folly.git
10 years agofolly/IndexedMemPool: correct an erroneous test for failed mmap
Jim Meyering [Tue, 5 Aug 2014 18:45:25 +0000 (11:45 -0700)]
folly/IndexedMemPool: correct an erroneous test for failed mmap

Summary:
* folly/IndexedMemPool.h (IndexedMemPool): Correct the test for
failed mmap.  Upon failure, it returns MAP_FAILED, not nullptr.

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

Reviewed By: ngbronson@fb.com

FB internal diff: D1480389

Tasks: 4846893
@override-unit-failures

10 years agosorted_vector containers have a shrink_to_fit() method
Marc Celani [Tue, 5 Aug 2014 13:04:24 +0000 (06:04 -0700)]
sorted_vector containers have a shrink_to_fit() method

Summary: Adds a shrink_to_fit() method to sorted vector types.

Test Plan: unit test

Reviewed By: mshneer@fb.com

FB internal diff: D1477864

10 years agoDon't check for thread running in terminateLoopSoon()
Marc Horowitz [Fri, 11 Jul 2014 23:48:20 +0000 (16:48 -0700)]
Don't check for thread running in terminateLoopSoon()

Summary:
Consider this use case:
thread 1:
make an evb
create and start thread 2
do some work
evb->terminateLoopSoon()
join thread 2 with a timeout
thread 2:
do some initialization
evb->loop()
Normally, this all works fine.  However, if the work thread 1 has to do is sufficiently small, and no external input occurs (one of the thing the evb is doing is listening on a socket), then sometimes, terminateLoopSoon() happens before loop() is called (or at least, before loopThread_.store() happens). isRunning() in terminateLoopSoon() is false, and so stop_ is never set, and event_base_loopbreak() is never called. The join times out, and thread 2 is still running.  Removing the isRunning() check gives the desired behavior.

Test Plan:
In my ad hoc tests, this fix caused my threads to exit when
I wanted them to exit in a situation like the one above.

Reviewed By: andrewcox@fb.com

FB internal diff: D1455885

Tasks: 2057547

10 years agorfc: move folly::json::ParseError type to folly/json.h
Ranjeeth Dasineni [Fri, 1 Aug 2014 03:01:24 +0000 (20:01 -0700)]
rfc: move folly::json::ParseError type to folly/json.h

Summary:
I don't know why this was in the cpp but seems pretty safe to move it to the .h
There are a few files that directly include json.cpp to get around this: https://fburl.com/30544625.
I need it for something else and if you dont see a problem I want to move it.

Test Plan: unittests

Reviewed By: pgriess@fb.com

Subscribers: simpkins, doug

FB internal diff: D1471958

Tasks: 3623725

10 years agoEliasFanoReader::goTo()
Philip Pronin [Fri, 1 Aug 2014 20:13:03 +0000 (13:13 -0700)]
EliasFanoReader::goTo()

Summary: Random lookup support.

Test Plan: fbconfig -r folly/experimental/test:eliasfano_test && fbmake runtests_opt -j32

@override-unit-failures

Reviewed By: soren@fb.com

FB internal diff: D1473244

Tasks: 4536072

10 years agoAPI for getTimeoutManager
Hitesh Khandelwal [Thu, 31 Jul 2014 21:47:46 +0000 (14:47 -0700)]
API for getTimeoutManager

Test Plan: Tested end-to-end with D1342452

Reviewed By: mshneer@fb.com

Subscribers: alandau, bmatheny

FB internal diff: D1470814

Tasks: 2682011

10 years agoinstall delayeddestruction header
Dave Watson [Thu, 31 Jul 2014 16:36:37 +0000 (09:36 -0700)]
install delayeddestruction header

Summary: added recently, needs to be installed

Test Plan: watch fbthrift contbuild

Reviewed By: noamz@fb.com

Subscribers: doug

FB internal diff: D1469760

Blame Revision: D1453095

10 years agoConstruct Later with exception_ptr
Rushi Desai [Thu, 31 Jul 2014 04:08:51 +0000 (21:08 -0700)]
Construct Later with exception_ptr

Summary:
It would be nice to be able to create a Later with pre-loaded
exception.

Test Plan: Unit test

Reviewed By: hans@fb.com

Subscribers: fugalh

FB internal diff: D1462810

10 years agoMove TDelayedDestruction to folly::DelayedDestruction
Alan Frindell [Wed, 23 Jul 2014 00:39:03 +0000 (17:39 -0700)]
Move TDelayedDestruction to folly::DelayedDestruction

Summary: This doesn't have many thrift dependencies and I need it for another async class I'm adding to folly.

Test Plan: unit tests

Reviewed By: alandau@fb.com

Subscribers: doug, mcduff, marccelani, mshneer, alandau, bmatheny

FB internal diff: D1453095

10 years agoIPAddress::empty()
Marc Celani [Fri, 25 Jul 2014 22:59:12 +0000 (15:59 -0700)]
IPAddress::empty()

Summary: Checks if IPAddress has not been initialized.

Test Plan: unit test

Reviewed By: tudorb@fb.com

Subscribers: tudorb

FB internal diff: D1459525

10 years agoBump version to 2:0
Chip Turner [Fri, 25 Jul 2014 16:07:33 +0000 (09:07 -0700)]
Bump version to 2:0

10 years agoTry again: folly::Singleton, a class for managing singletons
Chip Turner [Fri, 23 May 2014 03:17:14 +0000 (20:17 -0700)]
Try again: folly::Singleton, a class for managing singletons

Summary:
Singletons are surprisingly tricky in a codebase where
libraries depend on one another.  folly::Singleton hopes to make this
process more reliable by ensuring object creation happens in a safe
order, that destruction is possible, and that singletons are created
on-demand.

The basic fbcode use intention is to invoke registration completion in
initFacebook, so users need only declare singletons via
Singleton<ClassName> in their .cpp files.

This diff ties the Singletons into the core Init process, but not hhvm
(which will be a separate diff).

Test Plan: runtests

Reviewed By: joelm@fb.com, hans@fb.com

Subscribers: fbcode-common-diffs@, hphp-diffs@, soren, anca, lins, aalexandre, ps, trunkagent, lucian, hannesr, yfeldblum, maxwellsayles

FB internal diff: D1453135

10 years agoAdd method to parse parameter list in query string to folly::Uri
Jun LI [Fri, 25 Jul 2014 00:07:36 +0000 (17:07 -0700)]
Add method to parse parameter list in query string to folly::Uri

Summary:
Add a method to folly::Uri to get parsed query string

e.g. http://localhost?key1=foo&key2=bar

We can get key value map containing:
"key1" => "foo"
"key2" => "bar"

Test Plan:
fbconfig folly/test
fbmake runtests_dbg

Reviewed By: tudorb@fb.com

Subscribers: wormhole-dev@

FB internal diff: D1455158

Tasks: 4768038

10 years agoAdding one more test for trailing dots
Tom Jackson [Thu, 24 Jul 2014 17:37:15 +0000 (10:37 -0700)]
Adding one more test for trailing dots

Summary: Sorry, forgot to include tests for the other side of the equation.

Test Plan: It is a test

Reviewed By: tudorb@fb.com

FB internal diff: D1455583

10 years ago(wangle) set* should not invalidate getFuture
Hans Fugal [Wed, 23 Jul 2014 22:12:19 +0000 (15:12 -0700)]
(wangle) set* should not invalidate getFuture

Summary: @jcoens observed in D1451114 that it was unintuitive that you have to retrieve the `Future` before fulfilling the `Promise`. That seemed wrong to me too, but sure enough when I wrote the unit tests that doesn't work (throws "promise already fulfilled" when you call `getFuture`). I think this is just a simple mistake, but I'm going to carefully look at the output of contbuild test suites before committing.

