Re: GACK!

[Bruce Lilly]

John Millaway wrote:

> That's not what's supposed to happen. Let's make sure we are on the same page
> when we speak about "prototypes" and "definitions", because it's easy to
> confuse the issue. To clarify, this is what you are supposed to get by 
> default:
> 
>      // prototype
>     int foo(int, char*);
> 
>      // C89 definition
>     int foo (x.y)
>          int x; char* y;
>     {
>      ....
>     }

It's the definitions which are the issue (which are K&R C, not
ANSI C89).  The compiler diagnostics were like:

lex.mail.c(9853) : warning C4131: 'yy_get_next_buffer' : uses old-style 
declarator

which refers to the definition (after preprocessing):

static int yy_get_next_buffer (yyscanner) yyscan_t yyscanner;

> > This is really, really bad.  In a project with 60 object files, it
> > is unreasonable to have to specify the extraordinarily verbose
> > -DYY_NO_TRADITIONAL_FUNC_DEFS for all compilations when it ought not
> > to be needed for any object files (i.e. the project in question is
> > not "bootstrapping gcc").  A separate compilation command in a
> > makefile for the flex object isn't possible with gmake, because
> > gmake overrides it with the recipe based on SUFFIXES.
> 
> Huh? Just add the following line in your gmake Makefile:
> 
>     CPPFLAGS +=  -DYY_NO_TRADITIONAL_FUNC_DEF

That's the "extraordinarily verbose ... for all compilations"
issue.

Thanks for the fix.