Dominik Gabi [Wed, 23 Dec 2015 23:43:26 +0000 (15:43 -0800)]
wrapper for nullable attribute
Reviewed By: yfeldblum
Differential Revision:
D2784472
fb-gh-sync-id:
84c7426cc82edabb04c662fa699764ffc0864b7e
Dominik Gabi [Wed, 23 Dec 2015 23:43:24 +0000 (15:43 -0800)]
clang attribute wrappers
Summary: adding GCC compatible wrappers for clang attributes to folly.
Reviewed By: yfeldblum
Differential Revision:
D2706913
fb-gh-sync-id:
17db34089d48a6a01c20904321a47ff98141c2fe
Michael Lee [Wed, 23 Dec 2015 20:38:29 +0000 (12:38 -0800)]
Change `SYS_gettid` to __NR_gettid instead
Summary: SYS_gettid is not available on all platforms, but __NR_gettid should be.
Reviewed By: dcolascione
Differential Revision:
D2787003
fb-gh-sync-id:
ee024437ac95281a3573e2440653847f6f7d1738
Giuseppe Ottaviano [Wed, 23 Dec 2015 20:07:16 +0000 (12:07 -0800)]
Fix test Dynamic.ArrayReassignment with Clang 3.7
Summary:
`dynamic` is one of the few unfortunate recursive types that have a constructor of the form `T(initializer_list<T>)`. After the Defect Report 95 (http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#1467) a statement `T t = {v};` where `v` is of type `T` invokes the copy constructor instead of the initializer list constructor. Clang 3.7 implements the new behavior, and GCC probably will soon.
This diff only fixes a test where the ambiguous syntax is used, but a better solution would be preferable.
Reviewed By: philippv
Differential Revision:
D2786455
fb-gh-sync-id:
7fa5235e4041cbc8dd3ce685b5a0c23c343f78d2
Mike Kolupaev [Wed, 23 Dec 2015 10:05:01 +0000 (02:05 -0800)]
Added missing instantiation for HistogramBuckets::computeTotalCount()
Summary:
D2078239 added a template method to HistogramBuckets but didn't add an instantiation for it, similar to getPercentileBucketIdx() and getPercentileEstimate() (see comment in Instantiations.cpp). This diff adds it, making `computeTotalCount()` usable without including `Histogram-defs.h`.
Also removes the weird `const` in return type.
Reviewed By: simpkins
Differential Revision:
D2783534
fb-gh-sync-id:
9226489820116e0cbcb1f6a631b389439558061e
Michael Lee [Tue, 22 Dec 2015 23:48:46 +0000 (15:48 -0800)]
Remove extra FLAGS_seed definition
Summary: This is not used within the file. Cleaning up.
Reviewed By: yfeldblum
Differential Revision:
D2784500
fb-gh-sync-id:
60270cb20c40b79f988c5536fa00e0a0f1d08e46
Michael Lee [Mon, 21 Dec 2015 15:55:45 +0000 (07:55 -0800)]
Move ThreadLocal.h into the implementation
Summary: Move ThreadLocal.h out of Random.h so we don't over include.
Reviewed By: yfeldblum
Differential Revision:
D2770085
fb-gh-sync-id:
e6934d606236c2b5cdde728639f372641ee716dc
Kosov Eugene [Fri, 18 Dec 2015 18:17:19 +0000 (10:17 -0800)]
fixed build with g++ 5+ and ASAN enabled
Summary:
Fixing by deleting unnecessary macro checks - folly already requires at least g++4.8
Closes https://github.com/facebook/folly/pull/338
Reviewed By: yfeldblum
Differential Revision:
D2742024
Pulled By: fredemmott
fb-gh-sync-id:
c0ecc7edceeebd1f972b0be4f4740214211f1c24
Andrii Grynenko [Thu, 17 Dec 2015 21:03:15 +0000 (13:03 -0800)]
Fix EventBaseLoopController destruction races
Summary:
Existing scheduleThreadSafe implementation had 2 potential races on destruction:
1. (very unlikely) insertHead is complete, but fiber loop is already running on another thread, so it finishes processing all of the fibers, destroys FiberManager or EventBase or both. By the time we get to scheduleThreadSafe EventBaseLoopController is already destoyed
2. (more likely) scheduleThreadSafe is complete, but FiberManager loop which is already running, picks complete fiber, finishes the processing. After that FiberManager may be destoyed. So when EventBase actually executes the callback FiberManager is already dead.
This solution fixes both races. Holding the alive shared_ptr when completing sheduleThreadSafe assures EventBase can't be destoyed until its completed (or it won't try to schedule anything after EventBase was destroyed). Locking alive weak_ptr in the EventBase loop callback ensures FiberManager and thus EventBaseLoopController were not destroyed yet (they can be destoyed only by the same thread which is running EventBase loop).
Reviewed By: spalamarchuk
Differential Revision:
D2763206
fb-gh-sync-id:
1972d6c0c11aa931747ebdaed4029a209130f69c
Michael Lee [Thu, 17 Dec 2015 18:57:47 +0000 (10:57 -0800)]
Allow override for unaligned reads in SpookyHashV2
Summary: Add configuration for disabling unaligned reads in SpookyHashV2
Reviewed By: yfeldblum
Differential Revision:
D2750885
fb-gh-sync-id:
9f48dcdfd5af05478a38e354f6fa0332b1332c14
Bin Liu [Thu, 17 Dec 2015 00:48:11 +0000 (16:48 -0800)]
Put optimized memcpy into folly
Summary: There is an optimized assembler version of memcpy that is showing 1.5% gain on TAO, add it to folly
Reviewed By: yfeldblum
Differential Revision:
D2218473
fb-gh-sync-id:
d5ac7f5ab30ff6febe7e94b017766c68dbd8934d
Lovro Puzar [Tue, 15 Dec 2015 16:13:53 +0000 (08:13 -0800)]
Support gzip in folly/io:compression
Summary: gzip files can be easier to work with operationally than raw zlib thanks to gzip+gunzip. zlib supports adding/stripping the gzip header; expose that through a new CodecType.
Reviewed By: chipturner
Differential Revision:
D2759554
fb-gh-sync-id:
67e0f5f88e087d363db1c349cabb909fb3f00417
Neel Goyal [Tue, 15 Dec 2015 14:35:14 +0000 (06:35 -0800)]
Update SSLContext to use folly::Random and discrete_distribution
Summary:
Change next pick logic to use `std::discrete_distribution`. Use folly::Random::DefaultGenerator as the generator.
Update the TARGETS to have sslcontext depend on random (which was the build break in
D2741855)
Reviewed By: siyengar
Differential Revision:
D2749166
fb-gh-sync-id:
7a6b400173b1e8c69a961f9e75e97d4777d5467a
Yang Chi [Mon, 14 Dec 2015 23:33:49 +0000 (15:33 -0800)]
Add a per-socket buffer callback
Summary: this is way simpler than
D2623385 +
D2709121. There will be a followup diff to clean the existing per-write call BufferCallback. The new one is on per-socket basis, much straightforward. I will only setup this in HTTPUpstreamSession.
