EventBase doesn't actually control the thread - for example, that approach would also free up memory for stack-allocated EventBases where they are used synchronously by clients.

This diff doesn't change IO scheduling at all - current IO work is round robin, so this probably only helps if the whole server is idle (at least until we add smarter scheduling)

Test Plan:
Based on top of D1585087.

fbconfig thrift/perf/cpp; fbmake dbg
_bin/thrift/perf/cpp/loadgen -num_threads=100 -weight_sendrecv=1 -cpp2 -async

Ran loadgen for a while, watched res memory in top.  Stopped loadgen.  After ~5 sec, res memory was much reduced.

Reviewed By:

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

FB internal diff: D1641057

Tasks: 5002425

7 years agoRevert "D1587625 [thrift->folly] AsyncSocket"
Viswanath Sivakumar [Tue, 28 Oct 2014 02:38:10 +0000 (19:38 -0700)]
Revert "D1587625 [thrift->folly] AsyncSocket"


Test Plan:

Reviewed By:


FB internal diff: D1642334

Blame Revision: D1587625

7 years agoimprove io::Appender functionality
Adam Simpkins [Sun, 28 Sep 2014 01:24:18 +0000 (18:24 -0700)]
improve io::Appender functionality

Add an operator()(StringPiece) method to Appender, so it can be used
directly with folly::Formatter objects.  Also add printf() and vprintf()
methods to Appender, for places that need to use existing printf-style

This also includes versions of push() and pushAtMost() that accept
ByteRange objects.  Previously we only had push() implementations that
took separate buffer and size arguments.

Test Plan: Added new unit tests to IOBufCursorTest.cpp

Reviewed By:

Subscribers: trunkagent, doug, net-systems@, exa, njormrod

FB internal diff: D1583684

7 years agoAdd missing headers
Dave Watson [Mon, 27 Oct 2014 21:25:00 +0000 (14:25 -0700)]
Add missing headers

Summary: Jenkins ci build was broken for a while, so we missed these  in contbuild results

Test Plan: watch ci build for fbthrift

Reviewed By:

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

FB internal diff: D1641515

7 years agoAsyncSocket
Dave Watson [Fri, 26 Sep 2014 20:49:11 +0000 (13:49 -0700)]

Move async socket to folly.

* Made an AsyncSocketException type instead of TTransportException: Some of the exceptions didn't fit nicely in to std::exception types (like TIMED_OUT).  There are some wrappers in thrift/lib/cpp/async to convert back to TTransportException, so all existing code still compiles.
* Moved read/write callbacks out of AsyncTransport: filters are going to want to do the read/write stuff separately (see revproxy/tunnel/filters, and discussions in D1483148).

Test Plan:
fbconfig -r thrift; fbmake runtests
contbuild should catch everything else - exception types shouldn't change for existing code

Reviewed By:

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

FB internal diff: D1587625

7 years ago(wangle) fix race in Core::detachOne()
Hans Fugal [Mon, 27 Oct 2014 15:53:23 +0000 (08:53 -0700)]
(wangle) fix race in Core::detachOne()

In D1618240 I introduced a race condition in `detachOne()`, where `detached_` is incremented and then tested. If the promise and future are racing, they can both see `detached_ == 2` in the conditional, and then they'll both try to free the Core object. This fixes that.

It also fixes a related problem (which actually showed up more often with the test I wrote), where we transition into the Done state before setting the value, and then `maybeCallback()` observes the state is Done (because we're just reading an atomic, not grabbing the lock, which is intentional), tries to execute the callback, but `folly::Optional` throws an exception because the value hasn't been set yet (at least in debug it does). I should have listened to my gut and kept the state assignment after the transition action in the first place.

Test Plan: New unit test

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@, mnd

FB internal diff: D1636490

Tasks: 5438209

Blame Revision: D1618240

7 years agounrevert "(wangle) express current Core functionality with a state machine""
Hans Fugal [Mon, 27 Oct 2014 15:53:20 +0000 (08:53 -0700)]
unrevert "(wangle) express current Core functionality with a state machine""

Summary: Reverts D1633874. Companion to D1636490 which fixes the bug.