Test Plan:
red-green
careful dependency unit test inspection

Reviewed By: jon.coens@fb.com

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

FB internal diff: D1453780

10 years agoVersioning for Thrift, remove thrift/lib/cpp/config.h
Tudor Bosman [Tue, 22 Jul 2014 03:33:17 +0000 (20:33 -0700)]
Versioning for Thrift, remove thrift/lib/cpp/config.h

Summary:
Also, regenerate thrift_config.h.

Move .gitignore to public_tld so I can add thrift_config.h while still
having it checked into fbcode.

Test Plan: fbconfig -r thrift && fbmake runtests_opt

Reviewed By: meyering@fb.com

Subscribers: alandau, bmatheny, jhj, kma, lesha

FB internal diff: D1449068

10 years agoSupport trailing decimals for floats
Tom Jackson [Thu, 24 Jul 2014 00:19:47 +0000 (17:19 -0700)]
Support trailing decimals for floats

Summary: So integer-like numbers can be formatted in a manner which disambiguates them from integers.

Test Plan: Unit tests

Reviewed By: tudorb@fb.com

Subscribers: jfh, cscheau

FB internal diff: D1454446

10 years agoMake ElfFile not crash on invalid ELF files
Tudor Bosman [Wed, 23 Jul 2014 22:16:32 +0000 (15:16 -0700)]
Make ElfFile not crash on invalid ELF files

Summary:
@rkroll wants to run this on more than just fbcode binaries; there are x86 (not
x86_64) binaries in there and probably a lot of other junk. So, if you call
openNoThrow explicitly, you get a pretty error message in this case.

Test Plan: folly/experimental/symbolizer/test

Reviewed By: lucian@fb.com

Subscribers: rkroll, jhj, lesha, kma

FB internal diff: D1453758

10 years agoEnforce that only one version of folly is loaded at the same time
Tudor Bosman [Mon, 21 Jul 2014 22:18:02 +0000 (15:18 -0700)]
Enforce that only one version of folly is loaded at the same time

Test Plan: folly/test, OSS build, check the macro in separate file

Reviewed By: dancol@fb.com

FB internal diff: D1448086

10 years agoFix a folly build failure under clang: ConvTest.cpp.
Yedidya Feldblum [Mon, 21 Jul 2014 23:37:40 +0000 (16:37 -0700)]
Fix a folly build failure under clang: ConvTest.cpp.

Summary: [Folly] Fix a folly build failure under clang: ConvTest.cpp.

Test Plan: Build the unit-test folly/test/ConvTest.cpp with clang.

Reviewed By: tudorb@fb.com, njormrod@fb.com

Subscribers: mathieubaudet, dougw

FB internal diff: D1446232

Tasks: 4723132

10 years agoFix a folly build failure under clang: MPMCQueueTest.cpp.
Yedidya Feldblum [Mon, 21 Jul 2014 21:31:09 +0000 (14:31 -0700)]
Fix a folly build failure under clang: MPMCQueueTest.cpp.

Summary:
[Folly] Fix a folly build failure under clang: MPMCQueueTest.cpp.

In clang-v3.4, there is a bug with the combination of a lambda expression inside a function template taking a template name (rather than a type name) as a template argument.

This diff, in the interest of building folly under clang-v3.4, extracts the lambda expression into a separate function so that the function template taking a template name as a template argument no longer has a lambda expression in it.

Test Plan: Build folly/test/MPMCQueueTest.cpp under clang.

Reviewed By: njormrod@fb.com

Subscribers: mathieubaudet, dougw

FB internal diff: D1446279

Tasks: 4723132

10 years agoFix a folly build failure under clang: FutexTest.cpp.
Yedidya Feldblum [Mon, 21 Jul 2014 21:10:43 +0000 (14:10 -0700)]
Fix a folly build failure under clang: FutexTest.cpp.

Summary:
[Folly] Fix a folly build failure under clang: FutexTest.cpp.

In clang-v3.4, there is a bug with the combination of a lambda expression inside a function template taking a template name (rather than a type name) as a template argument.

This diff, in the interest of building folly under clang-v3.4, extracts the lambda expression into a separate function so that the function template taking a template name as a template argument no longer has a lambda expression in it.

Test Plan: Build folly/test/FutexTest.cpp under clang.

Reviewed By: njormrod@fb.com

Subscribers: mathieubaudet, dougw

FB internal diff: D1446237

Tasks: 4723132

10 years agofix D1422343 ("make `folly::Formatter` extendible") for clang
Tyler MacDonald [Mon, 21 Jul 2014 21:08:47 +0000 (14:08 -0700)]
fix D1422343 ("make `folly::Formatter` extendible") for clang

Summary:
This is D1422343, but with a one-line change to make it clang-compatible. I authored this diff by first copying the original, then updating, so it's easy to see what I had to change.

> on advice of @tudorb, move most of `folly::Formatter` into `folly::BaseFormatter` so that we can use compile-time polymorphism to provide different types of Formatters.

Test Plan:
```
fbmake clean
fbconfig -r thrift folly cold_storage && fbmake dbg && fbmake runtests
fbmake clean
fbconfig -r hphp --clang && fbmake dbgo && fbmake runtests
```

Reviewed By: tudorb@fb.com, pt@fb.com

Subscribers: mathieubaudet, tudorb

FB internal diff: D1440310

Tasks: 46242684667712

10 years agoFix the unaligned string toLowerAscii test
Brian Pane [Mon, 21 Jul 2014 18:26:19 +0000 (11:26 -0700)]
Fix the unaligned string toLowerAscii test

Summary:
* The test computed nonaligned inputs but then copied them into temporary
buffers to compare different implementations. The temporary buffers
were word-aligned.
* This diff keeps the temp buffers but ensures that the data in them
keeps the original input's alignment.

Test Plan:
* Ran the test case with a modified String.cpp containing an assert to
catch unaligned reads.  The assert failed, as expected, on a copy
of the code without the fix from D1434585

Reviewed By: meyering@fb.com

Subscribers: ruibalp

FB internal diff: D1435028

Tasks: 4696800

10 years agoA generic line-reading callback for communicate()
Alexey Spiridonov [Fri, 25 Apr 2014 22:21:49 +0000 (15:21 -0700)]
A generic line-reading callback for communicate()

Summary: There are a couple of places where this behavior is useful, and it's not 100% trivial to implement it from scratch. Adding it to Folly to save people code & bugs.

Test Plan: unit tests

Reviewed By: tudorb@fb.com

Subscribers: tjackson, folly@lists, tudorb

FB internal diff: D1297506

10 years agoFactor out string stream re-splitting as StreamSplitter
Alexey Spiridonov [Thu, 8 May 2014 00:43:51 +0000 (17:43 -0700)]
Factor out string stream re-splitting as StreamSplitter

Summary: This way I can reuse it in Subprocess. It also makes it easy to make a bunch of other convenient tokenization routines (e.g. delimiter-preserving folly::gen tokenizers, file tokenizers, etc, etc).