Reviewed By: afrind
Differential Revision:
D2723493
fb-gh-sync-id:
6b1c21a719281b9693330b6a4074f7149d7c342a
Yang Chi [Mon, 14 Dec 2015 23:33:46 +0000 (15:33 -0800)]
Remove per-write buffer callback from AsyncSocket
Summary: Remove per-write buffer callback from AsyncSocket
Reviewed By: afrind
Differential Revision:
D2733142
fb-gh-sync-id:
91214a8c833bbd479bf049c2bb72d660e0c30f50
Louis Brandy [Fri, 11 Dec 2015 17:21:43 +0000 (09:21 -0800)]
folly::dynamic::operator= for std::initializer_list
Summary:
If we're going to have a constructor for it, we should have operator=.
Otherwise, gcc-4.9 goes via the copy constructor effectively "ignoring" the attempted nesting for e.g..
```
d = { other_dynamic }; // Should be ARRAY containing dynamic
```
NOTE: this only fixes gcc-4.9, there's still issues in clang.
Reviewed By: yfeldblum
Differential Revision:
D2745180
fb-gh-sync-id:
667787c788fc7c131d8a34c608c355f5b875be50
Andrii Grynenko [Fri, 11 Dec 2015 02:16:51 +0000 (18:16 -0800)]
Fix RefCountTest and RCURefCount race
Reviewed By: alikhtarov
Differential Revision:
D2741459
fb-gh-sync-id:
c4bd068cf735ae25364edba40960096fb35e8c43
Philip Pronin [Fri, 11 Dec 2015 00:15:37 +0000 (16:15 -0800)]
setup signal handler with SA_ONSTACK
Summary:
By default signal handlers are run on the signaled thread's stack.
In case of stack overflow running the `SIGSEGV` signal handler on
the same stack leads to another `SIGSEGV` and crashes the program
Use `SA_ONSTACK`, so alternate stack is used (only if configured via
`sigaltstack`).
Reviewed By: luciang
Differential Revision:
D2747021
fb-gh-sync-id:
48388acd6147e2919412ec32acfca1ca76f22a16
Yedidya Feldblum [Thu, 10 Dec 2015 20:15:41 +0000 (12:15 -0800)]
Use SharedMutex as the default mutex type in Synchronized
Summary:
[Folly] Use `SharedMutex` as the default mutex type in `Synchronized`.
`folly::SharedMutex` is, overall, a faster mutex than `boost::shared_mutex`. Let's upgrade.
Reviewed By: nbronson
Differential Revision:
D2743132
fb-gh-sync-id:
e36881495e8e1002cd481607e05e555c5da19b9a
Neel Goyal [Thu, 10 Dec 2015 15:10:49 +0000 (07:10 -0800)]
D2741855 broke my wangle. Reverting
Summary: Revert
D2741855
Reviewed By: mzlee
Differential Revision:
D2744015
fb-gh-sync-id:
b1e9b0a5ab95cb988d2b5c08c86139452b092465
Barney Moss [Thu, 10 Dec 2015 10:37:18 +0000 (02:37 -0800)]
Unit test showing array assignment problem on gcc 4.9
Summary: Unit test showing array assignment problem on gcc 4.9
Reviewed By: lbrandy
Differential Revision:
D2739900
fb-gh-sync-id:
6b0fde956672b8248fbd4e620fd112195c45c646
Andrii Grynenko [Thu, 10 Dec 2015 04:03:34 +0000 (20:03 -0800)]
Add ReadMostlySharedPtr and TLRefCount to Makefile.am
Reviewed By: pavlo-fb
Differential Revision:
D2743230
fb-gh-sync-id:
4a52d1758a43fd103318bbe0913776340eaa6af7
Arjen Roodselaar [Thu, 10 Dec 2015 01:51:27 +0000 (17:51 -0800)]
Fix automake build
Summary: There is slight drift in external deps between the TARGETS and automake files, keeping tests from linking in the 'make check' step when building folly outside of fbcode.
Reviewed By: yfeldblum
Differential Revision:
D2740482
fb-gh-sync-id:
ae6243f9bb112a7d2680017b3d577a1f23ef4f60
Neel Goyal [Thu, 10 Dec 2015 00:21:19 +0000 (16:21 -0800)]
Update SSLContext to use discrete_distribution
Summary: Update the protocol pick logic to use discrete_distribution
Reviewed By: siyengar
Differential Revision:
D2741855
fb-gh-sync-id:
244bd087124a7a9584a1108fe8f8150093275878
Andrii Grynenko [Wed, 9 Dec 2015 23:09:05 +0000 (15:09 -0800)]
Switch folly::Singleton to ReadMostlySharedPtr
Summary:
This also introduces a new try_get_fast() API which returns ReadMostlySharedPtr.
We should eventually migrate all users onto this API.
Reviewed By: elsteveogrande
Differential Revision:
D2706745
fb-gh-sync-id:
d558d705f431fc6147334c188fd0c6156644ba5c
Alan Frindell [Wed, 9 Dec 2015 19:55:40 +0000 (11:55 -0800)]
Add support for ALPN
Summary: With openssl-1.0.2 and later add support for ALPN. Clients can request NPN only, but the default is to support either (client will send ALPN list, server will send NPN advertisement if ALPN is not negotiated).
Reviewed By: siyengar
Differential Revision:
D2710441
fb-gh-sync-id:
a8efe69e1869bbecb4ed9e0a513448fcfdb21ca6
Shijin Kong [Tue, 8 Dec 2015 23:29:19 +0000 (15:29 -0800)]
notification queue read -> readNoInt
Summary: EINTR should be retried
Reviewed By: afrind
Differential Revision:
D2735879
fb-gh-sync-id:
6f25cd26e29ab8e7e13a3d63f24bf2d104c14809
Chad Parry [Tue, 8 Dec 2015 22:37:00 +0000 (14:37 -0800)]
Avoid cost of cancelAll
Summary:
This is an experimental diff that may improve the performance of `HHWheelTimer::cancelAll`. If there are no timers to cancel, then it can bail early.
Since perflab is unlikely to be conclusive, and since this diff looks like a strict improvement, I'll proceed with landing it.
Reviewed By: djwatson
Differential Revision:
D2735297
fb-gh-sync-id:
9f5a811ee6d9fa9434576e9abd3ef3443a7579b7
Andrii Grynenko [Tue, 8 Dec 2015 20:51:40 +0000 (12:51 -0800)]
Fix TLRefCount race around thread local destruction and fix RefCount unit test
Reviewed By: pavlo-fb
Differential Revision:
D2708425
fb-gh-sync-id:
665d077210503df4f4e8aa8f88ce5b9b277582f3
Chip Turner [Tue, 8 Dec 2015 15:52:02 +0000 (07:52 -0800)]
Update zstd to 0.4.2
Summary: New version.
Reviewed By: cyan
Differential Revision:
D2730844
fb-gh-sync-id:
4305bdfba2e8d25eba295bbf3f13a140c0d04a1c
Neel Goyal [Tue, 8 Dec 2015 13:18:02 +0000 (05:18 -0800)]
Mark some methods virtual on SSL Context.
Summary:
Mark some methods as virtual. This makes writing some
better unit tests on the associated task possible.
Reviewed By: siyengar
Differential Revision:
D2730126
fb-gh-sync-id:
f79f7632e873cb218bb49883d9b53beea7bb96e7
John Sherwood [Mon, 7 Dec 2015 20:05:28 +0000 (12:05 -0800)]
add type info to broken promise what()
Summary: when a promise is broken, add type info to the message. This will allow for some small improvements to debugging errors from broken promises, namely giving developers a slightly more refined target to grep for.
