Test Plan: folly/test/fbstring_test in debug mode

Reviewed By: sdoroshenko@fb.com

FB internal diff: D661622

Summary: Extra '#' in there.

Test Plan: fbconfig nagios/plugins && fbmake opt

Reviewed By: carl.moran1@fb.com

FB internal diff: D660977

Summary:
handle at least clang. It will be better to turn this into
configure check, though.

Test Plan: .

Reviewed By: tudorb@fb.com

FB internal diff: D660139

Summary: Tests not compiling due to a copy and paste error in Makefile.am.

Test Plan: .

Reviewed By: tudorb@fb.com

FB internal diff: D660138

Summary: use proper type for index.

Test Plan: .

Reviewed By: tudorb@fb.com

FB internal diff: D660142

Summary:
Using perfect forwarding here prevents the ability to create
dynamic arrays.

Test Plan: Added unit test case that compiles now with this.

Reviewed By: tudorb@fb.com

FB internal diff: D660662

Summary:
The current UpgradedHolder implementation may lead to a deadlock when upgrading from a reader lock, because it is blocking.

A reader that has failed to set the UPGRADED bit may block the
winner (upgrader) that has successfully set the UPGRADED bit, while
waiting to upgrade in an infinite loop without releasing its own
reader lock. The upgrader needs to wait for all reader locks to be
released before it can move forward.

This is the code pattern:

{
ReadHolder reader(lock);
UpgradedHolder upgrader(std::move(reader));
WriteHolder writer(std::move(upgrader));
}

To avoid this to happen, removed UpgradedHolder(ReadHolder&&)
constructor and the function that impelments it
unlock_shared_and_lock_upgrade. This would require a programmer explicitly
release a read lock before acquiring an upgrade lock, therefore
avoid the above mentioned deadlock.

In addition, the current folly::RWSpinLock::try_lock_shared()
implementation does not check the UPGRADED bit at all. The UPGRADED bit can be used to avoid starving writers.  This diff fixed this by correctly checking the UPGRADED bit.

Test Plan:
Passed folly unit tests.
Tested in a Facebook service that uses the
folly::RWSpinLock::UpgradedHolder.

Reviewed By: xliux@fb.com

FB internal diff: D659875

Summary: .

Test Plan: .

Reviewed By: soren@fb.com

FB internal diff: D660143

Blame Revision:

Summary: I wouldn't even bother reporting this if github didn't make it so easy!

Test Plan: .

Reviewed By: soren@fb.com

FB internal diff: D660136

Summary:
Added benchmarks.

Before:

folly/test/RangeFindBenchmark.cpp               relative  time/iter  iters/s
============================================================================
string length 10:
FindFirstOfRange                                             1.36ns  733.07M
FindFirstOfOffsetRange                                       2.15ns  464.16M
============================================================================
string length 256:
FindFirstOfRange                                             1.36ns  733.07M
FindFirstOfOffsetRange                                       1.42ns  704.22M
============================================================================
string length 10240:
FindFirstOfRange                                             1.36ns  733.07M
FindFirstOfOffsetRange                                       3.72ns  268.70M
============================================================================
string length 10485760:
FindFirstOfRange                                             1.36ns  733.07M
FindFirstOfOffsetRange                                       5.00ns  199.96M

After:

string length 10:
FindFirstOfRange                                             1.36ns  732.92M
FindFirstOfOffsetRange                                       1.36ns  732.61M
============================================================================
string length 256:
FindFirstOfRange                                             1.36ns  732.93M
FindFirstOfOffsetRange                                       1.38ns  727.16M
============================================================================
string length 10240:
FindFirstOfRange                                             1.36ns  732.93M
FindFirstOfOffsetRange                                       1.79ns  558.40M
============================================================================
string length 10485760:
FindFirstOfRange                                             1.36ns  732.93M
FindFirstOfOffsetRange                                       2.73ns  366.44M

Test Plan: fbconfig folly && fbmake runtests

Reviewed By: delong.j@fb.com

FB internal diff: D660125

Summary:
Fixes instances of the following problems:
error: offset of on non-POD type 'folly::fbstring_core<char>::RefCounted'
[-Werror,-Winvalid-offsetof]
Solution: Since the atomic is not a POD I used sizeof() instead to calculate
the offset

warning: C++11 requires lambda with omitted result type to consist of a single
return statement
Solution: Specify a return type

error: in-class initializer for static data member is not a constant expression
Solution: Move initializer

Test Plan: - Compiled folly and ran fbmake runtests

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D656963

Summary:
Everything worked except for getting properly aligned full blocks
because 1ULL << 64 is invalid (the shift amount must be strictly less than
the value size in bits)

Test Plan: test added

Reviewed By: philipp@fb.com

FB internal diff: D657800

Test Plan: compiles

Reviewed By: philipp@fb.com

FB internal diff: D657623

Summary: Mechanical.