Test Plan:
git reverting code that was git reverted and hasn't changed in the interim
Won't be checked in without the companion bugfix diff (D1636490)

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1636487

Tasks: 5438209

Blame Revision: D1633874

7 years agoRemove some leftover apache paths
Dave Watson [Thu, 23 Oct 2014 22:21:23 +0000 (15:21 -0700)]
Remove some leftover apache paths

Summary: EventBaseManager was moved to folly, we shouldn't have deps on apache

Test Plan: fbconfig -r folly/experimental/wangle/concurrency; fbmake runtests

Reviewed By:

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

FB internal diff: D1636006

7 years agoRevert "(wangle) express current Core functionality with a state machine"
Hans Fugal [Thu, 23 Oct 2014 01:11:34 +0000 (18:11 -0700)]
Revert "(wangle) express current Core functionality with a state machine"


Test Plan:

Reviewed By:

Subscribers: net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1633874

Tasks: 5438209

7 years agofix Future<const T>::value()
Hans Fugal [Wed, 22 Oct 2014 17:14:45 +0000 (10:14 -0700)]
fix Future<const T>::value()

Summary: I'm still a little mystified how one makes a Future<const T> (I tried to make one explicitly in a test and failed), but this is clearly the bug @reedriley sees in

Test Plan:
fbconfig titan/cachius/test

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, reedriley, folly-diffs@

FB internal diff: D1630378

Blame Revision: D1620805

7 years agomove asyncserversocket to folly
Dave Watson [Thu, 25 Sep 2014 17:17:20 +0000 (10:17 -0700)]
move asyncserversocket to folly

* namespace to folly
* some std::chrono replacesments
* Moves shutdownSocketSet stuff to thriftserver instead of in the socket itself.
* Changes exception types.  I don't see anywhere that uses the TAsync*SERVER*socket exceptions depending on the TTransport type, but I could be wrong? I don't really know what to do about the exception types

unittests still postponed overnight.

Test Plan:
fbconfig -r thrift; fbmake runtests
fbconfig -r proxygen; fbmake runtests

Reviewed By:

Subscribers: hphp-diffs@, ps, folly-diffs@, anca, trunkagent, jsedgwick, fugalh, doug, alandau, bmatheny, njormrod

FB internal diff: D1579187

7 years agoadd PortableSpinLock.h to Makefile, unbreak fbthrift build
James Sedgwick [Wed, 22 Oct 2014 01:04:44 +0000 (18:04 -0700)]
add PortableSpinLock.h to Makefile, unbreak fbthrift build

fix fbthrift build, broken in D1621717

I wonder if it'd be possible to lint against this somehow. The jenkins failures are easy to miss especially on folly diffs with lots of failure noise (in this case it was still postponed)

Test Plan: wait for jenkins

Reviewed By:

Subscribers: njormrod, folly-diffs@

FB internal diff: D1630833

7 years agoSchedule destroyInstances in registrationComplete
Andrii Grynenko [Tue, 21 Oct 2014 20:55:56 +0000 (13:55 -0700)]
Schedule destroyInstances in registrationComplete

Summary: This makes it more likely that SingletonVaultDestructor won't be created after atexit calls.

Test Plan: unit test

Reviewed By:

Subscribers: hphp-diffs@, ps, njormrod, folly-diffs@

FB internal diff: D1629804

Tasks: 5353022

7 years ago(wangle) fix after-delete assert
Hans Fugal [Tue, 21 Oct 2014 22:50:02 +0000 (15:50 -0700)]
(wangle) fix after-delete assert

This would cause debug builds to do a bad thing (access the variable `this->detached_` within an assert, after `delete this`).

Test Plan: unit tests

Hopefully now that we have a dummy cpp file in `folly/wangle/detail` contbuild
will pick it up and all the dependencies will also run their tests.
Right now, we suspect maybe maelstrom (@wez) and adinvoicer and zookeeper
(@jying) and probably others are seeing this in unit test failures (esp. if
they use asan, which is rightly detecting read after free). Hoping contbuild
will catch them.

