Double-checked Locking for Shared Pointers - c++

Disclaimer: I come from a Java background, and, as such, I have no clue on how a lot of the internals of C++ (and associated libraries) work.
I have read enough to know that double-checked locking is evil, and a correct and safe implementation of a singleton pattern requires the appropriate tools.
I believe the following code may be unsafe, subject to compiler reordering and assignment of uninitialized objects, but I am not sure if I am missing something that I do not understand about the language.
typedef boost::shared_ptr<A> APtr;
APtr g_a;
boost::mutex g_a_mutex;
const APtr& A::instance()
{
if (!g_a)
{
boost::mutex::scoped_lock lock(g_a_mutex);
if (!g_a)
{
g_a = boost::make_shared<A>();
}
}
return g_a;
}
I believe the actual code is compiled under C++03.
Is this implementation unsafe? If so, how?

Yes, the double-checked locking pattern is unsafe in archaic languages. I can't do any better than the explanation in What's wrong with this fix for double checked locking?:
The problem apparently is the line assigning instance -- the compiler is free to allocate the object and then assign the pointer to it, OR to set the pointer to where it will be allocated, then allocate it.
If you were using C++11 std::shared_ptr, you could use atomic_load and atomic_store on the shared pointer object, which are guaranteed to compose correctly with each other and with mutexes; but if you're using C++11 you may as well just use a dynamically-initialized static variable, which is guaranteed to be thread-safe.

All of this is absolutelty unneccesary in modern C++. Singleton code should be as simple as this for anybody but dinosaurs:
A& A::instance() {
static A a;
return a;
}
100% thread safe in C++11 and up.
However, I have a mandatory statement to make: Singletons are evil antipattern and should be banned forever.
On thread safety of original code
Yes, provided code is unsafe. Since the reading is done outside the mutex, it is totally possible that the modifications to g_a itself would be visible, but not that modifications to the object g_a is pointing to. As a result, the thread will bypass mutex locking and return a pointer to non-constructed object.

Related

Possible solution to initialization order fiasco?

Reading a lot about this well known problem I discovered and used the "Creation on first use", but it main problem is that it produces memory leaks.
Working a bit, I've done some light modification trying to solve that problem, and I got a solution:
1) Using a global static pointer instead a local static pointer
2) Using a separate function to delete the object.
Example:
static myObject* myObjectPtr=nullptr;
static myObjectDel() {delete myObjectPtr;}
inline myObject& myObjectRef()
{
if(myObjectPtr==nullptr)
{
myObjectPtr=new myObjectPtr;
atexit(myObjectDel);
}
return(*myObjectPtr);
}
So, I can ensure that:
1) The object is created before first use.
2) The object is deleted after last use.
3) There are no memory leaks.
It works with MS Visual Studio 2013 due the static pointer is zero-initialized at program startup, but I'm not sure if all compilers and operating systems does it.
In other words, Is it portable?
Thanks you and sorry for my poor english (I'm not english native)
Your solution is portable, but it's not thread-safe.
It's also needlessly complicated.
Don't use dynamic allocation:
inline myObject& myObjectRef()
{
static myObject theObject;
return theObject;
}
Leak-free (since the dawn of time) and thread-safe (since C++11).
I'd prefer static myObject* myObjectPtr; since all static pointers will be initialised to nullptr anyway. The C++ standard guarantees that.
But your code is vulnerable to two threads calling inline myObject& myObjectRef() for the first time which could lead to multiple new to be balanced with only one delete. You ought to use a mutex in that function to obviate that.

How to enable Rust Ownership paradigm in C++

The system programming language Rust uses the ownership paradigm to ensure at compile time with zero cost for the runtime when a resource has to be freed.
In C++ we commonly use smart pointers to achieve the same goal of hiding the complexity of managing resource allocation. There are a couple of differences though:
In Rust there is always only one owner, whereas C++ shared_ptr can easily leak ownership.
In Rust we can borrow references we do not own, whereas C++ unique_ptr cannot be shared in a safe way via weak_ptr and lock().
Reference counting of shared_ptr is costly.
My question is: How can we emulate the ownership paradigm in C++ within the following constraints:
Only one owner at any time
Possibility to borrow a pointer and use it temporarily without fear of the resource going out of scope (observer_ptr is useless for this)
As much compile-time checks as possible.
Edit: Given the comments so far, we can conclude:
No compile-time support for this (I was hoping for some decltype/template magic unknown to me) in the compilers. Might be possible using static analysis elsewhere (taint?)
No way to get this without reference counting.
No standard implementation to distinguish shared_ptrs with owning or borrowing semantic
Could roll your own by creating wrapper types around shared_ptr and weak_ptr:
owned_ptr: non-copyable, move-semantics, encapsulates shared_ptr, access to borrowed_ptr
borrowed_ptr: copyable, encapsulates weak_ptr, lock method
locked_ptr: non-copyable, move-semantics, encapsulates shared_ptr from locking weak_ptr
You can't do this with compile-time checks at all. The C++ type system is lacking any way to reason about when an object goes out of scope, is moved, or is destroyed — much less turn this into a type constraint.