Test Plan: fbconfig unicorn/test unicorn/diskindex3 unicorn/diskindex4/test common/datastruct:css_tree_test folly/test:padded_test dragon/test:posting_list_test && fbmake opt

Reviewed By: soren@fb.com

FB internal diff: D649539

Summary: TSIA

Test Plan: Unit tests

Reviewed By: kedarb@fb.com

FB internal diff: D652180

Summary:
The test doesn't run when GCC < 4.7, however our test framework
needs some output anyways. I've added a placeholder test that does
nothing, but does cause something to be printed.

Test Plan: run fbmake runtests on folly/test/stl_tests/

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D652161

Summary:
Add specialized dynamic converter for containers of pairs,
which can convert from a list of pairs, or from a object

Test Plan: added a unit test

Reviewed By: njormrod@fb.com

FB internal diff: D650730

Summary:
This is FBVector 2.0. It supports all of the original FBVector
optimizations and is standard compliant.
Accompanying this diff are two suites, StlVectorTest and Benchmark.
StlVectorTest runs an extensive battery of tests against a vector
implementation per the N3337 standard. In addition to checking normal
correctness, StlVectorTest checks the use of allocators, exception
safety, memory leaks, and type requirements.
Benchmark run a wide range of speed tests between std::vector, the
original fbvector, and the new fbvector.

Test Plan:
First test: run StlVectorTest.
Second test: run Benchmark.
Third test: compile and run some fbcode (e.g. multifeed/).

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D566719

Summary:
Added IsBits (has value semantics; can be copied by memcpy and
deleted by free. Implies IsRelocatable) and Is0Initializable (new (ptr)
T() is equivalent to memset(ptr, 0, sizeof(T))).
Converted the boost types for IsRelocatable to std types (now available
from <type_traits>).
Added a new way to specify IsRelocatable, IsBits, and Is0Initializable:
typedef std::true_type <property-name> in your class. No namespace
exiting required. This method also supports inheritance, and may be
overriden in base classes.
Added a test file to test Traits.

Test Plan:
Run new test file. Compile some real-world code (in my case,
multifeed).

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D610996

Summary: I forgot fbmake runtests doesn't build all that's configured, oops.

Test Plan: Build folly/experimental/test, run benchmark.

Reviewed By: tudorb@fb.com

FB internal diff: D647191

Summary:
It makes sense to return subpiece of the original data
from ##rest()##.

Test Plan: gv tests

Reviewed By: soren@fb.com

FB internal diff: D647179

Summary:
It was previously breaking any time a generator included a comma, which
is pretty stupid. Now, it's modelled after `ScopeGuard`, using a little operator
overloading to get rid of the trailing `)` and sidestepping preprocessor issues.

Test Plan: Unit tests

Reviewed By: tudorb@fb.com

FB internal diff: D646825

Summary:
Code that aids in storing data aligned on block (possibly cache-line)
boundaries, perhaps with padding. There's class Node which represents one
block, and Iterator which, given an iterator to a container of Nodes, gives you
an iterator to the underlying elements.  There's also Adaptor, which converts a
sequence of Node into a sequence of underlying elements. (with enough
functionality to make it useful, although it's not fully STL compatible)

Split off from https://phabricator.fb.com/D641114

Also includes changes to make TypedIOBuf container-like so it can be used with padded_sequence::Adaptor.

I plan to rename this to Padded.h / folly::padded in a separate diff.

Test Plan: test added

Reviewed By: soren@fb.com

FB internal diff: D646249

Summary:
I noticed that push_back(Char) was slower for fbstring than
for std::string for long strings and found that I could make
it faster by inlining the call to mutable_data() and
exploit that it's always non-small and non-shared.

Benchmarks before:

./folly/test/FBStringTestBenchmarks.cpp.h       relative  time/iter  iters/s
BM_push_back_string(1)                                      69.42ns   14.41M
BM_push_back_string(23)                                    582.31ns    1.72M
BM_push_back_string(127)                                     1.47us  682.12K
BM_push_back_string(1024)                                    5.52us  181.07K
BM_push_back_fbstring(1)                                     9.55ns  104.74M
BM_push_back_fbstring(23)                                  212.45ns    4.71M
BM_push_back_fbstring(127)                                 864.00ns    1.16M
BM_push_back_fbstring(1024)                                  6.73us  148.52K

and after:

BM_push_back_fbstring(1)                                     9.55ns  104.74M
BM_push_back_fbstring(23)                                  212.45ns    4.71M
BM_push_back_fbstring(127)                                 783.08ns    1.28M
BM_push_back_fbstring(1024)                                  4.03us  248.05K

Test Plan: fbconfig folly/test && fbmake runtests

Reviewed By: tudorb@fb.com

FB internal diff: D646081

Summary:
Added closing curly braces at 173.
Sorry, must make diff if it's a new commit.

