> Yes, I know. But that really isn't a solution. I don't want to compile
> without optimisation. What I want is the bug in ecgs to be fixed.
It is not a bug in egcs, it is a bug in glibc. With -O2, the
preprocessor expands your code to
int func (char *array)
{
return ({
int __res;
if (sizeof (array[0]) > 1){
if (__builtin_constant_p (array[0])){
int __c = (array[0]);
__res = __c < -128 || __c > 255 ? __c : __ctype_tolower [__c];
}else
__res = tolower(array[0]);
}else
__res = __ctype_tolower [(int) (array[0])];
__res;
});
}
This is clearly not valid ISO C, so with -pedantic, the compiler must
complain. The fix is to tell it not to complain here, with
int func (char *array)
{
return __extension__({
...
In glibc 2.1, this translates to
--- ctype.h.old Sun Aug 15 07:54:39 1999
+++ ctype.h Sun Aug 15 07:55:06 1999
@@ -165,7 +165,7 @@
#if __GNUC__ >= 2 && defined __OPTIMIZE__ && !defined __cplusplus
# define __tobody(c, f, a) \
- ({ int __res; \
+ __extension__({ int __res; \
if (sizeof (c) > 1) \
{ \
if (__builtin_constant_p (c)) \
Regards,
Martin
