Re: [Bug-gnulib] Re: fnmatch.c includes alloca.h conditionally?

[Paul Eggert]

Derek Robert Price <address@hidden> writes:

> It strikes me that this ties the alloca.m4 file and the alloca_.h
> file rather tightly, maintenance-wise,

That's already the case for alloca, and anyway is pretty much the
tradition in gnulib these days.

> Actually, that is a "#ifndef".  Since IN_LIBINTL won't be defined in
> GNU applications, alloca.h _will_ be included.

But the code looks like this:

  #ifndef IN_LIBINTL
  # include <alloca.h>
  #endif

which means that <alloca.h> _won't_ be included in GNU applications.

> You yourself commented that it was odd that <alloca.h> was included
> dependant on IN_LIBINTL and then the use of alloca() only switched on
> HAVE_ALLOCA here:
>
> http://mail.gnu.org/archive/html/bug-gnulib/2004-04/msg00063.html

Yes, but that was a mild stylistic recommendation.  The code works
as-is; it's really up to Bruno (who maintains vansprintf.c).

> Could we change the switch to #ifdef HAVE_ALLOCA_H and assume that
> HAVE_ALLOCA_H will be defined or undefined properly, even when
> IN_LIBINTL is defined?

Yes, that would also work, and perhaps be a bit more straightforward.
But again, it's up to Bruno.

Similar stylistic cleanups are possible in argp too, now that
"#include <alloca.h>" always works (if you assume the alloca module).
I don't have time to pursue this right now, though.