NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/std/c++/668

Text File | 1996-08-06 | 3.3 KB | 73 lines

Path: chronicle.mti.sgi.com!austern
From: [e-mail] (J. Kanze)
Newsgroups: comp.std.c++
Subject: Re: Streams and eof()
Date: 08 Mar 1996 09:08:56 PST
Organization: GABI Software, Sarl.
Approved: [e-mail]
Message-ID: <[e-mail]>
References: <4hgap6$[e-mail]> <4hkean$[e-mail]>
NNTP-Posting-Host: isolde.mti.sgi.com
X-Original-Date: 08 Mar 1996 11:41:26 GMT
In-reply-to: [e-mail]'s message of 07 Mar 96 13:40:54 GMT
X-Auth: PGPMoose V1.1 PGP comp.std.c++
    iQBVAwUBMUBpuEy4NqrwXLNJAQHkFAH/eWdYdZrolAeCL59HFgV4TYX8D8dkximn
    D4XvlT8tEhZf/KqC25fmNpW4zCALHTQ90w+WPuIcRLYqkq3oGqG4BA==
    =oxej
Originator: [e-mail]

In article <4hkean$[e-mail]> [e-mail] (Steve
Clamage) writes:

|> No, it must work. EOF is required to have a value different from every
|> character value.

Where does it say this?  (It's what I always thought too, until a recent
discussion in comp.std.c proved me otherwise, at least for C.)

It is indirectly required if INT_MAX >= UCHAR_MAX, at least for fgetc in
C, but if int's and char's are the same size, all bets are off.

By `character value', I presume you mean a value that can be returned by
a successful read, since the original question concerned using
istream::peek and istream::get for binary input.  Although even for
character (non-binary) input, I can see no guarantees once one leaves
the "C" locale.

|> The character-reading functions that can return either
|> a character or EOF return a type that that can represent the EOF value
|> as distinct from any character value.

This is definitly not the case in ISO C, and I can find no such
guarantee in the current draft either, although in this case, it might
be a simple oversight.  Again, there is some ambiguity in your use of
the word character: imagine a 16 bit machine with sizeof( int ) ==
sizeof( char ) == 1, using Unicode as its native character code.
Unicode explicitly doesn't use 0xFFFF (-1) as a legal code, so if EOF is
-1 (a frequent implementation), there is no problem with reading
`characters'.  Reading a `char' from a file opened in binary mode is
still a problem, however.

|> In a traditional implementation, EOF has the int value -1, and characters
|> have 8-bit values. The peek function returns the character value promoted
|> to type int without sign extension, as if by the cast sequence
|>     (int)(unsigned char)the_value
|> Such a value cannot compare equal to the int value -1.

I suspect that a lot of code developed on a traditional implementation
will break on a machine where all integral types are the same length.
(I know that a lot of mine would.  Even in the parts I thought I wrote
portably.)
--
James Kanze           ([phone]          email: [e-mail]
GABI Software, Sarl., 8 rue des Francs Bourgeois, 67000 Strasbourg, France
Conseils, Θtudes et rΘalisations en logiciel orientΘ objet --
              -- A la recherche d'une activitΘ dans une region francophone
---
[ comp.std.c++ is moderated.  To submit articles: Try just posting with your
                newsreader.  If that fails, use mailto:[e-mail]
  comp.std.c++ FAQ: http://reality.sgi.com/austern/std-c++/faq.html
  Moderation policy: http://reality.sgi.com/austern/std-c++/policy.html
  Comments? mailto:[e-mail]
]