[RFC] C++1x breaking the ABI in one more place :(

[Mark Mitchell]

Jason Merrill wrote:

> As I mentioned before, I believe it will only break if something in the
> binary both derives from time_get and adds additional virtual functions,

Right.

> I don't know how likely this scenario is.

I don't know either, but the consequences are severe if it does occur.
That's why I think this should require a .so bump; fundamentally, it is
a break with binary compatibility.  If there were some kind of magic
shim so that old stuff still worked, I wouldn't make that claim.  And,
perhaps, if the class were so broken that no use of it had been
possible, then I wouldn't make that claim.  But, I don't think we should
be guessing what valid use people might not be making of the library.

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713