Reviewed By: fugalh
Differential Revision:
D2725575
fb-gh-sync-id:
b92edbde874e63eaeab97395da7bb52f76968800
Kyle Nekritz [Mon, 7 Dec 2015 19:04:13 +0000 (11:04 -0800)]
Fix segfaults in AsyncSSLSocket when used in unencrypted mode.
Reviewed By: djwatson
Differential Revision:
D2720027
fb-gh-sync-id:
1d56aabedeaf65a5d48e152eaee5c81695be7620
Bartek Ryniec [Fri, 4 Dec 2015 21:07:15 +0000 (13:07 -0800)]
json: Add option to parse numbers as strings
Summary: Adding a flag in `folly::json::serialization_opts` that allows you to request all numbers to be parsed into strings. This saves you from exceptions being thrown by `folly::parseJson` if it encounters numbers that wouldn't fit in int64.
Reviewed By: yfeldblum
Differential Revision:
D2713039
fb-gh-sync-id:
4f3d0d20f3012efd6229adf895200a3901bb4493
Kyle Nekritz [Fri, 4 Dec 2015 17:18:44 +0000 (09:18 -0800)]
Make getWrappedTransport/getUnderlyingTransport const.
Reviewed By: mzlee
Differential Revision:
D2709222
fb-gh-sync-id:
67a00a49bbcef5572b6092d7bfba1842d985e643
Kyle Nekritz [Fri, 4 Dec 2015 17:18:39 +0000 (09:18 -0800)]
Remove getBytesRead and getBytesWritten from folly::AsyncSSLSocket.
Summary: getRawBytesReceived and getRawBytesWritten should be used instead.
Reviewed By: mzlee
Differential Revision:
D2708684
fb-gh-sync-id:
2de87dc2b8c3466f94293a3d7df89fc5fce3a00f
Alexey Spiridonov [Fri, 4 Dec 2015 09:35:41 +0000 (01:35 -0800)]
Add an "after fork, before exec" callback
Summary:
In rare cases, it is required to run code **in the child process**, before it calls `exec()` because you need to change the state of the process **before** the child binary gets to run. This diff adds a callback to permit this. The callback is deliberately harder to use than an `std::function`, because you **REALLY HAVE TO KNOW WHAT YOU ARE DOING** to use this. And you have to check your work three times. Even glog `LOG()` must not be called from inside this callback. Your random library of choice is also probably unsafe.
This diff is primarily applicable to job managers and shells. For example, these tasks benefit from this callback: adding a child to a Linux `cgroup`, twiddling its various POSIX process attributes.
Implementing a correct callback for the post-vfork environment is fraught with peril. Read http://ewontfix.com/7/, and the docstrings in this diff.
Reviewed By: yfeldblum
Differential Revision:
D2688323
fb-gh-sync-id:
aae49e2b3957ca845895acca26e9cb44df1afc07
Umair Sadiq [Fri, 4 Dec 2015 05:38:07 +0000 (21:38 -0800)]
fix for negative delay in runAfterDelay
Summary: Currently, the tryRunAfterDelay implicitly converts a negative delay in int to unsigned int. With this change, the api makes it clear than only unsinged int is accepted as valid delay
Reviewed By: yfeldblum
Differential Revision:
D2710034
fb-gh-sync-id:
cb3c6b91bb86bbe07fa0e590b2ea871242d19745
Steve O'Brien [Thu, 3 Dec 2015 03:13:22 +0000 (19:13 -0800)]
ThreadLocalDetail: OSX fixes
Summary:
Change mutex to MicroSpinLock, for constexpr-ctor-ness on OSX.
Fixes this error seen on mac:
ThreadLocalDetail.h:202:38: note: non-literal type 'std::mutex' cannot be used in a constant expression
Reviewed By: ldemailly
Differential Revision:
D2713511
fb-gh-sync-id:
bc1c34b4a0ee21347278aa368b408f286345e050
Steve O'Brien [Wed, 2 Dec 2015 17:17:38 +0000 (09:17 -0800)]
ThreadLocalDetail: fix bug just introduced w/ recent change to constexpr-ctor style class
Summary: Small, silly bug which is triggered when there are no keys to unregister, invalidating the current size of the structure.
Reviewed By: yzhan
Differential Revision:
D2712604
fb-gh-sync-id:
f6f031936d8e4e458b49c08714f42cb2eec4c4fe
Tudor Bosman [Wed, 2 Dec 2015 16:57:22 +0000 (08:57 -0800)]
Add Optional::value_type
Summary: Comes in handy, and std::experimental::optional has it.
Reviewed By: yfeldblum
Differential Revision:
D2708324
fb-gh-sync-id:
87cdf2449f356d3baf9fb56a608e9c43d8fe79ae
Steve O'Brien [Wed, 2 Dec 2015 14:06:32 +0000 (06:06 -0800)]
folly: ThreadLocalDetail: make PthreadKeyUnregister constexpr-constructible, avoid SIOF
Summary:
Since this is used in so many places during the program's static-initialization (at startup), this class itself could be (has been demonstrated to be) a point of SIOF problems itself.
Made this class constexpr-constructible, so it doesn't need to be part of static initialization, making it SIOF-proof.
Reviewed By: luciang
Differential Revision:
D2709231
fb-gh-sync-id:
f248c9f2848c09045e000cfdc03636d847e522c9
Chad Parry [Wed, 2 Dec 2015 07:36:55 +0000 (23:36 -0800)]
Recommend a different exception_handler idiom
Summary: This is a //tiny// documentation change. I noticed that there are lots of places in the codebase where people follow this recommended usage of `with_exception` in an if–else chain, (e.g., https://fburl.com/
181606454). IMHO, a short-circuiting `||` operator is easier to read and write.
Reviewed By: yfeldblum
Differential Revision:
D2687840
fb-gh-sync-id:
6354555aeba81dcfe7adf17e7e573de0b1206b37
Andrii Grynenko [Wed, 2 Dec 2015 01:23:48 +0000 (17:23 -0800)]
Properly init collectDone_
Summary: facepaw
Reviewed By: djwatson
Differential Revision:
D2709030
fb-gh-sync-id:
20fe50bc70e97c4fedeae8fd271a4de70fa0e805
Yedidya Feldblum [Tue, 1 Dec 2015 20:04:24 +0000 (12:04 -0800)]
Avoid requiring a suppression (unevaluated-expression)
Summary: [Folly] Avoid requiring a suppression (`unevaluated-expression`).
I am not a fan of suppressing warnings. This warning comes from the internals of gtest, but anyway.
Replace with something that does not trigger the warning.
Reviewed By: Gownta
Differential Revision:
D2706279
fb-gh-sync-id:
2ecac1d230308a8fd21a69565193c52585526448
Daniel Andersson [Tue, 1 Dec 2015 15:46:00 +0000 (07:46 -0800)]
Enable find/emplace for key types other than KeyT.
Summary: Add template parameters to support arbitrary types when looking up a key.
This is useful to avoid the cost of 'materializing' a key which is likely to be present in the array (or, alternatively, switching on a tagged union KeyT).
Use the new capability to greatly simplify the lookup logic in HHVMs static string table.
Reviewed By: nbronson
Differential Revision:
D2662451
fb-gh-sync-id:
707fa033f350b80ca8080af17f1a8a74c59f2e88
Jim Meyering [Tue, 1 Dec 2015 15:27:54 +0000 (07:27 -0800)]
folly/Conv.h: suppress -Wfloat-conversion warnings
Summary: Using -Werror and -Wfloat-conversion was causing failure in code that is deliberately
performing that conversion. Add explicit casts to avoid the warnings/errors.