Test Plan: fbconfig folly/gen/test && fbmake runtests

Reviewed By: tjackson@fb.com

Subscribers: vkatich, tjackson

FB internal diff: D1317973

10 years agoextend folly::split
Marcin Pawlowski [Fri, 18 Jul 2014 04:50:01 +0000 (21:50 -0700)]
extend folly::split

Summary:
see task:
https://our.intern.facebook.com/intern/tasks/?t=4723861

Test Plan: unit tests

Reviewed By: marcelo.juchem@fb.com

FB internal diff: D1443223

Tasks: 4723861

10 years agoBump version to 1:0
Tudor Bosman [Mon, 21 Jul 2014 19:22:08 +0000 (12:22 -0700)]
Bump version to 1:0

10 years agomake folly::Range literal type
Philip Pronin [Thu, 17 Jul 2014 20:05:02 +0000 (13:05 -0700)]
make folly::Range literal type

Summary:
7.1.5 [dcl.constexpr] / 9 (N3337) requires type to be literal to use it
in `constexpr` object declaration (seems to be not enforced by GCC 4.8?)

Currently `folly::Range<>` fails one of the requirements:

* is an aggregate type, or has at least one constexpr constructor or
constructor template that is not a copy or move constructor,

Test Plan: fbconfig folly/test:range_test && fbmake runtests_opt -j32

Reviewed By: lucian@fb.com

Subscribers: chaoyc, search-fbcode-diffs@, unicorn-diffs@

FB internal diff: D1441646

Tasks: 4720575

10 years agoCheck for self-assignment in move assignment
Zejun Wu [Thu, 17 Jul 2014 16:49:39 +0000 (09:49 -0700)]
Check for self-assignment in move assignment

Summary:
Check for self-assignment in move assignment. Otherwise
Optional<Neko> cat = whatever;
cat = std::move(cat);
cat.hasValue(); // always returns false

Test Plan: fbmake runtests

Reviewed By: tjackson@fb.com

FB internal diff: D1440633

10 years agoreserve capacity in toAppend<StringType>(...)
Marcin Pawlowski [Tue, 15 Jul 2014 22:30:57 +0000 (15:30 -0700)]
reserve capacity in toAppend<StringType>(...)

Summary:
I modified the toAppend(args..., StringType* dest) so that
before appending it reserves enough space for data to be appended. It is
still work in progress (floats and doubles are really bad: we only do
very naive approximation of the size). On float like workload we gain
~10% perf, on strings/ints/chars we gain as much as 25% of perf.
Probably on bigger strings it will be even faster. We only modify the
case when toAppend() has more than 1 element to append as it would be
just overhead in case of one argument.

Test Plan:
with this change:

============================================================================
folly/test/ConvTest.cpp                         relative  time/iter
iters/s
============================================================================
preallocateTestNoFloat                                       1.59us 627.85K
preallocateTestFloat                                         1.09us 920.70K
----------------------------------------------------------------------------
============================================================================

without the change:

============================================================================
folly/test/ConvTest.cpp                         relative  time/iter
iters/s
============================================================================
preallocateTestNoFloat                                       2.12us 471.43K
preallocateTestFloat                                         1.22us 818.25K
----------------------------------------------------------------------------
============================================================================

Reviewed By: marcelo.juchem@fb.com

FB internal diff: D1420588

Tasks: 4632421

10 years agoadded replaceAt and replaceAll to MutableStringPiece
Marcin Pawlowski [Tue, 15 Jul 2014 22:20:29 +0000 (15:20 -0700)]
added replaceAt and replaceAll to MutableStringPiece

Summary:
To make MutableStringPiece more usable (and make the
implementation of replace* methods easier) I made sure that
all the const methods (like find) work for MutableStringPiece
in a fashion similar to how they work with StringPiece
(work with everything auto convertible to StringPiece). As
a result this change is pretty susbstatial in LOC, even
though now much happens here.

Test Plan: unit tests

Reviewed By: marcelo.juchem@fb.com

FB internal diff: D1420965

Tasks: 4632424

10 years agofolly: toLowerAscii: avoid unaligned access; also correct 3 conditions
Jim Meyering [Mon, 14 Jul 2014 23:21:07 +0000 (16:21 -0700)]
folly: toLowerAscii: avoid unaligned access; also correct 3 conditions

Summary:
* folly/String.cpp (toLowerAscii): Fix two errors: the most important
would cause unaligned accesses.  This would cause a performance loss
in general, but would also result in segfaults on ARM processes.
In addition, three conditionals were wrong, further limiting
the performance of this code: switch those "<" to "<=".

Test Plan:
Run this to exercise existing tests:
fbconfig folly/test:string_test && fbmake runtests_opt
Run this to generate timing stats (before and after this change), e.g.,
fbconfig folly/test:string_benchmark && fbmake opt
_bin/folly/test/string_benchmark > TIMING-toLower-old

These numbers show a 1.6% speed increase with this change:

--- TIMING-toLower-old  2014-07-14 16:51:12.793523778 -0700
+++ TIMING-toLower-new  2014-07-14 16:49:45.815119145 -0700
@@ -1,6 +1,6 @@
============================================================================
folly/test/StringBenchmark.cpp                  relative  time/iter  iters/s
============================================================================
-libc_tolower                                                 1.06us  941.91K
-folly_toLowerAscii                                          89.99ns   11.11M
+libc_tolower                                                 1.06us  941.90K
+folly_toLowerAscii                                          88.57ns   11.29M
============================================================================

Reviewed By: brianp@fb.com

Subscribers:

FB internal diff: D1434585

Tasks: 4696800

Blame Revision: D1421056

10 years agooops, remove stray libgtest
Tudor Bosman [Tue, 15 Jul 2014 00:49:34 +0000 (17:49 -0700)]
oops, remove stray libgtest

Reviewed By: lesha@fb.com
Test Plan: no

10 years agoUpdate folly/thrift README with tested deps for Ubuntu 13.10 and 14.04
Tudor Bosman [Mon, 14 Jul 2014 20:21:55 +0000 (13:21 -0700)]
Update folly/thrift README with tested deps for Ubuntu 13.10 and 14.04

Test Plan: no

Reviewed By: lesha@fb.com

Subscribers: alandau, jhj, kma

FB internal diff: D1434163

10 years agoAllow undefined LZ4_MAX_INPUT_SIZE
Tudor Bosman [Mon, 14 Jul 2014 20:24:27 +0000 (13:24 -0700)]
Allow undefined LZ4_MAX_INPUT_SIZE

Summary:
Because the version of lz4 that ships by default with Ubuntu 13.10 doesn't
define it (but has lz4_decompress_safe, so it's usable)

Test Plan: built on Ubuntu 13.10

Reviewed By: meyering@fb.com

Subscribers: jhj, lesha, kma

FB internal diff: D1433864

@override-unit-failures

10 years agoFix IPAddress operator==
Pavlo Kushnir [Mon, 7 Jul 2014 04:03:36 +0000 (21:03 -0700)]
Fix IPAddress operator==

Summary: IPAddress operator== throws an exception when comparing ip v6 mapped address with unitialized (default constructed) address. With this diff it returns false.

Test Plan: folly unit tests

Reviewed By: bmatheny@fb.com

FB internal diff: D1421280

10 years agoFix another race in Notification Queue
Gunjan Sharma [Thu, 10 Jul 2014 00:50:19 +0000 (17:50 -0700)]
Fix another race in Notification Queue

Summary: Consider a case we found that the queue is empty and unlocked and before our setActive(false) from SCOPE_EXIT gets called (or gets the lock) putMessageImpl got the lock. Now putMessage thinks that we donot want to add another signal but actually we do.

