On Mon, Mar 13, 2017 at 04:38:17PM +0000, Alan Third wrote:
> On Fri, Mar 10, 2017 at 12:41:17PM +0000, Alan Third wrote:
> > The NS port doesn’t use SIGIO to signal for input (I think it polls
> > using ns_select), and if I comment these two defines out I can remove
> > block_input/unblock_input from various places with no ill effects.
> 
> It looks like this isn’t right. The NS port works without SIGIO most
> of the time, but if wait_reading_process_output is called with
> read_kbd=0 then it goes into an infinite loop. (At least, I think
> that’s what’s happening.)

So I’ve been using the attached patch for a few weeks now with no
issues.

I’ve stopped the signal handler from responding to SIGIO, and the two
places where the NS GUI used to raise SIGIO have been replaced with
direct calls to the SIGIO handler code.

I’m not sure if this is a sensible approach. It fixes at least one
crash, and we could remove some HAVE_NS‐only code, so I think it’s
desirable. I’m just not sure about my implementation. Any thoughts,
anyone?
-- 
Alan Third