deque fails to meet 23.2.1.3.4-5 causing invalid iterators and extra assignments

[Paolo Carlini]

Steve LoBasso wrote:

>Speaking relative to 4.2.0 is fine, as far as I can tell, this bug exists
>in all versions of gnu deque as far back as I can see. In fact I see the
>same problem in the SGI implementation. It seems to be an old bug and must
>have been copied MANY times into MANY implementations.
>  
>
Exactly, personally I saw that pattern so many times...

>These seem to be the main issues:
> the one extra copy and at the same time copying in the wrong direction,
> thus invalidating the iterators. I can't see any other problems.
>
>Although it is a problem more often than when n = size() - 1
>ie. size() = 8 erase(begin() + 3, begin() + 4)
>  
>
I see, in such cases we don't have a problem with 23.2.1.3/4, still we 
have one with /5.

>I believe your fix is just as valid as mine. Either one seems to fix the problem.
>  
>
Excellent, I will soon commit a fix together with your 2nd testcase, 
which is very small (we have to pay attention to your Copyleft rules...) 
and fully deterministic, should be enough.

Thanks again,
Paolo.