Test Plan: nothing really

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D644925

Summary: Added two samples of use for FBVector in docs

Test Plan: just check if FBVector.md looks good.

Reviewed By: njormrod@fb.com

FB internal diff: D636813

Summary:
This code is not strictly correct per c++ spec section 14.2.4:

> When the name of a member template specialization appears after . or -> in a
> postfix-expression or after a nested-name-specifier in a qualified-id, and the
> object or pointer expression of the postfix-expression or the
> nested-name-specifier in the qualified-id depends on a template parameter
> (14.6.2) but does not refer to a member of the current instantiation
> (14.6.2.1), the member template name must be prefixed by the keyword template.
> Otherwise the name is assumed to name a non-template.

This causes clang to complain, but gcc is lax about this, as usual.

Test Plan: compile stuff that use this.

Reviewed By: delong.j@fb.com

FB internal diff: D643515

Summary:
the old code is ill-formed per spec (section 14.6.8):

> If a type used in a non-dependent name is incomplete at the point at which a
> template is defined but is complete at the point at which an instantiation is
> done, and if the completeness of that type affects whether or not the program
> is well-formed or affects the semantics of the program, the program is
> ill-formed; no diagnostic is required.

GCC is lax and allows this, clang gives an error.

Test Plan: compiles.

Reviewed By: delong.j@fb.com

FB internal diff: D643431

Summary:
curiously, std::out_of_range doesn't have a zero-argument constructor according
to the spec. This was causing issues in my clang setup...

Test Plan: compiles.

Reviewed By: delong.j@fb.com

FB internal diff: D643363

Test Plan: looked at the generated html in a browser

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D642998

Summary:
Added specializations of folly::to<String> for __int128_t
and __uint128_t.

Test Plan: Added tests of the above to the integral to string tests.

Reviewed By: andrei.alexandrescu@fb.com

FB internal diff: D636992

Summary:
Returns number of bytes "really used by the user", i.e. number of bytes
allocated on the arena via calls to `allocate()`.

Test Plan: compiled, unit tests

Reviewed By: azzolini@fb.com

FB internal diff: D636976

Summary:
Supressing the warning for GCC4.7, and supressing the error supressing
the warning in GCC4.6.

Test Plan: Compile, run unit tests on GCC4.6 and GCC4.7

Reviewed By: ldbrandy@fb.com

FB internal diff: D635982

Summary:
Noticed these TODOs in Conv.h and decided to fix them.
Discovered that the underlying_type implementation also
covers enum classes, which didn't work with the pre-existing code.

Test Plan:
fbconfig -r folly --platform=gcc-4.7.1-glibc-2.14.1-fb
fbmake dbg _bin/folly/test/conv_test
_bin/folly/test/conv_test

--platform=gcc-4.6.2-glibc-2.13
fbmake dbg _bin/folly/test/conv_test
_bin/folly/test/conv_test

Reviewed By: tudorb@fb.com

FB internal diff: D634309

Summary: D633912 broke unicorn because it relied on preallocate having a hard max.  Add an optional hard max parameter and rename maxHint to newAllocationSize.  Pass the hard max in for unicorn

Test Plan: folly and unicorn unit tests

Reviewed By: tudorb@fb.com

FB internal diff: D634894

Summary: IOBufQueue::preallocate currently takes a min and max size, and will return to the caller a buffer with N writable bytes with min <= N <= max.  This is a bit wasteful, since there may be more than max bytes available at the end if the queue with no extra cost.  Now preallocate will return the full amount of contigious space to the caller, even if it is larger than the requested max.

Test Plan: folly and proxygen unit tests

Reviewed By: simpkins@fb.com

FB internal diff: D633912

Summary:
cursor.clone() will clone a length of the chain.  insert(std::move(buf)) will splice in a length of chain
I want this to change thrift2 binary type to IOBuf: we will clone() the network data for zero-copy userspace data, and insert() it if the return value is a IOBuf.

Test Plan: added unittest

Reviewed By: afrind@fb.com

FB internal diff: D632073

Summary:
Add a version of IOBuf::copyBuffer() which accepts a std::string, as
well as a maybeCopyBuffer() function which returns a null pointer if the
input string is empty.

In proxygen we have a few places where we construct an IOBuf from a
string configuration parameter, and we almost always want a null pointer
rather than an empty IOBuf chain if the string is empty.

Test Plan: Included unit tests for these new functions.

Reviewed By: tudorb@fb.com

FB internal diff: D630547

Summary: ##byLine## already yields StringPieces, no neat to ##eachAs()## then.

Test Plan:
fbconfig folly/experimental/symbolizer folly/test folly/experimental/io folly/experimental/io/test tupperware/agent
fbmake opt

Reviewed By: tudorb@fb.com

FB internal diff: D630422

Test Plan: build

Reviewed By: dneiter@fb.com

FB internal diff: D630324