Reviewed By:

Subscribers: net-systems@, fugalh, exa, njormrod, folly-diffs@, wez, dcapel, jying, cgheorghe

FB internal diff: D1630301

Tasks: 54245465435720

Blame Revision: D1618240

7 years agowaitFor race workaround
Hans Fugal [Tue, 21 Oct 2014 18:55:52 +0000 (11:55 -0700)]
waitFor race workaround

Summary: Fixing the race will take some more thought. Probably a condvar. For now, I'm turning waitFor into a spinlock instead.

Test Plan: new unit test that deadlocks before the workaround

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1628015

Tasks: 5427828

7 years ago(wangle) Fix a couple compilation issues
Hans Fugal [Tue, 21 Oct 2014 17:24:14 +0000 (10:24 -0700)]
(wangle) Fix a couple compilation issues

Summary: If you try to include files in a weird (but not incorrect) order you get compilation errors. No longer!

Test Plan: new dummy `cpp_binary` target

Reviewed By:

Subscribers: trunkagent, folly-diffs@, net-systems@, fugalh, exa, njormrod

FB internal diff: D1621083

7 years ago(wangle) Interrupts (and therefore, cancellation)
Hans Fugal [Tue, 21 Oct 2014 17:24:10 +0000 (10:24 -0700)]
(wangle) Interrupts (and therefore, cancellation)

Modeled very closely after Finagle's interrupts. Compare with if you like.
The basic idea is the promise holder can register an interrupt handler, and then interrupts will call that handler. A typical handler would fulfil the promise with an exception (or special value) indicating that it was interrupted (if it was interrupted in time).
Raising an interrupt does not prevent setting a value or callbacks executing or any of that - it is only advisory to the promise holder.

Test Plan: I wrote some unit tests.

Reviewed By:

Subscribers: folly-diffs@, net-systems@, fugalh, exa, hannesr, njormrod

FB internal diff: D1620805

Tasks: 4618297

7 years ago(wangle) express current Core functionality with a state machine
Hans Fugal [Tue, 21 Oct 2014 17:24:06 +0000 (10:24 -0700)]
(wangle) express current Core functionality with a state machine

This is a refactor of the current functionality to use a state machine (inheriting from `FSM`). This will make it easier to extend to do cancellation and Future collapsing. Performance is the same, maybe slightly faster (abt 1%). (I might be a little conservative with the atomics, it might be worth going through and reasoning about whether they all need to be atomic).

The state machine is two states, Waiting (no value), and Done (has a value). Transitioning to Done will execute the callback if it exists and we are active. Otherwise the callback will happen when it is set and active is true.

There is a subjective balancing act in place here, between making a state for every single mutable bit of information (which results in an explosion of states: hasValue X hasCallback X isActive X isCancelled …), and finding a sweet spot of expressivity. This isn't too far from the way Twitter did it, though we don't (yet) have all the states they have (and they don't have the concept of hot/cold futures). But I got there by way of replacing the `mutex_` with the state, after changing all those variables to atomics so they didn't need mutex protection (resulting in only `callback_` and `result_` needing it). I expect the state machine will morph as the rest of the functionality is added, but hopefully it will be easier to understand and keep correct (that's the idea anyway).

Test Plan: Tests still pass (and not by accident, I made several awesome mistakes along the way).

Reviewed By:

Subscribers: net-systems@, fugalh, exa, njormrod

FB internal diff: D1618240

Tasks: 4618297

7 years ago(wangle) Core cleanup
Hans Fugal [Tue, 21 Oct 2014 17:24:03 +0000 (10:24 -0700)]
(wangle) Core cleanup

Just some cleaning hose in `detail::Core`.
- rename `fulfil` to `setResult`
- and `value_` to `result_`
- remove unnecessary helper methods (I can be convinced to keep them if you like them)

Test Plan: mechanical changes, tests still pass

Reviewed By:

Subscribers: trunkagent, folly-diffs@, net-systems@, fugalh, exa, njormrod

FB internal diff: D1618161

7 years ago(wangle) deprecate Later and ThreadGate
Hans Fugal [Mon, 20 Oct 2014 19:11:19 +0000 (12:11 -0700)]
(wangle) deprecate Later and ThreadGate

Later is now superfluous. Just hold a deactivated future (e.g. one returned by `Future::via`), and then activate (or destruct) it when you're ready.
ThreadGate has proven to be too rigid and too heavyweight to be really useful. And is superfluous now with `Future::via`. e.g.

