Fwd: Bug#39071: libc6-dev: ctype.h contains "-pedantic -O" warning

[Joel Klecker <espy@debian.org>]

[ This is a forwarded bug from the Debian Bug Tracking System, please 
keep 39071-forwarded@bugs.debian.org in any replies. Thanks.]

Debian version 2.1.1-10 is based on the final release of glibc 2.1.1.

The first case is considered a bug, yes?

I'm not sure of the second.
--- begin forwarded text
Return-path: <debian-bugs-dist-request@lists.debian.org>
Subject: Bug#39071: libc6-dev: ctype.h contains "-pedantic -O" warning
Reply-To: Kirk Hilliard <kirk@ghoti.com>, 39071@bugs.debian.org
To: submit@bugs.debian.org
From: Kirk Hilliard <kirk@ghoti.com>
Date: 06 Jun 1999 22:58:22 -0400
Message-ID: <87iu90n2cx.fsf@storm.x.ghoti.com>

Package: libc6-dev
Version: 2.1.1-10

Consider:
   $ cat toupper.c
   #include <stdio.h>
   #include <ctype.h>
   int main(void) {
     char c;
     c = toupper('a');
     printf("toupper('a') = '%c'\n", c);
     return 0;
   }
   $ gcc -pedantic -O toupper.c
   toupper.c: In function `main':
   toupper.c:5: warning: ANSI C forbids braced-groups within expressions
   $

This warning here is new with libc6 2.1.

Other header files use __extension__ before braced-groups within
expressions, so here is a patch for ctype.h to suppress the warning.

------------------------------ 8< ------------------------------
--- /usr/include/ctype.h.orig   Mon May 31 06:18:37 1999
+++ /usr/include/ctype.h        Sun Jun  6 17:18:23 1999
@@ -165,6 +165,7 @@
 
  #if __GNUC__ >= 2 && defined __OPTIMIZE__ && !defined __cplusplus
  # define __tobody(c, f, a) \
+  __extension__                                                         \
    ({ int __res;                                                         \
       if (sizeof (c) > 1)                                                \
         {                                                                \
------------------------------ >8 ------------------------------

[I you really want to patch this, use -l since I messed with the
  whitespace to make it fit.]

Even pedants should be able to optimize calls to toupper().

Kirk Hilliard


P.S.  Here is another warning that is new for libc6 2.1.

   $ cat strcmp.c
   #include<string.h>
   int main(void) {
     void *v = "Hello, world.";
     strcmp(v,v);
     return 0;
   }
   $ gcc -Wpointer-arith -O strcmp.c
   strcmp.c: In function `main':
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   strcmp.c:4: warning: pointer of type `void *' used in arithmetic
   $

At first this one did not bother me because it can be easily fixed by
explicitly casting the arguments of strcmp() to (string *).  But the
warning doesn't say a thing about implicit casting.  (Nor do -Wall or
-pedantic (or a slew of other warnings the upstream maintainer likes
to use) complain about the implicit casting.)

The source of the warning is in /usr/include/bits/string2.h:

   #define __string2_1bptr_p(__x) \
     ((size_t)(const void *)((__x) + 1) - (size_t)(const void *)(__x) == 1)
                             ^^^^^^^^^

I think that ``__extension__'' only works with -pedantic -- at least I
couldn't get it to work its magic here.  I don't know how to fix this.
Is it even worthy of being fixed?


--
To UNSUBSCRIBE, email to debian-bugs-dist-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
--- end forwarded text
-- 
Joel Klecker (aka Espy)       <URL:mailto:espy@debian.org>
Debian Package Maintainer for the GNU C Library.