What you could do is have a variant of unique_ptr that keeps a counter of how many "borrows" are active at run time. Instead of get() returning a raw pointer, it would return a smart pointer that increments this counter on construction and decrements it on destruction. If the unique_ptr is destroyed while the count is non-zero, at least you know someone somewhere did something wrong.
However, this is not a fool-proof solution. Regardless of how hard you try to prevent it, there will always be ways to get a raw pointer to the underlying object, and then it's game over, since that raw pointer can easily outlive the smart pointer and the unique_ptr. It will even sometimes be necessary to get a raw pointer, to interact with an API that requires raw pointers.
Moreover, ownership is not about pointers. Box/unique_ptr allows you to heap allocate an object, but it changes nothing about ownership, life time, etc. compared to putting the same object on the stack (or inside another object, or anywhere else really). To get the same mileage out of such a system in C++, you'd have to make such "borrow counting" wrappers for all objects everywhere, not just for unique_ptrs. And that is pretty impractical.
So let's revisit the compile time option. The C++ compiler can't help us, but maybe lints can? Theoretically, if you implement the whole life time part of the type system and add annotations to all APIs you use (in addition to your own code), that may work.
But it requires annotations for all functions used in the whole program. Including private helper function of third party libraries. And those for which no source code is available. And for those whose implementation that are too complicated for the linter to understand (from Rust experience, sometimes the reason something is safe are too subtle to express in the static model of lifetimes and it has to be written slightly differently to help the compiler). For the last two, the linter can't verify that the annotation is indeed correct, so you're back to trusting the programmer. Additionally, some APIs (or rather, the conditions for when they are safe) can't really be expressed very well in the lifetime system as Rust uses it.
In other words, a complete and practically useful linter for this this would be substantial original research with the associated risk of failure.
Maybe there is a middle ground that gets 80% of the benefits with 20% of the cost, but since you want a hard guarantee (and honestly, I'd like that too), tough luck. Existing "good practices" in C++ already go a long way to minimizing the risks, by essentially thinking (and documenting) the way a Rust programmer does, just without compiler aid. I'm not sure if there is much improvement over that to be had considering the state of C++ and its ecosystem.
tl;dr Just use Rust ;-)
What follows are some examples of ways people have tried to emulate parts of Rust's ownership paradigm in C++, with limited success:
Lifetime safety: Preventing common dangling. The most thorough and rigorous approach, involving several additions to the language to support the necessary annotations. If the effort is still alive (last commit was in 2019), getting this analysis added to a mainstream compiler is probably the most likely route to "borrow checked" C++. Discussed on IRLO.
Borrowing Trouble: The Difficulties Of A C++ Borrow-Checker
Is it possible to achieve Rust's ownership model with a generic C++ wrapper?
C++Now 2017: Jonathan Müller “Emulating Rust's borrow checker in C++" (video) and associated code, about which the author says, "You're not actually supposed to use that, if you need such a feature, you should use Rust."
Emulating the Rust borrow checker with C++ move-only types and part II (which is actually more like emulating RefCell than the borrow checker, per se)
I believe you can get some of the benefits of Rust by enforcing some strict coding conventions (which is after all what you'd have to do anyway, since there's no way with "template magic" to tell the compiler not to compile code that doesn't use said "magic"). Off the top of my head, the following could get you...well...kind of close, but only for single-threaded applications:
Never use new directly; instead, use make_unique. This goes partway toward ensuring that heap-allocated objects are "owned" in a Rust-like manner.
"Borrowing" should always be represented via reference parameters to function calls. Functions that take a reference should never create any sort of pointer to the refered-to object. (It may in some cases be necessary to use a raw pointer as a paramter instead of a reference, but the same rule should apply.)
Note that this works for objects on the stack or on the heap; the function shouldn't care.
Transfer of ownership is, of course, represented via R-value references (&&) and/or R-value references to unique_ptrs.
Unfortunately, I can't think of any way to enforce Rust's rule that mutable references can only exist anywhere in the system when there are no other extant references.
Also, for any kind of parallelism, you would need to start dealing with lifetimes, and the only way I can think of to permit cross-thread lifetime management (or cross-process lifetime management using shared memory) would be to implement your own "ptr-with-lifetime" wrapper. This could be implemented using shared_ptr, because here, reference-counting would actually be important; it's still a bit of unnecessary overhead, though, because reference-count blocks actually have two reference counters (one for all the shared_ptrs pointing to the object, another for all the weak_ptrs). It's also a little... odd, because in a shared_ptr scenario, everybody with a shared_ptr has "equal" ownership, whereas in a "borrowing with lifetime" scenario, only one thread/process should actually "own" the memory.
I think one could add a degree of compile-time introspection and custom sanitisation by introducing custom wrapper classes that track ownership and borrowing.
The code below is a hypothetical sketch, and not a production solution which would need a lot more tooling, e.g. #def out the checks when not sanitising. It uses a very naive lifetime checker to 'count' borrow errors in ints, in this instance during compilation. static_asserts are not possible as the ints are not constexpr, but the values are there and can be interrogated before runtime. I believe this answers your 3 constraints, regardless of whether these are heap allocations, so I'm using a simple int type to demo the idea, rather than a smart pointer.