Here are two of the diagnostics:
./folly/Conv.h: In instantiation of "typename std::enable_if<std::is_floating_point<_Tp>::value, Tgt>::type folly::to(folly::StringPiece) [with Tgt = float; typename std::enable_if<std::is_floating_point<_Tp>::value, Tgt>::type = float; folly::StringPiece = folly::Range<const char*>]":
./thrift/lib/cpp2/protocol/SimpleJSONProtocol.tcc:687:28: required from "T apache::thrift::SimpleJSONProtocolReader::castIntegral(const string&) [with T = float; std::string = std::basic_fbstring<char>]"
./thrift/lib/cpp2/protocol/SimpleJSONProtocol.tcc:680:26: required from "uint32_t apache::thrift::SimpleJSONProtocolReader::readJSONKey(T&) [with T = float; uint32_t = unsigned int]"
./thrift/lib/cpp2/protocol/SimpleJSONProtocol.tcc:654:16: required from "uint32_t apache::thrift::SimpleJSONProtocolReader::readInContext(T&) [with T = float; uint32_t = unsigned int]"
./thrift/lib/cpp2/protocol/SimpleJSONProtocol.tcc:1035:34: required from here
./folly/Conv.h:1222:31: error: conversion to "float" from "std::enable_if<true, double>::type {aka double}" may alter its value [-Werror=float-conversion]
./folly/Conv.h: In instantiation of "typename std::enable_if<((std::is_integral<_Tp2>::value && std::is_floating_point<_Tp>::value) || (std::is_floating_point<_DInputType>::value && std::is_integral<_Tp>::value)), Tgt>::type folly::to(const Src&) [with Tgt = long int; Src = double; typename std::enable_if<((std::is_integral<_Tp2>::value && std::is_floating_point<_Tp>::value) || (std::is_floating_point<_DInputType>::value && std::is_integral<_Tp>::value)), Tgt>::type = long int]":
./folly/dynamic-inl.h:636:51: required from "T folly::dynamic::asImpl() const [with T = long int]"
./folly/dynamic-inl.h:384:64: required from here
./folly/Conv.h:1245:16: error: conversion to "long int" from "double" may alter its value [-Werror=float-conversion]
This change deserves a little more explanation:
- return ceil((double(sizeof(IntegerType) * CHAR_BIT) * M_LN2) / M_LN10);
+ return (unsigned int)(ceil(sizeof(IntegerType) * CHAR_BIT * M_LN2 / M_LN10));
In addition to adding the cast to destination type, I have also removed the unnecessary cast to double of the first product (there is not risk that it will overflow, and no need to promote to double, there), and I have also removed the grouping parentheses, since the standard left-to-right evaluation works just fine here.
Reviewed By: Gownta, yfeldblum
Differential Revision:
D2703928
fb-gh-sync-id:
23b49281d9120d106f1fdbd30921e8f39c01367d
Yedidya Feldblum [Tue, 1 Dec 2015 06:54:20 +0000 (22:54 -0800)]
CodeMod: Use the ExceptionWrapper::with_exception variant that deduces exception types
Summary: CodeMod: Use the `ExceptionWrapper::with_exception` variant that deduces exception types.
Since we must specify the exception type in the lambda arg, and there is a variant of `ExceptionWrapper::with_exception` that deduces the type of the exception from the type of the lambda arg, we don't need to specify the exception type again as a template parameter anymore.
Reviewed By: meyering
Differential Revision:
D2694895
fb-gh-sync-id:
505469f9008973a315e836f356e5db97df4ec921
Jim Meyering [Tue, 1 Dec 2015 05:17:20 +0000 (21:17 -0800)]
folly/Conv.h: estimateSpaceNeeded: avoid undefined behavior
Summary: Do not negate signed numbers like INT_MIN or INTMAX_MIN, since
that would evoke undefined behavior. Otherwise, the test (below)
would fail with this run-time error:
[ RUN ] Conv.Integral2String
folly/Conv.h:521:47: runtime error: negation of -
2147483648 cannot be represented in type 'int'; cast to an unsigned type to negate this value to itself
Reviewed By: markisaa
Differential Revision:
D2704195
fb-gh-sync-id:
4036437fb972109672004163880078127e7df797
Yedidya Feldblum [Mon, 30 Nov 2015 23:29:32 +0000 (15:29 -0800)]
Avoid deprecated Singleton<T>::get() in folly/futures
Summary: [Folly] Avoid deprecated `Singleton<T>::get()` in `folly/futures`.
Changes `getTimekeeperSingleton()` to return a `shared_ptr<Timekeeper>`, and patches its call-sites.
Additionally, have the call-sites keep the singleton instance alive for the duration that it is being directly used.
Reviewed By: ddrcoder
Differential Revision:
D2702361
fb-gh-sync-id:
82b72ee514dc4f2a7f7522af8b2e92b34df063d6
Andrii Grynenko [Thu, 26 Nov 2015 20:33:16 +0000 (12:33 -0800)]
Deprecate get_weak()
Summary: get_weak() is always used with .lock(), but try_get() is actually more performant than get_weak().lock().
Using get_weak() to store a weak_ptr and keep locking is not safe in fork scenarios.
Reviewed By: yfeldblum
Differential Revision:
D2694223
fb-gh-sync-id:
908d44293ffd9b3782152d43e28d5de172d1654a
Yedidya Feldblum [Thu, 26 Nov 2015 05:48:38 +0000 (21:48 -0800)]
Fix ExceptionWrapper::with_exception to support lvalue ref functors
Summary: [Folly] Fix `ExceptionWrapper::with_exception` to support lvalue ref functors.
Ex:
auto handler = [&](const std::runtime_error& e) { handle_runtime_error(e); };
exception_wrapper wrap = get_some_exception_wrapper_from_context();
wrap.with_exception(handler); // broken before this diff, fixed after
Reviewed By: markisaa
Differential Revision:
D2698680
fb-gh-sync-id:
4976ba08e3601e22891d00d79a5dae5118887b71
Andrii Grynenko [Thu, 26 Nov 2015 04:03:55 +0000 (20:03 -0800)]
New ReadMostlySharedPtr implementation
Summary: This changes ReadMostlySharedPtr API to have 3 types: MainPtr, WeakPtr, SharedPtr. MainPtr and SharedPtr are equivalents of std::shared_ptr, and WeakPtr is an equivalent of std::weak_ptr. The only difference is that it can only be a single MainPtr, and while it's alive copying SharedPtr/WeakPtr or WeakPtr doesn't require atomic operations (and thus can be more performant than std::shared_ptr). Unlike original ReadMostlySharedPtr API, there're no thread-safety guarantees between reset() and getShared() for ReadMostlySharedPtr.
ReadMostlySharedPtr can work with different RefCount implementations. This diff introduces RCURefCount (which is currently using liburcu) and TLRefCount.
Reviewed By: djwatson
Differential Revision:
D2683572
fb-gh-sync-id:
a7a03af4b1cf5f81a613368c6eebe70b2eaef064
Mark Isaacson [Wed, 25 Nov 2015 22:25:16 +0000 (14:25 -0800)]
Fix proxygen OSS build
Summary: So far this is a speculative fix.. I am waiting to see the results from the Ubuntu machine, but I suspect this is the problem.
