Re: [PATCH][RFC] Remove volatile from data members in libstdc++

[Paolo Carlini <pcarlini at suse dot de>]

Benjamin Kosnik wrote:

> > To summarize, we have no idea why the (respected, would be Hans Boehm, I 
> > guess) author of that code decided to add the volatiles... I'd like to 
> > understand the issue a little better before going ahead...
> >    
> I don't have a problem with this patch, as long as it is sufficiently
> tested.
... the real problem is the latter requirement: in my opinion our 
current testsuite doesn't stress sufficiently well all those facilities 
in the multithreaded context which really matters. Without strong 
"theoretical" assurance that it's safe, I fear we may  damage something 
without noticing (and without a huge benefit)...

> I suspect that some of this is just the accumulation of patches over
> the years. Some of this may be to fix various compiler behavior, to
> avoid warnings, or to coax cetain output (and may not be relevant now).
>  
Personally, I have no clues. For sure those volatile were already there 
*a lot* of time ago and the semantics of volatile is sufficiently 
underspecified by the current standard that I would be interested in a 
clarification from the  original author... What can I say, assuming Hans 
has nothing to add I propose to commit the patch to mainline as soon as 
4.2 branches and hope for the best... ;)

Paolo.