Try uncommenting the use cases in main() below (run in compiler explorer with -O3 to see boilerplate optimise away), and you'll see the warning counters change.
https://godbolt.org/z/Pj4WMr
// Hypothetical Rust-like owner / borrow wrappers in C++
// This wraps types with data which is compiled away in release
// It is not possible to static_assert, so this uses static ints to count errors.
#include <utility>
// Statics to track errors. Ideally these would be static_asserts
// but they depen on Owner::has_been_moved which changes during compilation.
static int owner_already_moved = 0;
static int owner_use_after_move = 0;
static int owner_already_borrowed = 0;
// This method exists to ensure static errors are reported in compiler explorer
int get_fault_count() {
return owner_already_moved + owner_use_after_move + owner_already_borrowed;
}
// Storage for ownership of a type T.
// Equivalent to mut usage in Rust
// Disallows move by value, instead ownership must be explicitly moved.
template <typename T>
struct Owner {
Owner(T v) : value(v) {}
Owner(Owner<T>& ov) = delete;
Owner(Owner<T>&& ov) {
if (ov.has_been_moved) {
owner_already_moved++;
}
value = std::move(ov.value);
ov.has_been_moved = true;
}
T& operator*() {
if (has_been_moved) {
owner_use_after_move++;
}
return value;
}
T value;
bool has_been_moved{false};
};
// Safely borrow a value of type T
// Implicit constuction from Owner of same type to check borrow is safe
template <typename T>
struct Borrower {
Borrower(Owner<T>& v) : value(v.value) {
if (v.has_been_moved) {
owner_already_borrowed++;
}
}
const T& operator*() const {
return value;
}
T value;
};
// Example of function borrowing a value, can only read const ref
static void use(Borrower<int> v) {
(void)*v;
}
// Example of function taking ownership of value, can mutate via owner ref
static void use_mut(Owner<int> v) {
*v = 5;
}
int main() {
// Rather than just 'int', Owner<int> tracks the lifetime of the value
Owner<int> x{3};
// Borrowing value before mutating causes no problems
use(x);
// Mutating value passes ownership, has_been_moved set on original x
use_mut(std::move(x));
// Uncomment for owner_already_borrowed = 1
//use(x);
// Uncomment for owner_already_moved = 1
//use_mut(std::move(x));
// Uncomment for another owner_already_borrowed++
//Borrower<int> y = x;
// Uncomment for owner_use_after_move = 1;
//return *x;
}
The use of static counters is obviously not desirable, but it is not possible to use static_assert as owner_already_moved is non-const. The idea is these statics give hints to errors appearing, and in final production code they could be #defed out.
You can use an enhanced version of a unique_ptr (to enforce a unique owner) together with an enhanced version of observer_ptr (to get a nice runtime exception for dangling pointers, i.e. if the original object maintained through unique_ptr went out of scope). The Trilinos package implements this enhanced observer_ptr, they call it Ptr. I have implemented the enhanced version of unique_ptr here (I call it UniquePtr): https://github.com/certik/trilinos/pull/1
Finally, if you want the object to be stack allocated, but still be able to pass safe references around, you need to use the Viewable class, see my initial implementation here: https://github.com/certik/trilinos/pull/2
This should allow you to use C++ just like Rust for pointers, except that in Rust you get a compile time error, while in C++ you get a runtime exception. Also, it should be noted, that you only get a runtime exception in Debug mode. In Release mode, the classes do not do these checks, so they are as fast as in Rust (essentially as fast as raw pointers), but then they can segfault. So one has to make sure the whole test suite runs in Debug mode.

Why is 'this' not volatile?

Having spent the last few days debugging a multi-threading where one thread was deleting an object still in use by another I realised that the issue would have been far easier and quicker to diagnose if I could have made 'this' volatile. It would have changed the crash dump on the system (Symbian OS) to something far more informative.
So, is there any reason why it cannot be, or shouldn't be?
Edit:
So there really is no safe way to prevent or check for this scenario. Would it be correct to say that one solution to the accessing of stale class pointers is to have a global variable that holds the pointer, and any functions that are called should be statics that use the global variable as a replacement for 'this'?
static TAny* gGlobalPointer = NULL;
#define Harness static_cast<CSomeClass*>(gGlobalPointer);
class CSomeClass : public CBase
{
public:
static void DoSomething();
private:
int iMember;
};
void CSomeClass::DoSomething()
{
if (!Harness)
{
return;
}
Harness->iMember = 0;
}
So if another thread deleted and NULLed the global pointer it would be caught immediately.
One issue I think with this is that if the compiler cached the value of Harness instead of checking it each time it's used.
this is not a variable, but a constant. You can change the object referenced by this, but you can't change the value of this. Because constants never change, there is no need to mark them as volatile.
It wouldn't help: making a variable volatile means that the compiler will make sure it reads its value from memory each time it's accessed, but the value of this doesn't change, even if, from a different context or the same, you delete the object it's pointing at.