Reviewed By: bugok
Differential Revision:
D2697600
fb-gh-sync-id:
e33d452a52c5ace2b7401d6495621ce68a5076b5
Giuseppe Ottaviano [Wed, 25 Nov 2015 21:49:55 +0000 (13:49 -0800)]
Update symlinks
Summary: See
D2696092
Reviewed By: meyering
Differential Revision:
D2696360
fb-gh-sync-id:
2cfa9258be899a93a58c59dae70a10fef6f457c8
Giuseppe Ottaviano [Wed, 25 Nov 2015 21:39:14 +0000 (13:39 -0800)]
Skip reallocation in shrink_to_fit for empty fbvector
Reviewed By: philippv
Differential Revision:
D2696314
fb-gh-sync-id:
105de2d99bb9f5ac82310e37f9302022c103befc
Mark Isaacson [Wed, 25 Nov 2015 19:38:54 +0000 (11:38 -0800)]
Make folly's T_CHECK_TIMEOUT/T_CHECK_TIME_LT use SKIP() on failure
Summary: Make these more spurious failure-aware/tolerant.
Reviewed By: yfeldblum
Differential Revision:
D2689775
fb-gh-sync-id:
1a9b247b97cc3529b12f6f7b76a4af2e32822d45
Jim Meyering [Wed, 25 Nov 2015 19:06:24 +0000 (11:06 -0800)]
folly/detail/ThreadLocalDetail.h: avoid UBSAN-detected memcpy abuse
Summary: [technically, the existing code is probably a no-op on
all systems we care about, but since it is officially UB, switching
to a more strict platform could cause trouble, so it's worth fixing]
Calling memcpy with "nullptr" as 2nd argument is undefined, even when
the third argument is zero, and causes a failure when testing with an
UBSAN-enabled binary (-fsanitize=undefined).
Before this change, the buck-run test below would evoke this failure:
[ RUN ] ThreadLocalPtr.BasicDestructor
folly/detail/ThreadLocalDetail.h:533:29: runtime error: null pointer passed as argument 2, which is declared to never be null
third-party-buck/build/glibc/include/string.h:47:45: note: nonnull attribute specified here
Ironically, the failure of the target-determinator-buck_push_blocking test (due to an unrelated proxygen dep problem) would block me from landing this, so I am adding this line to override it.
Reviewed By: luciang, alexshap
Differential Revision:
D2692625
fb-gh-sync-id:
8bdc5cd2899705f39c9565d640921de1f363807d
Adrien Conrath [Wed, 25 Nov 2015 16:34:09 +0000 (08:34 -0800)]
Remove undefined behavior in goodMallocSize()
Reviewed By: ot
Differential Revision:
D2695741
fb-gh-sync-id:
d33263c1ffa5651d66f37992ce365dae157ed449
Giuseppe Ottaviano [Wed, 25 Nov 2015 08:26:31 +0000 (00:26 -0800)]
Make goodMallocSize always use nallocx
Summary: `goodMallocSize` is used extensively in `folly` data structures,
especially for containers optimized for small contents, such as
`fbstring` and `small_vector`.
However, it makes the design decision to align the allocation size to
a x86 cache line, forcing a minimum allocation size of `64` bytes,
despite jemalloc can provide smaller size classes (8, 16, 32,
48). This causes a large discontinuity between small contents that can
be inlined and heap-allocated contents:
- For `fbstring`, a string of 23 bytes (including terminator) occupies
24 bytes (`sizeof(fbstring)`), a string of 24 bytes occupies 24 + 64
+ allocation overhead when it could be 24 + 32 + allocation
overhead. The waste is more than 50%.
- For `small_vector<uint32_t, 1, uint32_t>`, for instance, a vector
with 1 element occupies 12 bytes, a vector with 2 elements occupies
12 + 64 + allocation overhead when it could be 12 + 8 + allocation
overhead. The waste is more than 250%.
With this diff we just trust jemalloc and always use `nallocx`. If a
data structure need cache-line alignment it should be implemented at
its level.
Reviewed By: elsteveogrande
Differential Revision:
D2688156
fb-gh-sync-id:
46548d4a91952e7c673d4f0997c4c067e03c190d
Yedidya Feldblum [Wed, 25 Nov 2015 04:26:15 +0000 (20:26 -0800)]
A well-known instance of folly::Unit
Summary: [Folly] A well-known instance of `folly::Unit`.
Reviewed By: jsedgwick
Differential Revision:
D2693426
fb-gh-sync-id:
fe595abc852500275a827b4f400cfda7091a55f7
Yedidya Feldblum [Wed, 25 Nov 2015 02:27:46 +0000 (18:27 -0800)]
ExceptionWrapper::with_exception variant, deducing exception type from functor type
Summary: [Folly] `ExceptionWrapper::with_exception` variant deducing exception type from functor type.
From the `README.md`:
// TODO(jsedgwick) infer exception type from the type of the function
Reviewed By: Gownta
Differential Revision:
D2685579
fb-gh-sync-id:
158bcd030019f28b4f392d17167ab0bd67cc1edb
Qinfan Wu [Tue, 24 Nov 2015 18:02:25 +0000 (10:02 -0800)]
Add sdallocx for jemalloc
Summary: Make jemalloc's sdallocx function available in folly and allow future code to
use sized deallocation.
Reviewed By: elsteveogrande, ot
Differential Revision:
D2687746
fb-gh-sync-id:
9ef8a16d2c63c4eb8cecab7639add423ef093c2d
Yedidya Feldblum [Tue, 24 Nov 2015 02:45:15 +0000 (18:45 -0800)]
Fix Build: folly/test/ThreadLocalTest
Summary: [Folly] Fix Build: `folly/test/ThreadLocalTest`.
The problem is that the to the shared lib used in the test is hardcoded. It's a relative path, so that's nice, but it has too many assumptions about where the build system deposits test binaries.
Instead, let the code simply look for the shared lib in the same directory as the test binary. This makes fewer assumptions on the build system.
Reviewed By: andriigrynenko
Differential Revision:
D2689047
fb-gh-sync-id:
648da079b7af5bc9893709d009e577347dd4a463
Haijun Zhu [Mon, 23 Nov 2015 22:25:47 +0000 (14:25 -0800)]
Fix HHWheelTimerTest.AtMostEveryN
Summary: If it takes too long to finish it is probably heavily loaded. Skip if
so.
Reviewed By: mmcduff
Differential Revision:
D2687346
fb-gh-sync-id:
282cac7e4df361f714de1089f11b098af79b1512
Chad Parry [Mon, 23 Nov 2015 20:00:54 +0000 (12:00 -0800)]
Simplify an exception handler
Summary: It looks like there is no reason to catch a `std::exception` and then `dynamic_cast` to a derived exception. We can just catch the derived exception directly.
whatcouldgowrong4
Reviewed By: yfeldblum
Differential Revision:
D2677284
fb-gh-sync-id:
0149e4d4aecc96257376d410f592620205441f66
Michael Oliver [Mon, 23 Nov 2015 19:50:20 +0000 (11:50 -0800)]
Benchmark std::to_string vs. folly::to<std::string>
Summary: We wanted to compare the benefits of using folly::to<std::string> directly, and decided this benchmarking might as well be contributed back to folly. Here are the results:
```
============================================================================
folly/test/ConvTest.cpp relative time/iter iters/s
============================================================================
...
