Re: lynx-dev Problem with ^Z suspending - so far...

[Klaus Weide]

On Thu, 28 Jan 1999, Henry Nelson wrote:

> > I'll have a look at some ncurses sources and see if there is a
> 
> Can you bang into my head why all this doesn't apply to slang? Or
> does it?

The getting-stopped-twice (I) doesn't happen with slang because it's
aware of the SIGTTOU signal, but ncurses isn't (wasn't) at all:
SIGTTOU doesn't occur at all in the 4.0 ncurses tar.

Speculation: When John Davis started with slang, he used Linux or
something else with POSIX signal calls, so it was natural to think
about blocking signals.  When ncurses started being developed, those
things may not have been (generally) available, or too different on
the various machines.

The cosmetic problem (II) does happen with slang.

Problem III doesn't happen because slang (as used by lynx) doesn't
doesn't try to be too clever.  It knows what tty modes it wants when
it is active, so it just sets those after being fg'd.  In contrast
the ncurses handler tries to pick them up by querying the tty file,
which is where the confusion came in.  It is perhaps understandable
that the ncurses handler has to try to be cleverer, since it's part
of the library and has to deal with more potential situations than
the slang handler (which is part of lynx).

I still have a situation where, even with slang, the tty ends up in
the wrong state after suspending (at the shell prompt), but I see
that only with ksh without line-editing; and who really needs that? :)

    Klaus