Re: possible of include of incorrect config.h file

[Patrice Dumas]

On Wed, Aug 19, 2015 at 03:04:06PM +0100, Gavin Smith wrote:
> On 19 August 2015 at 14:53, Patrice Dumas <address@hidden> wrote:
> >> Why wouldn't the config.h exist?
> >
> > It should exist in general, but if for some reason it does not, there
> > will be no easy way to understand that a change in behaviour or an error
> > happens because the config.h file included is the perl one and not the
> > one done by configure.
> >
> > It is not for the common cases, but an extra security, and I only raise
> > that issue because the config.h file from perl is in a directory
> > searched for by the preprocessor, if not I wouldn't propose that.  Also
> > I raise that issue because it happened to me, when doing the standalone
> > module, it compiled while it shouldn't have.
> 
> I thought it shouldn't happen because of the HAVE_CONFIG_H guard that
> you added. If you don't think that's enough, feel free to change it.

The HAVE_CONFIG_H guards is required for the standalone module as there
won't be any config.h in that case.  But it doesn't solve the case of
someone reusing the compilation line output by the Makefile created by
configure, with -DHAVE_CONFIG_H on it while the config.h file is not
there (manually, or in a script or for a build system).  This is clearly
an improbable situation, but if it happens it will be very confusing.

-- 
Pat