Re: libstdc++/4433

[Paolo Carlini <pcarlini@unitus.it>]

The following reply was made to PR libstdc++/4433; it has been noted by GNATS.

From: Paolo Carlini <pcarlini@unitus.it>
To: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, nobody@gcc.gnu.org
Cc:  
Subject: Re: libstdc++/4433
Date: Tue, 20 Nov 2001 09:29:17 +0100

 Hi,
 
 I think that the behaviour you are reporting is not a bug.
 
 I suggest you to confer to page 634 of Josuttis (ISBN 0-201-379-260)
 where it is stated that "this value (that reported by tellp/tellg) is
 *not* an integral value or simply the position of the character as an
 index". So, in general, it *cannot* be directly compared with the
 parameter given to a preceding seekp/seekg. In particular, the values
 returned by tellp/tellg after a relative seek (which you are using) may
 *differ* if you referred the seek to beg, cur, or end.
 
 Even if some implementations happen to pass your test this does not mean
 that all the implementations have to pass it to be conforming (in fact
 among the 5 or 6 I have tried more than half of them do *not* pass it)
 
 Personally, I prefer to use either only relative seeks, or absolute
 seeks and tells (according to the scheme explained, f.i., in
 Langer/Kreft (ISBN 0-201-18395-1)), never a mix.
 
 Cheers,
 Paolo Carlini.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=4433&database=gcc