Re: '#define inline' patch for Bison on Compaq/DEC Alpha OSF/1 4.0

[Paul Eggert]

> From: Akim Demaille <address@hidden>
> Date: 13 Nov 2002 10:43:42 +0100
> 
> But then, we are releasing glr.c for C((, not C89 extended, as was
> intended.  How about
> 
> #if ! defined (__GNUC__) && ! defined inline
> # define inline
> #endif
> 
> instead?

Currently, "inline" has approximately the same status that "const" had
in 1992.  It's required by the current C standard, and is supported by
most current C compilers, but there are still many older C compilers
still in use that don't handle "inline" properly.

However, I doubt whether many GLR users will employ those old
compilers, and the few who do can easily "#define inline /* empty */"
themselves (or via Autoconf); so it's not clear to me that we need to
worry about this porting issue.

However, if we do need to worry about it, why not remove "inline"
entirely from glr.c?  Even the above patch won't work correctly for
sufficiently-old GCC versions; also, the patch isn't compatible with
how Autoconf handles "inline", as Autoconf doesn't define "inline" if
the compiler supports it.