Re: compile warning with ./gnulib-tool --with-tests

[Bruno Haible]

Hi Eric,

> Right now, './gnulib-tool --with-tests --test xstrtol' issues this warning 
> when 
> run on cygwin 1.5.24:
> 
> gcc -DHAVE_CONFIG_H -I. -I../../gltests  -I. -I../../gltests -I.. 
> -I../../gltests/.. -I../gllib -I../../gltests/../gllib   -g -O2 -MT 
> test-wchar.o -MD -MP -MF .deps/test-wchar.Tpo -c -o test-wchar.o 
> ../../gltests/test-wchar.c
> In file included from ../../gltests/test-wchar.c:20:
> ./config.h:155:1: warning: "__STDC_LIMIT_MACROS_TRIGGER" redefined
> In file included from ./config.h:4,
>                  from ../../gltests/test-wchar.c:20:
> ./../config.h:147:1: warning: this is the location of the previous definition

Interesting: The warning goes away when the macro is renamed so that it does
not start with "__".

The appended patch fixes it.

> And while both files contain similar #defines (for example, both 
> did #define HAVE_DECL_IMAXABS 1), it is only the dual listing of 
> __STDC_LIMIT_MACROS_TRIGGER that caused gcc 3.4.4 to warn, even though both 
> definitions were to the same value.

Indeed, gcc seems to treat __ prefixed macros differently than those in the
user's namespace.

> Something weird is going on; gnulib-tool created the testdir in such a manner 
> that both build/config.h and build/gltests/config.h, with the latter 
> #including 
> the former.  ...
> Is there any reason why gnulib-tool uses AH_TOP([#include "../config.h"]) for 
> gltests/configure.ac?

There are comments about this in gnulib-tool.

> Could this be reworked to have the tests dir share the 
> same config.h as the rest of the project, rather than having two files with 
> duplicate definitions?

This would not be good. It would have the consequence that
"gnulib-tool --create-testdir $modules" could fail to build on some platform
although "gnulib-tool --create-testdir --with-tests $modules" would build
fine on the same platform. (This could happen if the module description of
a module lacks a dependency, and its tests bring in the needed dependency.)
This would mean that testing with --with-tests could overlook some failures -
which is against the primary goal of --with-tests.

Bruno


2007-08-07  Bruno Haible  <address@hidden>

        * m4/inttypes.m4 (gl_INTTYPES_H): Use GL_TRIGGER_STDC_LIMIT_MACROS
        instead of __STDC_LIMIT_MACROS_TRIGGER. This avoids a redefinition
        warning from gcc.
        Reported by Eric Blake.

--- m4/inttypes.m4      2 Jul 2007 23:28:02 -0000       1.20
+++ m4/inttypes.m4      7 Aug 2007 22:22:00 -0000
@@ -1,4 +1,4 @@
-# inttypes.m4 serial 10
+# inttypes.m4 serial 11
 dnl Copyright (C) 2006-2007 Free Software Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
 dnl gives unlimited permission to copy and/or distribute it,
@@ -150,12 +150,12 @@
     dnl Also be careful to define __STDC_LIMIT_MACROS only when gnulib's
     dnl <inttypes.h> is going to be created, and to avoid redefinition warnings
     dnl if the __STDC_LIMIT_MACROS is already defined through the CPPFLAGS.
-    AC_DEFINE([__STDC_LIMIT_MACROS_TRIGGER], 1,
+    AC_DEFINE([GL_TRIGGER_STDC_LIMIT_MACROS], 1,
       [Define to make the limit macros in <stdint.h> visible.])
     AH_VERBATIM([__STDC_LIMIT_MACROS_ZZZ],
 [/* Ensure that <stdint.h> defines the limit macros, since gnulib's
    <inttypes.h> relies on them.  */
-#if defined __cplusplus && !defined __STDC_LIMIT_MACROS && 
__STDC_LIMIT_MACROS_TRIGGER
+#if defined __cplusplus && !defined __STDC_LIMIT_MACROS && 
GL_TRIGGER_STDC_LIMIT_MACROS
 # define __STDC_LIMIT_MACROS 1
 #endif
 ])