Sender: crough45 AT amc DOT de
Message-Id: <97Aug2.172909gmt+0100.17037@internet01.amc.de>
Date: Sat, 2 Aug 1997 16:32:13 +0100
From: Chris Croughton <crough45 AT amc DOT de>
Mime-Version: 1.0
To: eldredge AT ap DOT net
Cc: djgpp AT delorie DOT com
Subject: Re: scanf? 
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Nate Eldredge wrote:

> I don't dare describe it as a bug, but IMHO it's at 
> least a misfeature. The problem is that scanf("%e",...) 
> (or %f, or %g) reads a float. But if you use upper case, 
> it reads a double instead (or a long double if you use 
> an `l' already). As a workaround, to read a double, use 
> the standard %le format.

> Can somebody with a copy of the ANSI standard enlighten 
> us as to whether this is supposed to happen? There is 
> already a bug with formats like %LG, for which I posted 
> a patch some time ago.

From Plauger's "The Standard C Library", which quotes the 
ANSI standard, in scanf only the letters 'e', 'f' and 'g'
(lowercase) are supported for floating point conversion.
If you want a double then you prefix them with 'l' (lower-
case 'L') and for a long double you use upper-case 'L'.
[ANSI Standard X3.159-1989, section 7.9.6.2, paraphrased]

It also says that "The conversion specifiers E, G and X
are also valid and behave the same as, respectively, e, g
and x" [7.9.6.2].  (But not, interestingly, F.)

So to be ANSI compliant, the options are:

  %e, %f, %g		float
  %le, %lf, %lg		double
  %Le, %Lf, %Lg		long double

and the uppercase conversion specifiers (%E, %F and %G) may
not be at all portable if you expect it to read a double or
long double (and it may also break ANSI-conforming programs,
if they expect E and G to input into a float).  In fact, this
is the way in which the OSF/1 C library (which is not the GNU
one) interprets it, as per the standard...

Chris C