Re: getch() waits even nodelay is set

[Fangtu Qiu]

I guess that your ncurses-5.2-31 spec-file is not from RedHat because they 
only have 5.2-28 in their ftp server.

I tried your recent patch which is ncurses-5.2-20020720.patch. It corrects 
my problem. That's great.

- Fangtu


On Sat, 20 Jul 2002, Thomas Dickey wrote:

> On Sat, Jul 20, 2002 at 10:05:42AM -0400, Fangtu Qiu wrote:
> > RedHat 7.3 uses gcc 2.96-110.
>                            ;-)
>  
> > More testing:
> > 
> > Same code works fine on RedHat 7.2 with both gcc and g++. The binary 
> > compiled by g++ on RedHat 7.2 didn't work on 7.3.
> > 
> > Replacing ncurses 5.2-26 on RedHat 7.3 with version 5.2-12 from 7.2 
> > distribution corrects the problem.
> 
> I have a copy of the spec-file for 5.2-31 (and don't see anything relevant).
> They've added several bugs in the terminfo entries, but haven't done
> any damage to the C code that I can tell by reading the patches.
> 
> > I checked the document of stream I/O. It says one should not mix with 
> > stdio. But there is no stream I/O routine in the code I provided. This 
> > confuses me.
> 
> I'm guessing that the stream initialization interferes with ncurses'
> code that changes the terminal input to raw mode.  You may be able to
> see this by using strace.
> 
> unrelated (I think - but related to C++):
> 
> One thing that comes to mind is that (noticed with gcc 3.x) the configure 
> check
> for C++ bool type doesn't work because someone decided to break the prototype
> for exit().  That's "fixed" in current autoconf basically by substituting a 
> new
> bug (making the test more gcc-specific than it should be).  (When I last put a
> patch for autoconf 2.5x out, someone immmediately found a problem in that
> version which is unrelated to my own changes, so I reverted that).
> 
> I don't recall what 2.96.x does for that.  You may want to try compiling
> ncurses (with the current rollup patch) and see if that part of the configure
> script fails for your compiler.  In that case it's using the wrong default
> (unsigned versus char) for bool, which could cause other subtle problems.
>  
>