----------------------------------------------------------------------------
u64ToStringClibMeasure(1) 157.88ns 6.33M
u64ToStringFollyMeasure(1) 212.52% 74.29ns 13.46M
----------------------------------------------------------------------------
u64ToStringClibMeasure(12) 158.14ns 6.32M
u64ToStringFollyMeasure(12) 212.55% 74.40ns 13.44M
----------------------------------------------------------------------------
u64ToStringClibMeasure(123) 152.26ns 6.57M
u64ToStringFollyMeasure(123) 210.95% 72.18ns 13.85M
----------------------------------------------------------------------------
u64ToStringClibMeasure(1234) 152.40ns 6.56M
u64ToStringFollyMeasure(1234) 204.82% 74.41ns 13.44M
----------------------------------------------------------------------------
u64ToStringClibMeasure(12345) 156.01ns 6.41M
u64ToStringFollyMeasure(12345) 208.90% 74.68ns 13.39M
----------------------------------------------------------------------------
u64ToStringClibMeasure(123456) 178.21ns 5.61M
u64ToStringFollyMeasure(123456) 238.77% 74.63ns 13.40M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
1234567) 161.38ns 6.20M
u64ToStringFollyMeasure(
1234567) 203.09% 79.46ns 12.58M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
12345678) 159.01ns 6.29M
u64ToStringFollyMeasure(
12345678) 192.42% 82.64ns 12.10M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
123456789) 169.07ns 5.91M
u64ToStringFollyMeasure(
123456789) 203.70% 83.00ns 12.05M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
1234567890) 166.56ns 6.00M
u64ToStringFollyMeasure(
1234567890) 187.21% 88.97ns 11.24M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
12345678901) 176.22ns 5.67M
u64ToStringFollyMeasure(
12345678901) 197.57% 89.20ns 11.21M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
123456789012) 174.74ns 5.72M
u64ToStringFollyMeasure(
123456789012) 182.52% 95.74ns 10.45M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
1234567890123) 178.47ns 5.60M
u64ToStringFollyMeasure(
1234567890123) 186.81% 95.54ns 10.47M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
12345678901234) 187.33ns 5.34M
u64ToStringFollyMeasure(
12345678901234) 184.05% 101.78ns 9.82M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
123456789012345) 198.16ns 5.05M
u64ToStringFollyMeasure(
123456789012345) 182.48% 108.59ns 9.21M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
1234567890123456) 182.56ns 5.48M
u64ToStringFollyMeasure(
1234567890123456) 164.19% 111.19ns 8.99M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
12345678901234567) 186.80ns 5.35M
u64ToStringFollyMeasure(
12345678901234567) 162.33% 115.07ns 8.69M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
123456789012345678) 195.08ns 5.13M
u64ToStringFollyMeasure(
123456789012345678) 165.51% 117.87ns 8.48M
----------------------------------------------------------------------------
u64ToStringClibMeasure(
1234567890123456789) 200.27ns 4.99M
u64ToStringFollyMeasure(
1234567890123456789) 164.62% 121.65ns 8.22M
----------------------------------------------------------------------------
u64ToStringClibMeasure(12345678901234567890U) 203.18ns 4.92M
u64ToStringFollyMeasure(12345678901234567890U) 162.38% 125.13ns 7.99M
----------------------------------------------------------------------------
...
============================================================================
```
Reviewed By: yfeldblum
Differential Revision:
D2680998
fb-gh-sync-id:
179fbd756c714c92b7230d3750af024865ecbb0a
Sara Golemon [Mon, 23 Nov 2015 19:40:18 +0000 (11:40 -0800)]
Add missing header files to folly-automake
Summary: Closes https://github.com/facebook/folly/issues/304
Reviewed By: yfeldblum
Differential Revision:
D2685289
fb-gh-sync-id:
563697fb01ee358472b9080058848d2b557ba466
Sara Golemon [Mon, 23 Nov 2015 19:39:26 +0000 (11:39 -0800)]
Fix minor typo in folly/Conv.h
Summary: 'flating point' => 'floating point'
Closes https://github.com/facebook/folly/issues/241
Reviewed By: yfeldblum
Differential Revision:
D2685280
fb-gh-sync-id:
2e9c2b5419d0c631af4a32c5ab6b7366033ca34c
Andrii Grynenko [Sat, 21 Nov 2015 05:25:11 +0000 (21:25 -0800)]
Fix folly/ThreadLocal with clang after PthreadKeyUnregister change
Summary: Make PthreadKeyUnregister a global singleton for all StaticMeta instances.
Reviewed By: luciang
Differential Revision:
D2678401
fb-gh-sync-id:
c0d58b483fa6f096b29aeb7df71897a75ea8c892
Lucian Grijincu [Fri, 20 Nov 2015 20:03:52 +0000 (12:03 -0800)]
Revert "Fix folly::ThreadLocal to work in a shared library"
Summary: This reverts commit
9d355f96c4fe073c889f28070d5952fa4eb007a0.
build-break
Reviewed By: sctaylor
Differential Revision:
D2680784
fb-gh-sync-id:
a2df233bcd3721ffbfcd6fd71ea3cabbebebd4c4
Haijun Zhu [Fri, 20 Nov 2015 18:20:54 +0000 (10:20 -0800)]
get back AsyncServerSocket's setMaxNumMessagesInQueue
Summary:
D2613681 codemod'ed a thrift server api but it also renamed that api
in AsyncServerSocket, which is a mistake. Fix that in AsyncServerSocket, and
all other places that calls this api.
Reviewed By: alandau, JoelMarcey
Differential Revision:
D2677837
fb-gh-sync-id:
0d91f1a623229e99be59ca9dcd27f1330a9a1b64
Anirudh Ramachandran [Fri, 20 Nov 2015 16:35:40 +0000 (08:35 -0800)]
Remove unnecessary resetClientHelloParsing callback in AsyncSSLSocket
Summary: Remove unnecessary resetClientHelloParsing callback which causes problems wiht
Openssl 1.0.2
Reviewed By: knekritz
Differential Revision:
D2664730
fb-gh-sync-id:
d1b55ae493b4c92627ad41e7bf85f1e1a777bd2b
Alexander Shaposhnikov [Fri, 20 Nov 2015 04:12:41 +0000 (20:12 -0800)]
Remove busy wait and support multiple wait
Summary: Remove busy wait from Future::wait.
If future.wait(timeout) has succeded we should
return a ready future, otherwise we should return a future
for the final result (not necessarily ready).
Reviewed By: yfeldblum
Differential Revision:
D2646860
fb-gh-sync-id:
62671d09073ad86e84df8c9257e961d2a8c2a339
Delyan Kratunov [Fri, 20 Nov 2015 00:25:09 +0000 (16:25 -0800)]
Expose move result from LockFreeRingBuffer::Cursor
Summary: Without feedback that a `moveForward` or `moveBackward` did nothing, a user
cannot implement this idiom safely:
```
while(rb.tryRead(entry, cursor)) {
doSomething(entry);
cursor.moveBackward();
}
```
In the case where the ring buffer has not overflowed and slot 0 is still
available, the reader will get stuck at it (read it continuously) until
the buffer overflows.
