Re: [PATCH] meson.build: Support ncurses on MacOS

[Daniel P . Berrangé]

On Tue, Nov 09, 2021 at 06:49:57PM +0100, Stefan Weil wrote:
> Am 15.06.21 um 03:53 schrieb Brad Smith:
> 
> > On 6/14/2021 1:45 AM, Philippe Mathieu-Daudé wrote:
> > > On 6/13/21 8:33 AM, Stefan Weil wrote:
> > > > Am 13.06.21 um 03:40 schrieb Brad Smith:
> > > > 
> > > > > This same problem also applies to OpenBSD as we have the same
> > > > > version of ncurses with support for wide characters. I have a similar
> > > > > patch in our QEMU port.
> > > > 
> > > > Then we should either extend the conditional statement to handle
> > > > OpenBSD
> > > > as well, or simply define both macros unconditionally:
> > > > 
> > > >      # Newer versions of curses use NCURSES_WIDECHAR.
> > > >      # Older versions (e. g. on MacOS, OpenBSD) still require
> > > > _XOPEN_SOURCE_EXTENDED.
> > > >      curses_compile_args = ['-DNCURSES_WIDECHAR=1',
> > > > '-D_XOPEN_SOURCE_EXTENDED=1']
> > > > 
> > > > Defining only _XOPEN_SOURCE_EXTENDED would also work with old and new
> > > > versions, so that's another option.
> > > It is simpler to ask Brad to upstream the OpenBSD patch :)
> > 
> > That doesn't answer his question and that's the part that actually
> > matters.
> 
> 
> The question is still unanswered: which alternative is preferred?
> 
> - define only _XOPEN_SOURCE_EXTENDED=1 unconditionally
> 
> - define DNCURSES_WIDECHAR=1 and _XOPEN_SOURCE_EXTENDED=1 unconditionally
> 
> - define DNCURSES_WIDECHAR=1 and _XOPEN_SOURCE_EXTENDED=1 for MacOS and BSD
> 
> All of them should work. We could also start and merge my commit which does
> not fix the issue for BSD but at least fixes it for MacOS.

Ignoring ncurses for a minute

 * QEMU sets _GNU_SOURCE=1
 * With GLibc on _GNU_SOURCE forces _XOPEN_SOURCE=700
 * _XOPEN_SOURCE_EXTENDED=1 is implied by _XOPEN_SOURCE >= 500

IOW for Linux/GLibC, none of this matters, we always get wide char
support via _GNU_SOURCE.

Setting NCURSES_WIDECHAR was added in

  commit b01a4fd3bd7d6f2ebd9eeba9cb6502d423c3bc85
  Author: Peter Maydell <peter.maydell@linaro.org>
  Date:   Fri Jun 2 15:35:38 2017 +0100

    configure: Define NCURSES_WIDECHAR if we're using curses

for the benefit for benefit of non-Linux / non_GLibC platforms
where _GNU_SOURCE doesn't take affect

Note it explicitly rejected the idea of using _XOPEN_SOURCE

[quote]
    We can't fix this by defining _XOPEN_SOURCE ourselves, because
    that also means "and don't provide any functions that aren't in
    that standard", and not all libcs provide any way to override
    that to also get the non-standard functions. In particular
    FreeBSD has no such mechanism, and OSX's _DARWIN_C_SOURCE
    doesn't reenable everything (for instance getpagesize()
    is still not prototyped if _DARWIN_C_SOURCE and _XOPEN_SOURCE
    are both defined).
[/quote]

Based on Stefan / Brad experiance patching in _XOPEN_SOURCE_EXTENDED=1
it seems it might not be too terrible in its effects, but perhaps we're
simply not noticing a silent bad effect.

I also don't like adding a global flag like _XOPEN_SOURCE_EXTENDED=1
in response to a feature that can be turned on/off via a configure
arg --without-curses.

Overall, I don't think we should be setting _XOPEN_SOURCE_EXTENDED=1
globally in CFLAGS in response to curses.

We could put it in any .c file that includes curses.h ie

  #if defined(__APPLE__) || defined(__OpenBSD__)
  #define _XOPEN_SOURCE_EXTENDED=1
  #endif
  #include <curses.h>

so that we isolate the effects of the global flag from the rest
of the QEMU codebase. The meson.build test would also need to
set this in its test program source.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|