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

Gabriel Dos Reis gdr@integrable-solutions.net
Fri May 21 02:37:00 GMT 2010


On Thu, May 20, 2010 at 9:19 PM, Mark Mitchell <mark@codesourcery.com> wrote:
> Gabriel Dos Reis wrote:
>
>>>>>> The bad reputation G++ has cultivated does not come from situation
>>>>>> remotely close to what the original patch was about.  G++ bad reputation
>>>>>> comes extensions that were introduced without much thought about
>>>>>> interactions, and that were removed later, etc.  The concrete
>>>>>> situation at hand comes from a change in the C++ standard itself.
>>>>>> We should not be conflating the two.  Otherwise, we lose credibility.
>>>>> You're entitled to your opinion, of course.  But, my opinion is that if
>
>>> Your statement, quoted above, implied that the binary incompatibility
>>> being contemplated in this thread was not a serious problem.  It is that
>>> opinion to which you are entitled.
>>
>> I am curious in the chains of logical inference you went through to get that
>> implication from the statement you indicated.
>
> Gaby, this is becoming a ridiculous conversation.

how is asking you to detail how you got to infer something (central
to your argument) ridiculous?

> Why don't you just
> state clearly whether you think that breaking the ABI by adding a
> virtual function to the class in question is acceptable without bumping
> the .so version?

I unambiguously answered that question in my reply to Joe's nessage.
Leave C++03 library alone, and make the binary breakage in C++0x mode.
Which of course is scheduled for a mjor bump in the .so version.

> Then you won't have to wonder about my thought process
> and I won't have to wonder what you mean.

the only reason I wonder about your thought process is the conclusion
you got and the way you make it central to your argument.



More information about the Libstdc++ mailing list