This diff allows the above example to become:
```
while(rb.tryRead(entry, cursor)) {
doSomething(etnry);
if (!cursor.moveBackward()) {
break;
}
}
```
Reviewed By: bryceredd
Differential Revision:
D2674975
fb-gh-sync-id:
a0c5857daf186ef19e203f90acc2145590f85c3b
Giuseppe Ottaviano [Thu, 19 Nov 2015 19:39:27 +0000 (11:39 -0800)]
Fix usingJEMalloc with Clang
Summary: Clang is too clever and in some contexts optimizes away the `malloc`, but we rely on a side-effect. Declaring the variable as static forces it to call `malloc`. We could free the pointer relying on the fact that the lambda is guaranteed to be called only once, but I feel more comfortable just leaking it (LSan won't complain).
Reviewed By: philippv
Differential Revision:
D2674769
fb-gh-sync-id:
1153a3ca226c6b7aa64c453bd61b036dcbf3ffcc
Anton Likhtarov [Thu, 19 Nov 2015 19:32:26 +0000 (11:32 -0800)]
Fix invalid DCHECK
Summary: There's a race between insert() and erase(): as soon as insert()
releases the lock (swaps kLockedKey_ with the actual key), an erase() might
jump in and invalidate the key.
As far as I can tell, this bug existed since the beginning.
Reviewed By: nbronson
Differential Revision:
D2673099
fb-gh-sync-id:
4721893d2ad4836e11acc0fb4ecb0dd7b2b69be1
Andrii Grynenko [Thu, 19 Nov 2015 19:30:20 +0000 (11:30 -0800)]
Fix folly::ThreadLocal to work in a shared library
Reviewed By: bmaurer
Differential Revision:
D2667499
fb-gh-sync-id:
463f86752240bd88761910de934ba25d6e62fafe
Kyle Nekritz [Wed, 18 Nov 2015 18:20:14 +0000 (10:20 -0800)]
Add getUnderlyingTransport to AsyncTransportWrapper.
Summary: Inspired by getSocketFromTransport from proxygen.
Reviewed By: siyengar
Differential Revision:
D2663937
fb-gh-sync-id:
f076215907cd06d6da3de033c57eec8a6a6ce320
Tom Jackson [Tue, 17 Nov 2015 19:01:28 +0000 (11:01 -0800)]
UTF8StringPiece, wrapping boost::u8_to_u32
Summary: For handling UTF8 strings better.
Reviewed By: yfeldblum
Differential Revision:
D1956771
fb-gh-sync-id:
e074f9f2c9b472f5e619fef25d8e17296847773c
Yedidya Feldblum [Mon, 16 Nov 2015 10:07:19 +0000 (02:07 -0800)]
Simplify some checks by using __CLANG_PREREQ
Summary: [Folly] Simplify some checks by using `__CLANG_PREREQ`.
Reviewed By: meyering
Differential Revision:
D2657979
fb-gh-sync-id:
80ff56bbab0e78465d71315b729ad14a09706ed5
Yedidya Feldblum [Sun, 15 Nov 2015 19:52:43 +0000 (11:52 -0800)]
Simplify some checks by using __CLANG_PREREQ
Summary: [Folly] Simplify some checks by using `__CLANG_PREREQ`.
Reviewed By: meyering
Differential Revision:
D2656842
fb-gh-sync-id:
e762fba423fb7cc08907d10dc10f8f46d93a9fe4
Andrii Grynenko [Fri, 13 Nov 2015 21:47:27 +0000 (13:47 -0800)]
Make collect work for types with no default constructors
Summary: This doesn't make the code less efficient, because RVO can't be used for CollectVariadicContext. Thus moving existing tuple vs constructing new tuple by moving all values from other tuple (where each value is wrapped in folly::Optional) should be pretty much the same.
Reviewed By: hannesr
Differential Revision:
D2650293
fb-gh-sync-id:
648a358bf093a0bb9d058a997af9bf59014ad77c
Yang Chi [Thu, 12 Nov 2015 23:34:45 +0000 (15:34 -0800)]
Allow a AsyncSocket to be corked the whole time
Summary: Add a new method to cork a socket in a persistent manner, instead of the current on-off manner. This is default to false. The liger part of turning this on will be in a separate diff.
I thought about whether I need to turn cork off based on some criteria to alleviate the perf degradation. The obvious things I can think off is just amount of data written as a threshold, or a timeout. But TCP is doing this already for us, unless we want the data threshold to be less than MSS, or we want the timeout to be less than 200ms. THoughts?
Reviewed By: shikong
Differential Revision:
D2639260
fb-gh-sync-id:
2821f669c9f72d5ac4c33195bb192fc4110ffe9d
Giuseppe Ottaviano [Thu, 12 Nov 2015 16:37:48 +0000 (08:37 -0800)]
Make Malloc.h self-contained
Reviewed By: philippv
Differential Revision:
D2643313
fb-gh-sync-id:
10b9f735725ce47fab4bbfaa5972b3863357365f
Giuseppe Ottaviano [Thu, 12 Nov 2015 16:37:14 +0000 (08:37 -0800)]
Disable assertions in FBString when used as std::string
Reviewed By: Gownta
Differential Revision:
D2643850
fb-gh-sync-id:
2c4bb844ea2006215b0637cb1ba08c636faefe05
Chad Parry [Thu, 12 Nov 2015 00:02:35 +0000 (16:02 -0800)]
Restore the definition of HHWheelTimer::UniquePtr
Summary: Changing the definition of `HHWheelTimer::UniquePtr` wasn't safe, because some clients were using that type outside of the `HHWheelTimer::newTimer` helper. I'm changing that part back. We'll still be able to proceed with my other codemod to `HHWheelTimer`, but we'll always have two different smart pointer types to manage: `UniquePtr` and `IntrusivePtr`.
Reviewed By: djwatson
Differential Revision:
D2644721
fb-gh-sync-id:
14685be62355f09d39c4139ef7186d60b5f48dcd
Stepan Palamarchuk [Wed, 11 Nov 2015 22:22:01 +0000 (14:22 -0800)]
Make Optional nothrow_default_constructible
Summary: It never throws.
Reviewed By: pavlo-fb
Differential Revision:
D2640886
fb-gh-sync-id:
cd643f8847f4bf5619415731484f91fb07116784
Yang Chi [Wed, 11 Nov 2015 21:32:04 +0000 (13:32 -0800)]
Add a buffer callback to AsyncSocket
Summary: This is probably easier than
D2612490. The idea is just to add a callback to write, writev and writeChain in AsyncSocket, so upper layer can know when data starts to buffer up
Reviewed By: mzlee
Differential Revision:
D2623385
fb-gh-sync-id:
98d32ca83871aaa4f6c75a769b5f1bf0b5d62c3e
Chad Parry [Wed, 11 Nov 2015 19:29:28 +0000 (11:29 -0800)]
Improve the DelayedDestruction smart pointer
Summary: There have been several ASAN bugs cropping up because the lifetime of an `HHWheelTimer` is not being manager properly. I think that people are too comfortable passing around a raw `HHWheelTimer*` even when it is difficult to prove correctness. The recommended solution used to be to create a `DestructorGuard` every time it was needed. There is enough friction there that not everyone is doing that like they should. We should make resource management easier---as easy as using raw pointers, in fact.
I've fixed the broken copy semantics of `DestructorGuard` and added the operators that allow it to be used as a smart pointer. I added the `IntrusivePtr` helper that can manage an arbitrary derived class of `DelayedDestructionBase`. Now, clients can all safely pass around an `IntrusivePtr` everywhere they used to use a raw pointer. They will get automatic resource management for free.
