Re: RFC: PATCH for thread-safe C++ static local initialization

[Mark Mitchell <mark at codesourcery dot com>]

Jason Merrill wrote:

> This is a patch to implement/use the C++ ABI interface for thread-safe
> initialization of local statics.  The use generates code something like
>  
Yay!

>           static <type> guard;
>           if (!guard.first_byte) {
> 	     if (__cxa_guard_acquire (&guard)) {
> 	       bool flag = false;
> 	       try {
> 	         // Do initialization.
> 	         flag = true; __cxa_guard_release (&guard);
> 	         // Register variable for destruction at end of program.
> 	       } catch {
> 	         if (!flag) __cxa_guard_abort (&guard);
> 	       }
>             }
>           }
>
> which is more complex than the old code which just checked the first byte
> of the guard function.  Seems like this should probably be controlled by a
> flag, but which?
>  
-pthread, and it's oh-too-many synonyms? 

It seems to me like if you've asked for threaded code, you probably want 
this level of thread-safety.  I suppose the GCC tradition would be to 
create yet another flag (-fuse-cxa-guard-acquire) and then another 
configuration switch (--enable-cxa-guard-acquire), but I that we should 
be trying to help our users by limiting their choices, rather than 
pretending to help them by expanding them. :-)

> How do people feel about extending the gthr interface to either support
> recursive mutexes (emulating in the gthr-* file if necessary) or thread_id,
> or both?
>  
I vote for both.

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com