Re: Five bugs (IBM-PC systems)

[Pavel Jan Lastovicka]

Tim Van Holder wrote:
> 
> > Pavel Jan Lastovicka <address@hidden> writes:
> >
> > > the following is produced:
> > >    for ac_exec_ext in '' $ac_executable_extensions;
> > >
> > > but ac_executable_extensions is nowhere defined.
> >
> 
> This was added so a DOSish system (or any other system where it is
> relevant) can set ac_executable_extensions in a config.site (as DJGPP
> does).

OK. But then README.DOS (or maybe README-non-Unix) for GNU packages 
should state that one should setup config.site with this variable.
Anyway IMHO the statement should be just
   for ac_exec_ext in $ac_executable_extensions;
so that files without an extension can't match. The variable would be
assigned '' in the case it wasn't defined.

> > >   if test "$CYGWIN" = yes || test "$MINGW32" = yes || test
> > "$DJ_GPP" = yes ; then
> >
> > I don't see this in CVS Autoconf.
> 
> That seems very very old code to me - those variables were the
> result of the old AC_CYGWIN etc checks (which mainly set EXEEXT);
> those were obsoleted when the 'smart' EXEEXT was added to the C
> compiler setup.

As I wrote it's from grep 2.5; configure input and script dated 2002-03-13.
For me it's not an old code. I have used much older grep 2.4 until now, which
configure script was far less problematic...

> > >   if test "$CYGWIN" = yes || test "$MINGW32" = yes || test
> > "$DJ_GPP" = yes ; then

This was generated not by AC_CYGWIN, but by 
  dnl check for the environ separator
  dnl sets SEP
  AM_SEP