This is stage one of removing these: mark them deprecated.
Stage two will be for me to change all existing usages in our code to not use them.
Stage three will be to rip these out altogether.

Test Plan:
Everything still builds and works, but now with deprecation warnings.

fbconfig -r \
common/smc/cpp/tests \
folly/wangle \
langtech/audiens \
messaging/avscan \
messaging/maelstrom \
neteng/traffic_manager \
notifications/nudge/service \
tao/client \
fbmake runtests

Reviewed By:

Subscribers: net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1626412

Tasks: 5409538

7 years agoremove fireAndForget
Hans Fugal [Mon, 20 Oct 2014 19:10:53 +0000 (12:10 -0700)]
remove fireAndForget

Summary: It was deprecated, now is unused.

Test Plan:
check contbuild

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod, folly-diffs@

FB internal diff: D1624221

7 years ago(wangle) dummy cpp files
Hans Fugal [Mon, 20 Oct 2014 19:10:34 +0000 (12:10 -0700)]
(wangle) dummy cpp files

Summary: for great contbuild

Test Plan: these projects still build. contbuild should start picking them up, based on discussion with @robbert

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, mwa, jgehring, fuegen, njormrod, folly-diffs@, robbert

FB internal diff: D1624206

Tasks: 5399744

7 years ago(FSM) updateState with unprotected action
Hans Fugal [Mon, 20 Oct 2014 19:07:36 +0000 (12:07 -0700)]
(FSM) updateState with unprotected action

Like the magic macros. As it says in the comment, this can lead to nicer code.
Also added/moved a couple examples to the top of `test/FSM.cpp`.

Test Plan: new tests

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod

FB internal diff: D1618184

7 years agosync wangle to makefile
James Sedgwick [Mon, 20 Oct 2014 18:21:35 +0000 (11:21 -0700)]
sync wangle to makefile

Summary: as above

Test Plan:

Reviewed By:

Subscribers: fugalh, njormrod

FB internal diff: D1589439

7 years agoSupport NotificationQueue on iOS/Android
Sean Cannella [Fri, 17 Oct 2014 20:38:30 +0000 (13:38 -0700)]
Support NotificationQueue on iOS/Android

This is the last of the changes to make folly and thrift portable. The old
thrift SpinLock implementation is exactly we
want here so port this into folly. If someone later makes efficeint
versions of MicroSpinLock for platforms other than x64 they can easily
be plugged in without changing the calling code.

Test Plan: compiled on multiple platforms

Reviewed By:

Subscribers: folly-diffs@, ldbrandy, trunkagent, alandau, bmatheny, njormrod, mshneer, shikong, benyluo, kmdent, fma, ranjeeth, subodh

FB internal diff: D1621717

Tasks: 5183325

7 years agoBump version to 13:0
dcsommer [Fri, 17 Oct 2014 18:46:07 +0000 (11:46 -0700)]
Bump version to 13:0

7 years agoFix mcrouter opensource build
Pavlo Kushnir [Fri, 17 Oct 2014 18:04:46 +0000 (11:04 -0700)]
Fix mcrouter opensource build

Summary: mcrouter now uses folly::Singleton.

Test Plan: visual

Reviewed By:

Subscribers: njormrod

FB internal diff: D1622647

7 years agoFix typo in folly::Optional
Max Wang [Thu, 16 Oct 2014 21:50:48 +0000 (14:50 -0700)]
Fix typo in folly::Optional

Test Plan: yes

Reviewed By:

7 years agoFix one missing conditional for Android
Sean Cannella [Thu, 16 Oct 2014 19:16:41 +0000 (12:16 -0700)]
Fix one missing conditional for Android

pthread_atfork isn't defined in the Android NDK (and therefore is not actually supported even though it's in bionic/libc) so don't call it there.

Test Plan: compiled on Linux, Android

Reviewed By:

Subscribers: njormrod, shikong, kmdent, fma

FB internal diff: D1620584

Tasks: 5183325

7 years agoFixing broken bits test
Tom Jackson [Thu, 16 Oct 2014 18:19:00 +0000 (11:19 -0700)]
Fixing broken bits test

Summary: Buffer size calculation had fallen out of sync with the actual test code. Good thing we have ASAN!

