Date: Sat, 6 Sep 1997 10:32:18 -0700 (PDT)
Message-Id: <199709061732.KAA08691@adit.ap.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: djgpp AT delorie DOT com
From: Nate Eldredge <eldredge AT ap DOT net>
Subject: scanf and upper-case format letters
Precedence: bulk

I think I may have mentioned this before, but it got no response.
I find it strange that the scanf family uses the case of the format letters
to determine the size of a floating-point number. For example, according to
the source, "%f" reads a float, but "%F" reads a double. There is already a
bug whereby "%LF" reads a double instead of a long double, for which I
posted a patch. However, I've been able to find no justification of this
behavior. It seems to me the ANSI standard wants "%f" and "%F" to be the
same, since in printf they just specify the case of the exponent 'e' (i.e.
"1.23e4" vs. "1.23E4"). Also, the Linux libc, which I assume is
ANSI-compliant, does nothing of the sort.
So, IMHO, scanf shouldn't care at all whether you use upper or lower-case
format letters. It should make no difference. Here's a patch to remove that
`feature':

*** doscan.c~	Sat May 18 21:40:02 1996
--- doscan.c	Sat Sep  6 10:23:34 1997
***************
*** 93,96 ****
--- 93,97 ----
      } else if (ch=='[')
        fmt = _getccl((const unsigned char *)fmt);
+ #if 0
      if (isupper(ch)) {
        /* ch = tolower(ch);
***************
*** 103,106 ****
--- 104,108 ----
          size = LONG;
      }
+ #endif
      if (ch == '\0')
        return(-1);

Please advise me on what you all think of this. It'd also be nice to know
exactly what the ANSI standard says.

Nate Eldredge
eldredge AT ap DOT net