Bruno Haible writes: > > Larry Jones wrote: > > I've found at least one otherwise-c89 compiler that > > keeps the old preprocessor behavior in its default mode. > > Which compiler is this, on which system? Please don't hide your knowledge.
Sorry, it's IBM C V6 for AIX. > > Since it's so > > easy to avoid the problem, it seems worthwile to do so. > > It's not "so easy" to support K&R C preprocessors. We have thousands of > macros in GNU packages. It's most certainly easier to support this > compiler by putting it into full ANSI C mode or by using the gcc-1.42 > preprocessor if that isn't possible. Indeed, I didn't mean to suggest that it's easy to support K&R preprocessors in general, just that this particular problem is easy to avoid in this particular case. Said compiler and its predecessors have been compiling CVS in the default mode for many years without any problems until the recent addition of the base64 module, so I don't think this particular problem is wide-spread, nor would I spend any time actively looking for other instances. I still think it's worthwhile to change base64 to avoid the problem, though, since we know about it. I've also submitted a fix to autoconf folks that will result in putting the compiler into a standard-conforming mode. -Larry Jones Philistines. -- Calvin
