[Libguestfs] [libnbd PATCH] tests: Make errors more robust under load

[Richard W.M. Jones]

On Tue, Jul 02, 2019 at 09:29:19PM -0500, Eric Blake wrote:
> +  if (dprintf (script_fd, "case $1 in\n"

This commit is already pushed, but the use of ‘dprintf’ which I'd
never heard of until yesterday triggers a bug (in our code) on
FreeBSD.

errors.c: In function 'main':
errors.c:103:7: warning: implicit declaration of function 'dprintf'; did you mean 'vdprintf'? [-Wimplicit-function-declaration]
   if (dprintf (script_fd, "case $1 in\n"
       ^~~~~~~
       vdprintf

This function was apparently standardized in POSIX.1-2008 and the
FreeBSD man page says it exists.  However:

     Applications that wish to use the dprintf() function described herein
     should either request a strict IEEE Std 1003.1-2008 ("POSIX.1")
     environment by defining the macro _POSIX_C_SOURCE to the value 200809 or
     greater, or by defining the macro _WITH_DPRINTF, prior to the inclusion
     of <stdio.h>.  For compatibility with GNU libc, defining either
     _BSD_SOURCE or _GNU_SOURCE prior to the inclusion of <stdio.h> will also
     make dprintf() available.

It was always my understanding that AC_USE_SYSTEM_EXTENSIONS would
define the right symbols, and indeed in <config.h> on FreeBSD we have:

/* Enable GNU extensions on systems that have them.  */
#ifndef _GNU_SOURCE
# define _GNU_SOURCE 1
#endif

The problem in fact is that we don't include <config.h> in all of the
tests uniformly.  I fixed a couple of these in my FreeBSD patch
yesterday, but we really must include <config.h> everywhere.

I'll try to make a patch in a minute.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-builder quickly builds VMs from scratch
http://libguestfs.org/virt-builder.1.html