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: meyering@fb.com

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

FB internal diff: D1660903

Signature: t1:1660903:1415215895:7f1f96ab7416b45c4dc18e78173fb59eb27bd03b

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

T1            T2
--            --
++ctx->count
++ctx->count
ctx->total
==
setValue
delete ctx
ctx->total
==
setValue
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: darshan@fb.com

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

FB internal diff: D1660663

Tasks: 5506504

Signature: t1:1660663:1415207632:49dc224363cec27736fc71fb211fa846be9e170f

Blame Revision: D1636487

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: sdoroshenko@fb.com, robot9@fb.com

Subscribers: fugalh, njormrod, folly-diffs@

FB internal diff: D1659558

Tasks: 5538418

Signature: t1:1659558:1415147560:2c5b0c996893854c3ea9d0ad02b006bcc5960ffa

Summary:
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
files.

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: davejwatson@fb.com

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

FB internal diff: D1648489

Tasks: 5486739

Signature: t1:1648489:1415035522:3d8bd9611eb7c7117b70d5e7f68de5768639a727

Summary:
Statics in headers are bad.

Test Plan: run unit tests

Reviewed By: robbert@fb.com

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

FB internal diff: D1646906

Tasks: 5486739

Signature: t1:1646906:1415042805:dc4d1cec54e9320f1e609808a73622c731a4cdc9

Summary:
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: davejwatson@fb.com

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

FB internal diff: D1649686

Tasks: 5486739

Signature: t1:1649686:1415035288:18efd2cf9aae8caab66d8303c22cdc26c6b54ae5

Summary:
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: davejwatson@fb.com

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

FB internal diff: D1604575

Tasks: 5002299

Signature: t1:1604575:1415034767:bc3b12fae726890aa6a55ed391286917ae23e56e

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

Test Plan: unit tests, review

Reviewed By: meyering@fb.com

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

FB internal diff: D1627280

Signature: t1:1627280:1414792755:004f5a737ece1e93bcf4331718a98afc57e4f80c

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

Test Plan:
unit, that's it :/

Reviewed By: davejwatson@fb.com

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

FB internal diff: D1585087

Summary: Put back previous liger diffs

Test Plan: It builds

Reviewed By: seanc@fb.com

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

FB internal diff: D1652754

Signature: t1:1652754:1414785984:df0fc7bf59dc2e89defd2c1a4ffe3b288238ba58

Summary:
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: robbert@fb.com

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644151

Tasks: 5486739

Signature: t1:1644151:1414779408:76bcec1d14daaa8ed071c715bf26b108c8fe4b87

Summary:
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: lucian@fb.com

Subscribers: trunkagent, njormrod, folly-diffs@

FB internal diff: D1639801

Tasks: 5464222

Signature: t1:1639801:1414784601:2f59d5a58e434f4cf9df5b25b917c5094c8b133f

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

Test Plan: build it.

Reviewed By: lakshmiganesh@fb.com

Subscribers: njormrod, folly-diffs@

FB internal diff: D1652341

Signature: t1:1652341:1414779013:9f363acba95fbac1988200081659edf71ac63eec

Summary:
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: andrei.alexandrescu@fb.com

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

FB internal diff: D1644493

Tasks: 5486739

Signature: t1:1644493:1414715317:491d0f631d8152a5b7ec66237e00c404530ab590

Summary: add missing ssl dep

Test Plan: build it

Reviewed By: dcsommer@fb.com

Subscribers: doug, njormrod, folly-diffs@

FB internal diff: D1652234

Signature: t1:1652234:1414778507:4504ea06c3076e0cdfa7772166cf48bc564c9076

Summary:
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: je@fb.com

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

FB internal diff: D1638370

Signature: t1:1638370:1414703429:f3ed2cc8dcb183ff4e8367e2c1f884e122605cba

Summary:
Better logging in SocketAddress

Test Plan: run it

Reviewed By: qwu@fb.com

Subscribers: ps, bmatheny, xning

FB internal diff: D1644573

Tasks: 5406091

Signature: t1:1644573:1414626398:1e23ee85ec8ee0be89cb6bdc07da23978b4d988a

Summary:
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
correctly

Test Plan: run unit tests

Reviewed By: robbert@fb.com

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

FB internal diff: D1650435

Tasks: 5486739

Signature: t1:1650435:1414713329:7844e7802ebabcac7ef21aef0becf71ba513afb0

Summary:
Avoid comparing-the-same-tokens error by parenthesizing one
side.