It can be. Just declare the member function as volatile.
struct a
{
void foo() volatile {}
};
volatile wouldn't help you.
It would make the access to a variable volatile, but won't wait for any method to complete.
use smart pointers.
shared_ptr
There's also an version in std in newer c++ versions.
If the object is being read whilst being deleted that is a clear memory error and has nothing to do with volatile.
volatile is intended to stop the compiler from "remembering" the value of a variable in memory that may be changed by a different thread, i.e. prevent the compiler from optimising.
e.g. if your class has a pointer member p and your method is accessing:
p->a;
p->b;
etc. and p is volatile within "this" so p->b could be accessing a different object than it was when it did p->a
If p has been destroyed though as "this" was deleted then volatile will not come to your rescue. Presumably you think it will be "nulled" but it won't be.
Incidentally it is also a pretty sound rule that if your destructor locks a mutex in order to protect another thread using the same object, you have an issue. Your destructor may lock due to it removing its own presence from an external object that needs synchronous activity, but not to protect its own members.

Understanding the meaning of the term and the concept - RAII (Resource Acquisition is Initialization)

Could you C++ developers please give us a good description of what RAII is, why it is important, and whether or not it might have any relevance to other languages?
I do know a little bit. I believe it stands for "Resource Acquisition is Initialization". However, that name doesn't jive with my (possibly incorrect) understanding of what RAII is: I get the impression that RAII is a way of initializing objects on the stack such that, when those variables go out of scope, the destructors will automatically be called causing the resources to be cleaned up.
So why isn't that called "using the stack to trigger cleanup" (UTSTTC:)? How do you get from there to "RAII"?
And how can you make something on the stack that will cause the cleanup of something that lives on the heap? Also, are there cases where you can't use RAII? Do you ever find yourself wishing for garbage collection? At least a garbage collector you could use for some objects while letting others be managed?
Thanks.
So why isn't that called "using the stack to trigger cleanup" (UTSTTC:)?
RAII is telling you what to do: Acquire your resource in a constructor! I would add: one resource, one constructor. UTSTTC is just one application of that, RAII is much more.
Resource Management sucks. Here, resource is anything that needs cleanup after use. Studies of projects across many platforms show the majority of bugs are related to resource management - and it's particularly bad on Windows (due to the many types of objects and allocators).
In C++, resource management is particularly complicated due to the combination of exceptions and (C++ style) templates. For a peek under the hood, see GOTW8).
C++ guarantees that the destructor is called if and only if the constructor succeeded. Relying on that, RAII can solve many nasty problems the average programmer might not even be aware of. Here are a few examples beyond the "my local variables will be destroyed whenever I return".
Let us start with an overly simplistic FileHandle class employing RAII:
class FileHandle
{
FILE* file;
public:
explicit FileHandle(const char* name)
{
file = fopen(name);
if (!file)
{
throw "MAYDAY! MAYDAY";
}
}
~FileHandle()
{
// The only reason we are checking the file pointer for validity
// is because it might have been moved (see below).
// It is NOT needed to check against a failed constructor,
// because the destructor is NEVER executed when the constructor fails!
if (file)
{
fclose(file);
}
}
// The following technicalities can be skipped on the first read.
// They are not crucial to understanding the basic idea of RAII.
// However, if you plan to implement your own RAII classes,
// it is absolutely essential that you read on :)
// It does not make sense to copy a file handle,
// hence we disallow the otherwise implicitly generated copy operations.
FileHandle(const FileHandle&) = delete;
FileHandle& operator=(const FileHandle&) = delete;
// The following operations enable transfer of ownership
// and require compiler support for rvalue references, a C++0x feature.
// Essentially, a resource is "moved" from one object to another.
FileHandle(FileHandle&& that)
{
file = that.file;
that.file = 0;
}
FileHandle& operator=(FileHandle&& that)
{
file = that.file;
that.file = 0;
return *this;
}
}
If construction fails (with an exception), no other member function - not even the destructor - gets called.
RAII avoids using objects in an invalid state. it already makes life easier before we even use the object.
Now, let us have a look at temporary objects:
void CopyFileData(FileHandle source, FileHandle dest);
void Foo()
{
CopyFileData(FileHandle("C:\\source"), FileHandle("C:\\dest"));
}
There are three error cases to handled: no file can be opened, only one file can be opened, both files can be opened but copying the files failed. In a non-RAII implementation, Foo would have to handle all three cases explicitly.
RAII releases resources that were acquired, even when multiple resources are acquired within one statement.
Now, let us aggregate some objects:
class Logger
{
FileHandle original, duplex; // this logger can write to two files at once!
public:
Logger(const char* filename1, const char* filename2)
: original(filename1), duplex(filename2)
{
if (!filewrite_duplex(original, duplex, "New Session"))
throw "Ugh damn!";
}
}
The constructor of Logger will fail if original's constructor fails (because filename1 could not be opened), duplex's constructor fails (because filename2 could not be opened), or writing to the files inside Logger's constructor body fails. In any of these cases, Logger's destructor will not be called - so we cannot rely on Logger's destructor to release the files. But if original was constructed, its destructor will be called during cleanup of the Logger constructor.
