[RFC] libstdc++/16956 vs interpretation of 27.7.1.3, p13

[Paolo Carlini]

Hi,

this PR basically questions our interpretation of the standard: we have 
*always* (in v3) interpreted 27.7.1.3, p13 as meaning that the return 
value is pos_type(newoff), where newoff has the value computed *before* 
changing xnext: in other terms, we return the "old position", as per 
Table 91.

According to submitter we should return instead the "new position", 
corresponding to a newoff evaluated *after* updating xnext.

Well, Icc/Dinkum appear to agree with this different reading and, in 
fact, the wording of the standard: "pos_type(newoff), constructed from 
the resultant offset newoff (of type off_type), that stores the 
resultant stream position, if possible." seems rather favourable to it...

Changing this behavior is very easy but it's important to reach a 
complete agreement about the issue.

Thanks,
Paolo.