Test Plan:
Running mcreplay2 without running into this problem on a box.
Benchmark:

Reviewed By: davejwatson@fb.com

FB internal diff: D1428249

10 years agoEnforce max uncompressed size; use safe LZ4 decompressor, which requires a newer...
Tudor Bosman [Thu, 10 Jul 2014 19:40:18 +0000 (12:40 -0700)]
Enforce max uncompressed size; use safe LZ4 decompressor, which requires a newer version of LZ4

Test Plan: folly/io/test, whatever contbuild dreams up for hphp

Reviewed By: meyering@fb.com

Subscribers: meyering, hphp-diffs@, ps, jhj, kma, sainbar, lesha

FB internal diff: D1429372

@override-unit-failures

10 years agoSmall changes for cygwin build
Elizabeth Smith [Thu, 10 Jul 2014 22:52:22 +0000 (15:52 -0700)]
Small changes for cygwin build

Summary:
Folly is almost "out of the box" working with cygwin

This has the proper ifdefs to include cygwin in two areas and a workaround for a cygwin bug when including headers

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: delong.j@fb.com

FB internal diff: D1413303

10 years agoMSVC includes for inet/sockets
Elizabeth Smith [Thu, 10 Jul 2014 22:19:57 +0000 (15:19 -0700)]
MSVC includes for inet/sockets

Summary:
MSVC has all the sockets and internet functionality required except for one missing item - which can safely be defined to a windows specific type
But it requires a different set of headers than on *nix systems

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: delong.j@fb.com

FB internal diff: D1413265

10 years agoMSVC intrinsics for bits and cpuid
Elizabeth Smith [Thu, 10 Jul 2014 22:15:26 +0000 (15:15 -0700)]
MSVC intrinsics for bits and cpuid

Summary: Use msvc intrinsics for cpuid, popcount, byteswap, and bit scan functionality

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: delong.j@fb.com

FB internal diff: D1413254

10 years agoExpression SFINAE issue in base.h
Elizabeth Smith [Thu, 10 Jul 2014 21:54:28 +0000 (14:54 -0700)]
Expression SFINAE issue in base.h

Summary:
MSVC 14 is still broken with expression sfinae - and the things that break are often strange
Moving this out into two expr templates solves the compilation issue

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: tjackson@fb.com

FB internal diff: D1413297

10 years agouse recursive_mutex to protect State
Wez Furlong [Thu, 10 Jul 2014 18:37:17 +0000 (11:37 -0700)]
use recursive_mutex to protect State

Summary:
maelstrom destructs a Promise during an indirect call from
maybeCallback and deadlocks on itself unless we use a recursive mutex.

There may be a smarter way to proceed but at the moment I can't build
and deploy a package from trunk because the service is non-functional
:-/

Test Plan:
run

```
fbconfig -r folly/wangle messaging/maelstrom
fbmake runtests
```

Reviewed By: hannesr@fb.com

Subscribers: fugalh, fsilva

FB internal diff: D1428504

10 years agoRemove unnecessary and broken hash specializations
Tudor Bosman [Wed, 9 Jul 2014 15:42:15 +0000 (08:42 -0700)]
Remove unnecessary and broken hash specializations

Summary:
No need to specialize std::hash for std::basic_string; the STL already does
this. Plus, the specializations were broken in multiple ways:

1. std::hash does not treat const char* specially, so it just hashes the
pointer value (contrasting the old __gnu_cxx::hash, which hashed a C-string)

2. Either way, using __gnu_cxx::hash<const char*> for std::string is broken,
as it won't hash anything past the first null byte.

Also, make sure fbstring gets the same (full) specializations as std::string
does in the standard.

Test Plan: fbconfig -r folly && fbmake runtests_opt, also tests in tupperware/agent which is still using __gnu_cxx::hash_map with string keys.

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

Subscribers: njormrod, jhj, kma, lesha

FB internal diff: D1426479

10 years agoFix race in Notification Queue
Gunjan Sharma [Wed, 9 Jul 2014 17:56:05 +0000 (10:56 -0700)]
Fix race in Notification Queue

Summary: When we are changing value of numActiveConsumers_ with setActive from handlerReady at the SCOPE_EXIT we have a race with reading of the same variable in putMessageImpl.

Test Plan: Had a local race which works fine now.

Reviewed By: davejwatson@fb.com

Subscribers: trunkagent

FB internal diff: D1424674

10 years agoSimplify pipe2 / O_CLOEXEC handling in Subprocess.cpp
Tudor Bosman [Tue, 8 Jul 2014 00:29:32 +0000 (17:29 -0700)]
Simplify pipe2 / O_CLOEXEC handling in Subprocess.cpp

Summary: per @rockyliu's suggestions

Test Plan: subprocess_test

Reviewed By: rockyliu4@fb.com

Subscribers: rockyliu, jhj, lesha, kma

FB internal diff: D1423157

10 years agoRevert "make `folly::Formatter` extendible"
Paul Tarjan [Wed, 9 Jul 2014 14:07:38 +0000 (07:07 -0700)]
Revert "make `folly::Formatter` extendible"

Summary:
This reverts commit 5ea0eb0c705224b82a5c284dc5f7722e1f71199f.

This breaks the HHVM clang build test.

Test Plan: clang will pass now

Reviewed By: smith@fb.com

Subscribers: dmitrys

FB internal diff: D1426399

Tasks: 4667712

10 years agomake `folly::Formatter` extendible
Tyler MacDonald [Wed, 9 Jul 2014 00:27:36 +0000 (17:27 -0700)]
make `folly::Formatter` extendible

Summary:
on advice of @tudorb, move most of `folly::Formatter` into `folly::BaseFormatter` so that we can use compile-time polymorphism to provide different types of Formatters.

I wasn't able to get the recursive formatter to be polymorphic -- whenever I tried to convert `class FormatValue<Formatter<containerMode, Args...>, void>` into `class FormatValue<BaseFormatter...`, `FormatTest.cpp:Test(Format, Nested)` wouldn't compile because it couldn't find the template. @tudorb, if you have an easy fix for this, lmk, otherwise I'm (reluctantly) okay with requiring that `Formatter`s define their own nesting `FormatValue`.

phew. the last time I did this sort of metaprogramming was over 5 years ago in perl. Doing it in C++ is... interesting.

Test Plan: `fbconfig -r thrift folly cold_storage && fbmake dbg && fbmake runtests`

Reviewed By: tudorb@fb.com

Subscribers: tudorb, dgp

FB internal diff: D1422343

Tasks: 4624268

10 years agoFix libc++ build errors
Tudor Bosman [Sat, 5 Jul 2014 15:50:38 +0000 (08:50 -0700)]
Fix libc++ build errors

Summary:
Reported externally:

https://github.com/facebook/folly/issues/70
https://github.com/facebook/folly/issues/71
https://github.com/facebook/folly/issues/72
https://github.com/facebook/folly/issues/73

Note that I can't test on libc++ myself, but the reports suggested fixes which
sounded good.

Test Plan: fbconfig -r folly && fbmake runtests_opt

Reviewed By: marcelo.juchem@fb.com

Subscribers: jhj, ntv, lesha, kma, fugalh, jdelong

FB internal diff: D1421029

10 years agoUse pipe2 in Subprocess; platform-specific config
Tudor Bosman [Mon, 7 Jul 2014 20:08:28 +0000 (13:08 -0700)]
Use pipe2 in Subprocess; platform-specific config

Summary:
We've wanted to use pipe2 in Subprocess for a while, but that's not supported
on glibc 2.5.1. This is not a problem in OSS (autoconf can solve this),
but, internally, we don't run autoconf; add an internal platform include file
with such per-platform differences.

