Re: Threadsafe singletons
Matthias Hofmann wrote:
"kanze" <kanze@gabi-soft.fr> schrieb im Newsbeitrag
news:1154436138.221933.71380@p79g2000cwp.googlegroups.com...
What exactly is the problem with creation of a local static
object and threads?
The code generated by most compilers isn't thread safe.
Typically, there will be a hidden static bool which says
whether the object has been initialized or not; the compiler
tests this, and if it is false, calls the constructor, and
then sets it to true. If no particular steps are taken by
the compiler, this isn't thread safe.
Is this hidden static bool shared by the threads?
Obviously.
I can imagine that there is a problem if the bool is shared
while the object is not or vice versa. Otherwise I don't
understand what could go wrong.
The fact that the initialization is not atomic. In the end, the
compiler generated code is more or less:
static bool initalized = false ;
if ( ! initialized ) {
doTheInitialization() ;
initialized = true ;
}
If another thread comes in while you are doing the
initialization, it too will do the initialization. If the
initiallization includes e.g. a constructor, you're going to
call the constructor on the object twice.
On a Posix system, the obvious solution is to do away with the
bool, and generate:
pthread_once( doTheInitialization ) ;
Of course, this becomes a bit hairy if the initiallization uses
local variables (like the function parameters). And who
guarantees that pthread_once is cheap.
[...]
Formally, there is no guarantee that static variables are
constructed before entering main, so you have no guaranteed that
your s_init object (or my ourInstance pointer) will be
initialized before entering main.
Really! Are you referring to the C++ standard?
It is described in 3.6.2/3. If I understand this part of the
standard correctly, it does, however, guarantee that the
ourInstance pointer will be initialized before it is first
used.
It guarantees that the pointer will be initialized before the
first use of anything defined in the translation unit. In
particular, it guarantees that the pointer will be initialized
before the first call to Singleton::GetInstance().
And what if nothing is used except for the pointer itself?
The pointer is an object in the translation unit. The standard
guarantees initialization before first use of anything in the
translation unit, not just functions.
Strangely, I found no text in the standard that explicitly
guarantees that an object will be initialized before it is
first used... Or maybe I just overlooked it?
No. This guarantee is explicitly not given; there are even
cases where the standard guarantees that the dynamic
initialization will NOT have taken place before the object is
used. (My solution to this problem, with:
Singleton* Singleton::ourInstance = &Singleton::instance() ;
is an example of code which counts on the function seeing the
zero initialization, before the dynamic initialization has
terminated.)
--
James Kanze GABI Software
Conseils en informatique orient?e objet/
Beratung in objektorientierter Datenverarbeitung
9 place S?mard, 78210 St.-Cyr-l'?cole, France, +33 (0)1 30 23 00 34
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]