Re: working with "good enough" functions

[Jim Meyering]

Mike Frysinger <address@hidden> wrote:

> the gnulib implementations of POSIX functions are pretty damn complete.  for
> most of my uses though, they're *too* complete :).
>
> for example, current gnulib will often enable printf functions on modern
> systems (such as Linux w/glibc 2.9).  this is because extended floating point
> support breaks from time to time.  in my case, my system C library has broken
> handling of long doubles.  however, i rarely use floating point code in
> projects i work on, so the system C library could have completely hosed
> support and i still wouldnt care.  thus using the replacement versions really
> only results in bloat.
>
> is there a standard way for addressing this ?  or should i cheat and set the
> vars to yes before calling gl_{EARLY,INIT} ?  if i add a line like this:
> gl_cv_func_printf_infinite_long_double=yes

Yes, that (seeding the cache) is the recommended approach.

> then the tests complete as i'd like ... the printf() implementation comes from
> the system C library rather than gnulib.
> -mike