Test Plan: fbconfig -r folly && fbmake runtests_opt

Reviewed By: meyering@fb.com

Subscribers: jhj, lesha, kma, agallagher

FB internal diff: D1422128

10 years agostride()
Tom Jackson [Thu, 3 Jul 2014 22:28:27 +0000 (15:28 -0700)]
stride()

Summary: thatwaseasy

Test Plan: iloveunittests

Reviewed By: lucian@fb.com

Subscribers: philipp

FB internal diff: D1419848

Tasks: 4636617

10 years agoMove the branchless, parallel ASCII toLower into folly
Brian Pane [Mon, 7 Jul 2014 17:27:30 +0000 (10:27 -0700)]
Move the branchless, parallel ASCII toLower into folly

Summary:
* Moved the fast toLower code into folly
* Updated the name to emphasize that it's ASCII-only

Test Plan: * Unit tests and benchmarks included.

Reviewed By: tudorb@fb.com

Subscribers: ruibalp, bmatheny

FB internal diff: D1421056

10 years ago(wangle) fix shadow
Hans Fugal [Mon, 7 Jul 2014 16:01:47 +0000 (09:01 -0700)]
(wangle) fix shadow

Summary: Not sure why this shadow error didn't turn up in my tests, but whatevs

Test Plan: traffic manager builds without warning-errors

Reviewed By: suhas@fb.com

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

FB internal diff: D1421495

Tasks: 4653938

10 years agofix Makefile.am
Tudor Bosman [Mon, 7 Jul 2014 15:59:17 +0000 (08:59 -0700)]
fix Makefile.am

Summary: wangle/detail.h -> wangle/detail/State.h

Test Plan: OSS build

Reviewed By: meyering@fb.com

Subscribers: fugalh, lesha

FB internal diff: D1421490

10 years agogflags now likes namespace gflags, not google
Tudor Bosman [Fri, 4 Jul 2014 05:18:06 +0000 (22:18 -0700)]
gflags now likes namespace gflags, not google

Summary:
Use namespace gflags going forward, import namespace google into it if
backward.

Also added a few "checking for..." messages in autoconf.

Test Plan: fbconfig -r folly && fbmake runtests_opt, OSS build

Reviewed By: meyering@fb.com

Subscribers: fjargsto, ntv, jhj, lesha, kma, davejwatson

FB internal diff: D1420575

10 years ago(wangle) cold via
Hans Fugal [Tue, 1 Jul 2014 00:29:05 +0000 (17:29 -0700)]
(wangle) cold via

Summary:
Instead of returning a Later, `via` returns a cold future.
This works without keeping a backreference like Later does, because an inactive Future will always activate on destruction. Alternatively we could have an extra Promise, a la Later, and pass that along like Later does, and require launch() at the end (though, implicit launching on destruction would be an option there too).

If you think this approach is viable I'll clean it up on Wednesday: make sure all the calling sites work, etc.

Test Plan:
new unit test
This may fail in contbuild, I haven't done the codemod for calling sites, if there are any.

Reviewed By: hannesr@fb.com

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

FB internal diff: D1412499

Tasks: 4480567

10 years ago(wangle) Future/Promise detachment
Hans Fugal [Mon, 30 Jun 2014 20:38:41 +0000 (13:38 -0700)]
(wangle) Future/Promise detachment

Summary:
Bring a bit more sanity to the lifetime. Now Future and Promise detach by calling the respective methods on State, and they do it during their destruction only. State manages its own lifetime.

Besides being cleaner, this also sets the stage for cancellation (where we'll need Future to hang on to its reference to State after setting the callback), and for folding in Later (we will have a bool for whether the future is hot and hold off executing the callback if it isn't).

Also cleans up various things I noticed while auditing the code for usages of `state_`.

Test Plan:
All the unit tests still pass.
Ran with ASAN (and found I had introduced a bug, then fixed it. yay)

Reviewed By: hannesr@fb.com

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

FB internal diff: D1412038

Tasks: 4618297

10 years ago(Folly/FBString) Enum class for category
Hannes Roth [Fri, 4 Jul 2014 15:11:36 +0000 (08:11 -0700)]
(Folly/FBString) Enum class for category

Summary: Suggested by @aalexandre.

Test Plan: Compile.

Reviewed By: andrei.alexandrescu@fb.com

Subscribers: folly@lists, njormrod, aalexandre

FB internal diff: D1246180

10 years agofolly: StringPiece: add skipWhitespace
Lucian Grijincu [Thu, 26 Jun 2014 23:26:33 +0000 (16:26 -0700)]
folly: StringPiece: add skipWhitespace

Test Plan: copied from folly::json

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

FB internal diff: D1417992

Tasks: 4527315

10 years agos/setCallback_/setCallback/ for detail::State
Hans Fugal [Mon, 30 Jun 2014 20:34:42 +0000 (13:34 -0700)]
s/setCallback_/setCallback/ for detail::State

Summary: This isn't really an internal-to-this-class method. :-/

Test Plan: builds

Reviewed By: hannesr@fb.com

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

FB internal diff: D1412008

10 years ago(wangle) s/continuation_/callback/ (missed some)
Hans Fugal [Mon, 30 Jun 2014 20:32:17 +0000 (13:32 -0700)]
(wangle) s/continuation_/callback/ (missed some)

Summary: In the spirit of D1406753

Test Plan: still builds and tests pass

Reviewed By: hannesr@fb.com

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

FB internal diff: D1412002

10 years ago(wangle) s/FutureObject/detail::State/
Hans Fugal [Mon, 30 Jun 2014 20:22:56 +0000 (13:22 -0700)]
(wangle) s/FutureObject/detail::State/

Summary:
codemod
I never was satisfied with this name. `State` feels a little better. More descriptive. But still pretty generic, so I'm open to alternative ideas.

Test Plan: It still builds and tests pass.

Reviewed By: hannesr@fb.com

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

FB internal diff: D1411506

10 years agoFix string split when last part is shorter than delimiter
Kevin Chou [Wed, 2 Jul 2014 16:14:24 +0000 (09:14 -0700)]
Fix string split when last part is shorter than delimiter

Summary: The bug is when ignoreEmpty is true, we use the wrong token size, which will be 0 when last token is shorter than delimiter.

Test Plan:
1. Add unit test
2. fbconfig -r folly && fbmake runtests

Reviewed By: philipp@fb.com

Subscribers: maxime, xiaol

FB internal diff: D1415803

10 years agoAdd Random-inl.h to Makefile.am
Tudor Bosman [Wed, 2 Jul 2014 01:03:26 +0000 (18:03 -0700)]
Add Random-inl.h to Makefile.am

Test Plan: OSS

Reviewed By: lesha@fb.com

FB internal diff: D1415190

@override-unit-failures

10 years agoupdate make check targets
Dave Watson [Fri, 27 Jun 2014 17:46:15 +0000 (10:46 -0700)]
update make check targets

Summary:
Update targets so that gtest isn't required unless you run 'make check'.

Simply make everything that uses gtest a 'check_PROGRAM' instead of a 'noinst_PROGRAM'

Test Plan: build folly with 'make -j32', gtest isn't required.

Reviewed By: andrewjcg@fb.com

Subscribers: pavlo, doug

FB internal diff: D1408288

10 years agoSimplify include path, make sure .. is in it
Tudor Bosman [Mon, 30 Jun 2014 23:12:47 +0000 (16:12 -0700)]
Simplify include path, make sure .. is in it

Summary: so #include <folly/File.h> works

Test Plan: built OSS folly, make check, built tiny program out of tree

Reviewed By: davejwatson@fb.com

FB internal diff: D1412123

10 years agoCodemod: use #include angle brackets in folly and thrift
Tudor Bosman [Mon, 30 Jun 2014 18:56:53 +0000 (11:56 -0700)]
Codemod: use #include angle brackets in folly and thrift

Summary: Also changed the thrift compilers to emit code with <...>.

