Re: Minor build fixes

[Eli Zaretskii]

> Date: Thu, 02 Oct 2014 18:24:49 -0400
> From: Ken Brown <address@hidden>
> Cc: Texinfo <address@hidden>
> 
> > If I understand this right, O_BINARY is being checked to see if it is
> > a Posix-style system, right, but Cygwin has O_BINARY defined? O_BINARY
> > is not actually being set in system.h for the benefit of source code
> > including it? I noticed in info/filesys.c O_BINARY is used for opening
> > a file, so will it matter if this value is 0 under Cygwin?
> 
> Cygwin *allows* files to be opened as "text" or "binary", but this is not 
> necessary.  Everything works fine if all files are viewed as "binary", as in 
> GNU/Linux and other Posix systems.  In particular, it's fine that O_BINARY is 
> 0 
> in info/filesys.c.

If O_BINARY is zero in filesys.c, it means Cygwin will display those
^M characters if the Info file was produced with CRLF end-of-line
format.  If this is not something Cygwin users will be annoyed about,
then indeed Cygwin doesn't need O_BINARY.

> > The problem with setmode is that it is being redefined as "_setmode",
> > correct? Does anyone know what systems need to use _setmode instead of
> > setmode?
> 
> I don't know about other systems, but on Cygwin, setmode is defined to be 
> _setmode, as in system.h.  That's why I was getting (harmless) compiler 
> warnings 
> about the redefinition of setmode.

I believe it's for MSVC and for MinGW when (for some reason I cannot
fathom) _NO_OLDNAMES is defined.

> >> --- info/termdep.h      (revision 5852)
> >> +++ info/termdep.h      (working copy)
> >> @@ -49,7 +49,7 @@
> >>   #  endif /* !HAVE_TERMIO_H */
> >>   #endif /* !HAVE_TERMIOS_H */
> >>
> >> -#ifdef GWINSZ_IN_SYS_IOCTL
> >> +#ifndef __MINGW32__
> >>   #  include <sys/ioctl.h>
> >>   #endif
> >>
> >
> > I see the same thing (conditional on __MINGW32__) is done when
> > including sys/ioctl.h in info/ioctl.h and info/man.h. It looks like
> > GWINSZ_IN_SYS_IOCTL is set by the AC_HEADER_TIOCGWINSZ macro in
> > configure.ac. However, your patch is probably what we want because
> > we're not just including <sys/ioctl.h> to get TIOCGWINSZ (although I
> > don't understand why you would want the definition of TIOCGWINSZ and
> > /not/ have the definition of ioctl, so maybe I'm missing something?)
> 
> I don't understand that either.  In any case, I think it's clear that we need 
> to 
> include <sys/ioctl.h> on most systems, regardless of the value of 
> GWINSZ_IN_SYS_IOCTL.

These considerations omit the MSDOS case, which also doesn't want
sys/ioctl.h.  Not sure about other Windows compilers, or whether they
are still supported.