If you are not convinced that `DestructorGuard` should be changed, then note that the existing behavior is dangerously buggy. Consider the following innocent code that accidentally uses the implicitly-defined copy constructor:
auto d = DestructorGuard(p);
This results in undefined behavior, because `p` can be accessed after it is destroyed! The bug happens because the default copy constructor copies the raw pointer but doesn't increment the count.
In a separate diff, I'll change all clients who pass around a raw `HHWheelTimer*` to use an `IntrusivePtr<HHWheelTimer>` instead. Then we can even entertain a long-term plan of switching from intrusive pointers to the standard `shared_ptr`.
Reviewed By: djwatson
Differential Revision:
D2627941
fb-gh-sync-id:
58a934d64540d0bbab334adc4f23d31d507692da
Shaft Wu [Wed, 11 Nov 2015 04:40:17 +0000 (20:40 -0800)]
UNSYNCHRONIZED does NOT unlock the mutex
Summary: My colleague tuomaspelkonen discovered a weird UNSYNCHRONIZED issue a few weeks ago and we ever since stopped using it. Now I finally have some time to root cause it. It turns out UNSYNCHRONIZED unlock the mutex then lock the mutex again, because it copy constructs LockedPtr for overriding the name within the scope, and copy construct locks the mutex again. A one character fix here is to take a reference of LockedPtr instead of copy construct it. However since this is my first time look at the code here, please advise if this is horribly wrong or propose better fix. Also added a test to reproduce the issue without the fix as well as verify the fix.
Reviewed By: yfeldblum
Differential Revision:
D2633028
fb-gh-sync-id:
a9e8d39b08d4d1265979f8bdaae83619566d10a0
Alexander Shaposhnikov [Wed, 11 Nov 2015 03:11:56 +0000 (19:11 -0800)]
Revert my change (which broke down the cont build)
Summary: hg backout -r
c9f7b5f3185a
Revert my change (which broke down the cont build t9048692)
Reviewed By: djwatson
Differential Revision:
D2640797
fb-gh-sync-id:
51f196ac5a3560fde4dc8fe7bb6ef278d74136e5
Alexander Shaposhnikov [Tue, 10 Nov 2015 23:42:18 +0000 (15:42 -0800)]
Remove busy wait
Summary: Wait uses baton & callback running baton.post
when the original future is ready. However wrapping baton.post
with a funciton call (preparing a new value) adds the following race: baton.wait wakes up
before that function has call actually finished.
The explanation is the following: to prepare the value of the new future it's necessary
1. baton.post() 2. set the value (move constructor, memory operations, ...)
and this code is executed in a separate thread.
The main idea of this fix is to avoid creating a new future
(whose value is calculated using that 2-step procedure)
and set a callback instead. This callback will be executed when the future is ready and actually
it either will be the last step of promise.setValue or it will run immediately if the future
we are waiting for already contains a value.
Reviewed By: fugalh
Differential Revision:
D2636409
fb-gh-sync-id:
df3e9bbcc56a5fac5834ffecc63f1bcb94ace02c
Shijin Kong [Tue, 10 Nov 2015 23:06:52 +0000 (15:06 -0800)]
change log error back to check when pipe read fail in notification queue
Summary: Revert
D2632752
Reviewed By: afrind
Differential Revision:
D2638974
fb-gh-sync-id:
43d8135421510db7840a99f3c197119a0fd26c09
Steve O'Brien [Tue, 10 Nov 2015 14:28:14 +0000 (06:28 -0800)]
FBString: fix constructors so it compiles with newer Clang's
Summary: Using Clang 3.7, this minimal test program:
#include <folly/FBString.h>
struct S { folly::basic_fbstring<char> x; };
int main(int argc, char *argv[]) {
S s {};
return 0;
}
... breaks with the following error output:
FBStringTest.cpp:5:8: error: chosen constructor is explicit in copy-initialization
S s {};
^
./folly/FBString.h:1009:12: note: constructor declared here
explicit basic_fbstring(const A& = A()) noexcept {
^
FBStringTest.cpp:3:40: note: in implicit initialization of field 'x' with omitted initializer
struct S { folly::basic_fbstring<char> x; };
... because this `basic_fbstring` is used in a struct and the struct is being default-constructed.
This patch splits the (nevertheless-correct) one-default-arg constructor into two, to deconfuse clang and have it select the correct constructor to avoid this issue.
Reviewed By: matbd
Differential Revision:
D2632953
fb-gh-sync-id:
2c75ae85732678c31543f5cccc73d58317982f07
Shijin Kong [Mon, 9 Nov 2015 21:09:58 +0000 (13:09 -0800)]
change CHECK to LOG(ERROR) when pipe read error on NotificationQueue
Summary: The check was originally added to track down cpu pinning issue. In reality the check did get triggered, so change to soft error to calm the crazy nag bot. what's wrong with you.
We need to see the soft error log to find out what the error is.
Reviewed By: mzlee
Differential Revision:
D2632752
fb-gh-sync-id:
87c11b186f97f0eb4a6c5ac13a1117b280198673
Alexey Spiridonov [Mon, 9 Nov 2015 19:16:11 +0000 (11:16 -0800)]
Add a tag type for in-place construction
Summary: Without this tag type, it's impossible to use `Synchronized` with types like:
```
struct A {
A(int, const char*);
A(const A&) = delete;
A& operator=(const A&) = delete;
A(A&&) = delete;
A& operator=(A&&) = delete;
};
```
In-place construction solves this problem. Usage:
```
Synchronized a(construct_in_place, 5, "c");
```
Reviewed By: nbronson
Differential Revision:
D2610071
fb-gh-sync-id:
251fe8f8f6a2d7484dda64cf04dcacb998145230
Chad Parry [Mon, 9 Nov 2015 18:39:22 +0000 (10:39 -0800)]
Make sure there are no outstanding timers in the destructor
Summary: This is the solution for our `HHWheelTimer` crashes, as suggested in
D2617966#25.
djwatson also mentioned that maybe there should be some logging in the destructor if there are outstanding callbacks. I couldn't think of anything that would add to the `assert` that already exists in `destroy`. I'm open to suggestions though.
Reviewed By: djwatson
Differential Revision:
D2628154
fb-gh-sync-id:
f3db6e9384517c9bf3cbb60af8c1e711703a07fa
Igor Sugak [Sun, 8 Nov 2015 08:32:52 +0000 (00:32 -0800)]
folly/portability: add clang version check
Summary: Define `__CLANG_PREREQ` macro to check version of clang.
Reviewed By: yfeldblum
Differential Revision:
D2630325
fb-gh-sync-id:
3d666e554e8ddfc2c1fecd439aaf93f015829025
Chad Parry [Fri, 6 Nov 2015 19:52:10 +0000 (11:52 -0800)]
Callbacks should ref the HHWheelTimer
Summary: Callbacks sometimes outlive the `HHWheelTimer` that they reference. Then the `Callback` tries to reference the dead `HHWheelTimer` and it could either misbehave or crash. This was caught reliably by ASAN tests.
Since `HHWheelTimer` already supports intrusive ref counting, the solution is to acquire a reference within the `Callback`.
Reviewed By: djwatson
Differential Revision:
D2617966
fb-gh-sync-id:
02be9ffc5851c269d5933288a17ad394b33ac2dd