Test Plan: fbconfig -r folly thrift && fbmake opt && fbmake runtests_opt

Reviewed By: davejwatson@fb.com

Subscribers: ruibalp, nli, shilin, tjackson, fugalh, alandau, bmatheny, njormrod

FB internal diff: D1411225

10 years agofix AsyncIO::doWait
Philip Pronin [Sun, 29 Jun 2014 11:49:47 +0000 (04:49 -0700)]
fix AsyncIO::doWait

Summary:
As it turns out, `io_getevents` may actually return less than
`min_nr` events.  According to the aio logic
(https://github.com/torvalds/linux/blob/10b5b5361a3c2a7fff9dbfa0f127adc2531e7732/fs/aio.c#L1634),
there may be a couple of rounds required to get at least `nr_min` events, and
if interrupted after the first one, incomplete results would be returned

Test Plan:
fbconfig -r folly/experimental/io/test && fbmake runtests_opt -32

and was no longer able to repro #4609062

Reviewed By: soren@fb.com

FB internal diff: D1410389

Tasks: 4609062

10 years agoBe explicit about what we're passing to Endian::big
Joel Marcey [Fri, 27 Jun 2014 21:47:54 +0000 (14:47 -0700)]
Be explicit about what we're passing to Endian::big

Summary:
This prevents a linker error on OSX:
```
Undefined symbols for architecture x86_64:
"folly::detail::EndianIntBase::swap(unsigned long)",
referenced from:
__GLOBAL__sub_I_MacAddress.cpp in libfolly.a(MacAddress.cpp.o)
ld: symbol(s) not found for architecture x86_64
```

We need folly and third-party changes in order to land a pull request for HHVM
that starts to get FastCGI running on OSX.

See the checklist of the HHVM pull request here: https://github.com/facebook/hhvm/pull/2944#issuecomment-47281003

Closes #68

GitHub Author: Daniel Sloof <goapsychadelic@gmail.com>

@override-unit-failures

Test Plan: fbmake runtests 100%

Reviewed By: pt@fb.com, njormrod@fb.com

FB internal diff: D1407426

10 years agoFix folly on OSX and BSD in prep for FastCGI on HHVM
Joel Marcey [Fri, 27 Jun 2014 21:46:19 +0000 (14:46 -0700)]
Fix folly on OSX and BSD in prep for FastCGI on HHVM

Summary:
A recent change in folly/MemoryMapping.cpp uses MAP_ANONYMOUS, which is
named MAP_ANON on OSX/BSD.

We need folly and third-party changes in order to land a pull request for HHVM
that starts to get FastCGI running on OSX.

See the checklist of the HHVM pull request here: https://github.com/facebook/hhvm/pull/2944#issuecomment-47281003

Closes #67

GitHub Author: Daniel Sloof <goapsychadelic@gmail.com>

@override-unit-failures

Test Plan: fbmake runtests 100%

Reviewed By: pt@fb.com, njormrod@fb.com

FB internal diff: D1407393

10 years ago(wangle) Return a Later from Future::via
Hans Fugal [Thu, 26 Jun 2014 23:23:11 +0000 (16:23 -0700)]
(wangle) Return a Later from Future::via

Summary: Stroke of brilliance, Hannes.

Test Plan:
unit tests, including a new one
Looked through `fbgs 'via('` and all the extant `via`s are attached to `Later`s already so it shouldn't break anything. But check contbuild before commit.

Reviewed By: hannesr@fb.com

Subscribers: net-systems@, fugalh, exa

FB internal diff: D1406976

Tasks: 4480567

10 years agofolly: futexWaitUntilImpl: avoid abort triggered via tao/queues tests
Jim Meyering [Wed, 25 Jun 2014 23:47:55 +0000 (16:47 -0700)]
folly: futexWaitUntilImpl: avoid abort triggered via tao/queues tests

Summary:
* folly/test/DeterministicSchedule.cpp (folly): Tao/queues tests exposed
an unhandled case in this code: when (futex->data == expected) happens,
we would call futexErrnoToFutexResult with a futexErrno value of 0,
which would cause an abort (unhandled valued in switch).
Here's a stack trace from the abort: https://phabricator.fb.com/P12558008
showing the invalid futexErrno value in frame #4.

Test Plan:
fbconfig -r --platform-all=gcc-4.8.1-glibc-2.17 --sanitize=address tao/queues:TimeoutWorkQueueTests
_bin/tao/queues/WorkQueueTests -fbunit_test_regexp '^mt\_stress\_deterministic$'

Reviewed By: mssarang@fb.com

FB internal diff: D1404572

Tasks: 4494871

10 years agoFix use-after-free in futexWaitUntilImpl
Sarang Masti [Thu, 26 Jun 2014 22:37:13 +0000 (15:37 -0700)]
Fix use-after-free in futexWaitUntilImpl

Summary: Handle wake-ups correctly in futexWaitUntilImpl.

Test Plan:
-- ran all folly unit tests
-- ran TimeoutWorkQueue test under tao/queues

Reviewed By: ngbronson@fb.com, meyering@fb.com

FB internal diff: D1406845

Tasks: 4494871

10 years ago(wangle) noexcept move constructor
Hans Fugal [Thu, 26 Jun 2014 22:22:55 +0000 (15:22 -0700)]
(wangle) noexcept move constructor

Summary: to make lint happy

Test Plan: builds

Reviewed By: hannesr@fb.com

Subscribers: net-systems@, fugalh, exa

FB internal diff: D1406766

Tasks: 4480567

10 years ago(wangle) s/continuation/callback
Hans Fugal [Thu, 26 Jun 2014 22:20:40 +0000 (15:20 -0700)]
(wangle) s/continuation/callback

Summary: The word "continuation" is too ambiguous. Prefer callback (except where we are actually talking about CSP continuations in the README).

Test Plan: Still builds. Nothing external is or should be calling `setContinuation`. But I'll wait for contbuild anyway.

Reviewed By: hannesr@fb.com

Subscribers: net-systems@, fugalh, exa

FB internal diff: D1406753

Tasks: 4480567

10 years agonuke executeWith
Hans Fugal [Thu, 26 Jun 2014 21:31:48 +0000 (14:31 -0700)]
nuke executeWith

Summary: Removing this crufty API. The only callsite for `executeWith` was `Later::via` so I just folded it in there.

Test Plan:
unit tests
contbuild

Reviewed By: hannesr@fb.com

Subscribers: net-systems@, fugalh, exa

FB internal diff: D1406592

Tasks: 4480567

10 years agoFix bizarre optimization
Hans Fugal [Thu, 26 Jun 2014 17:44:20 +0000 (10:44 -0700)]
Fix bizarre optimization

Summary:
I have no idea why the compiler is gettings its britches in a bunch because a unit test has a predictable output. Really, gcc?!

But whatever, avoiding inlining by putting it in the cpp file solves it.

Test Plan: fbmake runtests_opt

Reviewed By: davejwatson@fb.com

Subscribers: net-systems@, fugalh, exa

FB internal diff: D1405811

Tasks: 4591823

10 years ago(folly) QueuedImmediateExecutor
Hans Fugal [Wed, 4 Jun 2014 22:38:24 +0000 (15:38 -0700)]
(folly) QueuedImmediateExecutor

Summary: Add the `QueuedImmediateExecutor` which behaves like `InlineExecutor` but with different (and usually better) ordering semantics for nested calls.
@override-unit-failures

Test Plan: unit tests

Reviewed By: davejwatson@fb.com

Subscribers: folly@lists, net-systems@, fugalh, exa

FB internal diff: D1364904

Tasks: 3789661

10 years agoOops, fix README
Tudor Bosman [Tue, 24 Jun 2014 01:23:03 +0000 (18:23 -0700)]
Oops, fix README

Reviewed By: meyering@fb.com
Test Plan: no

10 years agofolly OSS fixes: add ThreadName.h and compression
Tudor Bosman [Sat, 21 Jun 2014 02:10:12 +0000 (19:10 -0700)]
folly OSS fixes: add ThreadName.h and compression

Summary: Also, optionalize dependencies on compression libraries.

