Re: Is there any notimeout-nodelay interaction?

[Thomas Dickey]

On Thu, Jun 09, 2016 at 08:55:21PM +0200, CLEX wrote:
> Dear ncurses maintainers,
> 
> I'm writing to you because a code that worked fine for about a
> decade recently broke down and I am unable to find the reason for
> it. It might be a bug in the ncursesw library or something that I
> have overlooked. Could you please kindly take a look?
> 
> The initialization basically looks like this:
> 
>     initscr();
>     raw();
>     nonl();
>     noecho();
>     keypad(stdscr,TRUE);
>     notimeout(stdscr,TRUE);
> 
> and the input routine uses get_wch(). It is a blocking read.
> 
> Few weeks ago users who installed a new version of Ubuntu Linux have
> reported problems. I have found out that get_wch() started to behave
> as non-blocking read and was immediately returning ERR value meaning
> no key.
> 
> Commenting out the notimeout() call fixed the problem.

I recall responding to a bug report (it would be nice if this discussion
referenced it, if it's the same one).
 
> However, according to the man page on my system, notimeout()
> controls escape sequence interpretation. The function changing the
> get_wch()'s behaviour from blocking to non-blocking is a different
> one. It's the nodelay() function and that one was not called.
> 
> In my understanding, calling notimeout() should not have the
> described effect, shouldn't it?

The change was here:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805618

and the manpage says:

       The nodelay option causes getch to be a non-blocking call.  If no input
       is ready, getch returns ERR.  If disabled (bf is  FALSE),  getch  waits
       until a key is pressed.

       While  interpreting an input escape sequence, wgetch sets a timer while
       waiting for the next character.  If  notimeout(win,  TRUE)  is  called,
       then  wgetch  does  not  set a timer.  The purpose of the timeout is to
       differentiate between sequences received from a function key and  those
       typed by a user.

What I noticed when investigating this, was that the notimeout value was
copied to a couple of places - but never tested.  Here's how it is used
now:

base/lib_getch.c:493:    if (win->_notimeout || (win->_delay >= 0) || 
(sp->_cbreak > 1)) {
base/lib_getch.c:500:       else if (win->_notimeout)
base/lib_screen.c:127:    DATA(_notimeout, pBOOL),
base/lib_screen.c:566:  nwin->_notimeout = tmp._notimeout;
base/lib_window.c:221:      nwin->_notimeout = win->_notimeout;
tinfo/lib_options.c:127:        win->_notimeout = f;

so... possibly the fix wasn't the best one (there may be a more suitable
way to apply notimeout).  I'll look into that.

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net

signature.asc
Description: Digital signature