Re: Singleton_pattern and Thread Safety
On Dec 12, 2:55 pm, Leigh Johnston <le...@i42.co.uk> wrote:
On 12/12/2010 02:27, Joshua Maurice wrote:
On Dec 11, 6:00 pm, Leigh Johnston<le...@i42.co.uk> wrote:
On 12/12/2010 01:23, Joshua Maurice wrote:
On Dec 11, 5:19 pm, Tiib<oot...@hot.ee> wrote:
On Dec 12, 2:30 am, Joshua Maurice<joshuamaur...@gmail.com> wrote:
On Dec 11, 11:57 am, "Chris M. Thomasson"<cris...@charter.net> wrote:
"Leigh Johnston"<le...@i42.co.uk> wrote in message
news:aI-dnTRysdoEVJ7QnZ2dnUVZ7sSdnZ2d@giganews.com...
On 11/12/2010 18:47, Chris M. Thomasson wrote:
[...]
One can instantiate multiple "Meyers Singletons" before creating any
threads to avoid any singleton related threading code. No object
destruction problems.
You continue to assert that a memory leak is bad a priori. The rest of
us in this thread disagree with that claim. Instead, we program to
tangible requirements, such as cost, time to market, meets (business)
use case. We also keep in mind less tangible but still important
considerations, like maintainability and reusability.
You are not listening. If you have multiple singletons using
Mr Kanze's method that "you all" agree with it is unspecified
as to the order of construction of these singletons across
multiple TUs; i.e. the method suffers the same problem as
ordinary global variables; it is no better than using ordinary
global variables modulo the lack of object destruction (which
is shite).
I'd suggest you read the code I posted first. It addresses the
order of initialization issues fully. (But of course, you're
not one to let actual facts bother you.)
Unspecified construction order is anathema to maintainability
as the order could change as TUs are added or removed from
a project.
Unspecified constructor order of variables at namespace scope is
a fact of life in C++. That's why we use the singleton pattern
(which doesn't put the actual variable at namespace scope, but
allocates it dynamically).
Unspecified destructor order of variables with static lifetime
(namespace scope or not) is also a fact of life in C++. That's
why we don't destruct the variable we dynamically allocated.
It's called defensive programming. Or simply sound software
engineering. It's called avoiding undefined behavior, if you
prefer.
[...]
As Chris pointed out the only problem with my version compared
to the version given in document by Meyers and Alexandrescu
that you seem so fond of is the lack of a memory barrier after
the initial fast check but this is only a problem for
a minimal number of CPUs as the load is dependent. If I had
to port my code to run on such CPUs I simply have to add this
extra barrier.
I'm not sure which problems in your code Chris tried to address,
or even how much of your code he actually studied; your code
definitely doesn't work on the machines I use (which includes
some with Intel processors, both under Linux and under Windows)
and the compilers I use.
In the real world people write non-portable code all the time
as doing so is not "incorrect".
Certainly not: until we get C++0x, all multithreaded code is
"non-portable". It's a bit disingenious, however, to not
mention the limitations before others pointed out what didn't
work.
--
James Kanze