RAII simplifies cleanup after partial construction.
Negative points:
Negative points? All problems can be solved with RAII and smart pointers ;-)
RAII is sometimes unwieldy when you need delayed acquisition, pushing aggregated objects onto the heap.
Imagine the Logger needs a SetTargetFile(const char* target). In that case, the handle, that still needs to be a member of Logger, needs to reside on the heap (e.g. in a smart pointer, to trigger the handle's destruction appropriately.)
I have never wished for garbage collection really. When I do C# I sometimes feel a moment of bliss that I just do not need to care, but much more I miss all the cool toys that can be created through deterministic destruction. (using IDisposable just does not cut it.)
I have had one particularly complex structure that might have benefited from GC, where "simple" smart pointers would cause circular references over multiple classes. We muddled through by carefully balancing strong and weak pointers, but anytime we want to change something, we have to study a big relationship chart. GC might have been better, but some of the components held resources that should be release ASAP.
A note on the FileHandle sample: It was not intended to be complete, just a sample - but turned out incorrect. Thanks Johannes Schaub for pointing out and FredOverflow for turning it into a correct C++0x solution. Over time, I've settled with the approach documented here.
There are excellent answers out there, so I just add some things forgotten.
##0. RAII is about scopes
RAII is about both:
acquiring a resource (no matter what resource) in the constructor, and un-acquiring it in the destructor.
having the constructor executed when the variable is declared, and the destructor automatically executed when the variable goes out of scope.
Others already answered about that, so I won't elaborate.
##1. When coding in Java or C#, you already use RAII...
MONSIEUR JOURDAIN: What! When I say, "Nicole, bring me my slippers,
and give me my nightcap," that's prose?
PHILOSOPHY MASTER: Yes, Sir.
MONSIEUR JOURDAIN: For more than forty years I have been speaking prose without knowing anything about it, and I am much obliged to you for having taught me that.
— Molière: The Middle Class Gentleman, Act 2, Scene 4
As Monsieur Jourdain did with prose, C# and even Java people already use RAII, but in hidden ways. For example, the following Java code (which is written the same way in C# by replacing synchronized with lock):
void foo()
{
// etc.
synchronized(someObject)
{
// if something throws here, the lock on someObject will
// be unlocked
}
// etc.
}
... is already using RAII: The mutex acquisition is done in the keyword (synchronized or lock), and the un-acquisition will be done when exiting the scope.
It's so natural in its notation it requires almost no explanation even for people who never heard about RAII.
The advantage C++ has over Java and C# here is that anything can be made using RAII. For example, there are no direct build-in equivalent of synchronized nor lock in C++, but we can still have them.
In C++, it would be written:
void foo()
{
// etc.
{
Lock lock(someObject) ; // lock is an object of type Lock whose
// constructor acquires a mutex on
// someObject and whose destructor will
// un-acquire it
// if something throws here, the lock on someObject will
// be unlocked
}
// etc.
}
which can be easily written as it would be in Java/C# (using C++ macros):
#define LOCK(mm_mutex) \
if(Lock lock{mm_mutex}) {} \
else
void foo()
{
// etc.
LOCK(someObject)
{
// if something throws here, the lock on someObject will
// be unlocked
}
// etc.
}
##2. RAII have alternate uses
WHITE RABBIT: [singing] I'm late / I'm late / For a very important date. / No time to say "Hello." / Goodbye. / I'm late, I'm late, I'm late.
— Alice in Wonderland (Disney version, 1951)
You know when the constructor will be called (at the object declaration), and you know when its corresponding destructor will be called (at the exit of the scope), so you can write almost magical code with but a line. Welcome to the C++ wonderland (at least, from a C++ developer's viewpoint).
For example, you can write a counter object (I let that as an exercise) and use it just by declaring its variable, like the lock object above was used:
void foo()
{
double timeElapsed = 0 ;
{
Counter counter(timeElapsed) ;
// do something lengthy
}
// now, the timeElapsed variable contain the time elapsed
// from the Counter's declaration till the scope exit
}
which of course, can be written, again, the Java/C# way using a macro:
void foo()
{
double timeElapsed = 0 ;
COUNTER(timeElapsed)
{
// do something lengthy
}
// now, the timeElapsed variable contain the time elapsed
// from the Counter's declaration till the scope exit
}
##3. Why does C++ lack finally?
[SHOUTING] It's the final countdown!
— Europe: The Final Countdown (sorry, I was out of quotes, here... :-)
The finally clause is used in C#/Java to handle resource disposal in case of scope exit (either through a return or a thrown exception).
Astute specification readers will have noticed C++ has no finally clause. And this is not an error, because C++ does not need it, as RAII already handle resource disposal. (And believe me, writing a C++ destructor is magnitudes easier than writing the right Java finally clause, or even a C#'s correct Dispose method).
Still, sometimes, a finally clause would be cool. Can we do it in C++? Yes, we can! And again with an alternate use of RAII.
##Conclusion: RAII is a more than philosophy in C++: It's C++
RAII? THIS IS C++!!!