Test Plan: fbconfig -r folly && fbmake runtests_opt

Reviewed By: meyering@fb.com

Subscribers: kma, jhj, simpkins, lesha, folly@lists

FB internal diff: D1396573

10 years agoFix for folly open source build on Ubuntu 12.04
Anton Likhtarov [Wed, 25 Jun 2014 01:52:54 +0000 (18:52 -0700)]
Fix for folly open source build on Ubuntu 12.04

Summary: On 12.04, there's both /usr/lib/libiberty.a and /usr/lib/libiberty_pic.a, and _pic is the one we want to build a Folly shared library.

Test Plan: build on Ubuntu 12.04

Reviewed By: meyering@fb.com

FB internal diff: D1402194

10 years agoScheduler interface of Executor
Hans Fugal [Thu, 19 Jun 2014 01:03:45 +0000 (18:03 -0700)]
Scheduler interface of Executor

Summary: and ManualExecutor implementation

Test Plan: unit tests, contbuild

Reviewed By: davejwatson@fb.com

Subscribers: bmatheny, folly@lists, net-systems@, fugalh, exa, marccelani, jsedgwick

FB internal diff: D1392999

Tasks: 4548494

10 years agofolly OSS fixes: add ThreadName.h and compression
Tudor Bosman [Sat, 21 Jun 2014 02:10:12 +0000 (19:10 -0700)]
folly OSS fixes: add ThreadName.h and compression

Summary: Also, optionalize dependencies on compression libraries.

Test Plan: fbconfig -r folly && fbmake runtests_opt

Reviewed By: meyering@fb.com

Subscribers: kma, jhj, simpkins, lesha, folly@lists

FB internal diff: D1396573

10 years agoMake randomNumberSeed read from /dev/urandom
Tudor Bosman [Fri, 20 Jun 2014 03:46:10 +0000 (20:46 -0700)]
Make randomNumberSeed read from /dev/urandom

Summary: Because @lesha asked "why not" and I couldn't give him an answer.

Test Plan: random_test

Reviewed By: bmaurer@fb.com

Subscribers: bmaurer, folly@lists, jhj, kma, lesha, sdoroshenko, soren

FB internal diff: D1394401

10 years agoAdd ThreadName.h to folly's Makefile.am
Peter Ruibal [Sun, 22 Jun 2014 20:40:11 +0000 (13:40 -0700)]
Add ThreadName.h to folly's Makefile.am

Summary:
fbthrift depends on <folly/ThreadName.h>, which isn't currently
getting installed as part of the autotools build.  Add it to Makefile.am

Test Plan:
Made this change to folly, re-autogen/configure/install, and
then was able to successfully compile fbthrift's compiler

Reviewed By: davejwatson@fb.com

Subscribers: doug, folly@lists

FB internal diff: D1397084

10 years agoExpose EVLOOP_NONBLOCK
Yunqi Zhang [Thu, 19 Jun 2014 01:40:19 +0000 (18:40 -0700)]
Expose EVLOOP_NONBLOCK

Summary:
This diff allows users to loop through EventBase without blocking if there are
not any events to process.

This is useful for sending and receiving requests on network, where users just
want to try if there are any events and do not want to block if not.

https://phabricator.fb.com/D1373887 is an example where we find this feature
useful, otherwise we have to add an empty callback before loop.
event_base_.runInLoop([] {});
event_base_.loopOnce();

@davejwatson, @fugalh, @simpkins, @stepan: Could you please take a look at the
proposed changes and let me know if there is any better ways of doing this.

Thank you!

Test Plan:
I think this would not break anything, but we might want to do some performance
profiling if needed.

Reviewed By: hans@fb.com

Subscribers: simpkins, davejwatson, fugalh, stepan, folly@lists

FB internal diff: D1383401

10 years agoMake fbstring libgcc-safe
Nicholas Ormrod [Wed, 18 Jun 2014 16:35:10 +0000 (09:35 -0700)]
Make fbstring libgcc-safe

Summary:
Some libgcc-incompatible code has been added to fbstring.
Removed/reorganized it so that we can drop fbstring right into libgcc.

Test Plan:
fbconfig -r folly && fbmake runtests

Copied FBString.h into libgcc's basic_fbstring.h, with no modifications.
Successfully tp2_build libgcc/4.8.1. Adjusted symlink, then fbmake clean
&& fbconfig -r folly && fbmake dbg. The fbmake dbg failed with an
assertion error, which is consistent with @lucian's observations in
D1373725; the important part is that the error was at runtime, so the
compile-time changes of this diff looks good.

Reviewed By: lucian@fb.com

Subscribers: folly@lists, sdwilsh, njormrod, lucian

FB internal diff: D1382873

10 years agoFBString conservative additions
Nicholas Ormrod [Sat, 14 Jun 2014 01:09:44 +0000 (18:09 -0700)]
FBString conservative additions

Summary:
Now that fbstring is conservative by default (D1373308), we can remove
the mutability of the data members and the call to c_str() in operator[].

Test Plan: fbconfig -r folly && fbmake runtests

Reviewed By: lucian@fb.com

Subscribers: folly@lists, sdwilsh, njormrod

FB internal diff: D1382644

10 years agoFix open source build
Anton Likhtarov [Sat, 14 Jun 2014 00:45:11 +0000 (17:45 -0700)]
Fix open source build

Test Plan: build it

Reviewed By: pavlo@fb.com

Subscribers: folly@lists

FB internal diff: D1383722

10 years agofolly: fbstring: make it conservative-only: write '\0' in ctor and drop the c_str...
Lucian Grijincu [Fri, 6 Jun 2014 21:48:15 +0000 (14:48 -0700)]
folly: fbstring: make it conservative-only: write '\0' in ctor and drop the c_str shenanigans

Test Plan: ran folly tests

Reviewed By: njormrod@fb.com

Subscribers: folly@lists, njormrod

FB internal diff: D1373308

10 years agofbstring conservative corruption
Nicholas Ormrod [Thu, 12 Jun 2014 18:11:55 +0000 (11:11 -0700)]
fbstring conservative corruption

Summary:
@lucian's D1373308 set fbstring to conservative by default.
This breaks, eg, ti/proxygen/httpclient tests, by failing an assertion
inside of c_str(). Specifically, the terminator is not '\0'.

The fbstring_core move constructor, when sourced by a MediumLarge
string, steals the source's internal data, then resets the source by
calling ##setSmallSize(0)##. That function sets the in-situ size of the
fbstring to zero, thus requalifying the string as a small string;
however, it does nothing to the data - the previous 23 bytes now contain
garbage.

Sources of a move must be in a consistent state after the move is
complete. The source, once a MediumLarge string whose first eight bytes
were a pointer, is now a small string of size zero whose first byte is
not necessarily '\0'. This breaks the FBSTRING_CONSERVATIVE invariant.

This can be fixed by writing a terminator after the setSmallSize call.
I have fixed all setSmallSize locations that do not writeTerminator.

fbstring_core's move constructor is called exclusively from
basic_fbstring's move assignment operator, hence the odd format of the
test case.

