Subj : Re: Can C++ local static objects be made thread safe?
To   : comp.programming.threads
From : Gianni Mariani
Date : Sun Feb 06 2005 11:05 pm

Giancarlo Niccolai wrote:
> Gianni Mariani wrote:
> 
> 
>>Giancarlo Niccolai wrote:
>>...
>>
>>>Having a reentrant mutex cheching for each initializer initialization 1)
>>>does not fix the generic problem of initialization during static init
>>>with MT already going and 2) would be ... crappy.
>>
>>Why ?  Honestly, why would this not solve the problem generically ?
> 
> 
> Because MUTEXING is not the heart of multithreading!!! You can mutex each
> and every single access to any variable, even from the stack, in all your
> program and still have thread unsafe code!!!
> 
> Mutexing the things up won't make your code SAFE. It may make it a little
> bit safer in some cases, but mutexing, although sometimes necessary, is
> rarely enough except in academic cases.
>

Stick to facts.

My proposal solves the problem of static locals being initialized in a 
generic way - period -.

PLEASE PLEASE, stick to the theme.  If you wish to do somthing else, go 
ahead, however there are definite merits to my original proposal.

a) Generic solution - threads and static locals are safe
b) Can be made fast (using thread specific variables if the compiler 
supports it)
c) No maintenance needed - works every time.

Sounds like a Good (TM) thing to me.

Tell me what I am missing (facts only please, no religion).

.