— C++ developer's outraged comment, shamelessly copied by an obscure Sparta king and his 300 friends
When you reach some level of experience in C++, you start thinking in terms of RAII, in terms of construtors and destructors automated execution.
You start thinking in terms of scopes, and the { and } characters become ones of the most important in your code.
And almost everything fits right in terms of RAII: exception safety, mutexes, database connections, database requests, server connection, clocks, OS handles, etc., and last, but not least, memory.
The database part is not negligible, as, if you accept to pay the price, you can even write in a "transactional programming" style, executing lines and lines of code until deciding, in the end, if you want to commit all the changes, or, if not possible, having all the changes reverted back (as long as each line satisfy at least the Strong Exception Guarantee). (see the second part of this Herb's Sutter article for the transactional programming).
And like a puzzle, everything fits.
RAII is so much part of C++, C++ could not be C++ without it.
This explains why experienced C++ developers are so enamored with RAII, and why RAII is the first thing they search when trying another language.
And it explains why the Garbage Collector, while a magnificient piece of technology in itself, is not so impressive from a C++ developer's viewpoint:
RAII already handles most of the cases handled by a GC
A GC deals better than RAII with circular references on pure managed objects (mitigated by smart uses of weak pointers)
Still A GC is limited to memory, while RAII can handle any kind of resource.
As described above, RAII can do much, much more...
RAII is using C++ destructors semantics to manage resources. For example, consider a smart pointer. You have a parameterized constructor of the pointer that initializes this pointer with the adress of object. You allocate a pointer on stack:
SmartPointer pointer( new ObjectClass() );
When the smart pointer goes out of scope the destructor of the pointer class deletes the connected object. The pointer is stack-allocated and the object - heap-allocated.
There are certain cases when RAII doesn't help. For example, if you use reference-counting smart pointers (like boost::shared_ptr) and create a graph-like structure with a cycle you risk facing a memory leak because the objects in a cycle will prevent each other from being released. Garbage collection would help against this.
I'd like to put it a bit more strongly then previous responses.
RAII, Resource Acquisition Is Initialization means that all acquired resources should be acquired in the context of the initialization of an object. This forbids "naked" resource acquisition. The rationale is that cleanup in C++ works on object basis, not function-call basis. Hence, all cleanup should be done by objects, not function calls. In this sense C++ is more-object oriented then e.g. Java. Java cleanup is based on function calls in finally clauses.
I concur with cpitis. But would like to add that the resources can be anything not just memory. The resource could be a file, a critical section, a thread or a database connection.
It is called Resource Acquisition Is Initialization because the resource is acquired when the object controlling the resource is constructed, If the constructor failed (ie due to an exception) the resource is not acquired. Then once the object goes out of scope the resource is released. c++ guarantees that all objects on the stack that have been successfully constructed will be destructed (this includes constructors of base classes and members even if the super class constructor fails).
The rational behind RAII is to make resource acquisition exception safe. That all resources acquired are properly released no matter where an exception occurs. However this does rely on the quality of the class that acquires the resource (this must be exception safe and this is hard).
The problem with garbage collection is that you lose the deterministic destruction that's crucial to RAII. Once a variable goes out of scope, it's up to the garbage collector when the object will be reclaimed. The resource that's held by the object will continue to be held until the destructor gets called.
RAII comes from Resource Allocation Is Initialization. Basically, it means that when a constructor finishes the execution, the constructed object is fully initialized and ready to use. It also implies that the destructor will release any resources (e.g. memory, OS resources) owned by the object.
Compared with garbage collected languages/technologies (e.g. Java, .NET), C++ allows full control of the life of an object. For a stack allocated object, you'll know when the destructor of the object will be called (when the execution goes out of the scope), thing that is not really controlled in case of garbage collection. Even using smart pointers in C++ (e.g. boost::shared_ptr), you'll know that when there is no reference to the pointed object, the destructor of that object will be called.
And how can you make something on the stack that will cause the cleanup of something that lives on the heap?
class int_buffer
{
size_t m_size;
int * m_buf;
public:
int_buffer( size_t size )
: m_size( size ), m_buf( 0 )
{
if( m_size > 0 )
m_buf = new int[m_size]; // will throw on failure by default
}
~int_buffer()
{
delete[] m_buf;
}
/* ...rest of class implementation...*/
};
void foo()
{
int_buffer ib(20); // creates a buffer of 20 bytes
std::cout << ib.size() << std::endl;
} // here the destructor is called automatically even if an exception is thrown and the memory ib held is freed.
When an instance of int_buffer comes into existence it must have a size, and it will allocate the necessary memory. When it goes out of scope, it's destructor is called. This is very useful for things like synchronization objects. Consider
class mutex
{
// ...
take();
release();
class mutex::sentry
{
mutex & mm;
public:
sentry( mutex & m ) : mm(m)
{
mm.take();
}
~sentry()
{
mm.release();
}
}; // mutex::sentry;
};
mutex m;
int getSomeValue()
{
mutex::sentry ms( m ); // blocks here until the mutex is taken
return 0;
} // the mutex is released in the destructor call here.
