Re: enable -Werror for lib/ in coreutils

[Bruno Haible]

Eric Blake wrote:
> > ++++ b/lib/isnan.c
> > +@@ -67,6 +67,8 @@
> > +   ((sizeof (DOUBLE) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
> > + typedef union { DOUBLE value; unsigned int word[NWORDS]; } memory_double;
> > +
> > ++int FUNC (DOUBLE x);
> 
> Why are there no header files declaring rpl_isnan?  Isn't a better fix to
> patch math.in.h to declare everything that we will be providing?

math.in.h does declare everything that it provides. But math.in.h is not
always present when isnan[fdl].c is compiled.

This should silence gcc and address your question.


2009-10-27  Bruno Haible  <address@hidden>

        * lib/isnan.c (rpl_isnan[fdl]): Repeat the specification declaration.
        Reported by Jim Meyering.

--- lib/isnan.c.orig    2009-10-28 01:22:06.000000000 +0100
+++ lib/isnan.c 2009-10-28 01:20:17.000000000 +0100
@@ -1,5 +1,5 @@
 /* Test for NaN that does not need libm.
-   Copyright (C) 2007-2008 Free Software Foundation, Inc.
+   Copyright (C) 2007-2009 Free Software Foundation, Inc.
 
    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -18,6 +18,18 @@
 
 #include <config.h>
 
+/* Specification.  */
+#ifdef USE_LONG_DOUBLE
+/* Specification found in math.h or isnanl-nolibm.h.  */
+extern int rpl_isnanl (long double x);
+#elif ! defined USE_FLOAT
+/* Specification found in math.h or isnand-nolibm.h.  */
+extern int rpl_isnand (double x);
+#else /* defined USE_FLOAT */
+/* Specification found in math.h or isnanf-nolibm.h.  */
+extern int rpl_isnanf (float x);
+#endif
+
 #include <float.h>
 #include <string.h>