Test Plan: Run the test with address santizer.

Reviewed By:

Subscribers: mpawlowski, trunkagent, njormrod

FB internal diff: D1619472

Tasks: 5325399

7 years agoAllow Optional<T> {=,!}= T comparisons in folly::Optional
Max Wang [Wed, 15 Oct 2014 20:00:18 +0000 (13:00 -0700)]
Allow Optional<T> {=,!}= T comparisons in folly::Optional

I can see how {<,>}{,=} comparisons might be ambiguous, but equality
should never be, so let's allow it.

Test Plan: tests

Reviewed By:

Subscribers: trunkagent, njormrod

FB internal diff: D1618088

7 years agomove shutdown socket set
Dave Watson [Fri, 3 Oct 2014 19:05:37 +0000 (12:05 -0700)]
move shutdown socket set

Move shutdownsocketset to folly, since it is a dep of the asyncsockets

Previoulsy tried moving it in to the server directly: D1583629, but had issues - close(fd) is called before the error callback, so we can't remove the fd before the close, which is essential to it working properly.

Just move it to folly instead.

Test Plan: fbconfig -r thrift/lib/cpp thrift/lib/cpp2; fbmake runtests

Reviewed By:

Subscribers: mshneer, trunkagent, fugalh, jsedgwick, doug, alandau, bmatheny, njormrod

FB internal diff: D1594950

7 years agoConditionals for iOS / Android compilation
Sean Cannella [Thu, 16 Oct 2014 17:09:06 +0000 (10:09 -0700)]
Conditionals for iOS / Android compilation

Conditional changes required to compile on iOS and Android.

Test Plan: compiled on all platforms

Reviewed By:

Subscribers: kmccray, trunkagent, shilin, njormrod, ranjeeth, subodh, bmatheny

FB internal diff: D1618028

Tasks: 5183325

7 years agoDon't throw in Singleton::get() if singleton is alive
Andrii Grynenko [Thu, 16 Oct 2014 00:40:15 +0000 (17:40 -0700)]
Don't throw in Singleton::get() if singleton is alive

This is a fix for situations where you know that something is keeping singleton alive (by getting a weak_ptr and locking it), and request a singleton instance via get() method (if e.g. it's hard to pass a pointer to singleton instance from a method which locked it). If shutdown was scheduled - an exception was previously thrown even though the object was not destroyed yet.

This simplifies conversion of legacy code to folly::Singleton.

Test Plan:
fbconfig -r gatehouse/usersets/tests
fbmake runtests

Reviewed By:

Subscribers: trunkagent, njormrod

FB internal diff: D1619311

7 years agoRemove global init of ThreadLocal in Random
Subodh Iyengar [Wed, 15 Oct 2014 23:51:46 +0000 (16:51 -0700)]
Remove global init of ThreadLocal in Random

Remove global initialization of ThreadLocal in Random

Static initializers and thread locals don't mix well, and we're seeing
some crashes which we think might be related to this.

Test Plan: Unit tests

Reviewed By:

Subscribers: trunkagent, seanc, njormrod

FB internal diff: D1617455

7 years agosupport for move-only types
James Sedgwick [Wed, 15 Oct 2014 22:58:44 +0000 (15:58 -0700)]
support for move-only types

Summary: universal references and perfect forwarding to the rescue

Test Plan: added demonstrative unit tests

Reviewed By:

Subscribers: trunkagent, fugalh, njormrod, bmatheny

FB internal diff: D1592032

Tasks: 5283342

7 years agomerge wangle/Executor.h and experimental/wangle/concurrent/Executor.h
James Sedgwick [Wed, 15 Oct 2014 22:57:47 +0000 (15:57 -0700)]
merge wangle/Executor.h and experimental/wangle/concurrent/Executor.h

the one in concurrent/ is a bit more generic, so I kept that as Executor and renamed the existing one ScheduledExecutor

because Hans is surfing I took the liberty of renaming Action->Func as an alias for std::function<void()>, because I think it's more reflective
also kept the version of add() that doesn't force rvalue-reference as it's more user friendly and probably not less performant in common cases (insert reference to "want speed? pass by value" here)