Also, are there cases where you can't use RAII?
No, not really.
Do you ever find yourself wishing for garbage collection? At least a garbage collector you could use for some objects while letting others be managed?
Never. Garbage collection only solves a very small subset of dynamic resource management.
There are already a lot of good answers here, but I'd just like to add:
A simple explanation of RAII is that, in C++, an object allocated on the stack is destroyed whenever it goes out of scope. That means, an objects destructor will be called and can do all necessary cleanup.
That means, if an object is created without "new", no "delete" is required. And this is also the idea behind "smart pointers" - they reside on the stack, and essentially wraps a heap based object.
RAII is an acronym for Resource Acquisition Is Initialization.
This technique is very much unique to C++ because of their support for both Constructors & Destructors & almost automatically the constructors that are matching that arguments being passed in or the worst case the default constructor is called & destructors if explicity provided is called otherwise the default one that is added by the C++ compiler is called if you didn't write an destructor explicitly for a C++ class. This happens only for C++ objects that are auto-managed - meaning that are not using the free store (memory allocated/deallocated using new,new[]/delete,delete[] C++ operators).
RAII technique makes use of this auto-managed object feature to handle the objects that are created on the heap/free-store by explcitly asking for more memory using new/new[], which should be explicitly destroyed by calling delete/delete[]. The auto-managed object's class will wrap this another object that is created on the heap/free-store memory. Hence when auto-managed object's constructor is run, the wrapped object is created on the heap/free-store memory & when the auto-managed object's handle goes out of scope, destructor of that auto-managed object is called automatically in which the wrapped object is destroyed using delete. With OOP concepts, if you wrap such objects inside another class in private scope, you wouldn't have access to the wrapped classes members & methods & this is the reason why smart pointers (aka handle classes) are designed for. These smart pointers expose the wrapped object as typed object to external world & there by allowing to invoke any members/methods that the exposed memory object is made up of. Note that smart pointers have various flavors based on different needs. You should refer to Modern C++ programming by Andrei Alexandrescu or boost library's (www.boostorg) shared_ptr.hpp implementation/documentation to learn more about it. Hope this helps you to understand RAII.

Thread safe lazy construction of a singleton in C++

Is there a way to implement a singleton object in C++ that is:
Lazily constructed in a thread safe manner (two threads might simultaneously be the first user of the singleton - it should still only be constructed once).
Doesn't rely on static variables being constructed beforehand (so the singleton object is itself safe to use during the construction of static variables).
(I don't know my C++ well enough, but is it the case that integral and constant static variables are initialized before any code is executed (ie, even before static constructors are executed - their values may already be "initialized" in the program image)? If so - perhaps this can be exploited to implement a singleton mutex - which can in turn be used to guard the creation of the real singleton..)
Excellent, it seems that I have a couple of good answers now (shame I can't mark 2 or 3 as being the answer). There appears to be two broad solutions:
Use static initialisation (as opposed to dynamic initialisation) of a POD static variable, and implementing my own mutex with that using the builtin atomic instructions. This was the type of solution I was hinting at in my question, and I believe I knew already.
Use some other library function like pthread_once or boost::call_once. These I certainly didn't know about - and am very grateful for the answers posted.
Basically, you're asking for synchronized creation of a singleton, without using any synchronization (previously-constructed variables). In general, no, this is not possible. You need something available for synchronization.
As for your other question, yes, static variables which can be statically initialized (i.e. no runtime code necessary) are guaranteed to be initialized before other code is executed. This makes it possible to use a statically-initialized mutex to synchronize creation of the singleton.
From the 2003 revision of the C++ standard:
Objects with static storage duration (3.7.1) shall be zero-initialized (8.5) before any other initialization takes place. Zero-initialization and initialization with a constant expression are collectively called static initialization; all other initialization is dynamic initialization. Objects of POD types (3.9) with static storage duration initialized with constant expressions (5.19) shall be initialized before any dynamic initialization takes place. Objects with static storage duration defined in namespace scope in the same translation unit and dynamically initialized shall be initialized in the order in which their definition appears in the translation unit.
If you know that you will be using this singleton during the initialization of other static objects, I think you'll find that synchronization is a non-issue. To the best of my knowledge, all major compilers initialize static objects in a single thread, so thread-safety during static initialization. You can declare your singleton pointer to be NULL, and then check to see if it's been initialized before you use it.
However, this assumes that you know that you'll use this singleton during static initialization. This is also not guaranteed by the standard, so if you want to be completely safe, use a statically-initialized mutex.
Edit: Chris's suggestion to use an atomic compare-and-swap would certainly work. If portability is not an issue (and creating additional temporary singletons is not a problem), then it is a slightly lower overhead solution.
Unfortunately, Matt's answer features what's called double-checked locking which isn't supported by the C/C++ memory model. (It is supported by the Java 1.5 and later — and I think .NET — memory model.) This means that between the time when the pObj == NULL check takes place and when the lock (mutex) is acquired, pObj may have already been assigned on another thread. Thread switching happens whenever the OS wants it to, not between "lines" of a program (which have no meaning post-compilation in most languages).