This diff only touches folly tests

Test Plan: run unit tests

Reviewed By: robbert@fb.com

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1650386

Tasks: 5486739

Signature: t1:1650386:1414713837:cbb578d23f759aa049f4246113d9da6d40504da4

Summary: Switch to noexcept

Test Plan: run unit tests

Reviewed By: robbert@fb.com, andrei.alexandrescu@fb.com

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

FB internal diff: D1644135

Tasks: 5486739

Signature: t1:1644135:1414715537:22e1baf91ab7e3250b0b2a460a12d56783f2baed

Summary:
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: robbert@fb.com

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

FB internal diff: D1644130

Tasks: 5486739

Signature: t1:1644130:1414715392:b6c783851aa030ad1148f84a98139a5dca207da0

Summary:
Move SSLContext to folly.

Test Plan: It builds.

Reviewed By: dcsommer@fb.com

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

Summary:
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: haijunz@fb.com, hans@fb.com

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

FB internal diff: D1650843

Tasks: 4698780

Signature: t1:1650843:1414711295:c7439733680ab4903eb9c05dcf8bf52100bf3f07

Summary: Looks like a bad git merge

Test Plan: builds

Reviewed By: hannesr@fb.com

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

FB internal diff: D1649773

Tasks: 5501131

Signature: t1:1649773:1414700920:cc691283884e9654803bba3ffe78f4553194752d

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

Test Plan: run unit tests

Reviewed By: andrei.alexandrescu@fb.com

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644176

Tasks: 5486739

Signature: t1:1644176:1414693471:e3a08e27ba27a5ae0e4b9a5e009acf0a16728c45

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

Test Plan: run unit tests

Reviewed By: delong.j@fb.com

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

FB internal diff: D1644380

Tasks: 5486739

Signature: t1:1644380:1414618757:fb910d8a3fe3e634da4215c432577edf7371be61

Summary:
Fixed unsigned/signed compares.

I had to suppress the warning in one place, but don't worry, clang
supports `#pragma GCC diagnostic ...` (see http://clang.llvm.org/docs/UsersManual.html#controlling-diagnostics-via-pragmas)

Test Plan: unit tests

Reviewed By: afrind@fb.com

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

FB internal diff: D1624249

Tasks: 5140804

Signature: t1:1624249:1414620209:5399e8d90d8ca32b30794a7b2a4a7c2d7d437dda

Summary:
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: andrei.alexandrescu@fb.com

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

FB internal diff: D1644182

Tasks: 5486739

Signature: t1:1644182:1414616294:4b6004d358cba865455366f5a943644b3542ae51

Summary:
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: andrei.alexandrescu@fb.com

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

FB internal diff: D1644170

Tasks: 5486739

Signature: t1:1644170:1414616595:7ac52f474c1312a0c28e89255b1151d56c680acf

Summary:
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: andrei.alexandrescu@fb.com

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

FB internal diff: D1644165

Tasks: 5486739

Signature: t1:1644165:1414616664:a6b9dc34660df84b33ed8faaf48ec048a02bad01

Summary:
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: delong.j@fb.com

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

FB internal diff: D1644296

Tasks: 5486739

Signature: t1:1644296:1414618605:d9a0db5193c82650b96e9c62b019d5da218b15c5

Summary:
variants of then() that bypass Try and forward any exceptions up to the next future. like next() from http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3865.pdf

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: hans@fb.com

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

FB internal diff: D1641776

Signature: t1:1641776:1414610434:c742563b8061a748fca9292fc2765081edcf9d52

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

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

Reviewed By: meyering@fb.com

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

FB internal diff: D1646519

Signature: t1:1646519:1414602292:cfb908ae094caaef02e64eb2c42544fd34fa1389

Summary: @override-unit-failures

Test Plan: n/a

Reviewed By: rhysparry@fb.com

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644139

Tasks: 5486739

Signature: t1:1644139:1414553949:626fdca07a583c3d5017abe19408a5e6f6d7674a

Summary:
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: robbert@fb.com

Subscribers: sdwilsh, njormrod, folly-diffs@

FB internal diff: D1644144

Tasks: 5486739

Signature: t1:1644144:1414599280:1e403ea3675d2173b249c4ce6db0dad621a139bc

Summary:
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
errors

Reviewed By: andrei.alexandrescu@fb.com

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

FB internal diff: D1644141

Tasks: 5486739

Signature: t1:1644141:1414607351:29740da758b75187f4f7b6b5a5ad28a523e7080d

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

Test Plan:
fbconfig -r folly && fbmake dbg

Reviewed By: hans@fb.com

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

FB internal diff: D1644190

Tasks: 5486739

Signature: t1:1644190:1414532704:87d5851f363cf37c58f7e7c3d57c8210a792d4a1

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

Test Plan: git diff

Reviewed By: jsedgwick@fb.com

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

FB internal diff: D1644550

Tasks: 4412111

Signature: t1:1644550:1414604128:8184e7ce1f3e417c170ef67346d553caecc1f013

Summary:

Test Plan:

Reviewed By: dcsommer@fb.com

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

FB internal diff: D1644071

Signature: t1:1644071:1414526899:c41dd55e2957a7e1fcc54508e20cdb4a9c8c30d4