Test Plan: compiled some major relevant bits, will let contbuild show me anything I missed

Reviewed By:

Subscribers: trunkagent, rushix, fbcode-common-diffs@, fugalh, msk, njormrod

FB internal diff: D1591237

Tasks: 5279196

7 years agoSchedule destroyInstances only when first singleton is created
Andrii Grynenko [Tue, 14 Oct 2014 03:08:15 +0000 (20:08 -0700)]
Schedule destroyInstances only when first singleton is created

Right now destroyInstances is scheduled when SingletonVault is requested. This may change singleton destruction order (folly::Singleton-managed vs unmanaged singletons) when new singleton is registered with folly::Singleton (no matter if it is even used).
This diff changes it to be more stable.

Test Plan: servicerouter unittests

Reviewed By:

Subscribers: njormrod, mshneer

FB internal diff: D1615213

Tasks: 5353022

7 years agoFix clang test compilation failures
Sean Cannella [Wed, 15 Oct 2014 21:30:46 +0000 (14:30 -0700)]
Fix clang test compilation failures

While working on other code I noticed folly stopped compiling
with clang. Fix that.

Test Plan: fbmake runtests (clang and gcc)

Reviewed By:

Subscribers: mathieubaudet, njormrod, bmatheny

FB internal diff: D1618189

7 years ago(wangle) Use MicroSpinLock
Hans Fugal [Wed, 15 Oct 2014 21:05:59 +0000 (14:05 -0700)]
(wangle) Use MicroSpinLock

Summary: For great speed

Test Plan:
folly/wangle/test/Benchmark.cpp                 relative  time/iter  iters/s
constantFuture                                             241.19ns    4.15M
promiseAndFuture                                 100.32%   240.42ns    4.16M
withThen                                          44.63%   540.47ns    1.85M
oneThen                                                    519.20ns    1.93M
twoThens                                          62.83%   826.41ns    1.21M
fourThens                                         36.80%     1.41us  708.75K
hundredThens                                       1.79%    29.05us   34.42K
no_contention                                                4.82ms   207.27
contention                                        62.91%     7.67ms   130.39

folly/wangle/test/Benchmark.cpp                 relative  time/iter  iters/s
constantFuture                                             159.79ns    6.26M
promiseAndFuture                                 101.23%   157.84ns    6.34M
withThen                                          41.78%   382.47ns    2.61M
oneThen                                                    358.23ns    2.79M
twoThens                                          63.07%   568.00ns    1.76M
fourThens                                         36.89%   971.07ns    1.03M
hundredThens                                       1.76%    20.34us   49.17K
no_contention                                                3.75ms   266.75
contention                                        59.83%     6.27ms   159.59

That's a 150% speedup.

Reviewed By:

Subscribers: net-systems@, fugalh, exa, njormrod

FB internal diff: D1617363

Tasks: 5278220

7 years agoDefer query string parsing from ctor to getQueryParams()
Jun LI [Wed, 15 Oct 2014 20:53:51 +0000 (13:53 -0700)]
Defer query string parsing from ctor to getQueryParams()

Query string parsing uses a lot of CPU, it happens in ctor killing CPU
even in users that are not interested in query string. Move it to

Test Plan:
fbconfig folly/test
fbmake runtests_dbg

Reviewed By:

Subscribers: trunkagent, njormrod, zellux

FB internal diff: D1604973

Tasks: 5304484

Blame Revision:

7 years agoFix opensource build
Pavlo Kushnir [Wed, 15 Oct 2014 18:46:22 +0000 (11:46 -0700)]
Fix opensource build

Summary: D1615707 moved State.h to Core.h, but didn't update

Test Plan: visual

Reviewed By:

Subscribers: fugalh, njormrod

FB internal diff: D1617562

Blame Revision: D1615707

7 years agoFix thread local random number generator static issue
Daniel Sommermann [Tue, 14 Oct 2014 23:39:00 +0000 (16:39 -0700)]
Fix thread local random number generator static issue

Thanks to @subodh for finding the first instance of this problem in the
proxygen codebase.