== TMI ==

Interestingly, the source will almost certainly* contain a '\0', which
prevents this simple ##str.size() != strlen(str.c_str())## bug from
turning into a memory-trampling monster. The new size of zero is not
what saves us - the 'size' byte of a small fbstring, through a very
clever trick to allow 23-byte in-situ strings, is actually 23 minus the
actual size (now 0), so is 23! Where, then, does the '\0' byte come? A
MediumLarge string's data layout is [pointer, size, capacity]. The
pointer is not guaranteed to contain a '\0', and neither are size or
capacity. However, the size of the string needs to be very large in
order to force the top byte of the size member to be non-zero; in that
case, the string is so large that malloc is returning memory by the
page. Since page sizes are a multiple of 2^8 (almost always, and if not
then I don't think your fbstring can support large enough sizes
anyways), and we use goodMallocSize, the capacity pointer would have a
least signfigicant byte of zero.

Why the (*)? Well, when reserving extra space on a non-refcounted Large
string, the reallocation does not yield its extra goodMallocSize space.
This could be fixed, though probably isn't worth the trouble. Anyways,
since we aren't goodMallocSize-ing the user-supplied requested capacity,
it probably won't contain a '\0'.

Test Plan:
fbconfig -r folly && fbmake runtests

Modify folly/test/FBStringTest.cpp to define FBSTRING_CONSERVATIVE, then
fbconfig folly/test/:fbstring_test_using_jemalloc && fbmake runtests
Note that this fails before the patch is applied.

Note that it is possible for the tests to pass even when this bug is
present, since the top byte of the heap pointer must be non-0 in order
for this error to be triggered.

Reviewed By: lucian@fb.com

Subscribers: folly@lists, njormrod, lucian, markisaa, robbert, sdwilsh, tudorb, jdelong

FB internal diff: D1376517

10 years agostatic-ify FBString asserts
Nicholas Ormrod [Thu, 12 Jun 2014 00:34:12 +0000 (17:34 -0700)]
static-ify FBString asserts

Summary: Some asserts could be static_asserts. Make it so!

Test Plan: fbconfig -r folly && fbmke opt && fbmake runtests_opt

Reviewed By: lucian@fb.com

Subscribers: folly@lists, sdwilsh, njormrod

FB internal diff: D1378670

10 years agofolly::gen::splitByLine
Vojin Katic [Wed, 11 Jun 2014 20:53:16 +0000 (13:53 -0700)]
folly::gen::splitByLine

Summary:
I made it work, but please send your feedback how to improve code quality.

splitByLine will split on \r, \n, and \r\n.

Test Plan: add new test, arc unit

Reviewed By: tjackson@fb.com

Subscribers: folly@lists, crawler-diffs@

FB internal diff: D1322212

10 years agofolly: do not disable RW_SPINLOCK_USE_X86_INTRINSIC_ for clang
Jim Meyering [Mon, 9 Jun 2014 20:32:26 +0000 (13:32 -0700)]
folly: do not disable RW_SPINLOCK_USE_X86_INTRINSIC_ for clang

Summary:
Without this, we'd see problems like this in tao, when building with clang:
With this change, this now works with clang-3.4 and clang.dev (3.4+).
This change reverts D950285, which change appears to have been made
to accommodate weakness in clang-3.3 or older.

In file included from tao/data_providers/common/simpledp.cpp:7:
./tao/data_providers/common/stats.h:175:18: error: no type named 'RWTicketSpinLockT' in namespace 'folly'
typedef folly::RWTicketSpinLockT<64, true> RWLockType;
~~~~~~~^
./tao/data_providers/common/stats.h:175:35: error: expected member name or ';' after declaration specifiers
typedef folly::RWTicketSpinLockT<64, true> RWLockType;
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^

Test Plan:
ensure that all of these pass:
fbconfig -r --clang folly/test:rw_spinlock_test && fbmake runtests
fbconfig -r --clang folly/test:rw_spinlock_test && fbmake runtests_opt
fbconfig -r         folly/test:rw_spinlock_test && fbmake runtests_opt
fbconfig -r --clang --with-project-version clang:dev \
folly/test:rw_spinlock_test && fbmake runtests_opt

Reviewed By: delong.j@fb.com

Subscribers: folly@lists, mathieubaudet

FB internal diff: D1370024

Tasks: 4090011

10 years agoSpecialize string to identical string conversion
Zejun Wu [Mon, 9 Jun 2014 17:16:31 +0000 (10:16 -0700)]
Specialize string to identical string conversion

Summary:
Avoid copying underlying char array in to<string>(const string&) and
to<fbstring>(const fbstring&).

Test Plan: fbconfig -r $redteamisthebestteam/$nekomikoreimu && fbmake

Reviewed By: ldbrandy@fb.com

Subscribers: jonp, folly@lists

FB internal diff: D1368183

Tasks: 4263125

10 years agoAdd support for returning number of benchmark iterations
Marcus Holland-Moritz [Fri, 6 Jun 2014 22:57:21 +0000 (15:57 -0700)]
Add support for returning number of benchmark iterations

Summary:
I'm looping through a large number of test cases in a benchmark and I'm
interested in the average time per test case rather than the total time,
which is what I'm currently seeing. In order to get the average time,
I need to be able to tell the benchmark module how many iterations have
been run.

This change adds _MULTI variants of the different BENCHMARK_ macros that
allow for returning the actual number of iterations that have been run,
e.g.:

BENCHMARK_MULTI(benchmarkSomething) {
std::vector<int> testCases { 0, 1, 1, 2, 3, 5 };
for (int c : testCases) {
doSomething(c);
}
return testCases.size();
}

Test Plan:
* fbconfig -r folly && fbmake runtests
* added new test cases to example benchmark code

Reviewed By: simpkins@fb.com

Subscribers: folly@lists, london_search@

FB internal diff: D1356437

10 years agorevert "conditionally write terminator in c_str()"
Philip Pronin [Fri, 6 Jun 2014 17:53:51 +0000 (10:53 -0700)]
revert "conditionally write terminator in c_str()"

Summary: D1318048#21

@override-unit-failures

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

Reviewed By: njormrod@fb.com

Subscribers: folly@lists, njormrod

FB internal diff: D1368939

Tasks: 4466412

Blame Revision: D1318048

10 years agoIntroduce destruction callbacks
Stepan Palamarchuk [Thu, 5 Jun 2014 05:33:39 +0000 (22:33 -0700)]
Introduce destruction callbacks

Summary:
This change allows users to track lifetime of EventBase and perform clean shutdown when EventBase gets destructed.

It is useful for users that rely on EventBase lifetime, but don't have any feedback mechanism with the owner of EventBase.

For instance some part of code might remain running in background on the EventBase after the main object was destroyed (e.g. it might be finalizing some async requests). In such case the original owner doesn't know that there's something still running and may try to destroy EventBase. In that case such background code will remain zombie forever.

AsyncMcClient changes are presented just as an example of usage.

@davejwatson, @simpkins: Could you please take a look at the proposed changes for the EventBase? If this is something not worth adding into EventBase, could you recommend a better way of doing things?

Test Plan: fbmake runtests

Reviewed By: alikhtarov@fb.com

Subscribers: folly@lists, simpkins, davejwatson

FB internal diff: D1353101

10 years agoAllowing additional arguments to be passed to split_step's functor
Marcelo Juchem [Tue, 3 Jun 2014 21:40:29 +0000 (14:40 -0700)]
Allowing additional arguments to be passed to split_step's functor

Summary: more flexibility for using functors with split_step

Test Plan: unit tests added + arc unit

Reviewed By: ldbrandy@fb.com

Subscribers: folly@lists

FB internal diff: D1362644