Re: Threadsafe singletons
Liviu wrote:
kanze wrote:
Liviu wrote:
kanze wrote:
David Barrett-Lennard wrote:
I think that's a shame. I like to be able to have code
[...]
You're not alone.
As a person who deals with legacy code which relies on this
kind of behaviour (not mandated by the standard as previously
pointed out) I must tell you that it can very quickly become a
mess from pov of both implementation and usage. There are more
robust techniques which can accomplish the same goal.
Such as? [...]
This could easily become a thread in its own. One possible
solution to the registry problem would be to use POSIX
pthread_once. The pthread_once_t variables can be initialized
statically and, if in the same translation unit as the
registry access functions - could be robustly used to
conditionally initialize the registry (once) in an MT environ.
(Unfortunately, I don't think there is any way to robustly
package that type and that POSIX function into a C++ class
type...)
Attention: pthread_once and exceptions don't mix well. And
registration can almost always trigger an exception, so can't be
done from pthread_once. (And of course, pthread_once isn't
present on Windows based systems. And boost::once explicitly
says that the called function may not throw.)
But I don't see where your approach solves anything. The
problem is to avoid having to list the objects to be registered
in any one C++ file. This may be a false problem in many
cases---I still have to list them somewhere, for example, in the
makefile, and it's pretty simple to write a makefile that uses
this list to generate the required C++ file. In other cases,
however... I've use the technique when configuring different
versions of a program, by linking in more or less different
modules. I've even used it for dynamically linked modules, some
of which weren't known when the program was built.
It's obviously possible to do away with automatic construction
of static variables completely, and just call an init function
for each module from main. This is the sort of thing dynamic
initialization of static variables was designed to do away with.
[...]
Hehe. Yes, following the standard does not give one the
warranty of a clean ride but not following the standard will
most certainly get you in trouble sooner or later.
Not always. It depends on what you are doing, and what your
portability goals are. As soon as you include <unistd.h>,
you're not following the C++ standard, but I do it all the time.
Anytime you convert a value read with istream::get() (or
fgetc()) to a char, you're risking undefined behavior. But
idioms along the lines of:
for ( int ch = source.get() ;
ch != EOF /* && something */ ;
ch = source.get()) {
someString.push_back( ch ) ;
}
are so common that I take the risk, considering that no compiler
would dare break it.
--
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! ]