Sorry, I'm lying, it is on by default without --enable-sigwinch. Still,
that doesn't mean it's a good idea.


On Sat, Apr 23, 2011 at 08:07:00AM +0100, Nicholas Marriott wrote:
> Hi
> 
> You need to explicitly configure upstream ncurses with --enable-sigwinch
> so many other platforms do not enable it either. They must also have
> been affected when ncurses made USE_SIZECHANGE depend on it. We have
> never had USE_SIGWINCH on.
> 
> 
> On Fri, Apr 22, 2011 at 08:24:34PM +0100, Mikolaj Kucharski wrote:
> > Hi,
> > 
> > I was troubleshooting issue with window resize in net/ekg. In the past
> > it was working properly, but stopped. I figured out that breakage
> > happened between OpenBSD 4.6 and 4.7, when ncurses was updated.
> > 
> > I've recompiled ncurses with -DTRACE and and used it with very simple
> > program to test (see screensize.c under url mentioned below).
> > 
> > Comparing tracing of above code from OpenBSD 4.6 and OpenBSD 4.9
> > (-current) show that in the latest code function _nc_update_screensize()
> > is not executed in doupdate() (lib/libcurses/tty/tty_update.c). It is
> > executed however in OpenBSD 4.6. Reason is that USE_SIZECHANGE is
> > defined as 0 (zero) bacause of following code (curses.priv.h from 4.9):
> > 
> > 
> > #if HAVE_SIZECHANGE && USE_SIGWINCH && defined(SIGWINCH)
> > #define USE_SIZECHANGE 1
> > #else
> > #define USE_SIZECHANGE 0
> > #undef USE_SIGWINCH
> > #define USE_SIGWINCH 0
> > #endif
> > 
> > 
> > HAVE_SIZECHANGE is defined to 1 in OpenBSD 4.6 and 4.9, but USE_SIGWINCH
> > is not defined to 1. In OpenBSD 4.6 only HAVE_SIZECHANGE was checked to
> > set USE_SIZECHANGE to 1 (curses.priv.h from 4.6):
> > 
> > 
> > #if HAVE_SIZECHANGE
> > #define USE_SIZECHANGE 1
> > #else
> > #undef USE_SIGWINCH
> > #endif
> > 
> > 
> > Is there any reason why USE_SIGWINCH is not defined to 1 in current
> > OpenBSD (in ncurses_cfg.h)?
> > 
> > 
> > You can find my debugging files under following location:
> > 
> >  http://www1.kucharski.name/pub/ekg/
> > 
> > - screensize.c -- code to test the issue
> > 
> > - patches for libcurses which enable tracing and adds some additional
> > logging
> > 
> > - trace logs, so you can compare how ncurses on OpenBSD 4.6, unpatched
> > 4.9 and with USE_SIGWINCH defined to 1 on 4.9 behave
> > 
> > 
> > 
> > -- 
> > best regards
> > q#

Reply via email to