Test Plan:
unit tests, tested proxygen unit tests on my laptop and I don't
get segfaults on program exit now.

Reviewed By:

Subscribers: trunkagent, doug, njormrod, subodh

FB internal diff: D1616149

Tasks: 5217022

7 years agowangle::detail::FSM
Hans Fugal [Wed, 15 Oct 2014 16:32:03 +0000 (09:32 -0700)]


Finagle uses a nifty state machine pattern in `Promise.scala`. Each state carries its data with it, then you switch statements with a simple state `cas`, within the functions (which are the transition inputs). See and for example.

I was thinking we could do something similar in C++, though we don't quite have the same luxury of type cases like scala. (Although as an aside I found this really cool paper on implementing type cases in C++, it's kind of evil but very cool.

I was looking at having a union of the different state classes, e.g.

union U {
State base;
Done done;
Waiting waiting;

and then you could use the memoized `dynamic_cast` trick in that paper to make a fairly efficient type case (and fairly readable, depending on how evil you want to get with the macros). However, `dynamic_cast<Done*>(&u.base)` blissfully segfaults. I'm not sure if this is something that should work and it's a compiler bug, or whether trying to (ab)use a union this way is against some arcane C++ rules. @hannesr suggested maybe a variant type might work. We can also do memory mangling on our own if it comes to it - however those are all more advanced techniques to play with at a later optimization time.

So instead, I went for a this-is-context helper base class. The mutable context is just `this`, and you inherit from `FSM`.

The magic macros make it more succint to lay out state transitions. See the tests for examples. Maybe in the future we can get really clever and find a way to generate state machine diagrams from code using this, especially when macros are being used.

Test Plan: unit tests were written and pass

Reviewed By:

Subscribers: meisner, trunkagent, net-systems@, fugalh, exa, njormrod, hannesr

FB internal diff: D1613497

7 years ago(wangle) s/State/Core/
Hans Fugal [Wed, 15 Oct 2014 16:31:52 +0000 (09:31 -0700)]
(wangle) s/State/Core/

`State` is such an overloaded term, and not really the best to describe this backing future/promise object. Yes, it holds the state but it's more than that and it gets in the way of calling the states of the state machines `State`s.

Test Plan: builds and tests pass

Reviewed By:

Subscribers: trunkagent, net-systems@, fugalh, exa, njormrod

FB internal diff: D1615707

7 years agoBump version to 12:0
Andrii Grynenko [Wed, 15 Oct 2014 01:00:43 +0000 (18:00 -0700)]
Bump version to 12:0

7 years agoDon't re-create singletons
Andrii Grynenko [Wed, 8 Oct 2014 01:12:33 +0000 (18:12 -0700)]
Don't re-create singletons

This forbids singleton creation/re-creation after destroyInstances() was called.
Only after reset() is called singletons can be created again.

registrationComplete() behavior is also slightly change. We disallow singleton registration after registrationComplete() is called even in Relaxed mode. Strict mode now only controlls whether singletons can be constructed before registation is complete.

Test Plan: unit test

Reviewed By:

Subscribers: hphp-diffs@, ps, trunkagent, njormrod, mshneer, lins

FB internal diff: D1605136

7 years agoBasic tests for EventHandler.
Yedidya Feldblum [Fri, 10 Oct 2014 19:44:26 +0000 (12:44 -0700)]
Basic tests for EventHandler.

[Folly] Basic tests for EventHandler.

Worthwhile adding some tests just for this class since it's rather fundamental. It is already tested indirectly via any number of things that make use of EventBase under the hood, but it's worth having a place for more direct tests.

Test Plan: Unit tests in `folly/io/async/test/EventHandlerTest.cpp`.

Reviewed By:

Subscribers: njormrod, chalfant, dougw

FB internal diff: D1606284

7 years agoBetter estimateSpaceNeeded(double)
Soren Lassen [Fri, 10 Oct 2014 14:12:31 +0000 (07:12 -0700)]
Better estimateSpaceNeeded(double)

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

Reviewed By:

Subscribers: njormrod

FB internal diff: D1604761

Blame Revision: D1420588