Furthermore, as Matt acknowledges, he uses an int as a lock rather than an OS primitive. Don't do that. Proper locks require the use of memory barrier instructions, potentially cache-line flushes, and so on; use your operating system's primitives for locking. This is especially important because the primitives used can change between the individual CPU lines that your operating system runs on; what works on a CPU Foo might not work on CPU Foo2. Most operating systems either natively support POSIX threads (pthreads) or offer them as a wrapper for the OS threading package, so it's often best to illustrate examples using them.
If your operating system offers appropriate primitives, and if you absolutely need it for performance, instead of doing this type of locking/initialization you can use an atomic compare and swap operation to initialize a shared global variable. Essentially, what you write will look like this:
MySingleton *MySingleton::GetSingleton() {
if (pObj == NULL) {
// create a temporary instance of the singleton
MySingleton *temp = new MySingleton();
if (OSAtomicCompareAndSwapPtrBarrier(NULL, temp, &pObj) == false) {
// if the swap didn't take place, delete the temporary instance
delete temp;
}
}
return pObj;
}
This only works if it's safe to create multiple instances of your singleton (one per thread that happens to invoke GetSingleton() simultaneously), and then throw extras away. The OSAtomicCompareAndSwapPtrBarrier function provided on Mac OS X — most operating systems provide a similar primitive — checks whether pObj is NULL and only actually sets it to temp to it if it is. This uses hardware support to really, literally only perform the swap once and tell whether it happened.
Another facility to leverage if your OS offers it that's in between these two extremes is pthread_once. This lets you set up a function that's run only once - basically by doing all of the locking/barrier/etc. trickery for you - no matter how many times it's invoked or on how many threads it's invoked.
Here's a very simple lazily constructed singleton getter:
Singleton *Singleton::self() {
static Singleton instance;
return &instance;
}
This is lazy, and the next C++ standard (C++0x) requires it to be thread safe. In fact, I believe that at least g++ implements this in a thread safe manner. So if that's your target compiler or if you use a compiler which also implements this in a thread safe manner (maybe newer Visual Studio compilers do? I don't know), then this might be all you need.
Also see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2513.html on this topic.
You can't do it without any static variables, however if you are willing to tolerate one, you can use Boost.Thread for this purpose. Read the "one-time initialisation" section for more info.
Then in your singleton accessor function, use boost::call_once to construct the object, and return it.
For gcc, this is rather easy:
LazyType* GetMyLazyGlobal() {
static const LazyType* instance = new LazyType();
return instance;
}
GCC will make sure that the initialization is atomic. For VC++, this is not the case. :-(
One major issue with this mechanism is the lack of testability: if you need to reset the LazyType to a new one between tests, or want to change the LazyType* to a MockLazyType*, you won't be able to. Given this, it's usually best to use a static mutex + static pointer.
Also, possibly an aside: It's best to always avoid static non-POD types. (Pointers to PODs are OK.) The reasons for this are many: as you mention, initialization order isn't defined -- neither is the order in which destructors are called though. Because of this, programs will end up crashing when they try to exit; often not a big deal, but sometimes a showstopper when the profiler you are trying to use requires a clean exit.
While this question has already been answered, I think there are some other points to mention:
If you want lazy-instantiation of the singleton while using a pointer to a dynamically allocated instance, you'll have to make sure you clean it up at the right point.
You could use Matt's solution, but you'd need to use a proper mutex/critical section for locking, and by checking "pObj == NULL" both before and after the lock. Of course, pObj would also have to be static ;)
.
A mutex would be unnecessarily heavy in this case, you'd be better going with a critical section.
But as already stated, you can't guarantee threadsafe lazy-initialisation without using at least one synchronisation primitive.
Edit: Yup Derek, you're right. My bad. :)
You could use Matt's solution, but you'd need to use a proper mutex/critical section for locking, and by checking "pObj == NULL" both before and after the lock. Of course, pObj would also have to be static ;) . A mutex would be unnecessarily heavy in this case, you'd be better going with a critical section.
OJ, that doesn't work. As Chris pointed out, that's double-check locking, which is not guaranteed to work in the current C++ standard. See: C++ and the Perils of Double-Checked Locking
Edit: No problem, OJ. It's really nice in languages where it does work. I expect it will work in C++0x (though I'm not certain), because it's such a convenient idiom.
read on weak memory model. It can break double-checked locks and spinlocks. Intel is strong memory model (yet), so on Intel it's easier
carefully use "volatile" to avoid caching of parts the object in registers, otherwise you'll have initialized the object pointer, but not the object itself, and the other thread will crash
the order of static variables initialization versus shared code loading is sometimes not trivial. I've seen cases when the code to destruct an object was already unloaded, so the program crashed on exit
such objects are hard to destroy properly
In general singletons are hard to do right and hard to debug. It's better to avoid them altogether.
I suppose saying don't do this because it's not safe and will probably break more often than just initializing this stuff in main() isn't going to be that popular.
(And yes, I know that suggesting that means you shouldn't attempt to do interesting stuff in constructors of global objects. That's the point.)