Ralf Wildenhues writes: >> 2) If signedness of char has to be known during configure time >> AC_C_CHAR_UNSIGNED is still useful. > > Exactly. So the developer who decides that AC_C_CHAR_UNSIGNED is useful > for him either cannot use autoconf -Wall any more, or is annoyed by the > warning which is (then) wrong for him. Since this warning hints at an > optimization rather than at a potential bug, IMVHO here a wrong warning > is worse than no warning. Well, it is a _potential_ bug in code which uses of the C macro defined by AC_C_CHAR_UNSIGNED, though hopefully an unlikely one. The reason I reacted to this in the first place was not that it can be "optimized" away, but that the macro invades the C implementation's namespace. Some implementation could e.g. define __CHAR_UNSIGNED__ as a boolean - 0 for signed, 1 for unsigned. That's different from the autoconf macros that e.g. #define const if missing, in the hope of compiling ISO C-like code on non-ISO C compilers. -- Hallvard _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
